Fluid Mechanics and Fluid Power: T. Prabu P. Viswanathan Amit Agrawal Jyotirmay Banerjee Editors

Lecture Notes in Mechanical Engineering
T. Prabu
P. Viswanathan
Amit Agrawal
Jyotirmay Banerjee Editors
Fluid
Mechanics
and Fluid
Power
Proceedings of FMFP 2019
Series Editors
Francisco Cavas-Martínez, Departamento de Estructuras, Universidad Politécnica
de Cartagena, Cartagena, Murcia, Spain
Fakher Chaari, National School of Engineers, University of Sfax, Sfax, Tunisia
Francesco Gherardini, Dipartimento di Ingegneria, Università di Modena e Reggio
Emilia, Modena, Italy
Mohamed Haddar, National School of Engineers of Sfax (ENIS), Sfax, Tunisia
Vitalii Ivanov, Department of Manufacturing Engineering Machine and Tools,
Sumy State University, Sumy, Ukraine
Young W. Kwon, Department of Manufacturing Engineering and Aerospace
Engineering, Graduate School of Engineering and Applied Science, Monterey,
CA, USA
Justyna Trojanowska, Poznan University of Technology, Poznan, Poland
Francesca di Mare, Institute of Energy Technology, Ruhr-Universität Bochum,
Bochum, Nordrhein-Westfalen, Germany
Lecture Notes in Mechanical Engineering (LNME) publishes the latest devel-
opments in Mechanical Engineering—quickly, informally and with high quality.
Original research reported in proceedings and post-proceedings represents the core
of LNME. Volumes published in LNME embrace all aspects, subfields and new
challenges of mechanical engineering. Topics in the series include:
• Engineering Design
• Machinery and Machine Elements
• Mechanical Structures and Stress Analysis
• Automotive Engineering
• Engine Technology
• Aerospace Technology and Astronautics
• Nanotechnology and Microengineering
• Control, Robotics, Mechatronics
• MEMS
• Theoretical and Applied Mechanics
• Dynamical Systems, Control
• Fluid Mechanics
• Engineering Thermodynamics, Heat and Mass Transfer
• Manufacturing
• Precision Engineering, Instrumentation, Measurement
• Materials Engineering
• Tribology and Surface Technology
To submit a proposal or request further information, please contact the Springer

Editor of your location:
China: Ms. Ella Zhang at ella.zhang@springer.com
India: Priya Vyas at priya.vyas@springer.com
Rest of Asia, Australia, New Zealand: Swati Meherishi at
swati.meherishi@springer.com
All other countries: Dr. Leontina Di Cecco at Leontina.dicecco@springer.com
To submit a proposal for a monograph, please check our Springer Tracts
in Mechanical Engineering at http://www.springer.com/series/11693 or contact
Leontina.dicecco@springer.com
Indexed by SCOPUS. All books published in the series are submitted for
consideration in Web of Science.
More information about this series at http://www.springer.com/series/11236

T. Prabu · P. Viswanathan · Amit Agrawal ·
Jyotirmay Banerjee
Editors
Fluid Mechanics and Fluid

Power
Proceedings of FMFP 2019
Editors
T. Prabu P. Viswanathan
Department of Mechanical Engineering Department of Mechanical Engineering
PSG College of Technology PSG College of Technology
Coimbatore, Tamil Nadu, India Coimbatore, Tamil Nadu, India
Amit Agrawal Jyotirmay Banerjee

Department of Mechanical Engineering Department of Mechanical Engineering
Indian Institute of Technology (IIT) S. V. National Institute of Technology Surat
Bombay Surat, Gujarat, India
Mumbai, Maharashtra, India
ISSN 2195-4356 ISSN 2195-4364 (electronic)

ISBN 978-981-16-0697-7 ISBN 978-981-16-0698-4 (eBook)
https://doi.org/10.1007/978-981-16-0698-4
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature
Singapore Pte Ltd. 2021
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
Contents
Transient Thermal Stratification in Sodium Pool of a Fast Reactor . . . . . 1

K. Natesan, K. Velusamy, and N. Kasinathan
Mean Flow Properties of a Three-Dimensional Wall Jet Developing
on Concave Cylindrical Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
D. V. S. Bhagavanulu and R. Balamurugan
Computational Analysis of Thrust Generation by an Elastic
Plate Undergoing Active Pitching and Large-Scale Flow-Induced
Deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Sujyesh Aanandh Manjunathan and Rajneesh Bhardwaj
Analysis of Entrance Region Flow of Bingham Nanofluid
in a Concentric Annuli with Rotating Outer Cylinder . . . . . . . . . . . . . . . . . 31
Selvam Mullai Venthan and Isaac Jayakaran Amalraj
Rheodynamic Lubrication of an Externally Pressurized
Converging Circular Thrust Bearing Using Viscoplastic Fluids . . . . . . . . 45
G. Alexander Raymand and I. Jayakaran Amalraj
Experimental Investigation of Flow Over a Backward-Facing Step
with Inclinations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
S. V. Abhijith and R. Abhilash
Experimental Investigations on Effect of Chevrons in Bunsen
Burners . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
Manasa Sadhasivan, V. Priyadarshini, and S. Soma Sundaram
A Computational Study of Entrained Flow Furnace with Swirl
Burner Configuration and Low Turbulence Intensity Flow . . . . . . . . . . . . 73
Nitesh Kumar Sahu, Mayank Kumar, and Anupam Dewan
Numerical Investigation on the Effect of Various Geometries
in a Solar Box-Type Cooker: A Comparative Study . . . . . . . . . . . . . . . . . . . 81
B. C. Anilkumar, Ranjith Maniyeri, and S. Anish
v
vi Contents
Sharp Interface Level Set Method on a Co-located Grid for High

Density Ratio Two-Fluid Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
C. S. Sanjid, Janani Srree Murallidharan, and Atul Sharma
Computational Studies on Combustion Instabilities of Afterburner
for Different Equivalence Ratios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Srinivasa Rao Gurrala and Andavan Shaija
Experiments and Numerical Simulation Sofa Francis Turbine
Model Reflecting Actual Site Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
Rajavamsi Gangipamula, Sasank S. Patnaik, Sudhan A. Kamat,
Shyam Shukla, and Ravindra Birajdar
Computational Analysis on Thermo-Hydrodynamic
Characteristics of Y-Shaped Multi-branched Micro Heat Pipe . . . . . . . . . 119
Asokan Bakhirathan and Lachireddi Gangadhara Kiran Kumar
Mixing in Oscillating Lid Driven Cavity—A Numerical Study . . . . . . . . . 129
Manjappatta Pazhiyottumana Neeraj and Ranjith Maniyeri
Using Dimensional Analysis for Development of Empirical
Correlations for Estimation of Liquid Holdup and Drop Size
in Rotating Disc Contactor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
K. Swetha Saraswathi and P. R. Naren
Effect of Eccentricity on the Thermal Behaviour of A Re-Entry
Vehicle With Concave Windward Surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
V. Vani Praneetha, M. Sudha Madhuri, V. Vijaya Lakshmi,
Karthik Sundarraj, and Prakash S. Kulkarni
Numerical Analysis of Liquid Mixing in a T-micromixer
with Convergent–Divergent Mixing Channel . . . . . . . . . . . . . . . . . . . . . . . . . 153
T. Manoj Dundi, V. R. K. Raju, and V. P. Chandramohan
Performance of Pool Boiling Heat Transfer on Annealed Nanofilm
Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
Sudev Das and R. Johnsan
Computational Study of the Unsteady Wave Drag Reduction
at Hypersonic Mach Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
D. Rajasekhar and Bibin John
Methodologies to Achieve High Performance in Cast Aluminum
Diesel Pistons to Meet BS6 and CAFE Norms . . . . . . . . . . . . . . . . . . . . . . . . 183
Hariprasath Arivukkarasu, Rajkumar Mani, and Selvaraji Muthu
Effect of the Pivot Point Locations on the Wake Dynamics
and Thrust Performance of a Flapping Elliptic Airfoil:
A Numerical Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
Jit Sinha, Nishanth Sreedharan, and Sunil Manohar Dash
Contents vii
Lattice Boltzmann Simulations of a Lid-Driven Cavity at Different

Moving Lengths of the Top Lid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
Prabir Sikdar, Sunil Manohar Dash, and Kalyan Prasad Sinhamahapatra
Shape Modification of Circular Cam Cylinder for Enhanced
Thermal Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
Refiya Hussain, B. Ganesh Kumar, and K. Arul Prakash
A Parametric Study on the Taylor Analogy Breakup Model . . . . . . . . . . . 223
Sumit Joshi, Yashodhan Kadam, and Sushrut Ranade
Effect of Slits in Diffuser Casing of Diffuser Augmented Wind
Turbines(DAWTs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
S. Surya, Jayakrishnan Radhakrishnan, and Amit Kumar
Transition of Stratified to Intermittent Flow Pattern . . . . . . . . . . . . . . . . . . 239
Sunny Saini, Nihar Shah, and Jyotirmay Banerjee
Effects of Partial Slip and Stretching on the Oscillations
of the Velocity Profiles of the Bödewadt Flow . . . . . . . . . . . . . . . . . . . . . . . . . 247
Dip Mukherjee and Bikash Sahoo
Flow and Heat Transfer Characteristics of Oscillatory
Thermo-Capillary Convection in Half Floating Zones . . . . . . . . . . . . . . . . . 255
R. Jayakrishnan and Shaligram Tiwari
Dynamic Contact Angle Formulation for Numerical Analysis
of Taylor Bubble Flow Over Obstacle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
Kathankumar N. Khalasi, Nishant M. Shah, and Jyotirmay Banerjee
Effect of Angle of Incidence on Unsteady Wake Characteristics
for Flow Past Surface Mounted Square Cylinder . . . . . . . . . . . . . . . . . . . . . 275
Prashant Kumar and Shaligram Tiwari
Investigations on Thermal Performance of Spiral Finned Latent
Heat Storage Unit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
Bhavesh V. Vaghela, Digant S. Mehta, Manish K. Rathod,
and Jyotirmay Banerjee
Iterative Three-Dimensional Tomographic Reconstruction
of Flame Using MENT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
Jay Patel, Nandini Rajguru, Bhumi Dhameliya, Preet Tejwani,
and Mitesh Shah
Computational Study of Aerodynamic Performance of a Spinning
Projectile with Different AFT-Body Configurations . . . . . . . . . . . . . . . . . . . 305
E. Aishwarya, T. J. Akhila, S. R. Amrutha, R. P. Sindhu,
Karthik Sundarraj, P. S. Kulkarni, Manoj Veetil, and R. Ganesh Pawar
viii Contents
Mixed Convection MHD Flow of a Nanofluid in a Rotating System

with Heat Generation/Absorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
P. Vimala and K. Manimegalai
RANS Simulations of Flow Past a NACA0012 Airfoil . . . . . . . . . . . . . . . . . 325
Deepak Kumar Singh, Rahul Prajapati, and Arjun Sharma
The Influence of Rotating Element on Flow and Heat Transfer
in the Two-Dimensional Confined Channel . . . . . . . . . . . . . . . . . . . . . . . . . . . 333
J. Ramarajan, D. Sathish Kumar, Ratnanjali Tiwari, and S. Jayavel
Hydrodynamic Effect of Tsunami Wave on Oscillating Water
Column (OWC) Type Wave Energy Converter (WEC) . . . . . . . . . . . . . . . . 343
Rujal D. Patel, Sagar G. Nayak, and Jyotirmay Banerjee
Effect of Orifice Lip Geometry on the Mean Flow Properties
of a Plane Wall Jet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
S. V. H. Nagendra, D. V. S. Bhagavanulu, and Prasant Nanda
Role of Inhibitor Flexibility in the Mitigation of Pressure
Oscillations in Large Solid Rocket Motors . . . . . . . . . . . . . . . . . . . . . . . . . . . 365
Kali Prasad V. A. S. N. Kallakunta, Sankaran Sathiyavageeswaran,
and Satyanarayanan R. Chakravarthy
Numerical Study of Fluid Flow and Heat Transfer in Stepped
Micro-Pin–Fin Heat Sink . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373
Prabhakar Bhandari and Yogesh K. Prajapati
Numerical Investigation of Cavitating Flow over S-Blade
in Acoustic Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383
Pankaj Kumar and R. Bala Anand
Study on the Effective Platooning of Indian Trucks Using
Computational Fluid Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 391
R. Praveen Kumar, S. A. Prabhaharan, and G. Vinayagamurthy
Review of Pool Boiling Heat Transfer During Power Transients . . . . . . . . 399
John Pinto and Janani Srree Murallidharan
Studies on Ignition Over Pressure Characteristics Related
to Typical Launch Vehicle Lift-off . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407
R. Venkatraman, Sathiyavageeswaran Sankaran,
V. Venkata Ramakrishna, Debopam Das, and T. Sundararajan
Revisiting T+ Versus Pry+ Scaling in Turbulent Flows With
0.01 ≤ Pr ≤ 760 Using RANS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 417
Kunal Kumar, Kuldeep Aswal, Manish Kumar Rajwade, Ankit Kumar,
S. S. Mondal, and L. Chandra
Contents ix
Double Diffusive Mixed Convection in a Lid-Driven Enclosure

with the Influence of Shape of Nanoparticle and Magnetic Field . . . . . . . 427
Ratnadeep Nath and K. Murugesan
Effect of Gust Intensity in Flow Past Cube with Built-In Wings
at Low Reynolds Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 435
Ramachandra Vijjapu and Shaligram Tiwari
Design and Fabrication of a Portable Stagnant Water Drainer . . . . . . . . . 445
R. Naveen, T. Prem Kumar, K. Sridharan, and D. Sandheep
Specific Heat of Nanofluids—An Experimental Investigation . . . . . . . . . . 453
Tushar Anand and Soumya Suddha Mallick
Real-Time Detection of Droplet Velocity Using Open-Source
Computer Vision on EWOD Device . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 463
Vandana Jain and Rajendra M Patrikar
Stability Analysis of Two-Phase Slug Flow Using OpenFOAM . . . . . . . . . 471
Rohit Singh Gulia, Siddharth Sharma, and Jyotirmay Banerjee
Characterisation of Steady Flow Regime and Drag Force
on the Forward and Backward Facing Trapezoidal Cylinders:
A Numerical Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 481
Akash Bhunia, Prabir Sikdar, Sunil Manohar Dash, and Kim Boon Lua
Forced Convection Analysis in a Horizontal Pipe in the Presence
of Aluminium Metal Foam—A Numerical Study . . . . . . . . . . . . . . . . . . . . . 491
Prakash H. Jadhav, Banjara Kotresha, N. Gnanasekaran,
and D. Arumuga Perumal
Mathematical Modeling of the Effects of CD147/EMMPRIN
on Tumor Angiogenesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 499
P. Vimala and S. Bhooma
Experimental Investigation on the Effects of Canalization
for Ignition Over Pressure Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 507
V. Venkata Ramakrishna, S. Sankaran, and R. Senthil Kumar
Numerical Investigation of Rotating Convection in a New
Configuration with Bidirectional Thermal Gradients . . . . . . . . . . . . . . . . . 517
Shivam Swarnakar, Ayan Kumar Banerjee, Amitabh Bhattacharya,
and Sridhar Balasubramanian
Thermal Hydraulic Investigations on Fast Reactor Fuel
Subassembly Under Blocked Condition—Using Porous Body
Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 525
S. Srihari, M. Naveen Raj, and K. Velusamy
Multiscale Simulation of Droplet Wall Interaction . . . . . . . . . . . . . . . . . . . . 533
M. Ananthan and Gaurav Tomar
x Contents
Numerical Study of Effect of Asymmetry on Performance

of Bio-mimetic Caudal Fin Shapes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 539
K. L. Vidhu Manohar and Ranjith Maniyeri
Inertial Migration of Neutrally Buoyant Particle in Confined
Channel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 547
Atul M. Bhagat and Partha S. Goswami
A Massively Parallel High-Order ADER-WENO Finite-Volume
Scheme on Unstructured Meshes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 561
Dipak Vaghani, Dasika Sunder, and Ratnesh K. Shukla
Experimental Investigation of Heat Transfer Enhancement
on Impinging Jet Cooling Using Surface Laser Treatment . . . . . . . . . . . . . 571
Maheandera Prabu Paulraj, Santosh Kumar Sahu,
Avadhesh Kumar Sharma, Anuj Kumar, Karthikeyan Sengadurai,
and I. A. Palani
Study on Pulsatile Blood Flow in Cerebral Stenosed Artery . . . . . . . . . . . . 581
Pawan Kumar Pandey, Raghav Agrawal, Parimal Mukul,
and Malay Kumar Das
Heat Transfer and Fluid Flow Characteristics of a Heated
Vibrating Square Cylinder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 589
Mohammad Athar Khan, Syed Fahad Anwer, and Saleem Anwar Khan
Development of Solver for Vibrating Bluff Bodies with Different
Geometrical Shapes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 597
Md.Abu Shahzer, Mohd.Athar Khan, and Syed Fahad Anwar
Floating Solid Waste Collection System Using Free Vortex Flow . . . . . . . 605
Shashikant S. Pawar, Glen L. Dsouza, Oyster J. D´souza,
and Alston R. Furtado
DEM Simulation of Finite Size Particle Impact on a Loose
Granular Bed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 613
B. Gokul and Dhiraj V. Patil
Using Automated Finite Element Framework to Analyze Offshore
Grid Turbine Blades Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 621
Utkarsh Thakre and Asim Tewari
Periodic Transient Performance Prediction of the Rotary Desiccant
Wheel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 631
Selvaraji Muthu and N. Sekarapandian
Study of Hybrid LES/RANS Models in OpenFOAM Applied
to Taylor–Couette Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 639
Sai Guruprasad Jakkala and S. Vengadesan
Contents xi
Investigation of Second Throat Supersonic Exhaust Diffusers

for the Altitude Testing of Rocket Motors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 647
R. Ashokkumar, Sathiyavageeswaran Sankaran, and T. Sundararajan
Aerodynamic Noise from Circular Cylinder in Low Subcritical
Flow Regime Using Large Eddy Simulations . . . . . . . . . . . . . . . . . . . . . . . . . 657
Joemon Jacob and S. K. Bhattacharyya
Review of Experimental Test Rigs for Pump as Turbine Applications . . . 667
Rahulkumar Painter, Ashish Doshi, and Mukund Bade
Assessment of Struts in Gas-Dynamic Steering of a Supersonic
Nozzle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 675
T. Thillaikumar and Mrinal Kaushik
Implementing Canonical CFD Techniques to Determine the Effect
of Optimised Aerodynamic Coefficients on the Performance
of a Car for a Given Race Track . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 683
K. Jayesh Balaji, K. Koushik Sundaram, S. P. Hareesh,
S. Arun Sundar, and R. Naveen
Flow Visualization and Computational Studies on Unsteadiness
Due to 3-D Shock-Induced Separation of Flow Over a Square
Protrusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 693
K. Ramachandra, S. Bhardwaj, and R. Sriram
Performance Analysis of Two-Stage Evaporative Cooler: A Review . . . . . 701
Shubha Deep Paul, Sheetal Kumar Jain, Ghanshyam Das Agrawal,
and Rohit Misra
A Review on Numerous Solar Drying Techniques . . . . . . . . . . . . . . . . . . . . . 709
Dilip Kumar, Sheetal Kumar Jain, Ghanshyam Das Agrawal, and Rohit Misra
Effect of Air Co-flow on Flickering Motion of LPG Laminar
Diffusion Jet Flame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 717
P. Satyanarayana Raju, Guguloth Mahesh Nayak, and S. Balusamy
Natural Convection Through High Porosity Metal
Foams—A Numerical Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 727
Banjara Kotresha, Prakash H. Jadhav, and N. Gnanasekaran
Optimization of Engine Operating Parameters to Meet Global
Emission Norms and CAFE Norms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 737
N. Sekarapandian and Selvaraji Muthu
Enhancement of Heat Transfer Capacity of Heat Pipe Heat Sink . . . . . . . 745
V. Saravanan, K. Rakshith, Vishakh, Nithesh Bhaskar, and Nataraj Badiger
The Forced Convective Heat Transfer from an Isoflux Sphere
for Liquid Metals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 753
Dipjyoti Nath and B. Hema Sundar Raju
xii Contents
Numerical Analysis of Coupled Heat and Mass Transfer in Solid

Desiccant for Stationary Hybrid Desiccants Beds . . . . . . . . . . . . . . . . . . . . . 761
Computational Analysis of Indoor Air Distribution Assessment
on Under-floor Air-Conditioning System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 769
Lakshmanan Ramesh Krishnan and Lachireddi Gangadhara Kiran Kumar
Review and Performance Evaluation of Artificially Roughened
Solar Air Heaters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 779
Ayush Jain, Sheetal Kumar Jain, Ghanshyam Das Agrawal, and Rohit Misra
Experimental Study of Gas-Solid Fluidized Bed Hydrodynamics
Using Optical Fiber Probe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 787
Mona Mary Varghese, Narasimha Mangadoddy, and Teja Reddy Vakamalla
Advection Stabilization Using Lower-Order Scheme Blending:
A Case Study of Rayleigh–Taylor Instability . . . . . . . . . . . . . . . . . . . . . . . . . 795
Rahul Jha, Ashish Arote, and Jyotirmay Banerjee
Modeling and Simulation of an Electro-Hydraulic System Using
Fuzzy Logic Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 807
A. Tony Thomas, S. K. Thangarasu, and T. Sowmithra
Experimental Investigation of Aerodynamic Coefficients of NACA
4412 Aerofoil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 823
Vikas Sharma, Sushanta Dutta, K. Shashi Kumar Naidu, and M. Ajay Kumar
Solar Energy Simulation of Fresnel Lens Concentrated System
for Thermal Electric Generator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 833
S. Babu, R. Sriram, S. Gopikrishnan, and A. Praveen
Flow Visualization of Two-Dimensional Bodies Using Hele-Shaw
Apparatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 841
Mayank Sharma, Shikhar Jaiswal, Sumit Singh Rajput,
Shreya Tripathi, Subiksha C, and G. Vinayagamurthy
Experimental Investigations on Heat Transfer Enhancement
in Double Pipe Heat Exchanger Using PT-SCA and PTT-SCA
Twisted Insert Profile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 853
A. P. Sivasubramaniam, K. Mayilsamy, and P. Murugesan
Characteristics of Upstream Flow in Vertical Inline Pump . . . . . . . . . . . . . 861
P. Srikar Sai Tej and Stephen Christopher
Design Optimisation of Split Case Centrifugal Pump . . . . . . . . . . . . . . . . . . 871
Manoj M. Ugale, Raju N. Patil, and Vishal V. Chavan
Contents xiii
Effect of Reynolds Number on Rotational Behavior of Low Speed

Wind Turbines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 879
P. Viswanathan, T. Prabu, V. Vijai Kaarthi, and J. Archana
Aerodynamic Performance of a Micro Wind Turbine Blade
with S-1223 Airfoil Ascribable the Bionic Bumps on Leading Edge . . . . . 887
T. Prabu, P. Viswanathan, V. Vijai Kaarthi, and J. Archana
About the Editors
Dr. T. Prabu is Professor in the Department of Mechanical Engineering, PSG

College of Technology, India. He has rich experience with academia, industry and
research. He has industry experience with top notch firms like Satyam Venture Engi-
neering Services and Emerson Innovation Centre. He has been an active participant in
conferences related to Purdue Compressor and published research articles in reputed
journals and conferences. Currently, he is involved in establishing state-of-the-art
wind tunnel facility and is pursuing research for government organizations. He is
Secretary of ISSE Coimbatore chapter and a member of ISHRAE and The Insti-
tution of Engineers (India). He is actively collaborating under different research
initiatives of Centre for Advanced Turbomachinery and Energy Research (CATER)
at the University of Central Florida.
Dr. P. Viswanathan is Associate Professor in the Department of Mechanical Engi-

neering at PSG College of Technology, India. He has over 15 years of teaching
experience. He served for over 10 years in machine building and manufacturing
sector and executed several projects in the paper, cement and textiles industry. His
current interests are in wind energy systems and energy efficient HVACR systems. Dr.
Viswanathan is actively involved in establishing a wind tunnel facility for the research
and study of systems related to aerodynamics. His research has been published in
several journals of national and international repute. He is the coordinator for Aero-
nautical Association, ISHRAE Chapter and CWC of ISHRAE Coimbatore Chapter.
Currently, he is working on the development of small wind turbines for the local
environment.
Dr. Amit Agrawal is currently Institute Chair Professor in the Department of

Mechanical Engineering, Indian Institute of Technology (IIT) Bombay, India. His
research interests include micro-scale flows, development of novel bio-microdevices,
theoretical fluid mechanics and turbulent flows. He has published close to 200 journal
articles and filed for a dozen patents with his students. His primary contributions are
in the development of a unique blood plasma separation microdevice and deriva-
tion of equations which are more general than the Navier–Stokes equations. He has
been elected Fellow of the Indian National Academy of Engineering (INAE) and the
xv
xvi About the Editors
National Academy of Sciences India (NASI). He has been awarded the Department of
Atomic Energy DAE-SRC Outstanding Investigator Award, Prof. K. N. Seetharamu
Medal by Indian Society of Heat and Mass Transfer, Prof. H. H. Mathur Award for
Excellence in Applied Sciences by IIT Bombay and the Shanti Swarup Bhatnagar
Prize for his work.
Dr. Jyotirmay Banerjee is Professor in the Department of Mechanical Engineering

at S. V. National Institute of Technology Surat, India. He obtained his Ph.D. from
Indian Institute of Technology (IIT) Kanpur in 2005. He has guided eight Ph.D.
students, and seven more students are presently working under his supervision for
their doctoral degree. He has guided more than 50 M.Tech. dissertations. He has more
than 60 journal publications of international repute. He has been actively involved in
research in the area of two-phase flow through macro and mini-scale channels. His
research group has established a state-of-the-art two-phase flow test rig established
with an earlier SERB sanctioned project.
Transient Thermal Stratification
in Sodium Pool of a Fast Reactor
K. Natesan, K. Velusamy, and N. Kasinathan
1 Introduction
Liquid sodium is used as coolant in fast reactors in view of its desirable neutronic and
thermo-physical properties. Reactor core in which nuclear power generation takes
place is located in a pool of sodium. The size of sodium pool is much larger in pool-
type fast reactors compared to that in a loop-type reactor. This is due to the need to
accommodate all the primary components, viz. pumps and heat exchangers within
the sodium pool itself in the pool-type concept. Pool-type reactors are preferred over
loop-type due to their safety advantages. Sodium coolant circulates in the primary
circuit through core and heat exchangers while transferring heat to other heat transport
circuits in the plant (Fig. 1). During normal operating condition of the reactor, the
upper plenum of the reactor (hot pool) contains sodium at average outlet temperature
of sodium from core. When an undesirable condition develops in the plant, control
rods are dropped into reactor core to stop the fission process and ensure safety. This
action immediately reduces the reactor power, and cold sodium starts emanating from
the reactor core due to loss of heating. Thus, adverse buoyancy conditions would
be developed on the cold sodium flow exiting from core which finds it difficult to
penetrate through the hot sodium prevailing in hot pool resulting thermal stratification
of pool. Due to the large size of pool, this phenomenon is more pronounced in
pool-type compared to loop-type. The thermal interface of the stratification layers is
unstable and leads to fatigue loading of adjoining structures. This issue is of utmost
importance in fast reactors due to the high heat transfer coefficient of sodium leading
to lesser attenuation of coolant temperature fluctuation in structures [1]. Several
theoretical and experimental studies have been carried out worldwide addressing
this phenomenon [2].
K. Natesan (B) · K. Velusamy · N. Kasinathan

Homi Bhabha National Institute, Indira Gandhi Centre for Atomic Research, Kalpakkam 603102,
India
e-mail: natesan@igcar.gov.in
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 1
T. Prabu et al. (eds.), Fluid Mechanics and Fluid Power, Lecture Notes in Mechanical
Engineering, https://doi.org/10.1007/978-981-16-0698-4_1
2 K. Natesan et al.
Fig. 1 Schematic of primary sodium circuit
Thermal stratification in sodium pool can be prevented if inertial force is made

to dominate over buoyancy or the development of adverse buoyancy conditions are
avoided. Typical plant transients involving reactor trip have been analysed using
plant dynamics code. Traditionally, plant dynamics codes adopt one-dimensional
modelling for entire plant with perfect mixing-based single control volume approx-
imation for sodium pools. Such an approach fails to simulate the multidimensional
thermal stratification of sodium pools. Therefore, a coupled modelling approach
by combining multidimensional model for hot pool along with one-dimensional
modelling for rest of the plant similar to those adopted in advanced plant dynamics
codes, viz. Cathare-2 [3], SSC-K [4], etc., have been adopted for the studies reported
in this paper.
2 Methodology
Plant dynamics codes intended to simulate the thermal hydraulic transient behaviour
of entire plant during various events are based on system dynamics approach which
adopts one-dimensional models. The main focus of these codes is the appropriate
representation of thermal hydraulic phenomenon involved in various systems and
components, viz. neutronic power generation in core, heat transfer across various
heat exchangers, heat and momentum transport in coolant circuits, performance of
pumping systems, etc. The codes which are intended to address thermal stratification
behaviour of sodium pools adopt multi-zone or CFD-based multidimensional models
for representing the thermal hydraulics of sodium pools. Detailed studies carried
Transient Thermal Stratification in Sodium Pool of a Fast Reactor 3
out earlier [5] have revealed that two-dimensional (2D axi-symmetric) modelling
approach for hot pool is adequate to accurately represent the thermal stratification
conditions in the hot pool of a fast reactor. Therefore, a coupled code system named
DYANA-HM with 2D model for hot pool and one-dimensional (1D) models for
rest of the plant has been developed for carrying out plant dynamics studies for a
pool-type fast reactor.
One-dimensional models in the code represent, (i) neutronic heat generation in the
core based on point kinetics model with prompt jump approximation, (ii) hydraulic
models based on integral momentum balance and mass balance, respectively, to esti-
mate flow and level evolution in primary and secondary sodium circuits, (iii) heat
balance models for core, intermediate heat exchangers (IHX) and steam generator
(SG) for simulating the heat exchange involved in them and (iv) torque balance
combined with pump characteristics modelling to simulate the operation of sodium
pumps. All the modules of the code are integrated to simulate thermal hydraulic
consequences in the whole plant during various transients. Axi-symmetric model
for hot pool is developed by solving governing equations for continuity, momentum
balance and energy balance along with adequate treatment of turbulence. The generic
k-ε-based turbulence model is adopted. Effect of buoyancy on sodium flow in
the pool is modelled through Boussinesq approximation with appropriate source
term considered in the axial momentum balance equation. The governing equations
are discretized through finite volume method adopting staggered mesh approach
for continuity and momentum balance equations. Pressure—velocity coupling is
achieved through SIMPLE algorithm. The schematic of axi-symmetric model of hot
pool is shown in Fig. 2. Pressure loss offered by porous structures in the hot pool
is simulated through negative source terms in the momentum balance equation, and
the flow entering the IHX through the inlet window is simulated thorough mass sink
in the continuity equation. Free surface is modelled as a slip wall, and other walls
are modelled as non-slip boundaries. The one-dimensional formulation adopted in
the code has been validated against commissioning experiments performed in the
Indian Fast Breeder Test Reactor and end of life studies carried out in the French
fast reactor PHENIX. The two-dimensional code has been validated against standard
benchmark problems reported in the literature and code to code validation (against
commercial CFD code Star CD) with respect to thermal hydraulic prediction of large
size pools. Flow and temperature evolution at the core outlet predicted by the 1 D
model are used as input boundary condition for the 2D model of hot pool and the
temperature evolution of sodium predicted at the inlet of IHX by the 2D model is
used as boundary condition for the 1D modelling of thermal hydraulics of IHX in
the coupled code.
3 Results and Discussion
Two typical plant transients, viz. reactor SCRAM and power failure, have been anal-
ysed as part of this study. In both the cases, reactor gets tripped. Continued cooling of
4 K. Natesan et al.
Fig. 2 Schematic model of hot pool
reactor core through steam generators will be available after reactor SCRAM event.
But, cooling of the core through steam generators for ½ h may not be available after
power failure. After ½ h, decay heat removal systems will get deployed and core
cooling will be possible through them.
3.1 Reactor SCRAM
During reactor SCRAM, control rods are dropped into core due to which power
reduces rapidly to decay power levels. Along with reactor SCRAM, automatic coast
down of speeds of sodium pumps to 20% level (known as sympathetic safety action)
is also adopted. Predicted velocity and temperature distribution in hot pool during
normal operating conditions of the reactor and at 600 s after SCRAM are shown in
Figs. 3 and 4, respectively. It can be observed that hot pool is in well-mixed condition
Fig. 3 Velocity and temperature profile in hot pool during normal operation
Fig. 4 Velocity and temperature profile in hot pool at 600 s after SCRAM
during normal operating conditions, and it gets stratified after SCRAM. The location
of the stratification interface is near the core top in the beginning of the transient, and
it moves upwards as time progresses. At 600 s, the stratification interface is located
near IHX inlet.
Hot pool gets stratified after SCRAM due to two reasons, viz. (i) adverse buoy-
ancy conditions developed due to cold sodium entering into sodium pool at higher
temperature and (ii) reduction sodium flow rate the circuit (sympathetic safety action)
resulting in reduced inertial effects. The first condition is unavoidable as reactor trip
action is essential to ensure safety when undesirable conditions develop in the plant.
6 K. Natesan et al.
Fig. 5 Velocity and temperature profile in hot pool at 600 s after SCRAM without Sy. SA
However, the sympathetic safety action (Sy. SA) is optional. Analysis has been carried
out without sympathetic safety action being implemented after SCRAM. It can be
observed from Fig. 5 that stratification is absent in this case. Higher inertial force
of sodium maintains the pool in well-mixed condition. However, Sy. SA retards the
rate of reduction of hot pool temperature, thereby reducing cold shock.
The predicted rate of reduction of sodium temperature at the inlet of IHX in the
cases with and without sympathetic safety action, which is a representative of the
cold shock seen by the components in hot pool, is shown in Fig. 6. It can be observed
that Sy. SA helps in reducing the thermal shock from ~110 °C/min to 30 °C/min. The
duration for which the rate of temperature reduction remains more than 10 °C/min
is the same in both the cases. Thus, though the Sy. SA results in development of
thermal stratification in hot pool for about 15 min after reactor SCRAM, a significant
reduction in thermal shock faced by components is the main benefit of this procedure.
Moreover, when Sy. SA is implemented, the location of stratification interface keeps
shifting from bottom to top of hot pool. Hence, the associated thermal fatigue effects
would not be significant. Thus, considering the thermo-mechanical benefits of Sy.
SA (reduced shock), this procedure is desirable to be implemented.
3.2 Power Failure
During power failure event, reactor trips and primary pumps run down to 17% speed
(driven by emergency power). Cooling through steam generator will not be available,
and secondary sodium circuit will be under natural convection mode during this
event. Predicted flow and temperature profiles in hot pool at 360 s after SCRAM are
shown in Fig. 7. It can be observed that pool is thermally stratified due to the adverse
Fig. 6 Rate of reduction sodium temperature at the inlet of IHX after SCRAM with and without
Sy. SA
Fig. 7 Velocity and temperature profile in hot pool at 360 s after power failure
8 K. Natesan et al.
Fig. 8 Velocity and temperature profile in hot pool at 360 s after power failure and pump trip
buoyancy effects developed on the flow exiting from core. This can be prevented if
the core flow rate is reduced in proportion to the decay power produced in the core.
Therefore, analysis has been carried out by tripping the sodium pump along with
power failure. The flow and temperature profile in hot pool at 360 s in this case are
shown in Fig. 8. The temperature evolution in the pool is governed by diffusion rather
than convection due to low flow rate (~3%) developed in the circuit. Sodium flow
coming out of core is hotter than that present in pool. Due to this, buoyancy causes
sodium flow from core to rise up along the control plug shell. Convective mixing
region in the pool is confined to a small volume close to the control plug shell.
Though thermal stratification is avoided, the core cooling after the event is delayed
in this case. Hence, from safety considerations, this scenario is not desirable and
hence emergency power provision is essential for pumps though it causes thermal
stratification of pool.
4 Conclusions
Analyses of typical plant transients in a sodium cooled fast reactor using a coupled
code combining 2D model for hot pool and 1D models for rest of the plant show
that the development of transient thermal stratification effects in hot pool can be
prevented by eliminating sympathetic safety action and emergency power provision
for primary sodium pumps. The above implementations lead to increased cold shock
and delayed cooling of core which are not desirable from safety considerations.
Hence, these are not desirable to be implemented in the plant.
References
1. Tenchine D (2010) Some thermal hydraulic challenges in sodium cooled fast reactors. Nucl Eng
Des 240:1195–1217
2. Bieder U et al (2018) CFD analysis and experimental validation of steady state mixed convection
sodium flow. Nucl Eng Des 326:333–343
3. Tenchine D et al (2012) Status of CATHARE code for sodium cooled fast reactors. Nucl Eng
Des 245(2012):140–152
4. Lee Y-B et al (2002) Development of a two-dimensional model for the thermohydraulic analysis
of the hot pool in liquid metal reactors. Ann Nucl Energy 29:21–40
5. Natesan K, Velusamy K (2019) Coupled system dynamics and computational fluid dynamics
simulation of plant transients in sodium cooled fast reactors. Nucl Eng Des 342:157–169
Mean Flow Properties
of a Three-Dimensional Wall Jet
Developing on Concave Cylindrical
Surfaces
D. V. S. Bhagavanulu and R. Balamurugan
1 Introduction
A wall jet is formed when a fluid issues on to a surface from any orifice (Glauert
[1]). Wall are used in the boundary layer control, effective film cooling of gas turbine
walls, window de-icing, paper drying and toughening of glass, etc. Wall jets are also
formed in many practical situations such as the flow over after-bodies downstream
from the nozzle of a fan jet engine, and the impingement of the fluid from the
burners on furnace walls and the surface discharge of warm water jets in cooling
water systems. In the field of hydraulic structures, wall jets occur in the flow under
deeply submerged sluice gates where the gate covers the whole or part of channel
width. The definition sketch of the wall jet is shown in Fig. 1.
The effect of curvature on wall jet is studied in two approaches: (1) The radius of
curvature of the surface is kept constant (Wilson and Goldstein [2]; Kobayashi and
Fujisawa [3]; Fujisawa and Shirai [4]; Gowda and Durbha [5] and (2) the curvature
parameter (B/R; Fig. 2) is maintained constant along the length of the curved surface
(Giles et al. [6]; Guitton and Newman [7]). In the present investigation, the radius of
the curvature is kept constant.
Wilson and Goldstein [2] carried out experiments on a circular cylindrical surface
and compared the results with the plane surface for a two-dimensional wall jet.
They varied the curvature parameter from a low value to a high value and ranges
from 0.067 to 1.27. The growth of halfwidth on the convex surface was found to be
more than twice to that on the plane surface in the longitudinal direction. Gowda and
Durbha [5] have made a similar observation in the case of three-dimensional wall jets
when developing on convex cylindrical surfaces. From all the studies, it is observed
that the shape of mean velocity profiles has not been affected with the curvature,
D. V. S. Bhagavanulu (B)
Civil Engineering, REVA University, Bengaluru, India
R. Balamurugan
Civil Engineering, Annamalai University, Annamalai Nagar, India
12 D. V. S. Bhagavanulu and R. Balamurugan
Fig. 1 Definition sketch of a wall jet
Fig. 2 Curvature parameter
but other properties such as growth of length scales, maximum velocity decay and
turbulent intensities have been affected. Fujisawa and Shirai [4] investigated two-
dimensional wall jets developing on concave curved surfaces. In their experiments,
they have provided an initial straight portion of 50d (d is width of the orifice) and
the remaining portion as curved. In their studies, the mean velocity profiles have
exhibited similarity in the longitudinal direction and a significant decrease in the
growth of halfwidth is observed. From the literature, it is observed that no work has
been done in the case of three-dimensional wall jets developing concave surface.
Hence, the present investigation is carried out to find the curvature effects on the
wall jet for concave curvature. The results obtained on the concave curved surfaces
compared with the results obtained on the plane surface.
Mean Flow Properties of a Three-Dimensional Wall … 13
Fig. 3 Experimental arrangement
2 Experimental Arrangement
Measurements were carried out using a jet tunnel as shown in Fig. 3. Air is supplied
from a centrifugal blower and the velocity of the flow is regulated using a dimmerstat
connected to blower. A number of screens have been provided in the settling chamber
to reduce the turbulence. An orifice plate made of mild steel is attached at the end
of the settling chamber and the diameter of the orifice is 10mm. A smooth polished
plate of size 1.4m × 1.7m × 20mm thick made of teak wood is used as a flat surface
to produce the wall jet. Similarly, convex and curved plates have been made with teak
wood. Both the curved surfaces are provided with a straight portion of 20d. Mean
velocities are measured using total pressure probe and three-dimensional velocities
(U, V and W) are measured using a 5-hole prism probe as shown in Fig. 4. It is
observed that the static pressure variation along the flow is negligible. The probe
was calibrated against a standard probe and the confidence level is about 99.2%. To
determine the uncertainity levels of various quantities measured (U, V & W), the
procedure adopted as explained by Kline and Mc-Clintock [8]. The method is based
on a careful specification of the uncertainties in the various primary experimental
measurements. The estimated overall uncertainty levels in the various quantities
measured with 20:1 odds are U = ±1.8%; V = ±2.4% and W = ±2.1%.
Initially, measurements have been made on the plane surface for comparison purposes
and to find the curvature effect on mean flow properties. Also, to find out the extent of
characteristic decay region [9] for a three-dimensional wall jet in the present study.
From the measurements on the plane surface, it is concluded that the characteristic
Fig. 4 .
decay region extending up to a distance of 20 times diameter of the orifice from the
exit of the orifice. The measurements have been extended up to an axial distance of
60d in the longitudinal direction both on the plane and concave curved surface in the
present investigation. Measurements have been carried out both in the longitudinal
and in the spanwise directions. The same procedure has been applied in the case of
curved surface also. In the present investigation, the following parameters have been
found; (a) the decay of the maximum velocity, (b) the three-dimensional velocity
profiles in the longitudinal and spanwise directions, (c) the growth of halfwidth of
wall jet. The results obtained at an exit Reynolds number are Re = 5.48 × 104 .
3.1 Decay of Maximum Velocity
The decay can be expressed in a power law form, i.e., (Um/Uj) ∝ (x/d)−n ; where
Uj is the jet exit velocity. Figure 5 shows the decay of the maximum velocity in the
longitudinal direction. It is found that the decay is slower on concave curved surface
when compared with the plane surface. The decay exponent (−n) is found to be 1.04
on the plane surface, 0.974 on concave curved surface. From the figure, it is observed
that the decay rate is similar up to x/d = 14, and later the decay is becoming slower
on concave curved surface. The lower value of decay exponent on concave curvature
is attributed to destabilizing nature of the surface in the inner layer of the wall jet.
Fig. 5 Decay of maximum 1

velocity in the longitudinal
direction
0.8
0.6
Um/Uj
PS
CVCYS
0.4
0.2
0
0 20 40 60
x/d
3.2 Mean Velocity Profiles in the Longitudinal Direction
The normalized mean velocity profiles on the curved surface in the longitudinal
direction are shown in Fig. 6 and compared with the mean curve on the plane surface.
It is seen that good similarity is observed between the mean curve on the plane surface
and the mean velocity profiles on the concave curved surface in the longitudinal
direction. And on the curved surfaces. Though there appears to be overall agreement
in the shape of the mean velocity profiles for the curved surface, a close observation
reveals that there is indeed some difference near the point of maximum velocity and
Fig. 6 Mean velocity

profiles in the longitudinal
direction
near the outer edge. Similar observation was made by Kobayashi and Fujisawa [10]
in the case of wall jet on concave curved surface. The effect of curvature makes the
profiles different in the outer region, but the overall profiles were not affected by the
curvature. It is also seen that the position of maximum velocity is slightly shifted for
concave curved wall jet. Measurements are also carried out to find the variation of
three-dimensional velocities V and W in the longitudinal directions using 5 whole
probe. Figures 7 and 8 show the variation of V and W in the longitudinal direction.
The position of maximum intensity of V is found to be at y/b= 0.8 and W is at y/b=
0.6. It is found that not much variation in the quantities of V and W is observed
in the longitudinal direction. It is also found that no effect of concave curvature on
the three-dimensional velocities. However, the trend of the profiles is remaining the
same both on the curved surfaces and plane surface. It is concluded that the effect
of wall is not felt on the V and W in the longitudinal direction, and the shape of the
Fig. 7 Variation of V in the

longitudinal direction
Fig. 8 Variation of W in the

longitudinal direction
Fig. 9 Mean velocity

profiles in the spanwise
direction
profiles resembles free jet profile and this is mainly due to stretching of vortices in
the spanwise direction.
3.3 Mean Velocity Profiles in the Spanwise Direction
The mean velocity profiles in the spanwise direction are shown in Fig. 9. It is observed
that in the spanwise direction, hardly any difference is observed between the results
for the curved surface and the plane surface. It is also observed that the results of
Fujisawa and Shirai [4] are well satisfied with the free jet solution. However, there
is a variation of U in the outer tip of the profile.
Figures 10 and 11 show the variation V and W in the spanwise direction. From
the figures, it is observed that minor variation is observed in V, whereas not much
variation, observed in W. Hence, it can be concluded that not much effect of curvature
on the mean velocities in the spanwise direction.
3.4 Variation of Length Scales
The variation of various length scales (b/d, ym/d and zm/2/d) is shown in Fig. 12. It is
seen that the growth of halfwidth (b) in the longitudinal direction on concave curved
surface is lower compared to the plane surface results. Similar observation is made
in the thickness of the inner region (ym). It is due to rolling of vortices in the outer
layer of the wall jet on concave curved surface. It is well-known fact that in the outer
region of concave curved wall jet more stability prevails, and hence, the growth of
halfwidth further reduced when compared to the results on the plane surface. The
Fig. 10 Variation of V in the

spanwise direction
Fig. 11 Variation of W in
the spanwise direction
growth of halfwidth in the spanwise direction is remaining same on both the plane
and concave surface. A similar feature is observed by Kobayashi and Fujisawa [10]
for two-dimensional wall jets. This is mainly attributed to destabilizing nature of the
concave surface [11, 12] in the inner layer of concave curved wall jet and the fluid
layer move close to the wall when compared to movement on the plane surface, and
hence, the reduction in the maximum velocity position is observed.
4 Concluding Remarks
1. The similarity of the mean flow velocity profiles normal to the surface in the
plane of symmetry appears to be not much affected by the curvature effects.
Fig. 12 Variation of length

scales in the longitudinal
direction
However, there is a marginal shift of the position of the maximum velocity

due to the concave curvature effects. The shape of the velocity profile in the
spanwise direction is seen to be unaffected due to curvature.
2. The decay rate of the maximum velocity is slower on concave surface when
compared to plane surface. The decay exponent is 0.0974 on concave curved
surface and it is 1.04 on the plane surface.
3. The three-dimensional velocities V is slightly higher on concave surface and no
change in the shape of the profiles and magnitude is observed in the case of W
on curved surface when compared to plane surface both in the longitudinal and
spanwise directions.
4. The growth of length scales normal to the plate on concave surface is lower
when compared to the growth on the plane surface. However, the growth of
the length scales in the spanwise direction remains unaltered due to curvature
effects.
References
1. Glauert MB (1956) The wall jet. J Fluid Mech. 1:625–643

2. Wilson DJ, Goldstein RJ (1976) Turbulent jets with cylindrical streamwise surface curvature.
J Fluids Eng Trans ASME 98:550–557
3. Kobayashi R, Fujisawa N (1983) Turbulence measurements in wall jets along strongly concave
surfaces. Acta Mech 47:39-52
4. Fujisawa N, Shirai H (1987) Theoretical and experimental studies of turbulent wall jet along a
strongly concaved surface. Trans Japan Soc Aero Space Sci 30:26–37
5. Gowda BHL, Durbha VSB (1999) Mean and turbulence characteristics of three dimensional
wall engineering. Trans ASME 121:596–604
6. Giles JA, Hays AP, Sawyer RA (1966) Turbulent wall jets on logarithmic spiral surfaces. Aero
Quart 17:201–215
7. Guitton DE, Newman BG (1977) Self-preserving turbulent wall jets over convex surfaces. J
Fluid Mech 81:155–185
8. Kline SJ, McClintock FA (1953) Describing uncertainties in single sample experiments
9. Sforza PM, Herbst GA (1970) A study of three- dimensional turbulent wall jet. J Fluid Mech
10:481–495
10. Kobayashi R, Fujisawa N (1983) Curvature effects on two–dimensional turbulent wall jet along
concave surfaces. Bull JSME 26:2074–2080
11. Fujisawa N, Kobayashi R, Shirai H (1985) Centrifugal instability of turbulent wall jets along
strongly concave surfaces. J Fluids Eng Trans ASME 52:492–493
12. Fujisawa N, Shirai H (1986) on the stability of turbulent wall jets along concave surfaces. Bull
JSME 29:3961–3766
Computational Analysis of Thrust
Generation by an Elastic Plate
Undergoing Active Pitching
and Large-Scale Flow-Induced
Deformation
Sujyesh Aanandh Manjunathan and Rajneesh Bhardwaj
1 Introduction
Interaction of an elastic plate with fluid flow has potential applications in designing
underwater vehicles, micro-aerial vehicles, and energy-harvesting devices. Here,
the plate may be subjected to active pitching and may undergo a large-scale flow-
induced deformation, which could be able to generate thrust on the plate. It is well-
established that the rigid pitching foils produce a thrust at a certain pitching frequency
and Reynolds number [1–3]. As pointed out by several researches, this occurs due
to the generation of jet-like flow along the foil with positive average velocity profile
and a reverse Karman vortex street in the wake of the pitching foil. The latter is the
signature of this thrust.
Studies have been further extended to the pitching of flexible plates, wherein the
influence of mass ratios, flexural rigidity, and reduced frequency had been inves-
tigated [4–6]. The resonant frequency for pitching was reduced significantly, as
compared to a rigid plate. The role of the flow-induced deformation and Reynolds
number, however, has not been fully explored in this context.
2 Methodology
2.1 Numerical Details
The governing equations are Navier–Stokes equations for viscous, incompressible,

and unsteady fluid flow and are given in dimensionless form as follows:
S. A. Manjunathan · R. Bhardwaj (B)

Department of Mechanical Engineering, Indian Institute of Technology, Bombay, India
e-mail: rajneesh.bhardwaj@iitb.ac.in
22 S. A. Manjunathan and R. Bhardwaj
∂νi
=0 (1)
∂ xi
∂νi ∂νi ν j ∂p 1 ∂ 2 νi
+ =− + (2)
∂τ ∂x j ∂xi Re ∂ x 2j
where i, j = 1, 2, 3 and ν, p, τ and Re denote dimensionless velocities, pressure,

time, and Reynolds number, respectively. The structural dynamics is governed by
Saint Venant Kirchhoff model, which accounts for geometric nonlinearity of an
elastic structure. An open-source finite element-based structural solver Tahoe [7]
was coupled with immersed boundary method based flow solver. This solver uses
an implicit coupling for faster solution convergence. The details of the coupling are
described in our previous works [8, 9] and the solver has also been validated in these
studies.
2.2 Simulation Setup
We study FSI of an elastic plate of uniform cross-section, pitching about its leading
edge (point O), as shown by a schematic in Fig. 1. The dimensionless thickness and
length of plate are 0.1 and 1, respectively. The pitching of the plate is governed by
the following equation of the motion,
Fig. 1 Schematic, computational domain, and boundary conditions of the problem considered in
the present work. Pitching of deformable plate of length L and thickness h about it leading edge is
considered
Computational Analysis of Thrust Generation by an Elastic Plate Undergoing Active … 23
θ (τ ) = θ0 sin(2π kτ ) (3)
where θ0 and k denote the amplitude of pitching (in radians) and reduced (dimen-
sionless) frequency, respectively. The computational domain along with the boundary
conditions is also shown in Fig. 1. Here, we used grid with 256 × 256 points in a 17L
× 12L domain. The uniform grid thickness in both x and y directions near the plate
is fixed to be 0.015. The plate thickness is discretized using 200 finite elements. The
Poisson ratio is fixed at 0.4 in the present study.
2.3 Performance Parameters
In order to quantify the propulsive performance of the plate, we use the following
quantities. The time-varying coefficient of thrust (CT ), coefficient of power (CP ), and
propulsive efficiency (η) is defined as follows,
2Fx
CT = − (4)
ρU∞
2 L
2P
CP = (5)
ρU∞
3 L
CT
η= 100% (6)
CP
where Fx and P is net force acting in the horizontal direction on the plate and power
fed for the plate motion. This force comprises of two components, namely pressure
(CT,P ) and shear (CT,S ) force. The time-averaged coefficients of thrust and power over
last 10 cycles of the plate oscillation are denoted by CT and CP , respectively.
3.1 Mechanism of Thrust Generation
The mechanism of thrust generation of rigid and elastic/flexible plate is explained

here, with the plates pitching at an amplitude of θ0 = 0.25 radians and at reduced
frequencies (k) of 6/π and 4/π, respectively. The mass ratio (m) and dimension-
less flexural rigidity (R) of flexible plate are fixed to be 0.5 and 8, respectively. Re
is considered to be 150. Other parameters are maintained the same throughout the
study. The temporal variations of the thrust as well as power coefficient for the rigid
and elastic plate have been plotted in Fig. 2 and Fig. 3, respectively.
Rigid Plate Pitching at k = 6/π .

As plotted in Fig. 2a, the temporal variation of trailing edge tip displacement in
Y-direction, YTip , is sinusoidal, as defined by Eq. (3). However, CT does not follow
a similar trend. CT increases initially from τ /T = 7 to 7.25, while the plate moves
from its mean position to the top extremum in this period. Thereafter, CT decreases
from τ /T = 7.25 to 7.5 and the plate returns back to the mean position from the top
extremum. At τ/T = 7.5, CT is minimum and is negative with dominance of shear
component, creating a drag on the plate. CT then increases to a positive maximum
around τ /T = 7.8 and further decreases when the plate returns back to the mean
position. CP signal plotted in Fig. 2b shows that the power fed to the plate increases
as thrust decreases and vice-versa.
CT (Eq. 4) produced is approximately + 1.4, exhibiting a net thrust on the plate.
CP (Eq. 5) obtained is around 20.5 and η (Eq. 6) produced is + 6.8%. The signature
of thrust production is present in vortex shedding pattern (Fig. 4, left column). The
developed vortex pair when shed along with progressive diffusion in fluid domain
imparts a jet-like flow along the pitching plate, thus imparting thrust to the plate due
to an excess momentum. The mode of vortices developed here due to the pitching is
S + P type [10] and is shown in Fig. 4a–d.
Elastic/Flexible Plate Pitching at k = 4/π .
In this case, CT , plotted in Fig. 3a, starts dropping from τ /T = 7.1 to 7.3, during
which the leading edge of the plate reaches the top extremum above the mean position.
Thereafter, CT attains maximum when the leading edge of the plate reaches its mean
position from top extremum. After this, the plate reaches the bottom extremum which
again reduces thrust till τ /T = 7.8. CT further increases when the plate reaches its
mean position at the end of the pitching cycle. The temporal variation of coefficient of
power for the flexible plate is plotted in Fig. 3b. CT produced is around +2.4 while
CP is computed to be around 24.9. The resulting η for this plate is around 9.7%.
The mode of vortices shed is also observed to be S + P type [10] and is shown in
Fig. 4e–h.
Rigid versus Elastic Plate
In this section, we discuss comparison of time-varying signals of CT for rigid
and elastic plate. In a typical pitching cycle, CT increases for the latter case when
the trailing edge of the plate returns back to mean position from top and bottom
extremum, as compared to that in the rigid case. The trend of time-varying CP for
the two cases is, however, similar and the power required to pitch is maximum when
the thrust reaches minimum in a cycle. Moreover, the ratio of maximum thrust to
maximum drag achieved is higher for flexible plate (≈1.9) when compared to that
of the rigid plate (≈1), showing dominance of thrust in the flexible plate. Both rigid
and flexible plate show similar trend of vortex shedding mode, with the difference
being the lag in the shedding—rigid plate leads flexible plate in vortex generation by
nearly τ /T = 0.2. A fully developed vortex pair sheds when the plate pitches from
mean position to top extremum. Development of a positive vortex starts when the
plate returns back from the top extremum. The intensification of the positive vortex
continues till the plate reaches the bottom extremum. During the return to mean
Fig. 2 a Temporal variation of CT , with CT,P and CT,S and YTip of rigid pitching plate pitching at k
= 6/π for three pitching cycles from τ /T = 6 to τ /T = 9. Dotted lines correspond to time instances
τ /T = 7, 7.2, 7.5 and 7.8; b Temporal variation of CP for three oscillation cycles from τ /T = 6 to
τ /T = 9
Fig. 3 a Temporal variation of CT with CT,P and CT,S and YTip of elastic pitching plate pitching
at a k = 4/π for three pitching cycles from τ /T = 6 to τ /T = 9. Dotted lines correspond to time
instances τ /T = 7.1, 7.3, 7.6 and 7.8; b Temporal variation of CP for three oscillation cycles from
τ /T = 6 to τ /T = 9
Fig. 4 Comparison of vorticity fields around rigid (left column) and elastic (right column) pitching
plate in a typical oscillation cycles at time instants (mentioned in Figure) where trailing edge of
both plated nearly match
position, a negative vortex starts developing and intensifying, finally creating an S

+ P vortex, ready for shedding.
The reason for a different trend of the transient variation of the thrust for rigid
and flexible plate is due to the deformation of the flexible plate. When flexible plate
passes above/below the mean position, it remains in a deformed state due to its inertia.
Once the plate starts returning to the mean position, the plate adapts and deforms
according to the flow direction.
3.2 Comparison Between Propulsive Performance of Rigid

and Elastic Plate at Different Frequencies
We compare the propulsive performance of the rigid and elastic plates as a function
of reduced frequency, k, varying it from 2/π to 8/π . As shown in Fig. 5a, trailing edge
displacement (ATip ) of the rigid plate remains constant and the edge displacement
of the elastic plate matches with that of the rigid plate till k = 3/π , since at these
low frequencies the flow-induced deformation of the elastic plate is smaller. ATip
increases steadily till k = 7/π and falls beyond this frequency due to the onset of
second mode of vibration. CT (Fig. 5b), CP (Fig. 5c) and η (Fig. 5d) show a similar
trend for the flexible plate. Note that the frequency at which the peak propulsive
efficiency is obtained, is lower for an elastic plate as compared to the rigid plate.
Fig. 5 Comparison between computed a ATip b CT ; c CP , and d η against k for the elastic and
rigid pitching plate. m and R for the elastic plate are 0.5 and 8, respectively
3.3 Impact of Reynolds Number
Simulations were further carried out for Re of 75 and 300, keeping the same variation
of k. A T i p decreases with an increase in Re due to an increase in magnitude of fluid
forces. The selected values of m and R also correspond to a fluid inertia dominant
structure [4, 5]. The increase in CT with the increase in Re is significantly higher, with
the percentage of increase in CT being nearly 40% and 15% on transition in Re from
75 to 150 and 150 to 300, respectively. This is due to presence of higher flow energy
for the generation of thrust at higher Re. CP remains almost the same at different k
and Re, with the percentage of deviation being within a range of 3.5–6%. Thus, η
increases with an increase in Re, with the percentage of increase nearly being 65 and
10% on increase in Re from 75 to 150 and 150 to 300, respectively. The variations
of ATip , CT , CP and η with respect to k at Re = 75 and 300 were similar to the trends
at Re = 150. The resonant k for all cases of Re was found to be the same (k = 4/π )
for the considered m and R.
4 Conclusions
We have investigated the thrust generation by pitching of a thin rigid and elastic plate
in a free stream laminar flow with an in-house sharp-interface immersed boundary
method-based FSI solver. A larger thrust and propulsive efficiency is achieved in
case of the rigid plate at larger reduced frequency. Whereas for the elastic plate,
a larger thrust and efficiency occurs at a smaller pitching frequency in which the
plate deformation is first natural mode. A peak propulsive efficiency is attained at
a optimal frequency of motion in both cases. The resonant frequency of pitching for
a particular mass ratio and flexural rigidity is found to be independent of Reynolds
number. However, the amount of thrust produced and the magnitude of attainable
propulsive efficiency are directly proportional to the Reynolds number.
Acknowledgements We gratefully acknowledge financial support from the Naval Research Board,
New Delhi, India (Grant No. NRB-403/HYD/17-18). S.A.M. was supported by a fellowship spon-
sored by IRCC, IIT Bombay. The research work reported in this chapter was presented in 46th
National Conference on Fluid Mechanics and Fluid Power (FMFP 2019) at P.S.G. Coimbatore,
Tamil Nadu, India in December 2019. Later, a detailed and extended version of this work appeared as
Ref. [11].
References
1. Das A, Shukla RK, Govardhan RN (2016) Existence of a sharp transition in the peak propulsive
efficiency of a low- Re pitching foil. J Fluid Mech 800:307–326
2. Pedro G, Suleman A, Djilali N (2003) A numerical study of the propulsive efficiency of a
flapping hydrofoil. Int J Numer Meth Fluids 42:493–526
3. Thekkethil N, Sharma A, Agrawal A (2018) Unified hydrodynamics study for various types of
fishes-like undulating rigid hydrofoil in a free stream flow. Phys Fluids 30:077107
4. Yin B, Luo H (2010) Effect of wing inertia on hovering performance of flexible flapping wings.
Phys Fluids 22:111902
5. Zhang Y, Zhou C, Luo H (2017) Effect of mass ratio on thrust production of an elastic panel
pitching or heaving near resonance. J Fluids Struct 74:385–400
6. Zhu X, He G, Zhang X (2014) Numerical study on hydrodynamic effect of flexibility in a
self-propelled plunging foil. Comput Fluids 97:1–20
7. “Tahoe, an open source c++ finite element solver, which was developed at sandia national labs,
ca.,” https://sourceforge.net/projects/tahoe/
8. Bhardwaj R, Mittal R (2012) Benchmarking a coupled immersed-boundary-finite-element
solver for large-scale flow-induced deformation. AIAA 50:1638–1642
9. Kundu A, Soti AK, Bhardwaj R, Thompson MC (2017) The response of an elastic splitter plate
attached to a cylinder to laminar pulsatile flow. J Fluids Struct 68:423–443
10. Zhang G, Ji L (2017) Investigation of two degrees of freedom on vortex-induced vibration
under the wake interference of an oscillating airfoil. Acta Astronaut 133:423–435
11. Manjunathan SA, Bhardwaj R (2020) Thrust generation by pitching and heaving of an elastic
plate at low Reynolds number. Phys Fluids 32(7):073601
Analysis of Entrance Region Flow
of Bingham Nanofluid in a Concentric
Annuli with Rotating Outer Cylinder
Selvam Mullai Venthan and Isaac Jayakaran Amalraj
1 Introduction
Recently, the study of non-Newtonian flow has been a tremendous attraction in the
fluid dynamics research. In particular, non-Newtonian fluid characterized by a yield
value such as Bingham plastics, Casson and Herschel–Bulkley fluids are the special
interest of the tribologists. The applications of non-Newtonian fluids are evident
in polymer devolatization and processing, wire and fiber coating, heat exchangers,
extrusion process, chemical processing equipment, etc. Many important industrial
fluids are non-Newtonian in their characteristics and are referred to as rheolog-
ical fluids. This includes blood, coal water or coal–oil slurries, glues, inks, foods,
polymer solutions, paints, and many others. In this analysis, the non-Newtonian fluid
considered here is the Bingham plastic.
The problem of entrance region flow of non-Newtonian fluids in an annular
cylinder has been studied by various authors. Huilgol [1] studied the minimum pres-
sure drop required to initiate the flow of viscoplastic fluids with constant yield stress,
such as Casson and Herschel–Bulkley fluids in the pipes of the following symmetric
cross sections: circular, annular, square, rectangle, equilateral triangle, elliptical, and
L-shape. Poole and Chhabra [2] numerically analyzed the development of laminar
pipe flow of yield stress fluids obeying the Bingham-type models. They have shown
the development lengths for various values of Reynolds number. Srinivasa Rao and
Kandasamy [3] numerically studied the effects of parameters, aspect ratio, and yield
value on the pressure drop and velocity profiles for the entrance region flow in concen-
tric annuli with rotating inner wall for Bingham fluid. Khali et al. [4] numerically
investigated the effect of porous layer on non-Newtonian fluids between rotating
concentric annular ducts using lattice Boltzmann method.
S. Mullai Venthan (B) · I. Jayakaran Amalraj

Department of Mathematics, SSN College of Engineering, Chennai, Tamil Nadu 603110, India
I. Jayakaran Amalraj
e-mail: jayakarana@ssn.edu.in
32 S. Mullai Venthan and I. Jayakaran Amalraj
Moreover, nanotechnology has been widely used in many industrial applications.

Nanofluids are engineered colloids made up of a base fluid and nanoparticles, i.e.,
nanofluid is the composite of nanosized particles, like copper, silver, gold, carbon
nanotubes, etc., and the base fluids, like water, engine oil, glycerin, etc. Choi and
Eastman [5] were the first one who gave the idea of utilizing composite fluid, namely
nanofluids, they used two different types of nanofluids, copper (copper with water)
and silver (silver with water), that were studied. The reasons for the use of water-
based Ag and Cu are generally regarded as a safe material for humans and animals
although it is recognized that this may change in the future with more fundamental
research on nanotoxicology.
Further, the experimental work on ethylene glycol-based nanofluids containing
with spherical TiO2 nanoparticles and the theoretical analyses on the high shear
viscosity, shear thinning behavior, and temperature dependence are carried out by
Haisheng Chen et al. [6]. This work aims at a more fundamental understanding of
the rheological behavior of nanofluids and the interpretation of the discrepancy. Ali
Esmaeilnejad et al. [7] investigated the advantages of using nanoparticles with partic-
ular size in non-Newtonian fluids. They have computed numerically the convection of
heat transfer and laminar flow of nanofluids with non-Newtonian base fluid through
rectangular microchannels. Hayat et al. [8] discussed the boundary layer flow of
a Casson fluid due to a stretching cylinder in the presence of nanoparticles and
thermal radiation. Here, all physical properties are taken as constant for the Casson
fluid, except the thermal conductivity. Usman et al. [9] investigated numerically the
velocity and thermal slip effects on Casson nanofluid with heat and mass transfer
phenomena over an inclined permeable stretching cylinder by using the collocation
method.
Even though many research works have been carried out separately in entrance
region study as well as in nanofluids by various researchers, the study of entrance
region flow using nanofluids has been very rare. Therefore, in this research article, the
entrance region flow of a Bingham fluid in the presence of nanofluids has been studied
in cylindrical concentric annuli. The focus will be on nanofluids containing TiO2 and
Al2 O3 by using water as a base fluid, and the theoretical analysis is based on our
current understanding of the rheology of Bingham fluid. The analysis has been carried
out under the assumption that the outer cylinder is rotating and the inner cylinder is
at rest. With Prandtl’s boundary layer assumptions, the equations of conservation of
mass and momentum are solved using linearized implicit finite difference technique.
The system of linear algebraic equations thus obtained has been solved by the Gauss-
Jordan method. The present study reports the development of axial, radial, tangential
velocity profiles and pressure drop in the entrance region for different values of
Bingham fluid characteristics, different volume fraction parameters for TiO2 -water
and Al2 O3 -water under various geometrical parameters. The effects of these on the
velocity profiles and pressure drop have been discussed.
Analysis of Entrance Region Flow of Bingham Nanofluid in a Concentric … 33
2 Analysis of the Problem
The flow of two different nanofluids is considered, namely TiO2 -nanofluid and
Al2 O3 -nanofluid, and they are coalescing with Bingham fluid and entering into the
horizontal rotating concentric circular cylinders, from a large chamber with a uniform
velocity v0 and pressure p0 initially. The radii of the inner and outer cylinders are r 1
and r 2 , respectively. Here, ω is the angular velocity for rotating inner cylinder where
the outer cylinder is assuming to be rest. The laminar flow is steady and incompress-
ible, with the physical properties being constants. We consider a cylindrical polar
coordinates system (r, θ , z) fixed at the entry section with r and z along the radial
and axial direction, respectively. The geometry of the problem is shown in Fig. 1.
The constitutive equation for nanofluid with the Bingham fluid is given as
τ0
τi j = (μn f + )εi j , τ ≥ τ0
ε
where

1 1
τ= τi j τi j and ε = εi j εi j (1)
2 2
where τ0 is the yield stress, μn f is the viscosity of the Bingham nanofluid.τi j and εi j
are the stress tensor and the rate-of-strain tensor, respectively. The flow is governed
by the equations
∂ ∂
(r vr ) + (r vz ) = 0 (2)
∂r ∂z
vθ2 1 ∂p
= (3)
r ρn f ∂r

∂(vθ ) ∂(vθ ) vvθ 1 ∂ ∂ vθ
vr + vz + = r 2
τ0 + kr (4)
∂r ∂z r ρn f r 2 ∂r ∂r r
∂(vz ) ∂(vz ) 1 ∂p 1 ∂ ∂vz
vr + vz =− + (r [τ0 + k ]) (5)
∂r ∂z ρn f ∂z ρn f r ∂r ∂r
Fig. 1 Geometry of the

concentric coaxial cylinders
where vr , vθ , vz are the velocity components in r, θ, z directions, respectively.ρn f

is density of the Bingham nanofluid. For the inner cylinder rotation, the boundary
conditions are
For z ≥ 0 and r = r 1 , vr = vz = 0 and vθ = 0.
For z ≥ 0 and r = r 2 , vr = vz = 0 and vθ = ωr2 .
For z = 0 and r 1 < r < r 2 , vz = v0
p = p0 at z = 0 (6)
Using the boundary conditions, the continuity Eq. (2) can be expressed in the
following integral form
r1
2 r vz dr = (r22 − r12 )v0 (7)
r2
The non-dimensionalization of the parameters is as follows

r vz ρn f vr r2 vθ
η= ,U = ,V = ,W = ,
r2 v0 μf ωr1
r1 p − p0 2vz (1 − R)
R= ,P = ,ζ = ,
r2 ρn f v02 r2 Re
τ0 r2 2ρn f (r2 − r1 )v0
B= , Re = ,
kv0 k

2ω2 ρn2 f r12 (r2 − r1 )3 r1
Ta = , where μ f = k
μ2f (r1 + r2 ) ωr2
μf
ρn f = (1 − ϕ)ρ f + ϕρs , μn f = (8)
(1 − ϕ)2.5
where ϕ is the solid volume fraction of the nanofluid, μf is the viscosity of the
fluid fraction, ρf is the reference density of the fluid fraction, and ρs is the reference
density of the solid fraction. By introducing the above non-dimensional parameters,
the governing Eqs. (2)–(5) as well as boundary conditions (6), they become

1 ∂V 1 V 1 ∂U
+ + =0 (9)
(A1 A2 ) ∂η (A1 A2 ) η (A1 A2 ) ∂ζ
W2 Re2 (1 − R) ∂ P
= (10)
η 2(1 + R)T a ∂η

1 ∂W ∂W 1 VW
V +U + =
(A1 A2 ) ∂ζ ∂ζ (A1 A2 ) η
2 (11)
1 ∂ W 1 1 ∂W 1 W 2B 1
+ − +
(A1 A2 ) ∂η2 η (A1 A2 ) ∂η (A1 A2 ) η2 η A2

∂U 1 ∂U ∂P 1 1 ∂U
V +U =− +
∂η (A1 A2 ) ∂ζ ∂ζ η (A1 A2 ) ∂η
2 (12)
1 ∂ U B 1
+ +
(A1 A2 ) ∂η2 η A2
1
2 ηU dη = (1 − R 2 ) (13)
R
where A1 = 1
(1−ϕ)2.5
and A2 = 1
.
(1−ϕ)+ϕ ρρs
f
For ς ≥ 0 and η = 1, V = U = 0 and W = 0.
For ς ≥ 0 and η = R, V = U = 0 and W = 1.
For ς = 0 and R < η < 1, U = 1
P = 0 at = 0. (14)
2.1 Numerical Solution
The numerical investigation and the scheme of evaluation can be considered as a

roundabout expansion work of Srinivasa Rao and Kandasamy [3]. It has been decided
to solve the above system of governing equations, using finite difference method.
The discretized solution space is shown in Fig. 2 in which η and ς correspond
to the grid size along the radial direction and also axial direction, respectively. By
assuming appropriate finite difference definitions, the non-dimensionalized system
of Eqs. (9)–(13) got reduced as

(n+1) 1 (n+1) 1 R + iη
V(i+1) = Vi
(A1 A2 ) (A1 A2 ) R + (i + 1)η

4η 1 2R + (2i + 1)η (15)
−
4ς (A1 A2 ) R + (i + 1)η

(n+1)
∗ Ui+1 + Ui((n+1) − Ui(n+1) − Ui(n)
(n+1) (n+1)
[W 2 ]i(n+1) (1 − R)Re2 Pi − Pi−1
= (16)
R + iη 2T a(1 + R) η
Fig. 2 Mesh work for the

finite difference technique
(n+1) (n) (n+1) (n)

Wi+1 + Wi+1 − Wi−1 − Wi−1 1
Vi(n) ∗ +
4η (A1 A2 )

(n+1)
(n) Wi − Wi(n) Vi(n) Wi(n) 1
Ui + ∗ =
ς R + iη (A1 A2 )

(n+1) (n)

Wi+1 + Wi+1 − 2Wi(n+1) − 2Wi(n) 1
∗ +
2(η)2 (A1 A2 )

(n) (n+1)

Wi−1 − Wi−1 1 Wi(n) 1
∗ − ∗ +
2(η)2 (A1 A2 ) (R + iη)2 (A1 A2 )

(n+1) (n) (n+1) (n)

Wi+1 + Wi+1 − Wi−1 − Wi−1 1
∗
(R + iη)4η (A1 A2 )
(17)
2B 1
+ ∗
R + iη A2

(n+1) (n+1)

(n+1)
(n) Ui+1 − Ui−1 1 (n) Ui − Ui(n)
Vi ∗ + Ui
2η (A1 A2 ) ς

(n+1)

P − Pi(n) (n+1)
Ui+1 (n+1)
− Ui−1 1
= − i+1 + ∗ +
ς (R + iη)2η (A1 A2 )

(n+1)

Ui+1 − 2Ui(n+1) − Ui−1
(n+1)
1
∗
(η)2 (A1 A2 )
(18)
B 1
+ ∗
R + iη A2
where i = 0 at η = R and i = m at η = 1. The application of trapezoidal rule to

Eq. (12) gives

η (n)m−1
1 − R2
RU0(n) + U(m)
(n)
+ η U1 (R + iη) = (19)
2 i=1
η
The boundary conditions (14) gives
U0(n) = U0(m) = 0
The above equation reduces to

m−1
1 − R2
η U1(n) (R + iη) = (20)
i=1
η
The above-mentioned Eqs. (15)–(18) and (20) have been solved using an iterative
process. At first, Eq. (17) has been solved at n = 0 column and i varies from 1 to m-1
to obtain a system of linear equations which is then solved using the Gauss-Jordan
method to obtain tangential velocity. Correspondingly Eqs. (16), (18) and (20) have
also been solved under similar conditions to determine the system of linear equations.
By Gauss-Jordan method, these systems are then solved to obtain the axial velocity
and pressure at adjacent column n = 1. Finally, the radial velocity is obtained from
Eq. (15) by Gauss-Jordan method using the known values of U. By repeating the
above procedure, we get the result column by column along the axial direction of
the annulus until the flow is fully developed, both axially and tangentially.
The computations have been done by fixing the Bingham number as B = 10 for
different aspect ratio R. The solid volume fraction of the nanofluid considered is ϕ
= 0.5 and 1.5 for both copper and silver nanofluids. Moreover, ratio between the
reference density of the solid fraction (ρs ) and the reference density of the fluid
fraction (ρ f ) for Al2 O3 and TiO2 nanofluids are 8.9590 and 10.5305. We fix the
ratio of Reynolds number to Taylors number as Rt = 20, 30 and the results have
been obtained at η = 0.02, 0.03 and ς = 0.007, 0.002, respectively. The velocity
Fig. 3 R = 0.2, Z = 0.02, B = 10 then φ = 0.5 and 1.5
profiles along axial, radial, tangential and pressure distribution along the radial direc-
tion have been computed for different aspect ratios R = 0.2 and 0.9, when the outer
cylinder is rotating and the inner cylinder is at rest.
The development of tangential velocity profiles in the entrance region is depicted
in Figs. 3 and 4 for the aspect ratio of 0.2 and 0.9, respectively. Here, the volume
fraction parameter ϕ influences the tangential velocity significantly, which is high for
a small value of ϕ. The tangential velocity gradually increases from the inner cylinder,
immaterial of the annular width and volume fraction. For a large annular space, the
tangential velocity increasing nonlinearly. But for a small annular space, the increase
in tangential velocity is linear. It has been observed that there is a marginal increase
in the tangential velocity value for a TiO2 -based Bingham nanofluid compared to
Al2 O3 nanofluid.
The axial velocity profiles are depicted in Figs. 5 and 6 for the above-mentioned
values of ϕ and R. It is observed that the axial velocity component increases from
the inner wall to reach its maximum at a particular radial position and then decreases
as it moves toward the outer wall of the annular cylinder, for any annular width and
volume fraction. For smaller annuli, the difference in the values of axial velocity is
almost negligible for both nanofluids irrespective of volume fraction. However, in
large annuli, for a high volume fraction, there is a significant difference in the values
of axial velocity between the two nanofluids. TiO2 -nanofluid velocity is increasing
marginally better than Al2 O3 -nanofluid, which may be due to its physical properties
with flow nature.
Fig. 4 R = 0.9, Z = 0.02, B = 10 then φ = 0.5 and 1.5
Fig. 5 R = 0.2, Z = 0.02, B = 10 then φ = 0.5 and 1.5

Fig. 6 R = 0.9, Z = 0.02, B = 10 then φ = 0.5 and 1.5
The radial velocity for the different aspect ratios and the volume fraction param-
eters are presented in Figs. 7 and 8. It has been realized that the radial velocity is
positive near the inner cylinder wall and negative near the outer cylinder for both
nanofluids. The radial velocity decreases as we progress in the axial direction. For
the small annuli, the radial velocity decreases from the inner wall to the outer wall
with the sinusoidal nature irrespective of volume fraction and nanofluids, and there
is marginal difference in the values. However, there is a significant difference in the
radial velocity values of two nanofluids for a high volume fraction in large annuli.
TiO2 -nanofluid takes high magnitude of radial velocity with the negligible difference
between the Bingham nanofluids.
The pressure distribution in the annuli is computed for various cases and is shown
in Figs. 9 and 10. For all the parameters considered, the pressure increases gradually
from inner stationary wall to outer rotating wall. However, this increase is very
marginal in magnitude. As volume fraction increases, pressure increases from the
inner wall to the outer wall irrespective of nanofluid. Rate of increase in pressure
decreases for TiO2 better than Al2 O3 . For large annular space, for both nanofluids, the
pressure increases significantly whereas for smaller annuli this increase in pressure
with respect to volume fraction is highly significant. Al2 O3 -nanofluid showed a
higher pressure drop when compared to TiO2 -nanofluid irrespective of aspect ratio
and geometrical parameters. TiO2 -nanofluid decreases the pressure drop better than
Al2 O3 -nanofluid, because of its inversely proportional to the viscous nature.
Fig. 7 R = 0.2, Z = 0.02, B = 10 then φ = 0.5 and 1.5
Fig. 8 R = 0.9, Z = 0.02, B = 10 then φ = 0.5 and 1.5

Fig. 9 R = 0.2, Z = 0.02, B = 10 then φ = 0.5 and 1.5
Fig. 10 R = 0.9, Z = 0.02, B = 10 then φ = 0.5 and 1.5

4 Conclusion
From this work, we observed that, in the smaller annuli, there is very negligible
difference in the increases in tangential, radial, axial velocities and pressure distribu-
tion. For the large space annuli, we observed the significant difference in increases
in velocities and pressure distribution. Low volume fraction does not have an impact
on velocity profiles irrespective of aspect ratio. From the results for low volume frac-
tion, Al2 O3 and TiO2 nanofluids have a similar trend of velocity profiles, which may
be because of Newtonian behavior and for higher volume fractions, TiO2 -nanofluid
showed a higher magnitude compared to Al2 O3 -nanofluid which may be due to its
physical properties. Pressure drop is more for higher volume fraction, because of the
viscous nature of nanofluids when compared with base fluid and lower volume frac-
tions of the nanofluids. From the results, higher volume fraction of Al2 O3 -nanofluid
showed a higher pressure drop when compared to TiO2 -nanofluids. Higher pressure
drop increases the pumping power and cost which is not suitable for empirical or
non-theoretical applications.
Acknowledgements The authors thank the Principal and management of SSN institutions for their
support in carrying out this work.
References
1. Huilgol RR (2006) A systematic procedure to determine the minimum pressure 8 gradient

required for the flow of viscoplastic fluids in pipes of symmetric cross section. J Non-Newtonian
2. Poole RJ, Chhabra RP (2010) Development length requirements for fully developed laminar
pipe flow of yield stress fluids. J Fluid Eng 132:1–4
3. Nadiminti SR, Kandasamy A (2016) Entrance region flow heat transfer in concentric annuli with
rotating inner wall for bingham fluid. J Comput Appl Mech 11(2):137–157 (2016)
4. Khali S, Nebbali R, Bouhadef K (2017) Effect of a porous layer on Newtonian and power-law
fluids flows between rotating cylinders using lattice Boltzmann method. J Braz Soc Mech Sci
Eng 39:3881–3895
5. Choi SUS, Eastman JA (1995) enhancing thermal conductivity of fluids with nanoparticles.
ASME International Mechanical Engineering Congress & Exposition, San Francisco, CA pp
12–17
6. Chen H, Ding Y, Tan C (2007) Rheological behavior of nanofluids. New J Phys 9:367(1–23)
7. Esmaeilnejad A, Aminfar H, Neistanak MS (2014) Numerical investigation of forced convection
heat transfer through microchannels with non-Newtonian nanofluids. Int J Thermal Sci 75:76–86
8. Hayat T, Asad S, Alsaedi A (2016) Flow of casson fluid with nanoparticles. Appl Math Mech
37(4):459–470
9. Usman M, Soomro FA, Ul Haq R, Wang W, Defterli O (2018) Thermal and velocity slip effects
on Casson nanofluid flow over an inclined permeable stretching cylinder via collocation method.
Int J Heat and Mass Transfer 122:1255–1263
Rheodynamic Lubrication
of an Externally Pressurized Converging
Circular Thrust Bearing Using
Viscoplastic Fluids
G. Alexander Raymand and I. Jayakaran Amalraj
1 Introduction
By principle, in externally pressurized thrust bearings, the bearing surfaces are sepa-
rated by a fluid film which is created and maintained by external means. As these
kinds of bearings have certain operational advantages such as low friction, high load-
carrying capacity and high stiffness, these bearings are widely used in mechanical
industries. Using externally pressurized thrust bearing, researchers like Khalil et al.
[1], Dinesh Kumar et al. [2], Elsharkawy et al. [3], Jaw Ren Lin [4] and Kandasamy
[5] analyzed the effects of lubricants under different flow conditions and medium.
The lubrication of modern machines has been a challenge and an emerging interest
in the tribologists. The pioneering researchers have laid out a foundation of hydrody-
namic lubrication. However, modern researchers intend to use non-Newtonian fluids
as lubricants. Also, the interest has been increasing to use non-Newtonian lubricants
characterized by a yield value. The modern aspect of tribology immensely focuses on
the application of the lubricants with variable viscosity namely non-Newtonian lubri-
cants. Hence, researchers have been keen on using non-Newtonian fluids with yield
stress as lubricants such as Bingham, Casson, Herschel-Bulkley fluids. A Casson
fluid is a shear-thinning liquid that has infinite viscosity at zero rate of shear, yield
stress below which no flow occurs, and a zero viscosity at the infinite rate of shear.
Batra and Kandasamy [6], Walika and Falucki [7], Haleh Amlimohamadi et al. [8],
Syed and Sheikh [9], Ganesh et al. [10] and Kandasamy and Vishwanath [11] used
viscoplastic fluids as a lubricant and analyzed the effect in bearing performance.
Very few researchers studied the performance of converging or diverging bearing.
Roy et al. [12] have discussed the effect of inertia forces in an externally pressurized
G. A. Raymand (B) · I. J. Amalraj

Department of Mathematics, SSN College of Engineering, Chennai, Tamil Nadu, India
I. J. Amalraj
e-mail: jayakarana@ssn.edu.in
46 G. A. Raymand and I. J. Amalraj
bearing with converging and diverging film thickness using a viscoelastic lubri-
cant. Vishwanath et al. [13] have analyzed the problem of a squeeze film bearing
with converging and diverging squeeze films. Further, Jayakaran Amalraj et al. [14]
worked on the effects of the angle of convergence on the thickness of the core in
a converging externally pressurized thrust bearing theoretically. Although a lot of
studies have been done on externally pressurized thrust bearings with uniform film
thickness, only a little work has been done with variable film thickness.
In this research work, we have studied the effects of non-Newtonian characteristics
and the angle of convergence on the performance of externally pressurized converging
thrust bearing. Numerical solutions have been obtained for the bearing performances
such as pressure distributions and load-carrying capacity for different Bingham and
Casson numbers and angle of convergence and found to be a significant increase in
load-carrying capacity.
2 Mathematical Formulation of the Problem
The geometry of the bearing is as shown in Fig. 1.

We consider an isothermal incompressible steady flow of the time-independent
Casson fluid between two converging circular plates, separated by a varying distance
“h”. Let R1 and R2 be the radius of the film inlet and outlet, respectively, p the
pressure of the film, pa the atmospheric pressure, and ρ the density of the fluid.
The origin is fixed at the center of the plate, and the cylindrical polar coordinates
with axial symmetry are considered. Let r and z be the distance along the radial and
axial direction, respectively. Let vr and vz be velocity components along with r and z
directions, respectively. The geometry of the problem is shown in Fig. 1. There will
be a region called “core region” where shear stress is less than the yield stress that
Fig. 1 Geometry of an externally pressurized convergent thrust bearing

Rheodynamic Lubrication of an Externally Pressurized Converging … 47
moves with the constant velocity vc . Let the boundaries of the core be z = − δ(r2)h
and z = δ(r2)h as shown in Fig. 1.
Applying the basic assumptions of lubrication theory for thin films and ignoring
inertia effects and the body forces, the equations of continuity and momentum for
the externally pressurized circular thrust bearing will be as follows.
Equation of continuity:
1 ∂ ∂ vz
(r vr ) + =0 (1)
r ∂r ∂z
Equations of momentum:
∂p ∂τ
− = rZ (2)
∂r ∂z
∂p
=0 (3)
∂z
the constitutive equation given for Bingham material and Casson fluid as

∂vr
τr z = η2 ± η1 (4)
∂z

1/2 ∂vr (1/2)
τr1/2 = η2 ± σ (5)
z
∂z
For
Here, η2 represent yield stress of the lubricant, η1 and σ 2 represent Bingham and
Casson’s viscosity, respectively.
The boundary conditions are:
h
vr = 0 at z = ± (6)
2
δh
vr = vc at z = ± (7)
2
p = pa at r = R2 (8)
∂ vr
is continuous, at τ = η2 (9)
∂z
where pa is atmospheric pressure.

3 Solution
The velocity distributions for the flow region and core region can be obtained by
solving governing Eqs. (1)–(5) and using the boundary conditions (6) and (7). We
obtain for Bingham and Casson lubricants.

1 dp δh 2 h δh 2
vr = − z− − −
2η1 dr 2 2 2
where
δh h
≤z≤ (10)
2 2
⎧

1 ⎨ dr
dp
δh 2 h δh 2
vr = 2 z− − −
σ ⎩ 2 2 2 2
1/ ⎛ 3/
h 4η2 2 ⎝ dp h δh 2
−2η2 z − + dp − − + η2
2 3 dr dr 2 2
⎞⎫
3/ ⎬
dp δh 2
⎠
− − z− + η2
dr 2 ⎭
where
δh h
≤z≤ (11)
2 2
And the velocity of the core region,
1. For Bingham material

1 dp h δh 2 δh
vc = − − − 0≤z≤ (12)
2η1 dr 2 2 2
2. For Casson fluid

⎧
2
1 ⎨ dr
dp
δh δh h2 δh 2
vc = 2 − δh − +
σ ⎩ 2 2 2 4 2

1/2 3/2
δh h 4η2 dp h δh 3/2
− 2η2 − + dp − − + η2 − η2 (13)
2 2 3 dr dr 2 2
The conservation of mass for externally pressurized bearing in integral form is
h/2
Q = 4π r vr dz (14)
0
where Q is the flow rate per unit width.

Using velocity distribution of Bingham material vr and vc from Eqs. (10)–(12)
into Eq. (14), we get

πr h 3 dp
Q= (2 + δ) 1 − δ 2 (15)
12η1 dr
Using velocity distribution of Casson fluid vr and vc from Eqs. (11)–(13) into
Eq. (14), we get

4πr h 3 dp 3 η2 h 2
Q= −δ + 3δ − 2 + 1 − δ2
σ2 48 dr 4
1/2 −1 3/2
4η2 dp h dp h δh 3/2
+ − (1 − δ) + η2 − η2
3 dr 2 dr 2 2
−1 5/2 −1
2 dp dp h 2 dp 5/2
+ − (1 − δ) + η2 − η2 (16)
5 dr dr 2 5 dr
Considering the equilibrium of an element in the yield surface − δ2h ≤ z ≤ δh

2
, it
is found that
dp 2 η2
=− (17)
dr δ (r ) h
Eliminating the pressure gradient by simplifying Eqs. (15), (16) and (17), we get
nonlinear algebraic equations to determine the core thickness
3 (2 + δ)(1 − δ)2
= (18)
πr η2 h2 δ
2η1 Q
For Bingham material and
3 −δ 2 24 2
= − + +3 (19)
πr η2 h 2 5 1/ δ
2σ 2 Q 5δ 2
For Casson fluid.

The variation in film thickness of the lubricant in the converging bearing can be
defined as

r
h(r ) = h 0 − h 0 tan φ (20)
R2
where h(r ) represents the varying film thickness between the plates, h 0 is the
maximum film thickness at the center of the bearing and ϕ is the angle of convergence.
The following non-dimensional parameters are introduced.
r h z
r∗ = ; δ∗ = δ (r ∗) ; h ∗ = ; z ∗ = ; h(r ) = h 0 (1 − r ∗ tan φ);
R2 h0 h
π R2 h 20 η22
B= , is the Bingham Number,
2 Qη1
π R2 h 20 η22
Yc (N ) = , is the Casson Number,
2Qσ 2
Using non-dimensional quantities in Eqs. (18) and (19), we get for Bingham
material and Casson fluid as
3 2
= + δ ∗2 −3 (21)
Br ∗ (1 − r ∗ tan φ) δ∗
−δ ∗
2
3 24 2
= − + ∗ +3 (22)
Yc (N )r ∗ (1 − r ∗ tan ϕ)2 5 1 δ
5δ ∗ /2
The non-dimensional form of pressure gradient.
1. For Bingham material
p dp ∗ 4B
p∗ = ; ∗
= ∗ (23)
Qη1 2
dr δ (1 − r ∗ tan ϕ)
π h 30
2. For Casson fluid

p dp ∗ 4Yc (N )
p∗ = ; ∗
= ∗ (24)
Qσ 2
dr δ (1 − r ∗ tan ϕ)
π h 30
The value of δ* has been determined for different values of i and Yc (N ) for
various values of r*, by iterative technique. Knowing δ*, the pressure distribution
for Bingham material and Casson fluid can be obtained by integrating (23), (24) and
using boundary condition (8) and it is given by
1
∗ d p∗
P − Pa∗ = dr ∗ (25)
dr ∗
r∗
The load-carrying capacity W for the externally pressurized thrust bearing can be
obtained from
1
W = (P ∗ −Pa ∗)r ∗ dr ∗ (26)
R∗
where R ∗ = RR21 is the ratio of inside to outside radius of the bearing. This integra-
tion is performed numerically for various values of the Bingham and Casson numbers
and values are given in the Tables.
Validation of results:
The pressure and load-carrying capacity for different values of Bingham numbers
and Casson numbers corresponding to φ = 0 are in good agreement with Kandasamy
[5].
Lubrication of an externally pressurized thrust bearing with (i) Bingham and (ii)
Casson lubricants has been investigated. The thickness of the core along radial direc-
tion for various values of Bingham and Casson number for a particular angle of
convergence ϕ = 0, 10 have been determined and tabulated in Tables 1, 2, 3 and 4.
The thickness of the core increases from the center to the periphery when the angle
of convergence is zero and it refers to uniform film thickness. For variable film thick-
ness, the angle of convergence increases the core thickness and reaches its maximum
and deteriorates as it moves toward the periphery. The position attaining maximum
depends only on the angle of convergence. Also, it has been observed that the plug
core increases for materials with larger values of Bingham and Casson numbers.
The pressure distribution in the region of the fluid film and the load-carrying
capacity of the bearing for various values of Bingham and Casson numbers and the
angle of convergence have been determined and are shown in Figs. 2 and 3.
It is evident from the pressure profiles that the pressure decreases gradually from
the center to the periphery of the bearing. It has been observed that the pressure is
found to increase as the Bingham number and the Casson number increase for a given
angle of convergence. The numerically computed values of the load-carrying capacity
of the bearing for various values of Bingham number and Casson number and the
angle of convergence are tabulated in Table 5. From the numerical solutions obtained,
we infer that the load-carrying capacity increases as the Bingham number and the
Casson number increase corresponding to any particular angle of convergence.
52
Table 1 Core thickness for various vales of Bingham number when φ = 0

ϕ=0 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
B=1 0.0601 0.1103 0.1530 0.1898 0.2220 0.2531 0.2782 0.3007 0.3212 0.3381
B=2 0.1104 0.1898 0.2504 0.2986 0.3381 0.3743 0.4023 0.4266 0.4480 0.4653
B=3 0.1531 0.2505 0.3193 0.3713 0.4124 0.4490 0.4767 0.5004 0.5211 0.5375
G. A. Raymand and I. J. Amalraj
Table 2 Core thickness for various vales of Bingham number when φ = 10

ϕ= 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
10
B = 1 0.0582 0.1039 0.1406 0.1706 0.1954 0.2180 0.2351 0.2496 0.2619 0.2712
B = 2 0.1071 0.1802 0.2334 0.2739 0.3056 0.3333 0.3537 0.3704 0.3842 0.3946
B = 3 0.1489 0.2392 0.3004 0.3449 0.3787 0.4075 0.4283 0.4451 0.4589 0.4692
Table 3 Core thickness for various vales of Casson number when φ = 0

ϕ=0 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
Yc (N ) = 0.0387 0.0644 0.0848 0.1020 0.1170 0.1315 0.1434 0.1542 0.1642 0.1725
1
Yc (N ) = 0.0644 0.1020 0.1303 0.1532 0.1725 0.1909 0.2056 0.2188 0.2307 0.2406
2
Yc (N ) = 0.0849 0.1303 0.1632 0.1893 0.2110 0.2313 0.2473 0.2616 0.2744 0.2850
3
Table 4 Core thickness for various vales of Casson number when φ = 10

ϕ = 10 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
Yc (N ) = 0.0377 0.0612 0.0790 0.0931 0.1046 0.1151 0.1231 0.1299 0.1357 0.1401
1
Yc (N ) = 0.0628 0.0976 0.1223 0.1414 0.1566 0.1702 0.1803 0.1889 0.1960 0.2015
2
Yc (N ) = 0.0829 0.1250 0.1541 0.1759 0.1932 0.2084 0.2197 0.2291 0.2369 0.2429
3
Fig. 2 Pressure distribution along the radial direction for Bingham material
Fig. 3 Pressure distribution

along the radial direction for
Casson fluid
Table 5 Load-carrying capacity for various values of Bingham number and Casson number
B = 0 = Yc (N ) B Yc (N )
1 2 3 1 2 3
ϕ=0 1.98 2.93 3.82 4.67 5.16 6.96 8.53
ϕ=5 2.27 3.29 4.25 5.17 5.82 7.82 9.54
ϕ = 10 2.67 3.76 4.81 5.81 6.69 8.91 10.8
ϕ = 15 3.23 4.42 5.57 6.67 7.86 10.3 12.5
ϕ = 20 4.08 5.40 6.68 7.91 9.56 12.5 15.0
ϕ = 25 5.53 7.02 8.46 9.87 12.2 15.8 18.8
Also, the increase in the angle of convergence leads to an increase in the load-
carrying capacity of the bearing for a particular Bingham number and Casson number.
We further observe that if the angle of convergence is marginal the increase in the
load-carrying capacity is also marginal whereas if the angle of convergence is appre-
ciable it is known that in general the non-Newtonian characteristics enhance the
performance of the bearing and it can be concluded that the similar trend continues
for converging bearing also. Moreover, for a given flow rate, the value of pressure
and thereby the load-carrying capacity for Bingham lubricant are less than that of
Casson lubricant.
5 Conclusions
The above investigation reveals that the Bingham number and Casson number and
angle of convergence significantly enhance the bearing performance such as pressure
distribution and load-carrying capacity of the bearing.
Acknowledgements The authors thank the Principal and management of SSN institutions for their
support to carry out this work.
References
1. Khalil MF, Kassab SZ, Ismail AS (1992) Influence of inertia forces on the performance of
turbulent externally pressurized bearing. Trib Int, Butterworth-Heinemann Ltd 25(1):17–25
2. Kumar D, Chandra P, Sinha P (1993) Ferrofluid lubrication of externally pressurized circular
plates and conical bearings. Int J Engng Sci 31(4):593–604
3. Elsharkawy AA, Kassab SZ, Nassar MM (1996) Lubrication analysis of externally pressurized
circular porous bearing. Appl Math Model 20:870–876
4. Lin J-R (1999) Static and dynamic characteristic of externally pressurized circular step thrust
bearing lubricated with couple stress fluids. Tribol Int 32:207–216
5. Kandasamy A (1996) Lubrication of an externally pressurized thrust bearing with non-
newtonian lubricants. Advances in Mechanical Engineering, NewDelhi, India, pp1015–1023
6. Batra RL, Kandasamy A (1989) Lubrication of a rotating circular thrust bearing with a Casson
lubricant. Appl Sci Res 46:127–140
7. Walicka, Falicki J (2015) Reynolds number effect in the flow of an electrorheological fluid of
a Casson type between the fixed surface of revolution. Appl Mathe Comput 250:636–649
8. Amlimohamadi H, Akram M, Sadeghy K (2016) Flow of a Casson fluid through a locally-
constricted porous channel: a numerical study. Korea-Australia Rheology J 28(2):129–137
9. Mohyud-Din ST, Irfanullah Khan, S (2016) Nonlinear radiation effects on squeezing flow of a
Casson fluid between parallel disks. Aerospace Sci Technol 48:186–192
10. Ganesh S, Kirubhashankar CK, Mohamed Ismail A (2015) Non-linear squeezing flow of casson
fluid between parallel plates. Int J Math Anal 9(5):217–223
11. Kandasamy, Vishwanath KP (2007) Rheodynamic lubrication of a squeeze film bearing under
sinusoidal squeeze motion. Comput Appl Math 26:381–396
12. Roy JS, Padhy S, Bhopa LK (1993) Inertia effect in externally pressurized thrust bearing with
converging and diverging film using visco-elastic lubricant. Acta Mech 96:1–12
13. Vishwanath KP, Kandasamy A (2008) Rheodynamic lubrication of a rectangular squeeze film
bearing with an exponential curvature variation using bingham lubricants. Int J Math Comput
Phys Electr Comput Eng 2(7)
14. JayakaranAmalraj I, Alexander Raymand G (2016) Core variation in an externally pressurized
converging thrust bearing with bingham lubricant. Appl Mech Mater 852:428–434
Experimental Investigation of Flow Over
a Backward-Facing Step
with Inclinations
S. V. Abhijith and R. Abhilash
1 Introduction
The flow separation and its subsequent reattachment to a solid surface occurs in
many aerodynamic systems. There have been cases in which the separated flows
could be used favorable for applications like mixing in combustors. But in almost all
practical situations the recirculations are to be minimized in order to reduce losses,
for example, flow separation along the body of a vehicle results in pressure drag
thereby increasing the aerodynamic losses.
Flow separation is a very complex topic due to the inherent instabilities present. In
order to simplify this, experiments are conducted on various geometries like fence,
rib, bluff body with splitter plate, expanding pipes, cavities, forward- and backward-
facing steps. These geometries simplify the study by either fixing the separation
point or by fixing the reattachment point. The backward-facing step is considered
as the most ideal geometry to study separated flow due to its single separation point
and the wake dynamics unperturbed by the downstream disturbances. The position
of the reattachment zone and its length from step edge determine the recirculation
region which is unstable. A number of experimental studies have been carried out
using various techniques like laser Doppler velocimetry [1], hot wire anemometer
[2], particle image velocimetry [3], etc. It is found from the literature that the factors
which affect the features of the backward-facing step are Reynolds number, expansion
ratio, aspect ratio, free stream turbulence intensity, and boundary layer thickness.
Reynolds number is found to play an important role in reattachment. In the laminar,
transition, and turbulent region, Reynolds number follows different trend on the
reattachment length. In the laminar region, the reattachment length has been found
to increase with increase in Reynolds number. A gradual decrease in reattachment
S. V. Abhijith · R. Abhilash (B)

Department of Mechanical Engineering, College of Engineering, Trivandrum 695016, Kerala,
India
e-mail: abhijithsv317@gmail.com; abhilashr@cet.ac.in
58 S. V. Abhijith and R. Abhilash
length has been observed in the transitional region. A small increase in length has been
reported in the turbulent regime [1]. In order to reduce pressure drag associated with
separation, various flow controlling methods such as dimples on step edge, curved
edge, step with inclinations, have been adopted. The present study aims to investigate
the effect of reattachment and recirculation by attaching the backward-facing step
with inclinations.
Drag reduction has been a hot topic of research in automobile industry as it can
be directly linked to the fuel efficiency. The authors were motivated by this aspect
and have carried out this work as a preliminary step toward the development of an
adjustable flap.
The Reynolds number range selected for this study (84,000 to 220,000) falls in
the turbulent regime, which is of interest to the automobile industry. The wind tunnel
facility available in the institution is able to attain velocities pertaining to the above
mentioned range only.
The objective of this work is to investigate the effects of downstream pressure
distribution, reattachment length, and recirculation region with step inclination angles
of 30°, 45°, 60°, 75°, 90°. Effect of Reynolds number is studied by varying it over a
range of 84,000 to 220,000.
2 Methodology
Present work involves experimental investigation of flow over a backward-facing

step with inclinations. The surface pressure distribution along downstream for five
inclinations has been obtained and compared. Experiments were carried out in a
subsonic wind tunnel. Pressure probes were placed at a number of points downstream
of the step and readings at each point were recorded for calculation of pressure
coefficient.
2.1 Experimental Details
The experiments were conducted in low-speed wind tunnel. The size of the test
section is 200 × 50 × 50 cm, bell-mouthed entry with contraction ratio of 9:1.
Honeycomb section 150 cm side square cross section with three screens of 8, 12, 16
stainless steel meshes is used to laminarize the inflow. Axial flow fan of diameter
100 cm with 6 blades attach to 5.5 kW AC motor with speed controlling device is
used in the wind tunnel.
For pressure measurement, multitube manometer is used. It involves 30 different
ports connected to a bottle that contains liquid ethyl for measuring pressures at
respective points simultaneously. Range of manometer is 0 to 40 cm. The least count
of the manometer is 0.1 cm. Pressure at 30 points downstream of the step has been
measured.
Experimental Investigation of Flow Over a Backward … 59
For measuring flow velocity pitot-static tube is placed upstream of backward-

facing step, and it consists of two ports: one port is placed in such a way that it is
perpendicular direction to the flow, for measuring static pressure and second port is
placed in axial direction of flow which measures the stagnation pressure; the two
ports are connected to manometer, and the difference in manometer height is used
to determine the flow velocity.
Figure 1 shows the BFS geometry with dimensions as function of step height h.
α is the step angle. The models used in this work are backward-facing steps with
inclination angles of 30°, 45°, 60° 75°, 90°, respectively. The model is fabricated
using acrylic sheet of 3 mm thickness, the length of all models is 110 cm, width of
models is 25 cm, step height is 22.2 cm, and thickness of the models is 0.3 cm. These
models contain pressure ports at center line of downstream with 0.4 cm diameter
holes each at a distance of 1 cm gap in order to measure the pressure acting on
its surface, adjustable inclination angle was made and fitted using hinge support at
beginning of downstream edge. The experiment is done for all models at different
Reynolds number of 84360, 111555, 152070, 183843, 219225 with various step
inclination angles 30°, 45°, 60°, 75°, 90°. The fabricated model of backward-facing
steps is shown in Figs. 2, 3, 4.
Fig. 1 BFS geometry
Fig. 2 BFS with 90° step inclination

The effect of step angle and Reynolds number on reattachment length is studied and
compared with 90° step angle. The step angles considered are 30°, 45°, 60°, 75°. The
surface pressure distribution C p versus dimensionless length is plotted for different
Reynolds number.
Figure 5 shows the pressure distribution for backward-facing step with step incli-
nation angle of 90° at Reynolds number of Rh = 84360, 111555, 152070, 183843,
219225. A sudden pressure drop occurs at the beginning of the downstream surface
and pressure recovery takes place and reaches a maximum value at positions of X r /h
= 2.62, 2.68, 2.72, 2.85, 3.0. A recirculating zone is identified at the beginning of
downstream surface, and the location of maximum pressure point is the reattachment
Fig. 5 Variation of C p corresponding to 90° inclination
point X r , the distance measured from the step edge to reattachment point is known
as reattachment length X r /h.
At the end of rear plate, small drop in pressure is observed and this indicates a
chance of formation of another recirculation region. It is observed that as Reynolds
number increases from 84,360 to 111,555 the reattachment length X r /h goes on
increasing.
The plot in Fig. 6 shows the pressure distribution for backward-facing step with
step inclination angle of 45° at Reynolds number of Rh = 84360, 111555, 152070,
183843, 219225. A sudden pressure drop occurs at the end of slanted height X/h
= 1 of the downstream surface, there recirculation region begins and then pressure
recovery takes place and reaches to a maximum value at positions of X r /h = 2.35,
2.38, 2.42, 2.53, 2.55. The position of maximum pressure points lies below initial
pressure that means reattachment effect is lower, after that flow is uniform, at the end
of downstream pressure remains constant, there is no chance of formation of another
recirculation as compared to BFS with 90° inclination, another observation is width
of low pressure region is reduced due to reduction in recirculation zone. It is observed
that as Reynolds number Rh increases from 84,360 to 219,225 reattachment length
X r /h increases with small increment, by comparing the effect of reattachment in BFS
with 90° inclination, it shows reduced reattachment length X r /h with deviation more
than 0.3 X r /h. According to BFS with 75° inclination and BFS with 60° inclination,
BFS with 45° inclination shows reduced reattachment length X r /h and recirculation
zone.

The plot in Fig. 7 shows the pressure distribution for backward-facing step with
step inclination angle of 30° at Reynolds number of Rh = 84360, 111555, 152070,
183843, 219225. A sudden pressure drop occurs at the end of slanted height X/h
= 1.5 of the downstream surface, there recirculation region begins and pressure
recovery takes place and reaches to a maximum value at position of X r /h = 2.32,
2.36, 2.40, 2.51, 2.54; the position of maximum pressure points after that flow is
uniform, at the end of downstream pressure remains constant, there is no chance of
formation of another recirculation region as compared to BFS with 90° inclination;
another observation is width of low pressure region which is reduced much higher
due to reduction in recirculation zone, and it is observed that minimum recirculation
is obtained at BFS with 30° inclination. It is observed that as Reynolds number Rh
increases from 84,360 to 219,225 reattachment length X r /h increases, by comparing
the effect of reattachment in BFS with 90° inclination, BFS with 75° inclination,
BFS with 60° inclination and BFS with 45° inclination, BFS with 30° inclination
shows lower reattachment length X r /h and lower recirculation zone.
Figure 8 shows the variation of reattachment length X r /h with various step incli-
nation angle α, at different Reynolds number. It is observed that lowest reattachment
length X r /h is obtained at BFS with 30° inclination. As step inclination angle α
increases, reattachment length of X r /h increases with increase in Reynolds number
Rh , up to Rh = 152,070 reattachment length X r /h increase with small increment but
for Rh = 183,843 & Rh = 219,225 reattachment length increases rapidly, for step
Fig. 8 Variation of X r /h with step inclination angle α

inclination of 75° and for 90°. For each range of reynolds number the reattachment
length Xr /h increases linearly with step inclination angle α from 30° to 75°.
4 Conclusions
In the present work, the experiments are conducted on backward-facing step with
five models with inclinations of 90°, 75°, 60°, 45°, 30°, respectively. The distribution
of coefficient of pressure downstream of the step at different Reynolds number is
determined. The effect of Reynolds number on reattachment length and recirculation
region has also been investigated. From the experiment, following conclusion is
made. For a backward-facing step
1. Reattachment length X r /h increases with increase in Reynolds number Rh .
2. Reattachment length X r /h increases with increase in step inclination angle α.
3. Recirculation region is maximum for backward-facing step with 90° inclina-
tion and recirculation zone is minimum for backward-facing step with 30°
inclination,
4. From this investigation, backward-facing step with 30° inclination is found to
minimize drag as the recirculation region is minimum.
References
1. Armaly B, Durst F, Pereira JCF, Schonung B (1983) Experimental and theoretical investigation
of backward-facing step. JFluidMech 127:473–496
2. Badran O, Bruun HH (1998) The effect of inlet conditions on flow over backward facing step.
J Wind Eng Ind Aerodyn 74:495–509
3. Biswas G, Breuer M, Durst F (2004) Backward facing step flow for various expansion ratios at
low and moderate Reynolds numbers. J Fluids Eng 126(3):362–374
Experimental Investigations on Effect
of Chevrons in Bunsen Burners
Manasa Sadhasivan, V. Priyadarshini, and S. Soma Sundaram
1 Introduction
Flames are being used in a wide range of places, from household to large industries
for power production. Depletion of the fossil fuels and increasing of the pollution
have resulted in the search of efficient methods of combustion. The process of burning
is to be modified such that it produces maximum power by minimal consumption
and pollution.
The effect of partial premixing on the flame characteristics was studied by Choi
et al. [1]. It was reported that the flame was found to contain two distinct reaction
zones, an inner rich premixed reaction zone and an outer partially premixed reaction
zone. The presence of two reaction zones was also observed by Mikami et al. [2]
in spray combustors. Spray combustors are usually used in oil fired furnaces, where
the liquid fuel is sprayed into the combustion chamber. Experimental analysis on the
spray combustor indicated that the existence of the internal premixed flame increased
the vaporization of droplets in the external flame.
Other methods of improving the efficiency of the burning process were also
studied by researchers. Lee et al. [3] experimentally examined a cylindrical multi-
hole premixed burner by varying the diameters and arrangement of the holes. While
a reduction in the NOx and CO emissions were observed, there was no signifi-
cant difference in the combustion area when fuel was changed between liquefied
petroleum gas (LPG) and liquefied natural gas (LNG). Experimental analysis, on
liquefied petroleum gas (LPG) burners modified with Al2 O3 porous particles and
SiSiC ceramic foam, was carried out by Herrera et al. [4]. A linear trend in the upper
and lower limits of the equivalence ratio was observed when the heat input rate was
M. Sadhasivan · V. Priyadarshini · S. Soma Sundaram (B)

Department of Mechanical Engineering, SSN College in Engineering, Chennai, India
e-mail: somasundarams@ssn.edu.in
66 M. Sadhasivan et al.
set below 154 kW/m2 . It was observed that as the heat input rate was increased, the
flow rate of fuel and air increased, causing lift of the flame along with increase in
CO emissions.
Kotb and Saad [5] compared the co-swirl and counter-swirl burners with inversion
diffusion flames (IDF). Two distinct regions, the entrainment zone and the turbulent
mixing and combustion zone, separated by the minimal cross-sectional area ‘neck’
were observed. However, downstream of the neck, two flame regions were formed
for co-swirl, while one was formed for counter-swirl. It was observed that the peak
values of temperature and concentration of CO2 were higher for the co-swirl burner,
indicating better combustion. However, the NOx emissions were also found to be
higher for the co-swirl burner when compared with counter-swirl burner. Studies
on IDF were also carried out by Rabee et al. [6]. Experimental analysis on co-axial
circular inverse flame was carried out. The effect of varying the burner diameter on
the flame structure was studied. The smaller diameter burners were found to produce
shorter flames with higher flame temperature.
Additionally, improving the combustor performance, with the introduction of
exhaust gases, was studied by Joannon et al. [7]. Experimental and analytical analysis
on the effect of exhaust gas recirculation in combustor was carried out. The effect of
preheating the fuel and air and increasing the concentration of inert (nitrogen) was
also studied. Four distinct regions namely ‘no ignition’, ‘incomplete combustion’,
‘sustainable’, and ‘hystereses were obtained for varying the preheating temperatures
and dilution levels. With increased dilution, a higher value of preheat resulted in
a stable combustion. Further increased dilution increased the kinetic characteristic
time and decreased the rate of temperature rise. Similarly, Kareyen et al. [8] studied
the effects of adding a different fuel to the combustor. Experimental and numerical
investigations on a low calorific value coal gas burn were carried out. The effect of
supplying additional hydrogen in the flame zone was analyzed. Based on the results,
it was observed that hydrogen addition resulted in a stable flame. Furthermore, the
performance of the combustor improved with the introduction of hydrogen.
Similar studies were carried out by researchers, for improving the performance
of Bunsen burners. The influence of burner diameter of a Bunsen burner on its
turbulent flame speed was studied by Tamadonfar et al. [9]. From the results of particle
image velocimetry (PIV) and Rayleigh scattering techniques, it was observed that,
the characteristic flame height increased with the increase of the burner diameter. The
structure of the flame indicated that, the wrinkling in the leading edge and surface
of the flame was higher, for larger diameter burner. However, the flame structure
at the burner tip was found to be unaffected, with the change in the diameter. Hu
et al. [10] carried out experiments and numerical simulations on a Bunsen burner,
operating at different pressures. The flame structure and the flame speed variation,
with the variation in pressure, were analyzed. Numerical analysis indicated a two-
reaction zone structure, an inner premixed flame, and an outer diffusion flame. It was
observed that with the increase in the operating pressure, the inner flame became
thinner and highly unstable.
It can be seen from the literature, that many of the improvement methods are
concerned with varying the geometry of the combustor, or changing the fuel used
Experimental Investigations on Effect of Chevrons … 67
and enhancing better mixing of the fuel and air. Chevrons are being used in aircraft
to mitigate the noise generated by jet engines. The function of chevrons is to allow
better mixing of the hot and cold jets of the aircraft, thereby reducing the noise.
This provided the motivation for carrying the present work, while chevrons are
found to enhance mixing and they are not used in combustors. Hence, in the present
work, experiments will be carried out by providing chevrons on the tip of Bunsen
burners. The effect of providing chevrons on both diffusion and premixed flame will
be studied.
2 Methodology
The details of the experimental setup and the procedure for determining the parame-
ters of interest namely flame height, flame width, and complete combustion area are
explained in this section.
The experiments have been carried out using standard Bunsen burners. Three such
identical Bunsen burners have been procured and chevrons three and four in number
have been cut in two of the Bunsen burners. Chevrons have been cut to a depth that
is equal to the radius of the Bunsen burner.
The outlet of the fuel pipeline is connected to the inlet of a Bunsen burner. The
fuel used is LPG. The LPG cylinder is placed on a weighing machine with a least
count of 1 g. The reduction in weight of the cylinder for every 30 s is noted and
is used to obtain the mass flow rate. All joints have been sealed with Teflon tape,
to prevent fuel leakage. A camera is mounted on a tripod and located at a distance
of 60 cm from the burner. The setup is then placed in front of a black backdrop, to
ensure contrast and clarity of the images obtained.
The charge is then ignited at the tip of the outlet using a gas lighter. Once the
flame is stable, the flame image is captured continuously using a DSLR camera over
a period of four minutes for each case. The experiment is started in the diffusion
flame mode (100% close of air opening). The airflow rate is increased by opening
the air supply in increments of 25%, till the flame becomes completely premixed (0%
close of air opening) and the flame image is captured using the same technique. The
procedure is then reversed to obtain data from hysteresis as well. After completing the
experiment, the fuel line is closed and residual charge in the apparatus is scavenged
completely. The Bunsen burner is allowed to cool and the experiment is repeated for
Bunsen burners with 3 Chevron and 4 Chevron.
From the images captured, frames exhibiting improper flame contours are filtered
and the rest are taken for further analysis. An average of 100 images of the flame
contour is chosen in each case, to develop a mean image of the flame. Once the mean
image is generated, a mesh of 1 mm is overlaid on the image. Such a mean image
determined for Bunsen burner with three chevrons is shown in Fig. 1. The maximum
flame height and flame width are determined by the number of grid points enclosed
within the flame. Similarly, the area of complete combustion and the grid points with
blue flame are also determined. The number of grid points across the diameter of
Fig. 1 Mean image of the flame overlaid with mesh for a Bunsen burner with three chevrons a 100%
close b 50% close and c 0% close of air supply
the Bunsen burner is also determined. With this data as reference, the flame height,
flame width, and complete combustion area are converted into physical dimensions.
The details of the results obtained from the aforementioned experiments and the
inferences got from them are discussed in this section. Based on the number of grids,
the height of the flame and the breadth of the flame are determined in mm and the
complete combustion area is determined in mm2 .
The variation of these parameters at various levels of air opening for different
Bunsen burners is shown in Fig. 2. While the 100% close of air supply indicates
diffusion flame, the 0% close of air supply indicates premixed flame.
Fig. 2 Variation of a flame

height, b flame width, and
c complete combustion area
with the air supplied for
different Bunsen burners
It can be observed that with the inclusion of chevrons the flame height tends to
increase. Also, there is an increase in the flame height with the closing of the air
supply. The increase in the flame height is more pronounced in the diffusion flame
case than the premixed flame case. The presence of chevrons at the tip of the burner
causes vertical structures to develop, which pushes the fuel further, thus increasing
the flame height. In case of the premixed flame, since the fuel and air are mixed inside
the Bunsen burner itself, the penetration of the mixture is hindered by the flame.
However, it can be visualized from the plots of flame width (Fig. 2b), that the
effect of chevrons on width of the flame is not as significant as that of the flame
height. In case of the diffusion flame, the value of the flame width for the normal
Bunsen burner (burner with no chevrons) lies between the flame width for Bunsen
burners with three and four chevrons. Also, the variation in the values is very small.
It is also seen that the flame width of the diffusion flame is higher than that of the
premixed flame. This arises as the fuel spreads out in a diffusion flame, to come into
contact with air and thus form the flame front.
In case of the compete combustion area given by the blue region of the flame, the
area increases with the usage of chevrons in diffusion flame but decreases in premixed
flame. For the diffusion flames, the introduction of chevrons enhances the mixing of
the fuel and air and thus increases the complete combustion area. However, in the
premixed flame, the air and fuel are mixed in the Bunsen burner tube and the chevrons
causes the mixture to dilute further. Thus, a reduction in the complete combustion
or blue region of the flame is observed.
4 Conclusions
Chevrons have been used to promote mixing of fluids for noise reduction. In this
work, chevrons are implemented in burners, to improve the mixing of air and fuel
contributing to complete combustion.
Experiments have been carried out to explore the effects of using chevrons in a
Bunsen burner and to quantify its effects, by studying the various flame characteristics
for varied airflow rates. The flame characteristics under investigation are the flame
height, flame width, and area of complete combustion. The experiments have been
carried out for a fixed fuel flow rate while varying the opening of the airflow in
increments of 25%. Thus, the experiments have been carried out for both diffusion
(100% closed air supply) and premixed flame (0% closed air supply).
Steady flame images have been captured for every case, and those exhibiting
distorted flame contours have been filtered. An average of 100 images of the flame
contour is chosen in each case, to develop a mean image, from which flame charac-
teristics are extracted. The effect of three and four chevrons is investigated. From the
work, it has been inferred that, the penetration of the flame is seen to increase with
the increase in number of chevrons. The effect is more pronounced in diffusion flame
than in premixed flame. Further, with the presence of chevrons, the area of complete
combustion is found to increase for the diffusion flame and show minimal variation
for premixed flames. Additionally, the spread of the flame is seen to be unaffected,
with the increase in number of chevrons.
From the above results, we conclude that implementing chevrons produce a longer
flame. The variation in the flow which causes the longer flame is to be studied
numerically. Additionally, the effect of varying the angle of the chevrons is also to
be studied. These studies will be carried out subsequently.
References
1. Choi CW, Puri IK (2000) Flame stretch effects on partially premixed flames. Combustion
123:119–139
2. Mikami M, Yamamoto K, Moriue O, Kojima N (2005) Combustion of partially premixed spray
jets. Proc Combust Inst 30:2021–2028
3. Lee S, Kum S-M, Lee C-E (2011) An experimental study of a cylindrical multi-hole premixed
burner for the development of a condensing gas boiler. Energy 36:4150–4157
4. Herrera B (2015) Karen Cacua, Luis Olmos-Villalba, Combustion stability and thermal effi-
ciency in a porous media burner for LPG cooking in the food industry using Al2 O3 particles
coming from grinding wastes. Appl Therm Eng 91:1127–1133
5. Kotb A, Saad H (2016) A comparison of the thermal and emission characteristics of co and
counter swirl inverse diffusion flames. Int J Therm Sci 109:362–373
6. Basem AR (2018) The effect of inverse diffusion flame burner-diameter on flame characteristics
and emissions. Energy 160:1201–1207
7. de Joannon M, Sabia P, Sorrentino G, Bozza P, Ragucci R (2017) Small size burner combustion
stabilization by means of strong cyclonic recirculation. Proc Comb Inst 36:3361–3369
8. Karyeyen S, Ilbas M (2018) Experimental and numerical analysis of turbulent premixed
combustion of low calorific value coal gases in a generated premixed burner. Fuel 220:586–598
9. Tamadonfar P, Gülder ÖL (2016) Effect of burner diameter on the burning velocity of premixed
turbulent flames stabilized on Bunsen-type burners. Exp Thermal Fluid Sci 73:42–48
10. Hu S, Gao J, Gong C, Zhou Y, Bai XS, Li ZS, Alden M (2018) Assessment of uncertainties
of laminar flame speed of premixed flames as determined using a Bunsen burner at varying
pressures. Appl Energy 227:149–158
A Computational Study of Entrained
Flow Furnace with Swirl Burner
Configuration and Low Turbulence
Intensity Flow
Nitesh Kumar Sahu, Mayank Kumar, and Anupam Dewan
1 Introduction
Coal remains an important resource to meet the near-future energy demands even
with the advent and widespread adoption of renewable energy. This is partly due to
the fact that coal is abundant and cheap in many countries, including India. There is
also a renewed focus on cleaner methods of coal utilization and hence on developing
a comprehensive understanding of the flow field inside the pulverized coal-fired
furnaces and its interaction with the heterogeneous combustion process.
The flow inside a pulverized coal-fired entrained flow furnace is inherently
unsteady and involves complex turbulent mixing of two phases, i.e., coal parti-
cles in solid phase and air in the gaseous phase. Other complex phenomena, such
as devolatilization, surface reactions, and complex gas phase chemistry, are also
present. Several numerical studies are available in the literature in which entrained
flow (EF) furnaces and gasifiers functioning at relatively high Re or fully turbu-
lent conditions [1–4] have been extensively explored. These researches presented
successful modeling of the physio-chemical processes, i.e., turbulence [1, 2], char
kinetics [3], gas phase chemistry [4], and radiation interaction between particles and
participating optically thick media [2, 4] occurring in the furnaces and gasifiers.
Recently, Sung et al. [5] performed experiments on pulverized coal-fired furnace
of Pusan National University, South Korea. It has two inlets, primary and secondary,
for feeding pulverized coal and air along with a pair of annular inlets to supply
N. K. Sahu · M. Kumar
Department of Mechanical Engineering, Indian Institute of Technology, Delhi, New Delhi
110016, India
e-mail: kmayank@mech.iitd.ac.in
A. Dewan (B)
Department of Applied Mechanics, Indian Institute of Technology, Delhi, New Delhi 110016,
India
e-mail: adewan@am.iitd.ac.in
74 N. K. Sahu et al.
liquified petroleum gas (Fig. 1). The air enters through the primary inlet axially along
with pulverized coal as the dispersed phase and through the secondary it enters with a
swirl, created by the swirl-vanes. The furnace was operated at atmospheric pressure
and low turbulence intensity conditions. No studies are reported in the literature
on modeling of a furnace with low turbulence intensity flows. Proper modeling
of turbulence is critical in turbulent multiphase reacting flows, because it directly
impacts the flow, mixing, and combustion processes. Thus, a computational model
for the Pusan National University furnace, providing due emphasis on the important
sub-models for the physio-chemical processes occurring in the furnace, has been
developed in the present study. Primary focus of the study is on the turbulence
modeling of the flow for which Reynolds-averaged Navier–Stokes (RANS) approach,
being computationally affordable, was used. Three turbulence models were tested for
simulating low turbulence intensity flow field and their performance was validated
by comparing the predictions of temperature and species concentrations against the
experimentally available data. The numerical study was performed using ANSYS
Fluent 19.2 commercial software.
Fig. 1 Schematic of Pusan

National University furnace
A Computational Study of Entrained Flow Furnace with Swirl Burner … 75
2 Simulation Conditions
The geometry and operating conditions of the pulverized coal combustion experi-
ments in a 16 kWth test facility [5] at the Pusan National University, South Korea,
were used in the present computational study.
3 Geometry
The furnace at the inlet has a co-axial structure with the main burner port (inner
diameter 10 mm) and the annular slit (width 5 mm). Secondary air entered through
the annular slit port after encountering an axial swirler with the vane angle of 60°.
A diverging section with 30° half-angle (burner-quarl) was located at the end of
the burner port. Based on the burner and the furnace geometry, a three-dimensional
hybrid mesh with 420,504 elements, minimum orthogonal quality of 0.16, maximum
aspect ratio 24, and maximum skewness of 0.81 was generated. The consideration
of y+ < 8 was strictly followed.
4 Operating Conditions
The experiment was performed using particles with the mean diameter of 107 μm.
Australian high volatile bituminous Glencore coal was used for the experimental
investigation. The remaining operating conditions related to the experiment are given
in Table 1. The numerical cases were solved with the pressure-based coupled solver
with the assumption of a statistically stationary flow.
The turbulence intensity for the primary and secondary inlets was set to 1% and
5%, respectively, and the intermittency for the transition SST k-ω model was set to
0.1 and 0.9, respectively.
Table 1 Operating
Value
conditions [5]
Coal mass flow rate (kg/h) 1.14
Primary air mass flow rate (kg/h) 0.77
Secondary air mass flow rate (kg/h) 22.14
Primary air inlet temperature (K) 338
Secondary air inlet temperature (K) 573
Secondary air swirl number 1.33
LPG mass flow rate (kg/h) 0.61
5 Modeling Approach
Mass, momentum, energy, species transport, and radiative transfer equations were
numerically solved for both the discrete and continuum phases. The two phases were
both physically and chemically coupled by two-way coupling of mass, momentum,
energy, and species. The coupling was modeled using source in cell approach due
to the sub-grid size of pulverized coal particles. Source terms in the continuum
conservation equations arising from source in cell approach were updated after each
of the 100 iterations of continuum conservation equations.
5.1 Gas Phase Models
5.1.1 Chemistry
The homogenous chemistry was modeled using Jones and Lindstedt (J&L) kinetics
mechanism for hydrocarbon fuels. The mechanism includes water gas shift reac-
tion both forward and backward, H2 oxidation, water dissociation, and reactions of
volatiles and LPG constituents with O2 and H2 O.
The turbulent chemistry interaction for the continuum was modeled using finite
rate/eddy-dissipation concept which evaluates the reaction rate using the minima of
kinetics and turbulent mixing rate. The turbulence mixing rate on the RANS scale
is governed and was thus modeled using large eddy mixing time scale [6] expressed
as:
t = C(k/ε) (1)
where t is large eddy mixing time scale, k is turbulent kinetic energy, ε is dissipation
rate of turbulence, and C is a constant derived for high turbulence intensity flows.
5.1.2 Radiation
Radiation dominates heat transfer in the combustion environment, especially within

the flame zone, and it was modeled by solving the radiative transfer equation (RTE).
In the present study, Discrete Ordinates (DO) model was used to solve the RTE. Each
octant of the angular space 4π was discretized into 4 × 4 solid angles, and a total of
64 RTE equations were solved in the three-dimensional space.
5.1.3 Turbulence
Various RANS models were used for modeling turbulence. The flow inside the
considered system has a low turbulence intensity and thus low-Re turbulence models;
that is, SST k-ω model with low-Re correction and transition SST k-ω model were
used for modeling turbulence. Further, a comparative study was also performed using
the standard SST k-ω model. The two low-Re turbulence models used here are modi-
fied versions of the existing SST k-ω model. The reason for their selection rests in
the ability of the SST k-ω model in simulating fully turbulent swirling flows in EF
systems with a reasonable accuracy [1]. The computational outputs obtained using
three different turbulence models, i.e., SST k-ω, SST k-ω with low-Re correction,
and transition SST k-ω were compared against the experimental results.
5.2 Solid-Phase Models
The solid-phase model accounts for the coal conversion via pyrolysis and char
consumption, and particle transport. The particle transport model solves the mass,
momentum, and energy equations for the particles along the jet trajectories. The
pulverized coal particles in EF systems are widely dispersed in the flow and are thus
tracked using Eulerian approach [4] for which discrete particle model (DPM) was
used.
The pulverized coal conversion was typically modeled as a sequential process
of multiple physio-chemical processes. The sequence begins with inert heating,
followed by evaporation, devolatilization, and finally char consumption. The
devolatilization was modeled using single kinetic rate model whose mathematical
expression is similar to the Arrhenius form. The values of frequency factor and
activation energy for the single kinetic rate expression were referred to from the
work of Chen et al. [2]. The product of coal devolatilization is called char. The
temperature profile around the char particle, char surface evolution, and kinetics of
char gasification reaction are the three factors that influence the char consumption
rate in a defined chemical environment. In the present study, the temperature and
species concentration profile around the char particle and char surface evolution
were modeled using single film model and shrinking core model, respectively. The
kinetics of char oxidation and gasification reactions were derived from the work of
Chen et al. [2].
Using the above-mentioned sub-models, a comprehensive computational model
was developed for simulating the Pusan National University furnace. Comparisons
of the results obtained from these simulations against the experimental available data
are discussed in the next section.
Figure 2 shows a comparison between the experimentally and computationally

obtained radial temperature profiles. Figure 2a, which presents temperature profiles
relatively close to the burner, shows overprediction of temperature in the annulus
region of furnace. Figure 2b and c presents temperature profiles at axial locations
further away from the burner and show that the transition SST k-ω model predicts
Fig. 2 A comparison of
simulated temperature
profile with simulated data
a at Lance-2 (210 mm), b at
Lance-3 (355 mm), and c at
Lance-4 (535 mm) from
burner port
temperature profile with better accuracy, whereas the other turbulence models signif-
icantly underpredict and overpredict the temperature in the furnace core and annulus
regions, respectively.
The turbulence models SST k-ω and SST k-ω with low-Re correction predict the
solution with a similar accuracy in the furnace core, and the predictions by these two
models in the annulus region are not significantly different either.
The transition SST k-ω model is a local correlation-based model which incorpo-
rates criteria required for the transition of a low turbulence intensity flow to a higher
turbulence intensity flow based on critical values of local variables, i.e., strain rate-
based Re and intermittency. This model, based on intermittency, classifies the flow
to be of low or high turbulence intensity. A high turbulence intensity region in flow
diffuses its fluctuations to low turbulence intensity regions and thereby enhances
the turbulent diffusivity. Our contention is that better prediction of the turbulence
intensity field leads to a better prediction of the temperature field as well. Thus, this
model predicts temperature with better accuracy as it diffuses turbulence intensity of
the secondary air to the near-wall region and the core of the furnace more effectively
as compared to the other turbulence models used.
Figure 3a and b shows variations of O2 and CO2 volume fractions along the
central axis of the furnace, respectively. The present results underpredict O2 and
Fig. 3 Axial variations of

a O2 and b CO2 percentage
volume fractions
CO2 fractions close and away from the burner, respectively. The transition SST k-
ω model still performs marginally better compared to other turbulence models as
it predicts O2 concentrations away from the burner with a marginal accuracy. The
reason for the deviation in prediction of O2 and CO2 concentration axial profiles
can be attributed to the implementation of high turbulence intensity-based turbulent
chemistry interaction model which defines the mixing rate based on Eq. (1).
7 Conclusions
The multiphase reactive flow has been investigated using the SST k-ω, SST k-ω with
low-Re correction, and transition SST k-ω model with two-way coupling of mass,
momentum, and energy. It can be concluded from the present study that transition
SST k-ω model performs better for low turbulence intensity flows compared to other
turbulence models used and there is a need to modify turbulence chemistry interaction
models in order to better describe multiphase reactive low turbulence intensity flows.
Acknowledgments The computational facilities for the simulations were provided by Department
of Mechanical Engineering, Indian Institute of Technology Delhi, New Delhi – 110016, India.
References
1. Kumar M, Ghoniem AF (2012) Multiphysics simulations of entrained flow gasification, part I:

validating the nonreacting flow solver and the particle turbulent dispersion model. Energy Fuels
26:451–463
2. Chen L, Ghoniem AF (2012) Simulation of Oxy-Coal combustion in a 100 kWth test facility
using RANS and LES: a validation study. Energy Fuels 26:4783–4798
3. Kumar M, Zhang C, Monaghan R, Singer SL, Ghoniem AF (2009) CFD simulation of EFG
with improved devolatilization and char consumption submodels. In: Proceedings of ASME
2009 international ME congress & exposition, USA, vol 3 pp 383–395
4. Abani N, Ghoniem AF (2013) Large eddy simulations of coal gasification in an entrained flow
gasifier. Fuel 104:664–680
5. Sung Y, Moon C, Eom S, Choi G (2016) Coal-particle size effects on NO reduction and burnout
characteristics with air-staged combustion in a pulverized coal-fired furnace. Fuel 182:558–567
6. Spalding D (1971) Mixing and chemical reaction in steady confined turbulent flames. Symp
(Intl) Comb 13:649–657
Numerical Investigation on the Effect
of Various Geometries in a Solar
Box-Type Cooker: A Comparative Study
B. C. Anilkumar, Ranjith Maniyeri, and S. Anish
1 Introduction
Solar cooker is one of the important solar energy utilization equipment which converts
the solar energy into useful heat energy. An increase in the prices of non-renewable
sources of energy demands the market for solar cookers. Solar cookers are clas-
sified as panel, parabolic and box cooker. Accordingly, various types of cookers
are designed, developed and tested by different researchers for the last four to five
decades. Box-type solar cookers are mostly used as it is of simple design and easy
to operate without much risk. The direct and diffuse radiations go directly through
the glass cover, and radiation is reflected from the internal surfaces of the cooker.
As of now, a lot of research works have been carried out in both developing
and developed countries which shows the need for advanced solar cookers, espe-
cially box-type with minimum heat loss from the cooker cavity. Thermal perfor-
mance study by experimental investigations on various types of solar box cookers
has been conducted and presented by different authors [1–7]. A lot of such works
have been carried out to study the thermal performances of different box-type cookers
by obtaining performance parameters such as first figure of merit (F 1 ), second figure
of merit (F 2 ), energy efficiency and exergy efficiency.
Verdugo [8], Guidara et al. [9] and Cuce [10] developed and numerically solved the
heat transfer processes in box-type solar cooker. Recently, some design modifications
and new thermal performance parameters were proposed by Sagade et al. [11] and
Collares-Pereira et al. [12].
B. C. Anilkumar · R. Maniyeri (B) · S. Anish

Department of Mechanical Engineering, National Institute of Technology Karnataka, Surathkal,
Mangalore 575025, Karnataka, India
e-mail: mranji1@nitk.edu.in
S. Anish
e-mail: anish@nitk.edu.in
82 B. C. Anilkumar et al.
Solar cookers used today have large amount of heat losses from the cooker cavity
due to conduction, convection and radiation which results in reduced effectiveness
of cooker. The heat losses taking place from solar box-type cooker are a cause of
concern and have to be minimized. As the fabrication and testing of solar box-
type cookers of various geometries are difficult, the analysis of the same is to be
done numerically. In the present work, the objective is to numerically investigate the
maximum absorber plate temperature and heat losses in solar box-type cookers of
different geometries using commercial software ANSYS16. The present study also
aims to experimentally investigate the thermal performance parameters proposed by
different authors.
2 Methodology
The description about experimental set-up, mathematical relations and numerical

analysis is presented in this section.
2.1 Experimental Study
The solar cooker used in the present study consists of rectangular box made up of
steel having outer dimensions (650 mm × 650 mm), inner dimensions (600 mm
× 600 mm) and height 250 mm with a double-glazed glass at the top (Fig. 1). The
absorber plate is made up of aluminium sheet painted with matte black. A cylindrical-
shaped aluminium vessel having a flat base and a lid of diameter 18 cm and height
6 cm is used as cooking vessel for the test.
The experimental procedure for the performance test is conducted as per IS 13429.
The stagnation test is conducted for finding the first figure of merit (F 1 ) of solar box
Fig. 1 Experimental set-up

Numerical Investigation on the Effect of Various Geometries in a Solar … 83
cooker. During this test, cooker is placed in an open area without load by covering the
reflector. Temperatures of absorber plate, ambient air and cooker cavity are measured
using thermocouples at each interval of time till the stagnation condition is obtained.
Intensity of solar radiation is measured using calibrated solar power meter.
The first figure of merit F1 is given by,
Tps − Tas
F1 = (1)
Hs
where ‘H s ’ is the solar radiation and ‘T as ’ is the ambient air temperature when the
absorber plate attains stagnation temperature, ‘T ps ’.
The thermal load test is conducted to find the second figure of merit (F 2 ). For this
test, cooking vessel having 1 kg of water is placed on the absorber plate of the cooker
and recorded the time for raising the temperature of water from T w1 (60–65 °C) to
T w2 (90–95 °C).
The second figure of merit is given by,

1 Tw1 −Ta

F1 (MC)w 1− F1 H
F2 = ln
1 Tw2 −Ta
(2)
Aτ 1− F1 H
where ‘A’ represents aperture area of the solar cooker, ‘τ ’ is the time interval for
which water temperature rises from T w1 to T w2 , ‘H’ is the average solar radiation
and ‘T a ’ is the ambient temperature.
2.2 Numerical Study
The different geometries of solar box cooker cavities used for the analysis are
rectangular, cylindrical, trapezoidal and frustum of a cone. These box geometries
are modelled and analysed using computational fluid dynamics (CFD) software
ANSYS16. The dimensions considered for the modelling of box geometries are in
such a way that aperture area of cooker is same. The models of different geometries
are shown in Fig. 2.
Discrete transfer radiation model (DTRM) is used to solve the heat losses in the
cavity which accounts for the radiation exchange in the cooker cavity along with
convection loss. The governing equation for the analysis is mass, momentum and
energy as given by,
∂u ∂v
+ =0 (3)
∂x ∂y
2
∂u ∂u 1 ∂p ∂ u ∂ 2u
u +v =− +v + (4)
∂x ∂y ρ ∂x ∂x2 ∂ y2
(a) Rectangular cavity (b) Cylindrical cavity
(c) Trapezoidal cavity (d) Frustum of a cone cavity
Fig. 2 Models of different box geometries
2
∂v ∂v 1 ∂p ∂ v ∂ 2v
u +v =− +v + + gβ T − Tg (5)
∂x ∂y ρ ∂y ∂x 2 ∂y 2
2
∂T ∂T ∂ T ∂2T
u +v =α + (6)
∂x ∂y ∂x2 ∂ y2
2.2.1 Validation of Results
The absorber plate temperature obtained from numerical analysis using ANSYS16
is compared with the readings indicated by thermocouple located on the absorber
plate of rectangular cavity solar cooker during experimentation. The plot of absorber
Fig. 3 Variation of absorber plate temperature with time
temperature with time as shown in Fig. 3 indicates that there is a close agreement
between the experimental and numerical results. This validates the simulation model.
3.1 Results of Experimental Study
The thermal performance test on box-type solar cooker is conducted as per IS 13429.
From stagnation test, first figure of merit is found to be 0.08, which is lower than 0.12
(A-grade). This indicates that there are higher convection and radiation losses from
side walls. The temperature profile recorded during the sensible heating test (load
test) is shown in Fig. 4. The peak absorber plate temperature is obtained between
1.30 and 2.30 PM as 106˚C, whereas water temperature is 98 °C between 2.30 and
3.30 PM. The second figure of merit is found to be 0.261. This indicates that cooker
has good thermal performance and able to cook food fast and at high temperature.
The energy and exergy efficiency are found to be 16.1% and 0.61%, respectively.
3.2 Results of Numerical Analysis
From numerical analysis, the steady-state solution is determined using the solver in
ANSYS16. Figure 5 shows the variation of absorber plate temperature with time for
Fig. 4 Variation of different temperatures with time
Fig. 5 Variation of absorber plate temperature for different geometries
various geometries of box-type solar cooker. For the trapezoidal geometry, the peak
temperature of absorber plate is 81.45 °C at 12 noon, which is higher than all other
geometries. This may be due to the lesser absorber plate area and the ability to absorb
a larger fraction of incident solar radiation falling on the aperture at larger incidence
angle. The frustum cavity-based design is found to achieve better result at times
when the solar radiations are less. This indicates that the heat losses from frustum
of cone cavity are lower than other geometries and yield better thermal performance
throughout the day.
The results also show that the maximum temperature is attained by the lid of
the cooking vessel when the cooking vessel is placed on the top of absorber plate.
Figures 6 and 7 shows the temperature contours of trapezoidal cavity cooker with
and without cooking vessel. The peak temperature of lid of the cooking vessel and
absorber plate is found to be 85.45 °C and 80.15 °C, respectively. This may due to
the effect of the cooking vessel blocking the solar irradiance from falling onto the
absorber plate. Also, the convective and radiative losses in trapezoidal cavity without
cooking vessel are found to be 33% and 67%, respectively, but it was 27% and 73%,
respectively, when a cooking vessel is placed. This is because, as the temperature of
black cooking pot differs considerably from the ambient temperature, the radiative
losses will increase more quickly.
Fig. 6 Temperature contour of trapezoidal cavity solar cooker without cooking vessel
Fig. 7 Temperature contour of trapezoidal cavity solar cooker with a cooking vessel
4 Conclusions
In this paper, we have investigated the effect of various box geometries such as
rectangular, trapezoidal, cylindrical and frustum of a cone on the performance of
solar cookers. There is good agreement between experimental and simulated results
for absorber plate temperature of rectangular cavity solar box-type cooker which
validates the computational model. From experimental investigation, performance
parameters such as first figure of merit (F 1 ) and second figure of merit (F 2 ) are found
to be 0.08 and 0.261 respectively, which indicates that there is large convection and
radiation heat loss from the cooker. The analysis of absorber plate temperatures of
solar cookers with various geometries was done by computationally, and trapezoidal
cavity attains highest temperature. This shows that trapezoidal cavity will give more
thermal performance than other geometries. The frustum of cone cavity achieves
consistently higher temperatures at times when solar radiations were less.
As the temperature of the black cooking pot is more than the absorber plate
temperature and differs considerably from their surroundings, the radiative heat loss
will increase more quickly. This is because radiative loss being the function of the
difference between the fourth power of the absolute temperatures of the cooking pot
and surroundings. Hence, the percentage of radiative losses in all geometries of solar
box cookers with cooking pot is found to be more than the cooker without cooking
pot.
Acknowledgements The authors gratefully acknowledge the support from Lal BahadurShastry
College of Engineering, Kasaragod, Kerala, India.
References
1. Vaishya JS, Tripathi TC, Singh D, Bhawalker RH, Hegde MS (1985) A hot box solar cooker:
performance analysis and testing. Energy Convers Manage 25(3):373–379
2. Tiwari GN, Yadav YP (1986) A new solar cooker design. Energy Convers Manage 26(1):41–42
3. Mullick SC, Kandpal TC, Saxena AK (1987) Thermal Test Procedure for Box-Type Solar
Cookers. Sol Energy 39(4):353–360
4. Nahar NM (2001) Design, development and testing of a double reflector hot box solar cooker
with a transparent insulation material. Renew Energy 23:167–179
5. Ekechukwu OV, Ugwuoke NT (2003) Design and measured performance of a plane reflector
augmented box-type solar-energy cooker. Renew Energy 28:1935–1952
6. El-Sebaii AA, Ibrahim A (2005) Experimental testing of a box-type solar cooker using the
standard procedure of cooking power. Renew Energy 30:1861–1871
7. Misra R, Aseri TK (2012) Thermal performance enhancement of box type solar cooker: a new
approach. Int J Sustain Energy 31(2):107–118
8. SoriaVerdugo A (2015) Experimental analysis and simulation of the performance of a box type
solar cooker. Energy Sustain Develop 29:65–71
9. Guidara Z, Souissi M, Stern AM, Maalej A (2017) Thermal performance of a solar box cooker
with outer reflectors: numerical and experimental investigation.Solar Energy 158(2017) 347-
359.
10. Cuce E (2018) Improving thermal power of a cylindrical solar cooker via novel micro/nano
porous absorbers: a thermodynamic analysis with experimental validation. Solar Energy
176:211–219
11. Sagade AA, Samdarshi SK, Panja PS (2018) Experimental determination of effective
concentration ratio for solar box cookers using thermal tests. Solar Energy 159:984–991
12. Collares-Pereira M, Cavaco A, Tavares A (2018) Figures of merit and their relevance in the
context of a standard testing and performance comparison methods for solar box-cookers. Solar
Energy 166:21–27
Sharp Interface Level Set Method
on a Co-located Grid for High Density
Ratio Two-Fluid Flow
C. S. Sanjid, Janani Srree Murallidharan, and Atul Sharma
1 Introduction
Co-located grid-based multiphase simulations are more challenging than simulations

performed in staggered grid. Numerical methods in co-located grid-based multiphase
simulation are incapable to handle multiphase flows with high density ratios.
Francois et al. [1] proposed that the inability on a co-located grid is due to the
dissimilarity prevailing during numerical discretization of interfacial forcing terms
and pressure force. Following this observation, Francois et al. [1] introduced a “Bal-
anced Force Method”, by estimating interfacial force at the face of control volume
which is consistent with the estimation of pressure force. Since both the forces are
the faces of control volume, these terms could be treated and discretized together.
Following this, majority of the numerical methods in co-located grid [2–5] utilized
balanced force approach. However, this approach involves compromising physical
reality (estimating interfacial force at the face of control volume instead of the inter-
face), in order to achieve a stable numerical approach. This physical inconsistency
prompted us to attempt the development of an alternate methodology.
Thus, in the present investigation, a sharp interface level set method (SI-LSM) is
implemented in co-located grid framework, which thereby enables us to evaluate the
interfacial forces at the interface location. The development of multiphase algorithm
in co-located grid enables multiphase simulation in complex geometries feasible.
C. S. Sanjid · J. S. Murallidharan · A. Sharma (B)

Department of Mechanical Engineering, IIT Bombay, Powai, Maharashtra 400076, India
e-mail: atulsharma@iitb.ac.in
J. S. Murallidharan
e-mail: js.murallidharan@iitb.ac.in
92 C. S. Sanjid et al.
2 Numerical Methodology
In this section, the governing equations adopted and its solution methodology for the
numerical simulation will be discussed in detail.
2.1 Flow Properties
The conservation equations solved in the present study are continuity equation and
momentum equation. A non-dimensional form of the continuity equation is given as,
=0
∇.U (1)
is the non-dimensional velocity vector, while the non-dimensional form of

where U
momentum equation is,
∂ U ∇P 1 1
+ ∇. U U = − + ∇.(2ηi D) − ĵ (2)
∂τ χi χi Re1 Fr 2
where τ , P, χ , Re, η, D and Fr denote non-dimensional time, non-dimensional pres-

sure, density ratio, Reynolds number, viscosity ratio, rate of deformation tensor and
Froude number, respectively. Using characteristic length (lc ), velocity scale (uc ) and
time scale (t c ), the non-dimensional form of variables is defined as,
x u t p
X = ; U = ; τ = ; P = (3)
lc uc tc ρ1 u 2c
where x, u, t and p represent dimensional variables like spatial vector, velocity
vector, time and pressure, respectively. In Eq. 2, subscript ‘i’ corresponds to the fluid
domain in which solution is calculated. For all the calculations, Fluid 1 is taken as
the reference fluid.
To calculate the bulk velocity, Eqs. 1 and 2 are solved here using a projection
method. The Navier–Stokes equation is converted into a linear algebraic equation
using finite volume method and a semi-explicit method of Sharma [6]. In Eq. 2, the
pressure and diffusive terms are taken implicitly, while rest of the terms are taken
explicitly. The methodology adopted by Shaikh et al. [7] is utilized in the present
study for co-located grid. A brief explanation of this methodology is presented here.
U P∗ − U Pn 1
= − ∇. U n U n + n ∇. 2ηin D n+1
τ χi Re1
Sharp Interface Level Set Method on a Co-located Grid … 93
1
− ĵ (4)
Fr 2
In the projection method, the cell center velocity is predicted by neglecting the
pressure term as shown in Eq. 4. Here, advective term is discretized using LinLin
TVD scheme given by Date [8], and diffusive terms are discretized using central
differencing scheme. Discretizing Eq. 2 similar to Eq. 4 with pressure term and
subtracting the resultant equation from Eq. 4 (by neglecting the diffusion term) will
result in an equation for pressure, given as,
U Pn+1 − U P∗ ∇ P n+1
=− (5)
τ χin
Taking divergence of Eq. 5 and applying continuity equation, a Poisson equation

for pressure is obtained as,

∇ P n+1 ∗
τ ∇. = ∇.U (6)
χin
A detailed discretization procedure regarding numerical treatment required for the

interfacial cell for the solution of pressure Poisson equation is available in Shaikh
et al. [7]. Since in this study a co-located grid is employed, the pressure–velocity
decoupling is avoided using momentum interpolation of Rhie and Chow [9].
2.2 Interface
Since a level set method is used to capture the interface, level set advection equation
and re-initialization equation as shown in Eqs. 7 and 8 are the required auxiliary
equations.
∂ϕ a .∇ϕ = 0
+U (7)
∂τ
where ϕ is the level set function and U a is the advection velocity with which level
a is same as the
set function is advected. Since there is no interfacial mass transfer, U
bulk velocity calculated from momentum equation. In this study, the spatial gradients
of the level set function are discretized using fifth order WENO scheme. Once the
level set function is advected using bulk velocity, this function loses its property of
being a normal distance function. This results in an inaccurate calculation of normal
and curvature of interface. To resolve this issue, as proposed by Sussmann [10], a
re-initialization equation needs to be solved.
∂ϕ
= Sε (ϕ0 )(1 − |∇ϕ|) (8)
∂τs

where τs = X/10 is the pseudo-time step and Sε (ϕ0 ) = ϕ0 ϕ02 + X2 is the
smoothened sign function.
In this section, the ability of above-described numerical methodology to handle high

density ratio cases is demonstrated by a verification study on several benchmark
problems.
3.1 Young–Laplace Law Test
Spurious currents present a major obstacle for codes trying to achieve accurate numer-
ical solutions. In order to test the present sharp interface level set code’s ability to
handle high density ratio cases with minimal spurious currents, the Young–Laplace
Law test case was selected.
An equilibrium rod problem proposed by Brackbill et al. [11] is investigated. In
this case, as shown in Fig. 1a, an infinitely long cylindrical rod with diameter ‘D’
is placed in a quiescent fluid. The fluids are considered to be inviscid placed in a
Fig. 1 a Computational setup used for Young–Laplace Law test case and b computed pressure
contour
Table 1 Non-dimensional
P/Pact U max V max
results obtained for
Young–Laplace Law test for a 1.06 6.861e−6 7.748e−6
grid size of 50 × 50
zero gravity environment. Hence, the pressure jump across the interface is balanced
only by the surface tension force, p = σ κ. Ideally, at equilibrium, the velocity
field throughout the computational domain should be zero. But the inaccuracy in
curvature estimation and numerical methodology results in generation of spurious
currents. In the present study, a cylindrical rod with radius 0.02 m and domain size of
1.5R × 1.5R is selected. Density is taken as 1000 kg/m3 for heavier and 500 kg/m3
for lighter fluid, and surface tension coefficient is taken as 0.0236 N/m. Figure 1b
shows the pressure contour with a sharp jump across the interface.
The results obtained for the Young–Laplace test case are shown in Table 1. The
non-dimensional maximum spurious velocities (U max and V max ) indicate that the
magnitude of the spurious currents is small and does not affect the solution signif-
icantly. The simulation was continued till 1000 iterations to compare the initial
and final location of interface. The interface location coincides showing minimal
influence of spurious currents.
3.2 Dam Break
In this section, the SI-LSM in a co-located grid is tested on the dam break problem, a
commonly adopted benchmark problem for multiphase simulation. In this problem,
as shown in Fig. 2a, a vertical fluid column is kept initially in a quiescent environ-
ment, physically made possible by placing a vertical plate. Once this vertical plate
is removed, due to gravity, the fluid column will collapse to the bottom with its
leading edge moving in the horizontal direction. Martin and Moyce [12] conducted
an experiment for this phenomenon by measuring the leading edge distance with
time.
Fig. 2 a Initial configuration of interface for simulation and b leading edge position at
non-dimensional time, τ = 2
Fig. 3 Validation of dam

break problem
For simulation as adopted by Javed et al. [7], a computational domain of non-

dimensional size 5.36 × 3 is selected with water column of dimension 2 × 1. Density
is taken as 1000 kg/m3 for heavier and 500 kg/m3 for lighter fluid, and surface
tension coefficient is taken as 0.0236 N/m. For solving non-dimensional conservative
equations, length scale is taken as initial height of fluid column, and velocity scale
is taken as (glc )0.5 . Figure 2b shows the final interface deformation, obtained from
SILSM on a co-located (red) and staggered grid (black). Figure 3 shows the validation
of leading edge distance with time, with an excellent agreement.
3.3 Falling Droplet
To evaluate the performance of surface tension model implemented in the code, a

falling droplet problem is considered. A water droplet with radius ‘R’ is placed at
certain height ‘H’ above a solid wall. This droplet, which is initially at a quiescent
condition, is allowed to drop. Kang et al. [13] performed a numerical simulation of
the aforementioned problem and presented the location and size of the droplet—at
various time instances. Hence, in the present study, the drop aspect ratio measured
from Kang et al. [13] at various time instances is used for verification.
For simulation, a computational domain with non-dimensional size 3 × 4.5 and
an initial droplet radius of 3.33 mm is selected. The thermo-physical properties of
the fluids are same as that in dam break problem. The initial diameter of the droplet
and (glc )0.5 are taken as length scale and velocity scale, respectively. Figure 4 shows
the prediction of the drop aspect ratio (ratio of drop diameter in vertical direction to
drop diameter in horizontal direction) at various time instances, as predicted by the
present code. As shown, excellent agreement is observed with the literature.
Fig. 4 Verification of falling

droplet problem
4 Conclusions
In the present study, a sharp interface level set method is implemented on a co-located
grid instead of a conventional “Balanced force method”. In conventional methods
for co-located grid, the interfacial forces are applied at the face of control volume
to ensure similarity in discretization of interfacial and pressure force. To ensure
physical consistency of the multiphase problem, i.e., the interfacial forces must act
at the interface, a SI-LSM is implemented. The capability of this methodology to
handle high density ratio cases in co-located grid is then established by validating
against several benchmark problems like Young–Laplace test case, dam break and
falling droplet. An excellent agreement between the present and published results is
demonstrated, ensuring the capability of this numerical methodology in co-located
grid to handle high density ratio flows.
Acknowledgments The authors would like to thank Nitin Goyal for the support during numerical
development.
References
1. Francois MM, Cummins SJ, Dendy ED, Kothe DB, Sicilian JM, Williams MW (2006) A
balanced-force algorithm for continuos and sharp interfacial surface tension models within a
volume tracking framework. J Comput Phys 213(1):141–173
2. Herrmann M (2008) A balanced force refined level set grid method for two-phase flows on
unstructured flow solver grids 227(4):2676–2706
3. Ni MJ (2009) Consistent projection methods for variable density incompressible Navier-Stokes
equations with continous surface forces on a rectangular collocated mesh 228(18):6938–6956
4. Popinet S (2009) An accurate adaptive solver for surface-tension-driven interfacial flows
228(16):5838–5866
5. Montazeri H, Zandavi SH, Bazylak A (2017) Sharp interface models for two-phase flows:
Insight towards new approaches. Comput Methods Appl Mech Eng 322:238–261
6. Sharma A (2017) Introduction to computational fluid dynamics: development, application and
analysis. Wiley & Athena Academic UK and Ane Books Pvt. Ltd., India
7. Shaikh J, Sharma A, Bhardwaj R (2018) On comparison of the sharp-interface and diffuse-
interface level set methods for 2D capillary or/and gravity induced flows 176:77–95
8. Date AW (2005) Introduction to computational fluid dynamics. Cambridge University Press,
New York
9. Rhie CM, Chow WL (1983) Numerical study of the turbulent flow past an airfoil with trailing
edge separation. AIAA J 21(11):1525–1532
10. Sussman M, Smereka P, Osher S (1994) A level set approach for computing solutions to
incompressible two-phase flow. J Comput Phys 114:146–159
11. Brackbill JU, Kothe DB, Zemach C (1992) A continuum method for modeling surface tension.
J Comput Phys 100(2):335–354
12. Martin JC, Moyce WJ (1952) An experimental study of the collapse of liuid columns on a rigid
horizontal plane., Philosophical transactions of the Royal society of London. Series A, Math
Phys Sci. 244(882):312–324
13 Kang M, Fedkiw RP, Liu XD (2000) A boundary condition capturing method for multiphase
incompressible flow. J Sci Comput 15(3):323–360
Computational Studies on Combustion
Instabilities of Afterburner for Different
Equivalence Ratios
Srinivasa Rao Gurrala and Andavan Shaija
1 Introduction
Military jet aircrafts require minimum thrust while cruise and maximum thrust during
take-off or manoeuvres and/or emergencies. Designing the afterburner, keeping in
view of the thrust required during take-off or emergency conditions, will increase the
engine size and specific fuel consumption. Hence, the best way is to have an after-
burner, which only works for a specific period, for supplying the required additional
thrust.
The disadvantages with the afterburner are the instabilities viz. buzz and screech.
The low-frequency acoustic waves, typically 50–200 Hz, travelling axially along
the engine afterburner and bounded by the fan or turbine and propelling nozzle, are
called buzz. The buzz frequency depends on the flight conditions and fuelling level.
Buzz occurs at low frequencies and hence are not of much concern to the safety or
performance of the aircraft. The high-frequency (500–5000 Hz) radial amplitudes,
called screech [1], travelling perpendicular to the fluid flow and the jet pipe axis, will
result in catastrophic failure of the jet aircraft, if proper steps are not taken, to mitigate
the same. Hence, the flow needs to be studied for the maximum production of thrust
and achieving the desired Mach number as well as to mitigate the screech combustion
instabilities. The experimental setups associated with the afterburners are very costly
and complex and are limited to Government Research Laboratories. Hence, very few
publications related to the experimental work are seen in the literature. Useller et al.
[2] experimentally investigated the influence of combustion chamber length on the
afterburner performance under conditions simulating altitude flight and found that the
most favourable combustion chamber length is 5 feet, considering both performance
S. R. Gurrala (B)
Indian Naval Academy Ezhimala, Dist Kannur, Kerala 670310, India
A. Shaija
Mechanical Engineering Department, NIT, Calicut, Kerala 673601, India
e-mail: shaija@nitc.ac.in
100 S. R. Gurrala and A. Shaija
and cooling requirement. Reacting flows in an aero-engine afterburner were analysed

using computational fluid dynamics by Unaune and Ganesan [3] for air–fuel ratios
of 30 and 46 and found that the air–fuel ratio of 30 gives better performance as
the pressure loss is within the design limit. The air–fuel ratio is also responsible
for causing the instabilities in the afterburner. Hence, in this chapter, an attempt is
made to test the validity of the argument that the best air–fuel ratio for mitigation of
instabilities in the afterburner is the rich in fuel–air ratio or otherwise.
Combustion instabilities are very difficult to capture and simulated. Carrying out
the transient analysis requires large computational facilities [4]. The reverse effect
of acoustics on flow and combustion could also not be considered due to limita-
tions of software. Hence, a novel method of carrying out steady-state CFD analysis
and importing the mean CFD parameters for acoustic analysis, with the objective to
capture the acoustic cavity modes for the afterburner models with different equiv-
alence ratios, is considered for the study to predict the optimum afterburner model
with minimum combustion instabilities or maximum absorption of acoustic energy.
Imposing the unit amplitude duct modes with time variations, on the mean flow
characteristics of CFD analysis, enabled studying the mean square pressure and
transmission loss for all these cases at various frequencies.
2 Governing Equations
2.1 Equations for CFD Analysis
The governing equations for CFD analysis are presented as given in ANSYS® Fluent
Theory Guide [5].
Mass conservation equation: -
∂ρ
+ ∇.(ρ v) = Sm (1)
∂t
Equation (1) is valid for compressible and incompressible flows. The source Sm
is the mass added to the continuous phase from the dispersed second phase and any
user-defined sources, v is the velocity vector, and t is the time.
2.1.1 Momentum Conservation Equation
The momentum conservation equation is given below: -
∂
(ρ v) + ∇.(ρ vv) = −∇p + ∇. τ + ρ −
→
g + F (2)
∂t
Computational Studies on Combustion Instabilities of Afterburner … 101
where p is the static pressure; τ is the stress tensor; ρ −

→
g and F are the gravitational
force and the external body forces (forces that arise from interaction with dispersed
phase), respectively. F also contains other model-dependent source terms such as
porous media and user-defined sources.
The stress tensor [5] is defined by

2
τ = μ ∇ v + ∇ v − ∇.
T
vI (3)
3
where μ is the molecular viscosity, I is the unit tensor, and the second term on the
right side is the effect of volume dilation.
2.1.2 Energy Conservation Equation
The equation can be written as

⎛ ⎞
∂
v (ρ E + p)) = −∇.⎝keff ∇T −
(ρ E) + ∇.( v ⎠ + Sh
h j j j + τ eff . (4)
∂t j
where keff is the effective conductivity (K + K t ), K is the thermal conductivity,K t

is the turbulent thermal conductivity, and j j is the diffusion flux of species J .
2.1.3 Turbulence Modelling with Realizable k − ε Model
The realizable k − ε contains the alternative formulation for the turbulent viscosity,
and a modified transport equation for the dissipation rate ε has been derived from
an exact equation for the transport of mean square velocity fluctuation. The term
realizable means that the model satisfies certain mathematical constraints on the
Reynolds stresses, consistent with the physics of turbulent flows.
2.1.4 Transport Equations for Realizable k – E Model
The applicable equations [5] are as given below: -

∂ ∂ ∂ μt ∂k
(ρk) + ρku j = μ+ + G k + G b − ρε − Y M + Sk (5)
∂t ∂ xi ∂x j σk ∂ x j

∂ ∂ ∂ μt ∂ε
(ρε) + ρεu j = μ+
∂t ∂ xi ∂x j σε ∂ x j
ε2 ε ε2
+ ρC1 Sε − ρC2 √ + C1ε (G k + C3ε G b ) − C2ε ρ + Sε (6)
K + νε k k

n k
C1 = max 0.43, , n = S , S = 2Si j Si j
n+5 ε
where G k is the generation of turbulence kinetic energy due to the mean velocity
gradients, G b is the generation of turbulence kinetic energy due to buoyancy, Y M is
the contribution of the fluctuating dilatation in compressible turbulence to the overall
dissipation rate, and C1ε , C2ε and C3ε are constants. σk and σε are the turbulent Prandtl
numbers for k and ε. Sk and Sε are user-defined source terms.
Turbulent (or eddy) viscosity is computed as a where Cμ is a constant. C1ε =
1.44, C2ε = 1.92, Cμ = 0.09, σk = 1.0,σε = 1.3 and C3ε = tanh uv , where v
is the component of flow velocity parallel to the gravitational vector, and u is the
component of flow velocity perpendicular to the gravitational vector. Hence, C3ε will
become 1 for buoyant shear layers for which the main flow direction is aligned with
the direction of gravity and become 0 for buoyant shear layers that are perpendicular
to the gravitational vector.
2.1.5 Species Transport Equations
ANSYS® Fluent predicts the local mass fraction of each species, Yi , through the
solution of a convection–diffusion equation for the ith species. The conservation
then can be written as
∂ −
→
(ρYi ) + ∇.(ρ vYi ) = −∇. Ji + Ri + Si (7)
∂t
where Ri is the net rate of production of species I by chemical reaction, and Si is the
rate of creation by addition from the dispersed phase plus any user-defined sources.
2.1.6 Discrete Phase Model
ANSYS® Fluent predicts the trajectory of a discrete phase particle (or droplet
or bubble) by integrating the force balance on the particle, which is written in a
Lagrangian reference frame. This force balance equates the particle inertia with the
forces acting on the particle [5]: -

d u p −→
g ρp − ρ
= FD u − u p + + F (8)
dt ρp

where F is an additional acceleration (force/unit particle mass) term, FD u − u p is
the drag force per unit particle mass, and FD = 18μC D Re
ρ p d 2p 24
.
u is the fluid phase velocity, u p is the particle velocity, μ is the molecular viscosity
of the fluid, ρ is the density,C D is the drag coefficient, ρ p is the density of the
particle,d p is the particle diameter, and Re is the relative Reynolds number which is
ρd |u −
u|
defined as Re ≡ p μp .
2.2 Equations for Acoustic Analysis
The equations solved in Actran™ for acoustic analysis are presented, as given in
Actran™ User Guide [6].
The continuity equation can also be written as
∂ρ
+ ∇.(ρv) = 0 (9)
∂t
where ρ is density, and v is velocity.

Momentum equation can be written in the form
∂v
ρ + ρ∇ B = ρT ∇s + ρv × (∇ × v) − ∇τ (10)
∂t
where T is temperature, τ is viscous stress tensor, s is entropy, h is enthalpy, and

B = h + γ2 = Total enthalpy.
2
Energy conservation
Dh T ∂p
ρ − = ∇(λ∇T ) + ∇(ν.τ ) (11)
Dt ∂t
where λ is material conductivity.

Neglecting the power dissipated by viscous forces and heat
Dh T ∂p
ρ =
Dt ∂t
Combining the mass conservation, momentum conservation and energy conser-
vation equations the Mohring equation for Convected acoustics obtained is used for
the acoustic analysis.

∂ ρ Db ρv Db ρ
+∇ 2 2 − 2 ∇b = R (12)
∂t ρT2 c2 Dt ρT c Dt ρT

∂p
where c2 = ∂ρ s=const
; ρT = total density and
Db DB
b = Scaled enthalpy defined = ρT
Dt Dt

∂ ρ ∂b ρvi ∂b ∂ ρvi ∂b ∂b ∂b ρ
+ + + vi − =R (13)
∂t ρT2 c2 ∂t ρT2 c2 ∂ xi ∂ xi ρT2 c2 ∂t ∂ xi ∂ xi ρT2
where

∂ ρ ∂τi j ∂ vi ∂ρ ∂s ρT ∂s
R=− (v X ω)i − + −
∂ x i ρT ∂x j ∂ xi ρT ∂s p ∂t ρT ∂ x i

∂ 1 ∂ρ ∂s ρvi ∂ρT
+ − 2
∂t ρT ∂s p ∂t ρT ∂ x i
In the present case, the propagation of acoustic perturbations within a homoge-

neous moving medium was considered for the study, taking into consideration, the
limitations of computational facilities and software capabilities. Hence, Eq. (13), in
terms of the propagation of acoustic perturbations within a homogeneous moving
medium, can be expressed in terms of acoustic pressure as follows
2

ik + M.∇ + p = 0 (14)
where k is the wave number, and M is the uniform mean velocity vector divided by
the speed of sound. This equation is also called as the convected Helmholtz equation,
which is used in the present case to calculate the acoustic pressure.
3 Modelling Methodology
The afterburner model is modelled in SolidWorks as explained by Gurrala and Shaija

[7]. The afterburner was computationally analysed for equivalence ratios of 0.58, 0.49
and 0.32 using ANSYS® Fluent, and the results were obtained. These results were
interpolated using Actran™ ICFD, onto the model already imported into Actran™.
The acoustic analysis for each model was then carried out. The modelling procedure
was explained below.
3.1 CFD Modelling with ANSYS® Fluent
The 60° sector of afterburner, was modelled, using unstructured grid with tetrahe-
dral cells for discretizing the domain. The mesh was generated with a maximum of
7,135,549 tetrahedral elements with 1,410,033 nodes. Fine mesh was adopted near
the flame stabilizer, where large gradients in flow can be expected. Course mesh is
adopted, where not much variation in the flow is expected, like the tailpipe region
and nozzle.
Pressure and mass flow boundary conditions for inlet and pressure outlet boundary
conditions for outlet were chosen for the analysis. Wall boundary conditions were
chosen for V-gutter, fuel manifold, diffuser, inner wall, top wall of bypass flow.
Periodic/cyclic boundary conditions were assigned for the side faces. The holes
provided in the liner for anti-screech and cooling rings were applied with interior
boundary conditions. The mass flow rate for main inlet was specified as 11.67 kg/s
for main inlet, pressure 353 kPa and temperature 873 K. The mass flow rate for
bypass inlet was taken as 2.33 kg/s, pressure 396 kPa and temperature 373 K. The
“Extended Volume” and “Outlet” zones were defined as pressure outlets. Kerosene
(liquid) (C12 H23 ) was taken as fuel. The injection points were defined virtually. A
total of 24 spray bars were created, each consisting of nine injection points. Hence,
for a 60° sector model, 36 injection points were created. The mass flow rate of fuel,
from each injection point, was defined, as per the calculated value based on the
air–fuel ratio (or equivalence ratio) of each afterburner model. Energy equation was
enabled for temperature calculations. The discrete phase model was selected. Species
transport with finite-rate/eddy-dissipation model was used for modelling combustion
and solving the multiple simultaneous chemical reactions. The simulations of the
reacting flow analysis was carried out, for the three afterburner models with different
equivalence ratios of 0.58, 0.49 and 0.32, by varying the amount of fuel injected, as
specified earlier, to obtain the simulations for different equivalence ratios of 0.58,
0.49 and 0.32.
The afterburner model was imported into Actran™ extended domain has been
removed and remeshing was carried out. The average size of the mesh is taken as
10 mm as explained below. The ANSYS® Fluent results like velocity, temperature,
speed of sound, pressure and density obtained for these three models with equivalence
ratios of 0.58, 0.49 and 0.32 were imported into Actran™ meshed model for further
acoustic analysis. The boundary conditions employed are also explained below.
3.2 Acoustic Mesh and Boundary Conditions
The average size of the mesh is decided based on the highest frequency range,
i.e. 3000 Hz and accordingly, the average element size has been taken as 10 mm.
Duct modes were used at the inlet and outlet. At the inlet, duct modes excitation
as a combination of incident and reflected wave (constrained) boundary conditions
with semi-infinite duct and at the outlet, non-reflecting circular duct (free) boundary
condition with semi-infinite duct were used to represent the exterior acoustic radi-
ation. Liners are represented as admittance boundary conditions and not modelled
physically in the mesh for acoustic analysis.
A full-scaled model of the afterburner was created in SolidWorks which was imported
into ANSYS® Fluent for computational analysis of fluid flow for the different
equivalence ratios of 0.58, 0.49 and 0.32.
4.1 CFD Results
Figure 1a shows the variation of velocity, along the axis of the afterburner, for
different equivalence ratios. The study increase in the velocity was observed for the
equivalence ratio of 0.58, and the maximum velocity of 1050 m/s was observed. The
velocity was observed increasing near the top region, where the air enters the after-
burner through screech rings. This increase in velocity was observed with increased
equivalence ratios. Figure 1b shows the variation of static temperature along the
axis of the afterburner for different equivalence ratios. The maximum temperature of
1480 K for the equivalence ratio of 0.49 which increased even after the exit. However,
for the equivalence ratio of 0.58, the maximum temperature was observed as 1400 K,
and the temperature decreased as the flow comes out of the exit of the afterburner,
indicating the maximum utilization of heat energy.
Maximum pressure at the inlet and minimum pressure at the exit, along the axis,
was observed for the equivalence ratio of 0.58 from Fig. 2a, which enabled the
breakup of fuel particles into droplets resulting in proper mixing and increase in the
density at the axis of the afterburner as shown in Fig. 2b. The pressure also reduced
to its minimum value at the exit in the case of equivalence ratio of 0.58 and thus
allowing the maximum conversion of this static pressure into velocity.
Fig. 1 Variation of a velocity b temperature along axis

Fig. 2 Variation of a pressure b density along axis
Fig. 3 Variation of a CO2 b O2 along axis
The variation of CO2 mass fraction along the axis of the afterburner was shown in
Figs. 3a and b, respectively, for comparison for the four air–fuel ratios. It was observed
that the maximum CO2 mass fraction of 0.075, and minimum O2 mass fraction of
0.145 was observed in the case of equivalence ratio of 0.58 which indicates proper
combustion with desired thrust.
4.2 Acoustic Results
The mean square pressure (dB power) variation for all these three equivalence ratios is
presented in Fig. 4 for the frequency ranges from 0 to 3000 Hz. The mean square pres-
sure of 134 dB power is observed at 1690 Hz for equivalence ratio of 0.49, whereas,
117 dB power is observed for the equivalence ratio of 0.58. Accordingly, less mean
square pressures of 15-20 dB power, is observed at almost all the frequency ranges
in screech region, for the equivalence ratio of 0.58 than the other two equivalence
ratios.
Transmission loss which is the ratio of incident power (at inlet face of core and
bypass) to transmitted power (outlet face of core and bypass) is shown in Fig. 5.
High transmission loss of 75.5 dB at 530 Hz, 68.3 dB at 1000 Hz and 50.9 dB at
1940 Hz was observed, for the equivalence ratio of 0.58. At all the frequencies, the
transmission loss is more in the case of equivalence ratio of 0.58 which indicates
that the absorption of more acoustic energy due to the proper combustion, formation
Fig. 4 Variation of means square pressure with frequency
Fig. 5 Variation of transmission loss with frequency
of recirculation zone, non-excitation of high amplitude duct modes due to the flame
resonant oscillations leading to more absorption of acoustic energy.
5 Conclusions
A 60° full-scaled afterburner model, with extended domain, is created, and the
reacting flow is evaluated with in the afterburner. The extended domain is created in
order to obtain more accurate results for the analysis of the flow. The acoustic anal-
ysis results carried out with the help of Actran™ software show that the maximum
equivalence ratio of is to be maintained for better absorption of acoustic energy in
order to avoid screech.
References
1. Kampa A (2007) Combustion instability screech in gas turbine afterburners. Ph.D. thesis,
Department of Aerospace Engineering, Indian Institute of Science, Bangalore
2. Useller JW, Braithwaite WM, Rudy CJ (1954) Influence of combustion chamber length on
afterburner performance. Lewis Flight Propuslion Laboratory, USA
3. Unaune SV, Ganesan V (2005) Analysis of reacting flows in an aero-engine afterburner using
computational fluid dynamics. Ind J Eng Mater Sci 11:31–37
4. Ebrahimi JB (2006) Overview of gas turbine augmentor design, operation and combustion oscil-
lation, ILASS Americas. In: 19th annual conference on liquid atomization and spray systems,
Toronto Canada, (2006)
5. ANSYS, Inc., “ANSYS Fluent Theory Guide,” Canonsburg, PA, U.S.A., Release 15.0, ANSYS,
Inc., November 2013.
6. Free Field Technologies. Actran 16.1 User Guide. 2014.
7. Gurrala SR, Shaija A Computational analysis of reacting flows in afterburner. Heat Transfer Eng
https://doi.org/10.1080/01457632.2018.1546946
Experiments and Numerical Simulation
Sofa Francis Turbine Model Reflecting
Actual Site Environment
Rajavamsi Gangipamula, Sasank S. Patnaik, Sudhan A. Kamat,

Shyam Shukla, and Ravindra Birajdar
1 Introduction
Hydraulic turbines are extensively used to meet electricity demands. In any

hydropower plant, the turbine is considered as the main component which converts
the available potential energy of water into electrical energy through the electrical
generator. Reaction turbines are best suited for higher flow rate and lower head situ-
ations. In this type of turbines, the rotation of runner or rotor (rotating part of the
turbine) is partly due to impulse action and partly due to changes in pressure over the
runner blades. Among the available hydraulic turbines, the broader operating regime
of Francis turbine enables it to be used for varying range of small to large hydropower
plants. These turbines can transform up to 95% of the available potential head into
electrical energy [1]. Computational fluid dynamics (CFD) techniques have been
used to study the flow conditions inside hydraulic turbines over the past four decades
[2]. The present-day available CFD technique is considered as an alternative tool to
provide insight into the flow characteristics of hydro turbine components. The CFD
tool can be used to predict the efficiency of the turbine and do the necessary design
modifications to improve the performance. Khare et al. [3] predicted the variation of
R. Gangipamula · S. S. Patnaik · S. A. Kamat · S. Shukla (B) · R. Birajdar

Kirloskar Brothers Limited, Pune 411045, India
e-mail: shyam.shukla@kbl.co.in
R. Gangipamula
e-mail: raja.vamsi@kbl.co.in
S. S. Patnaik
e-mail: Sasank.patnaik@kbl.co.in
S. A. Kamat
e-mail: Sudhan.kamat@kbl.co.in
R. Birajdar
e-mail: ravindra.birajdar@kbl.co.in
112 R. Gangipamula et al.
flow angles with guide vane opening of a Francis turbine. Yaping et al. [4] presented a
performance study for Francis-99 by using different turbulence models for a steady-
state case with the Frozen rotor interface. From experiments, it was concluded that
the result obtained by the standard k-ε turbulent model has a relatively small differ-
ence with the experimental results. Aakti et al. [5] investigated the performance of a
high-head Francis turbine at design and off-design conditions using OpenFOAM. In
this study, it was concluded that the CFD codes overestimated the efficiency because
the simulation domain does not consider the rotor side spaces and the associated
losses. Laouari et al. [6] presented a study on simulations of steady cavitating flow in
a small Francis turbine. It explained the prediction of the hydraulic performance and
cavitating flows in the full passage of a small model of a Francis turbine at different
operating conditions using ANSYS-CFX solver [7].
In most of the literature survey, the procedure adopted to evaluate the hydraulic
performance numerically is to specify the mass flow rate at the inlet and pressure at
the outlet. The robustness of the boundary condition has encouraged to use instead
of the specific conditions like hydraulic energy difference. But, at most of the sites,
the hydraulic energy difference is the variable parameter because of which the guide
vane opening is decided. Moreover, cavitation behaviour is predicted at specific unit
speed. The existing numerical boundary condition may be used for the performance
prediction, but it may not be a proper initial condition for cases where cavitation
performance is predicted for specific hydraulic energy. In the present case, numerical
studies are conducted to predict the hydraulic performance of a Francis turbine model
using hydraulic energy difference as a boundary condition. The procedure has been
applied for two different guide vane openings (GVO) which correspond to the best
efficiency point (BEP) and overload and compared with the model test parameters
as per site data.
2 Experimental Data
The model test has been conducted for a specific turbine and the experimental data
presented in terms of the hill curve (shown in Fig. 1). The turbine model test consists
of a runner with a diameter of 350 mm and net head about 20 m. The discharge
and output power have been, respectively, measured for each case, and efficiency is
calculated. Iso efficiency curves are plotted for each GVO with unit speed (N11 ) and
unit power (P11 ) as abscissa and ordinate, respectively.
In the present paper, importance has been given to prediction of the performance
of a model Francis turbine for two different guide vane openings (GVO), GVO60 and
GVO80, keeping constant unit speed.GVO60 represents 60 mm guide vane opening
which also corresponds to the best efficiency point (BEP) for this turbine model.
GVO80 represents the overload condition of the turbine model with 80 mm guide
vane opening. The turbine is supposed to be operating at best efficiency condition and
an overload condition. Hence, performance prediction analysis has been carried out
for these two cases. Numerical simulation results are validated with the experimental
Experiments and Numerical Simulation Sofa Francis Turbine Model … 113
Fig. 1 Hill curve of Francis

turbine
results. With the help of available experimental data for the present hydraulic turbine
model, the necessary output parameters like hydraulic efficiency for a given unit
speed are obtained from the analysis. The deviation of efficiency is also evaluated
for the cases with two different guide vane openings as mentioned above.
3 Numerical Modelling
3.1 Geometry
The three-dimensional geometry has been prepared using a 2D drawing of the hydro
turbine model. Turbine assembly includes the spiral casing, stay ring with 20 stay
vanes, distributor with 20 guide vanes, a runner with 15 blades and draft tube. The
three-dimensional model of each individual component has been prepared and assem-
bled using CREO-2. Figure 2 shows the three-dimensional model of the Francis
turbine assembly. To analyse the turbine at different guide vane openings (GVO), the
geometry of the guide vane domain in the spiral casing has been changed for each
opening.
3.2 Mesh Independence Study
The unstructured mesh has been generated using triangular elements for 2D surfaces
and tetrahedron elements of a 3D flow domain using ANSYS ICEM CFD V18.0.
The hydraulic turbine is divided into three domains: spiral casing with distributor,
runner and draft tube. The mesh has been independently created in all domains and
Fig. 2 a Three-dimensional
model of Francis turbine
assembled for analysis purpose as shown in Fig. 3. Prism layers are generated around
runner blade wall surfaces to resolve the boundary layer near the rotating wall. With
this, the Y+ magnitude of the impeller is limited to nearly 100–200. Denser mesh
created near stay vanes and guide vanes regions. To minimize the influence of grid
number on the numerical results, a grid independence test has been carried out with
four different sets of mesh.
The different grids are named as grid 1, grid 2, grid 3 and grid 4 varying from
coarse grid number to fine grid number, respectively. The hydraulic efficiency is
used as a parameter to verify the mesh sensitivity for present grids for which the
efficiency correlation coefficient (ξ) is defined. ξ = |ηmη−η
m
s|
∗ 100%, where ηm and ηs
are the values of predicted efficiency, for large grid number and small grid number,
respectively [8]. The value of the correlation coefficient is found to be less than 1%
in all cases. But grid 3 and grid 4 are showing almost the same efficiency values and
relatively lower correlation coefficient. So, grid 3 has been considered as an optimized
mesh with the node number of 1.93 million for simulation. Figure 4 shows the plot of
Fig. 3 Turbine meshing at

the component level
Optimized Grid number

(grid 3)
Fig. 4 Grid independence study of the turbine
efficiency with respect to grid numbers. The quality of mesh in terms of orthogonality
(>0.4) and skewness (>0.8) is found to be within permissible limits.
3.3 Boundary Conditions
The boundary conditions have been set in the geometry as per unit speed. The net
head difference is provided as a boundary condition at inlet and outlet for a given
unit speed. All other components are specified as a wall with a specified roughness
of 8 microns. Casing, stay vane, guide vane and draft tube are assigned as stationary
domain, and the runner is assigned as a rotating domain with a speed of 830 rpm.
The k-ε turbulence model is used with scalable wall function, a convergence criterion
of 1E-05 and discretization scheme is selected as high resolution. Simulations have
been carried out for two guide vane openings GVO60 and GVO80. The analysis is
conducted for unit speeds of 45, 55, 65, 75 and 85 for each GVO case.
As described in the previous section, numerical simulations have been carried out
for two different guide vane openings (GVO60 and GVO80) with a variety of unit
speed using ANSYS-CFX V18.0. For both, the openings of the computed values are
given in Tables 1 and 2.
From the values, it is observed that the maximum hydraulic efficiency achieved
is 90.4% from CFD, and the experimental value is 93.8% at a unit speed of 65. The
numerical analysis also conducted for another 4 unit speeds (45, 55, 75 and 85) to
plot the trend of the curve. From all five cases, it is found that the maximum deviation
in the efficiency is less than 3.7%.
Table 1 Comparison of efficiency (GVO 60)

S. No. N11 η (%) % of deviation
Experiment CFD
1 45 81 78.3 −3.2
2 55 89.5 87.1 −2.5
3 65 93.8 90.4 −3.7
4 75 89 86.2 −2.8
5 85 74 71.7 −3.0
Table 2 Comparison of efficiency (GVO 80)

S. No. N11 η (%) % of deviation
Experiment CFD
1 45 78 77.9 −0.1
2 55 88 87.0 −1.0
3 65 91 89.7 −1.3
4 75 89 88.6 −0.4
5 85 78 76.6 −1.4
Similarly, numerical simulations are also conducted for the GVO80 condition
which is overload condition for this turbine model. CFD results for the different unit
speeds are given and compared in Table 2.
For GVO60 case, maximum deviation observed in the efficiency is −3.7% points
while for GVO80 case the maximum deviation observed in the % point efficiency
is −1.4%. The percentage deviation is lower at overload conditions. The oversizing
of the draft tube could be the reason which makes it more efficient for an overload
condition. This can be further verified qualitatively by streamline plots comparison
for GVO 60 and GVO 80 as shown in Fig. 5. From the streamline plots, it is understood
Fig. 5 Streamline plots a GVO 60 b GVO 80

that the flow in the draft tube is relatively streamlined at off-design point causing the
lower head loss in the draft tube at the off-design condition. But the combined losses
in runner and casing are relatively more, in this case, causing the higher total head
loss compared to the best efficiency point.
The achieved numerical results for all the cases are plotted with experimental
values to understand the variations in CFD results. It is observed that the efficiency
values for both the guide vane openings are well aligned with the experimental values.
Figure 6 shows the plots of efficiency for GVO 60 and 80 cases. The CFD prediction
reflected a similar trend of experiments curve. The variation could be attributed to
the error in discretizing the governing equations or in experimental studies.
Fig. 6 Efficiency plots a GVO 60 b GVO 80

5 Conclusions
The present study focused on a methodology for predicting turbine hydraulic perfor-
mance using numerical simulations with replication of specific site conditions. In a
Francis turbine, the use of net head difference as a boundary condition and k-ε turbu-
lence model with scalable wall function method provides well-established results
which align with the test data. The achieved analysis results can be further used
for numerical studies of cavitation. The maximum % point deviation in efficiency
observed for GVO corresponding to best efficiency point (GVO60) is −3.7% points.
Similarly, for higher guide vane opening (GVO80), the maximum deviation in effi-
ciency is observed to be −1.4% points. The maximum % point deviation in unit flow
is observed to be below 2.5%. The deviations of efficiency are also found to be on the
lower side of the experimental values. The profile of the efficiency and unit discharge
with unit speed was well aligned with the experimental results. Hence, the present
simulation method is a more realistic approach which suits to site conditions.
Acknowledgements Authors would like to place on record their gratitude to the management
of Kirloskar Brothers Limited, Pune, India for the encouraging attitude towards Research and
Engineering activities, which led to this paper.
References
1. Anup KC, Thapa B, Lee YH (2014) Transient and numerical analysis of rotor-stator interaction
in a Francis turbine. Renew Energy 65:227–235
2. Keckand H Sick M (20082) Thirty years of numerical flow simulation in hydraulic turboma-
chines, Acta Mech, 201 (2008)211–229
3. Khare R, Prasad V, Kumar S (2010) CFD approach for flow characteristics of hydraulic Francis
turbine. Int J Eng Sci Technol 2(8):3824–3831
4. Zhao Y, Liao W, Ruan H, Xingqi L (2015) Performance study for Francis-99 by using different
turbulent models. J Phys 579:012012
5. Aakti B, Amstutz O, Casartelli E, Romanelli G, Mangani L (2015) On the performance of a
high head Francis turbine at design and off-design conditions. J Phys 579:012010
6. Ahmed L, Ghenaiet A (2016) Simulations of steady cavitating flow in a small Francis turbine.
Int J Rotating Mach 15
7. ANSYS-CFX Users Guide, V 18.0, By M/S Ansys Inc USA, Mar 2017
8. Gohil PP, Saini RP (2016) Numerical study of cavitation in Francis turbine of a small hydropower
plant. J Appl Fluid Mech 9:357–365
Computational Analysis
on Thermo-Hydrodynamic
Characteristics of Y-Shaped
Multi-branched Micro Heat Pipe
Asokan Bakhirathan and Lachireddi Gangadhara Kiran Kumar
Nomenclature
ρ Density
μ Viscosity
k Thermal conductivity
Cp Specific heat
λ Latent heat of vaporization
σ Surface tension
Q Heat input (W)
R Thermal resistance (K/W
Subscripts
l Of liquid phase
v Of vapor phase
m Of mixture
A. Bakhirathan · L. Gangadhara Kiran Kumar (B)

CFD lab, Mechanical Engineering Department, National Institute of Technology Calicut,
Kozhikode, Kerala, India
e-mail: ganga@nitc.ac.in
120 A. Bakhirathan and L. Gangadhara Kiran Kumar
1 Introduction
In the electronic cooling application, generally, the passive heat transfer devices
are used to transport the generated heat from the source, which does not use any
externally driven drives or shaft to get operated. There are many passive heat transfer
devices such as thermosyphon, heat pipe, vapor chamber, etc., among them the heat
pipe is the most preferable option. The working principle of the heat pipe is based on
the transmission of heat from the source by the latent heat of vaporization [1]. The
heat pipe has a working fluid filled and sealed, which has evaporator section where
the heat is been given, then the working fluid is filled under the particular pressure
get vaporizes and the vapor travels to the condenser section, due to the operating
condition of the condenser the vapor gets condenses to a fluid. The condensate
returns to the evaporator by the presence of the wick or sintered structure by the
effect of a capillary. In 1984, micro heat pipe (MHP) [2] was instigated mainly for
the microscale application, the non-circular channels were preferred the most since
the liquid meniscus formed by the sharp edges are responsible for the capillary action.
Quite a lot of researchers all over the world worked on the enhancement of the
heat transfer in the MHP by changing the material, working fluid, properties of the
working fluid, geometric dimensions of the microchannel, fill ratios, evaporator and
condenser conditions [3]. The enhancement method like increasing the heat transfer
are carried out by multi-branched or multi-source by having two condensers and
two evaporators respectively based on the applications [4, 5]. Generally, the multi-
evaporator heat pipes were used in the cooling the notebook, where the heat pipe can
collect the heat from the processor as well as the graphical card with one heat sink
set up to enhance the heat transfer.
In this paper the Y-shaped MBMHP is designed based on the electronic cooling
applications with 500 µm as the hydraulic dimension and the thermal characteristics
of the Y-shaped MBMHP found based on the thermal resistance modelling [6] and the
hydrodynamic characteristics of the Y-shaped MBMHP deals the major part which
is responsible for the heat transfer and is the function of the fill ratios, heat inputs
and the condenser conditions. The inter-relation of the thermal and hydrodynamic
characteristics is discussed for various parameters and the dry out condition of the
Y-shaped MBMHP is discussed by the liquid volume fractions in transient condition.
2 Modelling and Simulation
The Y-shaped MBMHP is investigated for various heat inputs, fill ratios and different
heat transfer coefficients. The dimension of the MBMHP has 10 mm long evaporator
section, 30 mm long condenser section with two limbs and in order to increase the
length of the MBMHP, 30 mm long adiabatic section was considered and channel
dimension of 500 µm as the hydraulic diameter is considered for flowing the working
fluid as shown in Fig. 1. The copper plate of thickness 1.5 mm and water as the
Computational Analysis on Thermo-Hydrodynamic Characteristics of Y-Shaped … 121
Fig. 1 Dimensions of
MBMHP
working fluid was considered. The 3-D model was created for Y-shaped MBMHP
and the complete computational domain was discretized into hexahedral cells for
accuracy. The thermal and hydrodynamic analysis of Y-shaped MBMHP was simu-
lated using ANSYS-FLUENT® . In the multiphase modelling, it is quite difficult to
track the interface of the phases which leads to the change in the thermophysical
property of the working fluid. In order to have a clear definition for the motion of the
interface, Volume of Fluid (VOF) model was considered which has more advantages
over the other model, especially it has less computational time.
The 3-Dimensional continuity equation for two-phase by considering the VOF
model, for the liquid phase is,
∂ρ ∂(αρl u) ∂(αρl v) ∂(αρl w)

+ + + = Sm (1)
∂t ∂x ∂y ∂z
Due to the mass interaction in the evaporation and condensation process Sm is

being added in the continuity equation of the liquid phase. In the case of the vapor
phase, the continuity equation is

n
αi = 0 (2)
i=1
The VOF modelling involves additional description for the surface tension effect
of the working fluid on the interface within the channel, the momentum equation
including the viscous dissipation and surface tension on the liquid–vapor interface
for this model is,

D 2
(ρm uu) = ρm g − ∇ p + ∇. μm ∇u + ∇u T − μm ∇.u I + FC S F (3)
Dt 3
where the surface tension effect at the interface is considered from the continuum
surface force model,

αρl Cv ∇(1 − α) + (1 − α)ρv Cl ∇α
FCSF = σlv (4)
2 (ρl
1
+ ρv )
and C is the considered curvature.

In the VOF model, the energy equation involves the internal energy of the liquid–
vapor mixture and the heat source term for the mass interaction of evaporation and
condensation. Where the Sh source term for the evaporation and condensation process
is calculated based on the weight average basis for the mixture of the liquid and the
vapor. The energy equation of the VOF model with the mixture internal energy and
the source term is,
D
(ρm em u) = ∇.(km ∇T ) − ∇(ρm u) + Sh (5)
Dt
In the ANSYS-FLUENT® , the VOF model is being considered for simulating the
MBMHP with Implicit scheme. The thermophysical property of the water-liquid and
water–vapor such as density, viscosity, thermal conductivity and specific heat were
considered as constant. The SIMPLE algorithm was considered for pressure–velocity
coupling for the implicit scheme in the VOF model. The second order upwind scheme
was used for solving momentum and energy equation and for solving pressure and
volume fraction PRESTO! and Compressive was used respectively. The flexible
under-relaxation factor was considered for solving the VOF model and the higher
order term are also relaxed. The element size maintained during meshing is 0.00005 m
where the residual of 1E-04 was considered as convergence criteria. The simulations
are carried for different fill ratios such as 0.5, 0.7 and 0.9 of the total volume of
the microchannel. The evaporator and the condenser section were also simulated by
varying different such, 0.2, 0.5, 1, 1.5 and 2 W heat inputs and 20 W/m2 -K convective
heat transfer coefficient was considered and the ambient temperature is maintained at
300 K. Based on varying these parameters the thermal investigation of the Y-shaped
MBMHP were done. For the hydrodynamic characteristics of the Y-shaped MBMHP,
the high heat input case i.e., 2 W is considered for all three fill ratios and discussed
in the following section.
3 Results and Discussions
The modelled Y-shaped MBMHP was simulated for investigating the thermal and
hydrodynamic analysis and 300 s of total computational time is considered. The
various heat inputs are given in the evaporator Section (0.2, 0.5, 1, 1.5 and 2 W)
where the input is in the form of heat flux which is calculated based on the contact
surface area of the evaporator. The condenser section is subjected to constant convec-
tive heat transfer coefficient of 20 W/m2 -K at 300 K. The filled working fluid in the
microchannel of the Y-MBMHP rises initially due to the heat applied in the evapo-
rator, the vapor plug is developed which accommodate the plug flow and agglomerate
with vapor in the condenser where the vapor releases the latent heat of vaporization.
The upward movement of the plug flow cases the slug to move downward because of
the density difference where the slug-plug separation occurs which leads to the oscil-
lation. The amplitude of the slug-plug oscillation increases with temperature rise.
The thermal resistance and the variation of liquid volume fractions are discussed
to support the thermal and hydrodynamic characteristics of the Y-shaped MBMHP
respectively. The thermal resistance is calculated based on the average tempera-
ture difference for the respective heat input. The average temperature difference is
measured between the average temperature of the evaporator and the condenser. The
liquid volume fraction for the various points as shown in Figs. 2 and 3 which clearly
states the hydrodynamic characteristics of Y-shaped MBMHP.
Figure 4 shows the average temperature difference for the various fill ratios for
higher heat input case. In the Y-shaped MBMHP, the average temperature difference
curve is been characterized by three different part trends, initially, the curve follows
the start-up trend then attains the transient trend followed by the steady-state trend.
The oscillation of the average temperature difference is seen in the transient trend.
Fig. 2 Point in the

evaporator section
Fig. 3 Points in the condenser section
Fig. 4 Average temperature difference for different fill ratios

The start-up trend seems to be accomplished earlier for the lower fill ratio where the
transient trend with oscillation in temperature starts earlier for the lower fill ratios and
curve shifts increases as the increase in the fill ratios. The steady-state trend appears
once after completion of the oscillation of the slug-plug within the microchannel. At
the time of the steady-state trend, the Y-shaped MBMHP has larger plug separated
by the smaller slug where the fraction of the liquid volume got converted to the vapor
which increase in time for the supplied constant heat input.
The oscillation in the average temperature difference is quite long for the 0.7
fill ratio, which portraits that the optimum liquid fractions are maintained for the
slug-plug to oscillate. The operating condition is clearly noted by the evaporator
temperature for the total computational time. Figure 6 shows the evaporator temper-
ature and the respective liquid volume fractions at the point in the evaporator specified
in Fig. 2. The 0.7 fill ratios have an optimum level of heat transfer than the other
2 fill ratios considered, to support that the oscillation of the average temperature
Fig. 5 Evaporator liquid 1

fraction for various fill ratios
0.8
Liquid Volume fraction
0.6
0.4
0.2
0
0 100 200 300
Time (s)
0.5 FR 0.7 FR 0.9 FR
Fig. 6 Variation of
Average temperature difference (K)
evaporator liquid volume 390 1

fraction for respective
Liquid volume fraction
evaporator temperature for 0.8

0.7 FR 360
0.6
0.4
330
0.2
300 0
0 100 200 300
Time (s)
Evaporator temperature Evaporator liquid fracƟon
is quite large, where the evaporator is well entrained due to the slug-plug oscilla-
tion. Figure 7 explains the motion of the liquid slug at the point of separation of
the condenser for 0.7 fill ratio. Within the operating temperature of 360 K for the
Y-shaped MBMHP with 0.7 fill ratio seems to be operating better with the optimum
level of liquid volume fractions in it in Fig. 5.
The slug-plug oscillation is key parameter noticed which controls the thermal and
hydrodynamic characteristics, where the retainment of the operating temperature
is also maintained. In the aspect of the thermal resistance based on the average
temperature difference, even the 0.7 fill ratio offers better thermal resistance than the
other considered fill ratios. The huge fluctuation in the thermal resistance is seen for
0.5 fill ratio and for the 0.9 fill ratio case the thermal resistance seems to be almost
the same for all heat inputs. The lower thermal resistance for the Y-shaped MBMHP
is 8.25 K/W at 1.5 W for 0.7 fill ratio are shown in Fig. 8.
Fig. 7 Liquid volume 1.0

fraction at the point of
separation of the condenser
0.8
Liquid volume fraction
for 0.7 FR
0.6
0.4
0.2
0.0
0 100 200 300
Time (s)
Point of seperaƟon of condenser
Fig. 8 Thermal resistance of 8.42

Y-shaped MBMHP for
Thermal resistance (K/W)
various heat inputs
8.36
8.3
8.24
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Heat input (W)
0.5 FR 0.7 FR 0.9 FR
4 Conclusion
The computational investigation for the thermal and hydrodynamic characteristics

was carried out for the Y-shaped MBMHP for different heat inputs, various fill
ratios and for constant convective heat transfer coefficient. The thermal and the
hydrodynamic characteristics of the Y-shaped MBMHP was better for 0.7 fill ratio
than other considered fill ratios, where the slug-plug oscillations and the retainment
of the liquid in the point of separation of the condenser section seemed to be very
long. Even the lowest thermal resistance was found to be 8.25 K/W at 1.5 W. The
operating temperature of the Y-shaped MBMHP with 0.7 fill ratio was about within
350 K, which showed the optimum level of liquid volume fraction retains inside the
microchannel of multi-branched MHP.
References
1. Reay DA, Kew PA, McGlen RJ (2006) Heat pipes: theory design and applications (2006)
2. Cotter TP (1984) Principles and prospects for micro heat pipes. In: Proceeding 5th international
heat pipe Conference, vol 1, pp 328–335
3. Cao Y, Faghri A (1994) Micro/miniature heat pipes and operating limitations. J Enhanc Heat
Transf 1:265–274
4. Cai Y, Li Z, Zhai J, Tang Y, Yu B (2017) Experimental investigation on a novel multi-branch
heat pipe for multi-heat source electronics. Int J Heat Mass Transf 104:467–477
5. Zhong G, Ding X, Tang Y, Yu S, Chen G, Tang H, Li Z (2018) Various orientations research
on thermal performance of novel multi-branch heat pipes with different sintered wicks. Energy
Convers Manag 166:512–521
6. Wang J-C (2008) Novel thermal resistance network analysis of heat sink with embedded heat
pipes. Jordon J Mech Ind Eng 2:23–30
Mixing in Oscillating Lid Driven
Cavity—A Numerical Study
Manjappatta Pazhiyottumana Neeraj and Ranjith Maniyeri
1 Introduction
The lid driven cavity with constant top wall motion is one of the standard problems
in the field of computational fluid dynamics. This particular problem addresses the
complex flow regimes in the cavity. The application of this problem mainly lies in
mixing applications which is highly important in chemical and process industries.
Usage of stirrers, cavity stretching are also good methods to improve mixing. The
speciality of oscillating lid is that they will make disturbances in the flow and make
the mixing faster.
Indukuri and Maniyeri [1] studied about different types of oscillating wall cavities
like top wall moving, parallel wall moving and anti-parallel wall moving for different
Reynolds numbers. They found the optimum Reynolds number, amplitude (A = 1)
and angular velocity (ω = 2π/6) for better mixing.
Koseff et al. [2] did a study on the end wall effects on the downstream of secondary
eddy in a lid driven cavity. Prasad et al. [3] studied about effects of different Reynolds
number on end wall effects. The study of Ghia et al. [4] is a standard for comparison
of lid driven cavity problems in higher Reynolds numbers in laminar range. Many
research works have been reported on the study of effects of aspect ratios on cavity
and vortex behaviour inside the cavity for wide range of Reynolds number using
lattice Boltzmann method, finite difference method and stream vorticity approach
[5–7].
Anderson et al. [8] found that an extra steady motion of the opposite cavity wall,
resulting in a small perturbation to the original flow, causes the chaotically mixed
region to be spread over almost the whole cavity. Carlsson et al. [9] investigated about
effect of vibrating walls on mixing in a cavity. Most effective mixing is observed for
M. P. Neeraj · R. Maniyeri (B)

Biophysics Laboratory, Department of Mechanical Engineering, National Institute of Technology
Karnataka, Surathkal, Mangalore, Karnataka 575025, India
130 M. P. Neeraj and R. Maniyeri
flows at channel half-widths of similar, or lower order than the vibratory wavelength
and for sufficiently high frequencies.
Van-Le et al. [10] studied about chaotic mixing in two dimensional square cavity
by using finite volume method. The effects of mixing in square cavity at two different
Peclet numbers 2000 and 10,000 were compared in this work. Both the Peclet
number and the initial conditions of concentration plays an important role in mixing
effectiveness.
Most of the studies showed the results as velocity stream lines. A study of concen-
tration profile in mixing phenomena in lid driven cavity has not been reported much.
Addressing the concentration variation in a lid driven cavity will be of significant
interest in the field of mixing.
In the present study we simulate two cases. The constant top wall motion and
oscillating top wall motion to study the mixing phenomena. A staggered grid is used
and Finite volume method is used for discretization. The continuity equation, Navier
Stokes equations and the concentration equation is solved using Euler’s explicit
scheme. A FORTRAN code is developed to simulate the flow and mixing.
2 Methodology
We consider the fluid dynamics in a square cavity with only top wall moving. For the
constant top wall velocity the horizontal velocity u and vertical velocity v are given
as
u = u ref , v = 0 (1)
whereas for oscillating top wall
u = u ref ∗ cos(ωt), v = 0 (2)
where ω is the angular velocity, t is the time period and uref is the constant velocity.
At all other walls no-slip boundary conditions are applied. The physical model is
shown in Fig. 1a for constant top wall velocity and Fig. 1b for oscillating top wall
velocity. Here we simulate one square cavity.
The concentration boundary conditions at every wall in both cases are given as.
Left and right wall
∂c
=0 (3)
∂y
Top and bottom wall
∂c
=0 (4)
∂x
Mixing in Oscillating Lid Driven Cavity—A Numerical Study 131
Fig. 1 Physical model a Constant top wall velocity and b Oscillating top wall velocity
The continuity, Navier–Stokes and the concentration equations are non-

dimensionalized and solved as below.
∇.
u=0 (5)
∂ u 1 2
+ u.∇ u = −∇ p + ∇ u (6)
∂t Re
∂c 1 2
+ u.∇c = ∇ c (7)
∂t Pe
where c, Re, Pe, −→u , p denotes concentration, Reynolds number, Peclet number,
velocity and pressure respectively.
The finite volume method is used to discretize all terms and the equations are
solved using Euler’s Explicit scheme.
The numerical model of the square lid driven cavity with both constant top wall
velocity and oscillating top wall velocity by applying all boundary conditions and
solving all the governing equations.
Fig. 2 Centreline velocity with respect to vertical distance y for a Reynolds number = 100
3.1 Validation of the Results
The square lid driven cavity with constant top wall velocity was simulated. Three
different grid sizes of 25 × 25, 65 × 65 and 129 × 129 are used and 129 × 129 is
found to be the optimum grid. The same grid size is also used by [4]. The numerical
results corresponding to the optimum grid size are validated using the findings of
[4] as shown in Figs. 2 and 3 (for different Re of 100 and 1000). The centreline u
velocity is used for the comparison. The validation is found to be perfectly matching.
The numerical study is conducted on both constant top wall velocity and oscil-
lating top wall velocity cases. The Reynolds number used is 100 [1] and the Peclet
number is 2000. The initial conditions of concentration are shown in Fig. 4. As
concentration gradients at all walls are zero, a gradient should be provided as the
initial condition for the mixing to start and propagate. So here, as seen from the Fig. 4
the concentration at the top half of the cavity is 1 and the bottom half concentration
is 0. This type of initial condition of concentration is called horizontal separation
(HS). Van-Le et al. [10] found that for HS condition best mixing was achieved at
Peclet number of 2000. So Peclet number is chosen as 2000.
The velocity streamlines and concentration plot after a time period of t = 20 is
shown in Figs. 5 and 6 respectively.
The same results are shown for oscillating top wall velocity also in Figs. 7 and 8.
It is seen from the results that the mixing is lesser in oscillating condition compared
to that in constant velocity condition. It may be due to the lower amplitude value
Fig. 3 Centreline velocity with respect to vertical distance y for a Reynolds number = 1000
Fig. 4 Initial concentration

conditions
Fig. 5 Velocity stream line

for constant top wall velocity
case
Fig. 6 Concentration
contour for constant top wall
velocity case
of the oscillating velocity. So, we changed the amplitude to 5 and the contour of
concentration after a time period of t = 20 for the higher amplitude is shown in
Fig. 9.
From Fig. 9 we can understand that more mixing occurred at a higher amplitude.
This may be due to the reason that at an amplitude of 1 the range of velocities applied
to the top wall is very much lesser compared to that of constant top wall velocity of
1 according to Eq. (2). When amplitude is equal to 5 that range increases and hence
mixing effect also increases comparing to that of constant top wall velocity case.
Fig. 7 Velocity stream line

for oscillating top wall
velocity case (A = 1)
contour for oscillating top
wall velocity case (A = 1)
4 Conclusions
The numerical simulations are conducted in a square lid driven cavity with both
constant top wall velocity and oscillating top wall velocity conditions. The stream
lines and concentration contours are plotted in both cases and compared. The mixing
is found to be higher in case of oscillating lid driven cavity while using a higher oscil-
lation amplitude. The lid driven cavity with oscillation amplitude (A = 5) performs
better as compared to oscillation amplitude A = 1.
contour for oscillating top
wall velocity case (A = 5)
Acknowledgements This research was supported by Science and Engineering Research Board, a
statutory body of Department of Science and Technology (DST), Government of India through the
funded project ECR/2016/001501.
References
1. Indukuri JV, Maniyeri R (2018) Numerical simulation of oscillating lid driven square cavity.
Alexandria Eng. J. 57:2609–2625
2. Koseff JR, Street RL (1984) On end wall effects in driven cavity flow. J Fluids Eng 106(4):385–
389
3. Prasad AK, Koseff JR (1989) Reynolds number and end wall effects on a lid driven cavity flow.
Phys Fluids A 1(2):208–218
4. Ghia UKNG, Ghia KN, Shin CT (1982) High-Re solutions for incompressible flow using the
Navier-Stokes equations and a multigrid method. J Comput Phys 48(3):387–411
5. Patil DV, Lakshmisha KN, Rogg B (2006) Lattice Boltzmann simulation of lid driven flow in
deep cavities. Comput Fluids 35:1116–1125
6. Lin LS, Chen YC, Lin CA (2011) Multi relaxation time Lattice Boltzmann simulations of deep
lid driven cavity flows at different aspect ratios. Comput Fluids 45(1):233–240
7. Perumal DA (2012) Simulation of flow in two-sided lid-driven deep cavities by finite difference
method. J Appl Sci Thermodyn Fluid Mech 6(1):1–6
8. Anderson PD, Galaktionov OS, Peters GWM, Van De Vosse FN, Meijer HEH (2000) Chaotic
fluid mixing in non-quasi-static time-periodic cavity flows. Heat and Fluid Flow 21:176–185
9. Carlsson F, Sen M, Lofdahl L (2005) Fluid mixing induced by vibrating walls. Eur J Mech
B/Fluids 24:366–378
10. Van-Le TH, Suh YK, Kang S (2006) Chaotic mixing in two-dimensional square cavity flow by
using finite volume method (FVM). Proceedings of the Korean fluid machinery Associations
2:935–938
Using Dimensional Analysis
for Development of Empirical
Correlations for Estimation of Liquid
Holdup and Drop Size in Rotating Disc
Contactor
K. Swetha Saraswathi and P. R. Naren
1 Introduction
Liquid–liquid extraction (LLE) involves preferential separation of components from

the liquid feed using another liquid phase as solvent.
The rotating disc contactor (RDC) is particularly used for liquid–liquid systems
with medium to low interfacial tension. It consists of a vertical column equipped with
a series of equally spaced stators and a shaft [1]. Rotor discs are mounted on a shaft
driven by a motor. This induces agitation in the column. Further, rotation of the discs
creates high shear in the liquid and thus leads to increased interfacial area. The rotor
discs accelerate the flow and the stator rings decelerates the flow. A counter-current
flow of the two phases is maintained in the contactor.
This contactor has many advantages: simplicity in construction, low power
consumption, high throughput, operational flexibility, lower cost for solvent inven-
tory and maintenance [2]. The performance of the RDC is mainly attributed to
the breaking down the light phase droplets and therefore increasing the column
performance. The mass transfer efficiency of the commercial RDCs is low due to its
structural limitations, high axial back-mixing and entrainment of small droplets.
Rotating disc contactor has drawn considerable attention as extraction equipment
in deacidification of vegetable oils [3], wastewater treatment and downstream opera-
tions to separate inhibitory fermentation. It is also widely used in the petrochemical,
chemical, hydrometallurgy [4] and biochemical industries for its low investment and
less maintenance [5].
K. S. Saraswathi
Chemical Engineering (Integ.), SASTRA Deemed To Be University, Tamil Nadu, Thanjavur
613401, India
P. R. Naren (B)
Chemical Engineering, School of Chemical and Biotechnology, SASTRA Deemed To Be
University, Tamil Nadu, Thanjavur 613401, India
e-mail: prnaren@scbt.sastra.edu
138 K. S. Saraswathi and P. R. Naren
Table 1 Significant variables

Fluid property Density (ρ), Viscosity (μ), Interfacial tension (σ )
Column dimension Column diameter (Dc ), Column height (H), Compartment height (hc ),
Rotor diameter (d r ), Stator diameter (d s )
Operating condition Velocity (V ), Rotor speed (N)
Others Acceleration due to gravity (g)
Hence, the knowledge of the hydrodynamic parameters is important for the design
and scale-up of extractors. Hydrodynamic behavior is a function of the operating
conditions of the system, fluid properties, and system properties.
Many empirical correlations are available in literature to predict the hydrodynamic
parameters of RDC [6–9]. However, the available correlations may not accurately
predict for all the liquid–liquid systems. It is desirable that the developed correlations
are validated with large experimental data sets to arrive at better predictions. In the
present work, a generalized empirical correlation was developed based on dimen-
sional analysis for predicting the dispersed phase holdup and drop size in a rotating
disc contactor.
2 Development of Empirical Correlation

for Hydrodynamic Parameters in RDC
2.1 Selection of Significant Variables
Hydrodynamic behaviour of RDC is a function of the operating conditions of the

system, fluid properties, and system properties. Table 1 gives a list of 11 significant
variables chosen for this study.
2.2 Development of Dimensionless Groups
To determine the dispersed phase hold and drop size in RDC, dimensional anal-
ysis was performed using the Buckingham pi theorem approach. The hydrodynamic
parameters were written as a function of one or many independent variables as
follows:
ϕ = f (a0 , a1 , a2 , a3 , . . . . . . an ) (1)
The list of significant variables was then grouped based on dimensional similarity.
The number of dimensionless groups (k) is equal to the difference between the
Using Dimensional Analysis for Development of Empirical Correlations … 139
independent variables (n) and the number of dimensions required to characterise the
variables (r).
k = n −r (2)
Equation (1) can be rewritten as follows:
ϕ = f (π1 , π1 , π2 , π3 , . . . . . . πn−r ) (3)
The difficulty in implementation of the Buckingham theorem is in the right selec-

tion of independent parameters that influence the output parameter. Dimensionless
groups (π1 , π1 , π2 , π3 , . . . . . . πn−r ) used in the empirical function, based on the list
of significant variables, are selected using a trial and error procedure.
2.3 Power Law Model
Power law model was used for the development of empirical correlations for
dispersed phase holdup and drop size in rotating disc contactor. Most correlations
reported in literature [6–9] followed power law model. Hence, it was adopted as the
first choice in this study.
80% of the collected experimental data from literature was used as training set
and the rest 20% was for validation or testing. The selection of training and testing
data sets were done randomly. Three different training and validation sets were used.
About 10 power law models were tested for both dispersed phase holdup and drop
size in RDC [10]. The best fit model was chosen and explained in the following
section.
3.1 Consolidated Experimental Data from Literature
Experimental data from 12 sources covering 20 liquid–liquid systems are consoli-

dated and used for developing a generalized correlation to predict dispersed phase
holdup and drop size. Plot Digitizer software (version 2.6.8) was used to extract the
experimental data from the literature.
Holdup data varying from 0.009–0.7270 for dispersed phase density range of
806–1300 kg/m3 were covered in this study. Similarly, experimental drop size values
ranging from 0.497–9.103 mm were documented for interfacial tension range of
8.7–50.0 mN/m [10]. The data obtained from literature for dispersed phase holdup
and drop size in RDC is made available at https://doi.org/10.7910/DVN/HMNWIC.
The different power law models were tested for its efficacy in predicting the
dispersed phase holdup in RDC by evaluating the average absolute relative error
(AARE). The AARE value was computed using the equation below [11]:

N calculated value−experimental value
experimental value
AARE = × 100
i=o
N
3.2 Dispersed Phase Holdup in RDC
The best empirical correlation developed for dispersed phase holdup is given by:
a1 a2 a3 a4 a5
hc ρ N dr ρc Vc Dc N 2 dr
ϕ = a0
Dc ρc Vc μc g
The pre-constant, exponent and AARE values of the training set is given in
Table 2. As compared to the available correlations for holdup in literature, the corre-
lation developed in this work gave an AARE value of 39.75% whereas the best fit
literature correlation Kasatkin et al. [6], Murakami et al. [7] gave AARE values of
77.82% and 70.68% respectively.
The parity plot for dispersed phase holdup is given in Fig. 1.
The correlation developed for holdup in this work is thus shown to predict better
over a wide range of data covering various systems and column geometry. From
holdup experiments, it was found that the dispersed phase holdup was not affected
by the continuous phase velocity. However, it was found to depend largely on rotor
speed. The holdup for large column geometry was not predicted very accurately
by the developed correlation. This is because, most of the experiments performed
before were done in laboratory scale RDC. Only a few authors [12, 13] had performed
experiments in pilot scale RDCs. However, the fit was satisfactory for columns with
smaller diameter.
Table 2 Pre-constant, exponent and AARE values for Holdup

Training set a0 a1 a2 a3 a4 a5 AARE (%)
1 0.3237 − 0.228 − 0.292 − 0.358 − 0.262 0.321 39.363
2 0.3642 − 0.17 − 0.5317 − 0.2732 − 0.185 0.2614 38.512
3 1.3585 − 0.1943 − 0.243 − 0.3473 − 0.2503 1.3123 38.216
4 0.8921 − 0.1808 − 0.3216 − 0.3179 − 0.2282 0.3094 39.747
Mean 0.7346 − 0.193 − 0.347 − 0.324 − 0.232 0.3011 –
Fig. 1 Parity plot for dispersed phase holdup in RDC
3.3 Drop Size in RDC
The best empirical correlation developed for drop size is given by:
a1 a
μc N dr a2 Vc a3 N 2 dr 4
d32 = a0 √ 1+
Dc ρc σ Vc Vd g
The pre-constant, exponent and AARE values of the training set is given in Table
3. The parity plot for drop size is given in Fig. 2.
Drop size was majorly influenced by the rotor speed and from experiments, it
was observed that drop size decrease with increase in rotor speed. Therefore, the
dimensional groups that were chosen for drop size correlation featured rotor speed
and its exponent value was negative. AARE value was 35.98% for the complete drop
size data set. The experimental data collected for drop size covered a wide range of
Table 3 Pre-constant, exponent and AARE values for drop size

Training set a0 a1 a2 a3 a4 AARE (%)
1 0.0029 − 0.8004 0.019 0.2771 − 0.1546 36.2556
2 0.0013 − 0.9064 0.0191 0.3133 − 0.1338 34.2682
3 0.0015 − 0.9003 0.0147 0.2816 − 0.1148 37.0315
4 0.003 − 0.8078 0.012 0.2646 − 0.1306 35.9822
Mean 0.0022 − 0.8537 0.0162 0.2841 − 0.1334 −
Fig. 2 Parity plot for drop size in RDC
column geometry and operating conditions. Hence, the developed correlation had a
wide margin of error.
It was also observed from Tables 2 and 3 that the developed correlation was not
sensitive to the experimental data as the coefficient values do not change significantly
with different training data sets.
4 Conclusions
In the present study, dispersed phase holdup and drop size were studied in Rotating
Disc Contactor. Power law model was used to develop a generalized correlation to
predict holdup and drop size in RDC. The developed correlations were tested for
its efficacy by evaluating the average absolute relative error (AARE) values. The
AARE values of the developed empirical correlation for dispersed phase holdup and
drop size was found to be 39.75% and 35.98% respectively. Correlations reported
in the present work are found to be better than the reported correlations available in
literature.
References
1. Moris MA, Díez FV, Coca J (1999) Solvent extraction of molybdenum and tungsten by
Alamine—336 and DEHPA in a rotating disc contactor. Sep Purif Technol 17:173–179
2. Vermijs HJA, Kramers H (1954) Liquid-liquid extraction in “rotating disc contactor. Chem
Eng Sci 8:55–64
3. Pina CG, Meirelles AJA (2000) Deacidification of corn oil by solvent extraction in a perforated
rotating disc column. J Am Oil Chem Soc 77:553–559
4. Pandey G, Mukhopadhyay S, Renjith AU, Joshi JM, Shenoy KT (2016) Recovery of Hf and Zr
from slurry waste of zirconium purification plant using solvent extraction. Hydrometallurgy
163:61–68
5. Kent E Goklen, Alan Hatton T (1987) Liquid-Liquid extraction of low molecular-weight
proteins by selective solubilization in reversed micelles. Sep Sci Technol 22
6. Kasatkin AG, Kagon SZ, Trukhanov VG (1962) The hydrodynamic characteristics of rotating-
disk extractors. J Appl Chem 35:1903–1910
7. Murakami A, Misonou A (1978) Dispersed phase holdup in a rotating disc extraction column.
Int J Chem Eng 18:16–22
8. Marr R, Husung, Moser F (1975) Hydrodynamics of rotating disc contactor (RDC). Chem Eng
Technol 47:203–210 (1975)
9. Magiera J, Zaldo J (1977) Droplet size of the dispersed phase in an extraction column with
rotating discs. Inzynieria Chemiczna 7:127–142
10. Swetha Saraswathi K (2019) Hydrodynamic studies on rotating disc and external—loop airlift
contactors, M.Tech (5-year Integrated) Thesis, SASTRA Deemed University, Thanjavur, India
11. Vaishali V (2018) Hydrodynamics of asymmetric rotary agitated columns and pulsed disc-
doughnut columns, M. Chem. Engg. Thesis, Institute of Chemical Technology, Maharashtra,
India
12. Olney RB (1964) Droplet characteristic in counter current contactors. A1ChE J 10:827–835
13. Jeffreys GV, Al-Aswad KKM, Mumford CJ (1981) Drop-size distribution and dispersed phase
holdup in a large rotating disc contactor. Sep Purif Technol 16:1217–1245
Effect of Eccentricity on the Thermal
Behaviour of A Re-Entry Vehicle With
Concave Windward Surface
V. Vani Praneetha, M. Sudha Madhuri, V. Vijaya Lakshmi,

Karthik Sundarraj, and Prakash S. Kulkarni
Nomenclature
δ = Boundary layer thickness

x = Boundary layer length
μ = Dynamic viscosity
u* = Friction velocity
M = Mach number
P = Density
e = Eccentricity
Re = Reynolds number
y = First node distance
1 Introduction
The major problem for re-entry space shuttle type spacecrafts is the aerodynamic
heating. All the re-entry vehicles including reusable spacecrafts, crew return vehicles
have to survive extreme aerodynamic heating during its descent through the dense
atmosphere. As the space programs and space tourism growing exponentially, the
travel to and from the space has to be made safer and reliable. During the re-entry
V. Vani Praneetha (B) · M. Sudha Madhuri · V. Vijaya Lakshmi

LBRCE, Mylavaram, India
K. Sundarraj
MSC Software Corporation, California, USA
P. S. Kulkarni
IISc, Bangalore, India
e-mail: psk@iisc.ac.in
146 V. Vani Praneetha et al.
into the earth’s atmosphere, the flat bottom of the spacecraft faces most of the wind
directly. In order to have the minimum deceleration when the reusable spacecraft with
lifting trajectory has to re-enter into the atmosphere at high angles of attack, the drag
should be high. Due to this deceleration, it may cause severe aero thermodynamic
heating. This is a huge risk which needs to be minimised with the people onboard
by making some feasible activity in order to reduce the aerodynamic heating. The
use of concave nose has been widely reported as means to reduce stagnation point
heat transfer rates. This research takes the concept forward and proposes the use of
concave windward surface for the re-entry space vehicle by means of aerodynamic
investigation over a flat plat and with concavity of different eccentricities under
re-entry conditions.
2 Computational Study
2.1 Geometry
A full three-dimensional geometric model has been generated using CATIA software.
The base geometry consists of a three-dimensional 1 m x 1 m x 1 cm as shown in
figure. The remaining geometries are obtained by bending the geometry along the
longitudinal axis to get a hyperbolic curve. The hyperbolic curve co-ordinates are
obtained using coding by considering the following equations of hyperbola;
x2 y2
− =1 (1)
a2 b2
Equation (1) represents the canonical form. The focal point of the hyperbola is
(−ea, 0) which represents the eccentricity of the hyperbola. The distance r from a
point (x, y) on the hyperbola to the focal point (−ea, 0) is given by;
r = −ex − a (2)
Polar co-ordinates (r, ѳ) of hyperbola are considered, and we get the (x, y) co-
ordinates for plotting are as follows:
X = −ae + r cos (3)
Y = r sin (4)
Solving Eqs. (3) and (4) the value for distance r in polar co-ordinates is given by;

a e2 − 1
r= (5)
1 + e cos
Effect of Eccentricity on the Thermal Behaviour … 147
2.2 Solver Setup
ScFLOW solver is used for all the simulations. Density-based solver is used along
with steady-state flow and steady time. All the simulations are done at hypersonic
flow and second-order upwind spatial discreitization. The air (fluid) is considered as
ideal gas with density 0.000288 kg/m3 , and specific heat Cp is 1006.43 J/kgK. The
thermal conductivity is 0.0242 W/m-K. The viscosity and molecular weight are 1.4
e−5 kg/m-s and 28.966, respectively. The material of the solid boundary is considered
as titanium with density 4850 kg/m3 , C p as 544.25 J/kgK and thermal conductivity
as 7.44 W/m-K. The boundary conditions are taken as wall (object), inlet, outlet and
symmetry for 3D condition. The object is stationary with no-slip boundary layer. The
adiabatic condition is assumed over the wall at T = 245.45 K. The pressure far-field
condition is given with a pressure of 20.314 Pa and M = 20.
The boundary layer thickness, δ is found to be 0.019369 m using the Formula:
5x
δ=√ (6)
Re
where x—boundary layer length

The y valve is calculated as 1.9 e−4 m considering y + as 1.
The corresponding Reynolds number is 2.6e5 . The y value is calculated by using the
Formula:
y = y + μ/ρu ∗ (7)
where μ—dynamic viscosity

u ∗ —Friction velocity
3 Mach Number Contours
The Mach number contours give us an insight of position of shock and the stagnation
region ahead of plate (Fig. 1).
4 Temperature Contours
Temperature contours are captured on the flat plate which shows the position of
the peak temperature on the surface of the flat plate and compared with that of the
different eccentric plates (Fig. 2).
a b
c d
e f
Fig. 1 a Flat plate mach contour, b Eccentricity 2 mach contour, c Eccentricity 3 mach contour,
d Eccentricity 5 mach contour, e Eccentricity 7 mach contour, f Eccentricity 10 mach contour
a b
c d
e f
Fig. 2 a Flat plate temperature contour, b Eccentricity 2 temperature contour, c Eccentricity 3

temperature contour, d Eccentricity 5 temperature contour, e Eccentricity 7 temperature contour,
f Eccentricity 10 temperature contour
Eccentricity Peak temperature

Flat plate 19744.94
Eccentricity 2 20433.10
5 Ongoing Works
Work is in progress on effects of varying Mach numbers on peak heat based on

different eccentricities. There are also studies going on in different eccentric shapes.
There is also plan to conduct unsteady simulations, which shall be made for final
paper submission.
6 Conclusions
Numerical simulations are conducted for different eccentric plates and compared
with flat plate. From the numerical simulations, it is understood that as eccentricity
increases, peak temperature is reduced. This proves to be a very useful finding as
higher eccentric geometries are closer to that of a fat plate configuration, whereas
the full surface is not exposed to stagnation flow field. Thus, surface heating at lower
surfaces of the object gets reduced. While majority of the heating might be at the
tips. Further investigations are in progress to understand the effects of varying Mach
numbers and also the unsteady nature of flow.
Acknowledgements Sincere thanks to all who have provided their timely guidance and support,
without which this research couldn’t have succeeded. I want to express my special thanks to Prof.
Prakash S Kulkarni for allowing us to work in his laboratory at IISc Bangalore.
Bibliography
1. Stallings RL, Burbank PB (1959) Heat transfer and measurements on a concave nose cylinder
for a mach number range of 2.49–4.44. NASA TM-X-221
2. Saravanan S, Jagadeesh G, Reddy KPJ (2009) Investigation of missile shaped body with forward
facing cavity at Mach 8. J Spacecraft Rockets 46(3)
3. Engblom et al (1996) Experimental and numerical study of the hypersonic forward facing cavity.
J. Spacecrafts Rockets 33(3)
4. Engblom WA, Goldstein DB (1996) Nose-tip surface heat reduction mechanism. J Thermo Phys
Heat Transf 10(4)
Numerical Analysis of Liquid Mixing
in a T-micromixer
with Convergent–Divergent Mixing
Channel
T. Manoj Dundi, V. R. K. Raju, and V. P. Chandramohan
1 Introduction
Micromixers are one of the most predominant equipments among miniaturized

devices. They have replaced the macro-reactors and are being widely utilized in
various applications in the field of chemical and bio-medical engineering. Consuming
very tiny amounts of costly reagents, fast analysis, better portability, and low cost
of manufacturing [1] are some of their various advantages in comparison to macro-
reactors. Micromixers are generally two types, active (employ an external energy
source to induce mixing) [2] and passive (no external energy except pressure head,
channel geometry itself induce mixing) [3] mixers. Passive mixers are desirable over
active mixers because it is easy to integrate them into complex microsystems, easy
to operate, and also incurs low cost.
A T-micromixer is most effective mixing device among the simplest passive
designs available in the literature. Particularly, the T-micromixer operated in the
engulfment regime (occurring at a relatively higher Re) yields a significant amount
of mixing. The study carried out by Engler et al. [4] is one of the earliest to report the
occurrence and inside phenomenon of engulfment regime. They found that when the
centrifugal forces reach a sufficient level at the junction of T-mixer at Re of 199, the
vortices already formed in the vortex regime at a Re of 60 are broken-up and signif-
icantly increase the interfacial area between samples. Bothe et al. [5] employed a
hybrid grid to obtain highly accurate simulations of mixing in T-mixer with chemical
reactions. They observed only engulfment regime in T-micromixer can give efficient
mixing and faster rates of reaction. Fani et al. [6] investigated the engulfment regime
in T-mixer in depth using direct numerical simulations and sensitivity analysis. The
non-fully and fully developed inlet velocity conditions lead to delayed and earlier
onset of critical Re for engulfment. Soleymani et al. [7] through numerical analysis
T. M. Dundi (B) · V. R. K. Raju · V. P. Chandramohan

Mechanical Engineering Department, National Institute of Technology Warangal, Warangal,
Telangana State 506004, India
154 T. M. Dundi et al.
investigated the effect of mixing angle, throttle size, aspect ratio, and flow rates on
mixing efficiency. It is found that the parameters of geometry can considerably affect
the formation of engulfment regime. Cherlo and Pushpavanam [8] through numerical
and experimental study proved that the depth of mixing channel has considerable
influence on the onset of engulfment regime and can be used to either delay or
promote the regime.
The inside phenomenon of engulfment regime was studied in detail [4–6] and
various studies [7, 8] showed the effect of geometric parameters on engulfment
regime. However, there is no study carried out on the effect of shape of mixing
channel on engulfment regime. Therefore, in the present study, mixing performance
is evaluated for a T-micromixer with convergent–divergent mixing channel and also
compared with simple T-micromixer in the Re range of 0–266.
2 Methodology (Uppercase TNR 11 Bold)
2.1 Mathematical Modeling
The governing equations to obtain the flow and concentration fields in the micromixer
are given by Eqs. (1), (2), and (3), respectively.
∇·V =0 (1)
1
V · ∇V = − ∇ P + v∇ 2 V (2)
ρ
(V · ∇)c = D∇ 2 c (3)
where V, ρ, p, and ν are velocity vector, density, pressure, and kinematic viscosity
of the fluid, respectively. And c represents concentration, and D represents diffusion
coefficient.
The mixing quality ‘α’ is evaluated by using the Eq. (4)

σ M2
α =1− (4)
σmax
2
where, σmax
2
is the maximum variance (0.5) and σ M2 is the variance of the mixture
at a cross section given by Eq. (5)
1
n
σ M2 = (ci − c M )2 (5)
n i=1
Numerical Analysis of Liquid Mixing in a T-micromixer … 155
(a)
(b)
Fig. 1 Schematic of a simple T-mixer b T-mixer with convergent–divergent mixing channel
where c M is the average of species concentration for n grid units at the cross
section considered. The mixing quality ‘α’ is ‘zero’ implies no mixing, and if it is
‘one,’ it implies complete mixing.
2.2 Numerical Modeling
The geometries of T-micromixers are created in ANSYS design modeler. Figure 1a

and b shows the 2D schematic of T-mixer with straight and convergent–divergent
mixing channels. The length of mixing channels for both mixers is taken as 3000 μm,
and the width and depth of both mixers are taken as 100 μm.
Validation of results (Subheading 2, Italic TNR 11)
Initially, numerical simulations are carried out for simple T-mixer and the mixing
quality at different Re has been obtained to compare the results with existing literature
and check the accuracy of present numerical model. Mixing quality at the outlet of
T-mixer is calculated using Eq. (4). Then it has been compared quantitatively with
the results reported in Kockmann et al. [9] as shown in Fig. 2. The results are in good
agreement with the reported data.
The mixing performance of simple T-mixer with straight mixing channel and T-mixer
with convergent–-divergent mixing channel has been evaluated for Re of 0 to 220.
The mixing quality at the outlet of both mixers has been calculated using Eq. (4) and
Fig. 2 Comparison of mixing quality at different Re for present simple T-mixer and Kockmann
et al. [9] T-mixer
plotted as shown in Fig. 3. It can be observed that the trends of mixing quality are
very similar for both mixers at different Re and the mixing quality remained nearly
same for both mixers in the laminar and vortex regimes (Re < 150). The mass fraction
contours at different planes in the flow direction in mixing channel for both mixers
are presented in Figs. 4a and b. It can be observed that mixing phenomenon of simple
T-mixer (Fig. 4a) and T-mixer with convergent–divergent mixing channel (Fig. 4a)
Fig. 3 Comparison of mixing quality between simple T-mixer and T-mixer with convergent–diver-
gent mixing channel at different Re
Fig. 4 Comparison of mass fraction contours at different planes at a Re of 106 a simple T-mixer
b T-mixer with convergent–divergent mixing channel
remained very similar. The convergent–divergent channel could not alter the flow
phenomenon much in the downstream from junction due to dominant viscous forces
in this Re range.
From Fig. 3, it is seen that the onset of engulfment regime has started pretty
much at a same Re (Re of 200) for both mixers. However, after the onset of engulf-
ment, for Re > 200, T-mixer with convergent–divergent mixing channel has shown
a significantly higher mixing as compared to simple T-mixer (Fig. 3). The conver-
gent–divergent mixing channel sustained and enhanced the engulfment effect until
far downstream of the mixing channel. In simple T-mixer, after a certain distance
from the junction of the mixing channel, the mixing phenomenon remained similar
throughout the downstream until the outlet of mixing channel as shown in Fig. 5a.
This is due to the fact that the flow behavior remains unaffected by the straight
wall channels. However, in the case of T-mixer with convergent–divergent mixing
channel, the mixing phenomenon continuously changed and improved in the down-
stream of the mixing channel as shown in Fig. 5b. The convergent–divergent shape of
Fig. 5 Comparison of mass fraction contours at different planes at a Re of 266 a simple T-mixer
b T-mixer with convergent–divergent mixing channel
Fig. 6 Comparison of pressure drop with Re for simple T-mixer and T-mixer with convergent–
divergent mixing channel
mixing channel improved the rolling-up of vortex structures further downstream and
therefore could significantly enhance the interfacial area and thereby mixing quality.
The pressure drop for both mixers at different Re has been plotted as shown
in Fig. 6. Higher pressure drops are observed in case of T-mixer with convergent–
divergent mixing channel at all Re, and the increase is continuously rising as the Re
is increased. The rise in pressure drop implies the presence of flow in the transverse
direction in the mixing channel of T-mixer with convergent–divergent channel. The
dominant viscous forces restricted and damped secondary flows at lower Re (Re <
150), and thereby, no improvement in mixing is obtained. The inertial forces due to
relatively higher Re in the engulfment regime sustained the secondary flows in T-
mixer with convergent–divergent mixing channel and thereby enhanced the mixing
performance.
4 Conclusions
The mixing performance of simple T-mixer and T-mixer with convergent–diver-

gent mixing channel has been evaluated in the Re range of 0–260. The following
conclusions have been made from the numerical results.
• At lower Re (Re < 150), the convergent–divergent channel could not alter the flow
phenomenon much in the downstream from junction due to dominant viscous
forces, and therefore, mixing remained nearly same for both mixers.
• The convergent–divergent mixing channel sustained and enhanced the engulf-
ment effect until far downstream of the mixing channel. Its shape has improved
the rolling-up of vortex structures further downstream and therefore could

significantly enlarge the interfacial area and thereby improve mixing quality.
• Pressure drops are higher for T-mixer with convergent–divergent mixing channel
as compared to simple T-mixer. This implies the presence of secondary flow in
the T-mixer with convergent–divergent channel. However, the dominant viscous
forces constrained and damped secondary flows at lower Re (Re < 150) which
resulted in a mixing phenomenon similar to simple T-mixer. At higher Re (Re ≥
200), the presence of inertial forces sustained the secondary flows generated due
to convergent–divergent channel and therefore considerably improved the mixing
performance as compared to simple T-mixer.
Therefore, employing a T-mixer with convergent–divergent mixing channel
instead of simple T-mixer in engulfment regime is more beneficial for better quality
of mixing at the expense of smaller rise in pressure drop.
Acknowledgements The authors are thankful to the Mechanical Engineering Department, NIT
Warangal, for their support.
References
1. Jeong GS, Chung S, Kim CB, Lee SH (2010) Applications of micromixing technology. Analyst
135(3):460–473
2. Oddy MH, Santiago JG, Mikkelsen JC (2001) Electrokinetic instability micromixing. Anal
Chem 73(24):5822–5832
3. Nguyen NT, Wu Z (2004) Micromixers—a review. J Micromech Microeng 15(2):R1
4. Engler M, Kockmann N, Kiefer T, Woias P (2004) Numerical and experimental investigations
on liquid mixing in static micromixers. Chem Eng J 101(1–3):315–322
5. Bothe D, Lojewski A, Warnecke HJ (2011) Fully resolved numerical simulation of reac-
tive mixing in a T-shaped micromixer using parabolized species equations. Chem Eng Sci
66(24):6424–6440
6. Fani A, Camarri S, Salvetti MV (2013) Investigation of the steady engulfment regime in a
three-dimensional T-mixer. Phys Fluids 25(6):064102
7. Soleymani A, Kolehmainen E, Turunen I (2008) Numerical and experimental investigations of
liquid mixing in T-type micromixers. Chem Eng J 135:S219–S228
8. Cherlo SK, Pushpavanam S (2010) Effect of depth on onset of engulfment in rectangular micro-
channels. Chem Eng Sci 65(24):6486–6490
9. Kockmann N, Engler M, Föll C, Woias P (2003) Liquid mixing in static micro mixers with various
cross sections. In: ASME 2003 1st International conference on microchannels and minichannels.
American Society of Mechanical Engineers, pp 911–918
Performance of Pool Boiling Heat
Transfer on Annealed Nanofilm Surfaces
Sudev Das and R. Johnsan
Nomenclature
q Heating rate (W)

Q Heat flux (W/M2)
h Heat transfer coefficient (W/m2 k)
k Thermal conductivity of metal (W/m.k)
V Voltage (V)
I Current (A)
D Diameter (m)
T Wall superheat (◦ k)
Ts Surface temperature (◦ k)
Tl Saturated temperature of liquid (◦ k)
A Area of cross section (m2 )
Uh Uncertainty of heat transfer coefficient
Uq Uncertainty of heat flux
UA Uncertainty of area
1 Introduction
The conventional thermal management systems are unable to meet the present world
demands due to their low heat transfer coefficients, low thermal conductivity, low
specific heat, and high specific volume pertaining to the coolant and system. The
improvement in heat transfer performance will result in reducing the size and cost of
S. Das (B) · R. Johnsan

Department of Chemical Engineering, National Institute of Technology Calicut, Calicut, Kerala
673601, India
e-mail: sudev@nitc.ac.in
162 S. Das and R. Johnsan
heat dissipation equipment by lowering the material and manpower needed. Among
the many techniques exploited in thermal management, pool boiling garnered wide
attention from researchers due to the transit of substantial amount of heat at minimal
driving force lacking moving parts [1–3]. With this credit, it plays an inevitable
role in thermal management of supercomputers, nuclear reactors, and spacecraft.
The manipulation of heating surfaces with nanostructures of metallic oxide for the
enhancement of pool boiling heat transfer is greatly studied; however, its structural
stability and compatibility is unexplored. Micro/nanostructured surfaces are garnered
wide attention for their tendency to ameliorate both critical heat flux (CHF) and heat
transfer coefficient (HTC) [4–10]. The current phase of work is dedicated to the
synthesis and analysis of the annealed TiO2 surface for the pool boiling heat transfer
enhancement by the physical vapour deposition technique.
2 Experimental Facility Details
2.1 Experimental Setup and Procedure
The experimental pool boiling rig is similar to previous work that has been used in this
work [13, 14]. Pool boiling setup is used for the measurement of heat transfer coef-
ficient on TiO2 and SiO2 nanostructure-coated (over Cu) surfaces, without and with
annealing, along with natural and smooth Cu surfaces. In Fig. 1, a schematic sketch
of the pool boiling apparatus along with its actual photographic view is displayed.
The boiling apparatus consists of various parts such as boiling chamber, heating
block, ceramic insulator block, condenser, liquid–vapour separator, and surge tank.
Fig. 1 Schematic representation of experimental apparatus with parts

Performance of Pool Boiling Heat Transfer … 163
A heating block with a heater unit is placed at the lower section of boiling rig and its
top surface was exposed to the test fluid at saturated condition.
The test surface (0.009 m) is mounted coaxially on the top surface of the heating
block which is exposed to the test fluid. Apart from the test surface, all other parts of
the heating block are insulated by Teflon covering to avoid heat loss to the ambience.
The outermost surface of the Teflon insulation is observed stationary ensuring the
steady-state condition prevailing. The magnitude of the outermost surface tempera-
ture is low to consider the heat loss to the ambience trivial. A tight contact between
the Teflon insulation and the heating block was ensured to avoid the seepage of
liquid, which will pose uncertainty in the surface temperature measured by three k
type thermocouples.
The discrete range of heat flux to the system was accomplished by an autotrans-
former by providing voltage to the heater in steps. The heating block was equipped
with three holes equidistantly for accommodating k type thermocouples of diam-
eter 0.001 m. The interface temperature (T s ) prevailing between the bulk liquid and
the nanostructured surface was calculated by Eq. 1, and the bulk liquid temperature
is continuously monitored to ensure adherence to the saturation state. The pres-
sure pertaining to the system is maintained at atmospheric pressure with the aid of
condensers (internal and external) to avoid fluctuation in the saturation temperature.
2.2 Test Surface
Copper material is preferred as a test surface due to its higher thermal conductivity
along with high electrical conductivity, low cost and also its exploitation in bulk in
industries. Five 1000 nm TiO2 and 1000 nm SiO2 -coated nanostructure without and
with treatment at 500 °C for 24 h were developed for the evaluation their pool boiling
performance. Electron beam physical vapour deposition (EBPVD) method (15F6,
HHV India) was employed for the formation of TiO2 nanostructures over copper
substrate to create micro/nanostructure.
2.3 Data Reduction Methodology
The interface or surface temperature (T s ) was calculated by dimensional Fourier law

of heat conduction. Equation 1 is obtained from the resistance diagram, 1(d) and used
to calculate the surface temperature (T s ) is measured from the thermocouple reading.
With the parameters V and I, the heat flux delivered to the system is obtained by the
relation 2. From Eq. 3, the Newton law of cooling, the heat transfer coefficient, h,
was calculated.

VI Z 3y Z x y Z ys VI
Ts = T3 − + + = T3 − δ (1)
K A3x Ax y Ays K
Table 1 The uncertainty involved in the measured parameters

Sl. no. Parameter Measuring instrument Uncertainty
1 Current Digital ammeter
2 Temperature Thermocouple ±0.5 °C
3 Diameter and width of heating surface Linear scale LC
2 = ±0.025 cm
(L.C. 0.5 mm)
4 Voltage Digital voltmeter

Z 3y Z x y Z ys
where δ = A3x
+ Ax y
+ Ays
q VI
Q= = π 2 (2)
A 4
d
q q
h= = (3)
T Ts−TL
2.4 Uncertainty Analysis
The uncertainties involved in the parameters pertaining to the system are estimated to
fathom the level of enhancement. Following the methodology of Kline and McClin-
tock, the uncertainties were calculated [10]. The range of uncertainty pertaining
each parameter was tabulated in Table 1 and the uncertainty in the calculation of heat
transfer coefficient (h) using Eq. (4) is 6.5–13.5%.

2
2
UQ −QU A
UH +
A(TS − TL )
A2 (TS − TL )
2 2 1/2
−QUTS QUTL
+ + (4)
A(TS − TL )2 A(TS − TL )2
2.5 Results and Discussion
The six surfaces, untreated, treated, EBPVD coated, and annealed, of copper are
characterized in possible means to enlist the parameters behind the enhancement
of pool boiling heat transfer when they are employed as the heating substrate. To
comprehend the amelioration of pool boiling performance, the changes in parameters
such as surface roughness, active nucleation sites, dimensions of the surface cavities,
and the structural stability are needed to be quantified. The surface morphology and
topography of the substrates were characterized by SEM and AFM techniques. The
cavities formed upon the deposition of SiO2 and TiO2 by EBPVD method is depicted
in the SEM images as shown in Fig. 2. The altitude of the peaks in the surface was
Fig. 2 SEM characterization of heating micro/nanostructure where a untreated, b polished, c and

d 1000 nm TiO2 thin film with and without annealed, e and f SiO2 1000 nm thin film without and
with annealed
Fig. 3 TEM characterization of a TiO2 and SiO2 indicating rod-like crystal
increasing with the increasing thickness of the coating, which facilitates the sponta-
neous evaporation of liquid. The TEM characterization was carried out to picturize
the individual structure of the crystal and its crystallinity. From Fig. 3, the crystals
were found to be in rod-like structure aiding in forming deeper cavities accounting
for abundant active nucleation sites. A uniform surface topography was observed for
the coated surfaces evident from the AFM images from Fig. 4. The components or
the composition of the surface coatings were analysed by characterizing the surface
by EDX techniques, which confirmed the presence of copper (Cu), titanium (Ti),
silicon (Si), and oxygen (O). The presence of peak corresponding to oxygen makes
us to fathom the presence of metallic oxides evidently (Figs. 5, 6).
The performance of the substrates was analysed by conducting a pool boiling
heat transfer study for a saturated water at different heat flux input and plotting
heat flux versus the wall superheat ( T ) to obtain what is called boiling curve
as shown in Fig. 7. The heat flux intakes (q) by the liquid increased monotonic
with the increase in the wall superheat. The vital point that has to be noted is the
extent of increase in heat flux intake for a prescribed wall superheat, interestingly,
the annealed TiO2 heating surfaces exhibited maximum heat flux intake for a lower
wall superheat among all other heating surfaces. In addition, the reduction rate of
wall superheat of TiO2 has increased comparing with SiO2 nanostructured coating
surfaces. Thus, it is ascertained that the heat transfer coefficient can be enhanced
with such surface fabrication technique (EBPVD) and modification (annealing) by
reducing wall superheat at elevated heat flux. The superior performance of 50%
dip in wall superheat is recorded for 1000 nm TiO2 -coated nanostructured copper
Fig. 4 AFM characterization of copper substrate with TiO2 and SiO2 NF surfaces where a untreated,
b polished, c and d 1000 nm SiO2 thin film without and with annealed e and f 1000 nm TiO2 thin
film with and without annealed
surface with annealing at maximum high flux intake as a corollary of surface wetting
mechanism evident from the previous work [11–15].
Thus, the substantial growth in the HTC was observed with TiO2 -coated copper
substrate. The amelioration of the nucleate boiling heat transfer coefficient (NBHTC)
exhibited by TiO2 -coated nanostructure surfaces are conducive of innumerable active
nucleation sites and also evident from the previous work [16–18]. Moreover, a notable
rise in NBHTC pertaining to TiO2 coating is the consequence of the synergic order
Fig. 5 EDX spectra for TiO2 nanofilm surface
Fig. 6 EDX spectra for SiO2 nanofilm surface
between the elevated bubble generation frequencies, abundant nucleation sites, and
the extravagant of bubble interaction.
The enhancement of heat transfer coefficient of pool boiling being the corollary
of surface characteristics, the structural stability and its intact nature needs emphasis.
Figure 8 depicts that the annealed TiO2 surface underperforms the untreated copper
substrate by a significant margin recording greater heat transfer coefficient at varied
heat flux intake. The NBHTC displayed monotonic increase with increasing heat flux,
alas and trivial degradation in NBHTC was observed compliant to successive boiling
Fig. 7 Boiling curve of heating surfaces
Fig. 8 Boiling dependency curves for repeated run of experiments
runs. The average degradation of the NBHTC compared with first run and fifth run is
closely 7.5% for the nanostructure surface TiO2 annealed surface and 11.5% for the
untreated surface. Promisingly, the NBHTC reached stationary value after several
runs. The negligible degradation of the surface morphology of the nanostructure due
to the rampant hydrothermal environment of pool boiling contributes to the decrease
in NBHTC. It is heralded that the annealed TiO2 is resistive to the degradation in
surface morphology and also promoting pool boiling heat transfer evident from the
reported results.
3 Conclusion
Nanoscale film of silicon and titanium oxides was deposited over copper substrate
for the enhancement of pool boiling heat transfer, and the effect of annealing was
investigated to improve the hydrothermal stability of nanostructure to extend the
lifetime. The nanostructures were crystalline with the formation of deep trenches
laden with innumerable active nucleation sites which are evident from SEM and
XRD analysis. Annealed TiO2 nanostructured surface of 1000 nm thickness recorded
the maximum and substantial enhancement of nucleate boiling heat transfer coeffi-
cient. This remarks that TiO2 annealed surface underperforms annealed SiO2 surface
of same thickness which has positive trend towards enhancement. The profound
improvement in the HTC is the corollary of enhanced hydrophilicity, abundant
bubble formation sites and drastic departure frequency of bubbles emanating from
the nanostructured surfaces. In essence, the effect of annealing of titanium oxide
nanostructure-coated copper substrate on pool boiling heat transfer enhancement
was studied elaborately to proclaim that annealing is a potent tool to enhance the
hydrothermal stability of nanostructured coating.
Acknowledgements This work was supported by the Faculty Research Grant-2019, National Insti-
tute of Technology Calicut. The authors thank NIT Agartala, SAIF (IITB), and SAIF (IITP) for
characterization of heating surfaces.
References
1. Mudawar, Assessment of high-heat-flux thermal management schemes. IEEE Trans Compo-

nents Packaging Technol 24(2):122–140
2. Components and Packaging Technologies (2001) 24(2):122–140
3. Tang YB, Tang Q, Li J, Qing L, Chen K (2013) Pool boiling enhancement by novel metallic
nanoporous surface. Exp Therm Fluid Sci 44:194–198
4. Zhang BJ, Kim KJ (2012) Effect of liquid uptake on critical heat flux utilizing a three
dimensional, interconnected alumina nano porous surfaces. Appl Phys Lett 101:054104
5. Wu W, Bostanci CH, Hong LC, Kizito JP (2010) Nucleate boiling heat transfer enhancement
for water nd FC-72 on TiO2 and SiO2 surfaces. Int J Heat Mass Transf 53:1773–1777
6. Das S, Bhaumik S (2014) Enhancement of nucleate pool boiling heat transfer on titanium oxide
thin film surface. Arab J Sci Eng 39:7385–7395
7. Lu MC, Chen R, Srinivasan V, Carey VP, Majumdar A (2011) Critical heat flux of pool boiling
on Si nanowire arraycoated surfaces. Int J Heat Mass Transf 54:5359–5367
8. Shi B, Wang YB, Chen K (2015) Pool boiling heat transfer enhancement with copper nanowire
arrays. Appl Therm Eng 75:115–121
9. Ahn HS, Sinha N, Zhang M, Banerjee D, Fang SK, Baughman RH (2006) Pool boiling
experiments on multi walled carbon nanotube forests. J Heat Transf 128:1335–1342
10. Ujereh S, Fisher T, Mudawar I (2007) Effects of carbon nanotube arrays on nucleate pool
boiling. Int J Heat Mass Transf 50:4023–4038
11. Das S, Johnsan R, Sujith Kumar CS, Datta A (2020) Experimental study of pool boiling heat
transfer on an annealed TiO2 nanofilm heating surface. J Therm Anal Calorim https://doi.org/
10.1007/s10973-020-09503-3
12. Gajghate SS, Das S, Saha BB, Bhaumik S (2020) Effect of ZrO2 nanoparticle deposited layer
on pool boiling heat transfer enhancement. Heat Transfer Eng. https://doi.org/10.1080/014
57632.2020.1777011
13. Gajghate SS, Barathula S, Das S, Saha BB, Bhaumik S (2019) Experimental investigation and
optimization of pool boiling heat transfer enhancement over graphene-coated copper surface.
J Therm Anal Calorim. https://doi.org/10.1007/s10973-019-08740-5
14. Das S, Saha B, Bhaumik S (2017) Experimental study of nucleate pool boiling heat transfer of
water by surface functionalization with SiO2 nanostructure. Exp Therm Fluid Sci 81:454–465
15. Das S, Saha B, Bhaumik S (2017) Experimental study of nucleate pool boiling heat transfer
of water by surface functionalization with crystalline TiO2 nanostructure. Appl Therm Eng
113:1345–1357
16. Das S, Majumder B, Bhaumik S (2017) Effect of wettability on micro and nanostructure
surface using sessile droplet contact angle for heat transfer application. IJST-T MECH ENG
41:129–135
17. Das S, Durbba SK, Bhaumik S (2016) Experimental study of nucleate pool boiling heat transfer
of water on silicon oxide nanoparticle coated copper heating surface. Appl Therm Eng 96:555–
567
18. Das S, Bhaumik S (2016) Experimental study of nucleate pool boiling heat transfer using water
on thin film surface. IJST-T MECH Eng 40:21-29
19. Das S, Swapan B (2016) The effect of coating thickness and roughness of nucleate pool boiling
heat transfer on nanoparticle coated surface. J Instit Eng (India): Series E 97:55–62
20. Schultz RR, Cole R (1979) Uncertainty analysis in boiling nucleation. Int AIChE Symp Ser
75:32–38
Computational Study of the Unsteady
Wave Drag Reduction at Hypersonic
Mach Number
D. Rajasekhar and Bibin John
1 Introduction
Hypersonic flow has some distinctive and important physical phenomena including
a thin shock layer, entropy layer, viscous interaction, high-temperature effect and
low-density effect [1]. In high-speed flights, the leading edge of the vehicle must be
of a large-angle blunt cone configuration, in order to address the severe effects of the
high levels of pressure drag and heat transfer encountered during the atmospheric
flight of hypersonic vehicles. Theoretically, when a high-speed flow moves past a
blunt body, it experiences a deceleration across a bow shock wave created in front
of the blunt body. The flow downstream of the shock wave loses most of its kinetic
energy and gains a large increase in the local pressure and temperature, which cause
a high drag and heat transfer on the body. So, the minimization of the wave drag has
gained significant interest in order to reduce the cost of space travel or to improve
the range of supersonic missiles.
Conventionally, many flow control techniques have been put forth to reduce the
wave drag by means of shock layer alterations. A widely studied passive technique
is the use of an Aero-spike [2] with different aero discs that are attached to the tip
of the hypersonic vehicle. The usage of the “physical spikes” proved to be capable
of offering substantial drag reduction rates, up to 30–45% [3]. Another active tech-
nique proposed in the literature is stagnation point injection [4]. In this case, the
fluid jet injected from the stagnation point of the blunt body acts as a fluid spike and
recasts the shock structure, thus reduces the stagnation point pressure and overall
drag. Moreover, this technique has a better potential of reducing drag and heat flux
simultaneously. However, the requirement of additional injection circuits overturns
its relative merits. Concentrated energy addition to the field ahead of the body is
another technique proposed for the drag reduction [5]. Although these three tech-
niques have been widely studied and proven as potential methods for reducing the
D. Rajasekhar · B. John (B)

School of Mechanical Engineering, Vellore Institute of Technology, Vellore 632014, India
174 D. Rajasekhar and B. John
wave drag, their real-time implementation poses lot of challenges. Moreover, the
viability of these techniques is not good, especially in case of stagnation point injec-
tion. Hence, the present work focuses on the investigation of an active flow control
technique centred on the concept of pulsed energy addition. This technique differs
from the concentrated energy addition, as the later offers a time-dependant drag
reduction rather than steady-drag reduction.
2 Methodology
The investigation of pulsed energy addition-based drag reduction is carried out using
computational fluid dynamics. Since the focus of the present work is study of wave
drag reduction, the viscous aspects of the flow are omitted from this study. This is
relatively a realistic assumption, since the frictional drag contribution to the overall
drag is very minimal in comparison with that of wave drag in hypersonic speeds.
Therefore, the flowfield of interest is resolved by solving inviscid compressible flow
governing Euler equations. The 2D-axisymmetric form of Euler equations [6] can
be written as:
∂U ∂F ∂G
+ + + αS = 0 (1)
∂t ∂x ∂y
where U is the conservative variable vector, F and G are fluxes in xand y-directions,
respectively, and S represents axisymmetric source term. U, F, G and S are given by,
⎡
⎤ ⎡ ⎤
ρ ρu
⎢ ρu ⎥ ⎢ 2 ⎥
U =⎢ ⎥ F = ⎢ ρu + p ⎥
⎣ ρv ⎦ ⎣ ρuv ⎦
ρE ρu H
⎡ ⎤ ⎡ ⎤
ρv ρv
⎢ ρuv ⎥ ⎢ ρuv ⎥
G=⎢ ⎥S = 1 ⎢ ⎥
⎣ ρv + p ⎦
2
y⎣ ρv 2 ⎦
ρv H (ρ E + p)v
The freestream and geometry conditions used for the present study are adapted
from the experimental investigation of drag and heat transfer by Sahoo et al. [7] and
are tabulated in Table 1.
In the present study, the flowfield is considered as 2D-axisymmetric due to the
axisymmetric nature of the blunt body and zero angle of attack of the flowfield. The
Computational Study of the Unsteady Wave Drag … 175
Table 1 Freestream and geometric conditions

M∞ T∞ (K) P∞ (Pa) Model base diameter (m) Model nose radius (Rn ) (m)
5.75 90 180 0.07 0.03/0.035
Fig. 1 Computational
domain used for the
simulation
computational domain meshed with 360 × 280 quadrilateral cells and marked with
boundary conditions is shown in Fig. 1.
The computational simulations were performed initially without energy addition
to obtain the steady-state shock structure of the blunt body. Later, this steady-state
solution is taken as the initial guess for the unsteady blast wave simulation. The
impact of added energy pulse is specified such that the temperature will increases
instantaneously in the location of the energy addition; thus, it increase the internal
energy of the fluid particles entrapped in the energy addition region. Therefore, the
new static temperature is defined by the equation [8] as:
T (r, z) = Ts (r, z) + T e−(r/ro )

2
(2)
where 0 ≤ r ≤ ro . Where r is the radial distance in the filament, ro = d/2 is the filament
radius, Ts (r, z) is the steady-state temperature , T is the temperature increase due
to energy deposition. Hence, the density and the velocity vectors can be assumed to
remain unchanged in the energy location initially. Similarly, the pressure is increased
according to the ideal gas equation, i.e.,
p(r, z) = ρs (r, z)RT(r, z) (3)
And therefore
p(r, z) = ps (r, z) + pe−(r/ro )

2
(4)
where
p = ρs (r, z)RT (5)
where ρs (r, z) is the steady-state density, ps (r, z) is the steady-state pressure, p

is the pressure rise due to energy deposition. For the present analysis, the energy is
assumed to be deposited into a spherical bubble of 2 mm radius. Parametric study is
performed by varying the energy bubble location. Two locations, viz., 30 and 50 mm
upstream of the stagnation point of the blunt body are considered.
A detailed grid independence study is carried out to ensure absence of grid resulted
error in the final solution. The solution obtained with above-mentioned freestream
conditions and an energy pulse of magnitude 0.1 J is used here to discuss the grid
independence study. Figure 2 shows that the solution is very identical for grid levels
of 320 × 240 and 360 × 280. Hence, a medium level grid refinement is followed for
the subsequent studies. Further time independent study is also performed to check
the effect of time step on the transient solution. The drag variation obtained with
different time step sizes is presented in Fig. 3. The present solver employs a CFL-
based global minimum t calculation; hence, it ensures a time step which is of the
order of fixed time step employed for the time independent study. Moreover, this
provision of in-house solver ensured stable time marching until the attainment of
specified total time.
Fig. 2 Grid independence

study
Fig. 3 Time independence

study
The simulations are carried out with above-discussed numerical setup. Energy bubble
location of 50 mm upstream of the stagnation point and energy bubble of 2 mm
radius are considered. It can be clearly seen from Fig. 4 that the pulsed energy
addition caused blast wave simultaneously grow in size and translate towards the
blunt body as time elapses. At t = 0, the centre of the blast wave was at 50 mm
upstream of the stagnation point. Later, the wave has grown in size and translated
towards the blunt body as represented by the flowfield at t = 9 μs. However, at t =
15 μs, the wavefront of blast wave starts to interact with the standing bow shock in
both the blunt body cases. At instance, at around t = 18 μs, the first high-pressure
region of moving shock (BW) hits the stagnation region of the blunt body. The
impingement of BW front on the wall of the blunt body leads to shrinkage of blast
wave and subsequent generation of high surface pressure in the stagnation region.
Due to this reason, the drag coefficient increases. Subsequently, the low-density core
of the blast wave enters the shock layer. Therefore, the shock layer gets pushed
upstream (lens effect) and the intensity of bow shock reduces. A maximum drop-in
stagnation region pressure obtains when the low-density core of the blast wave hits
the wall. Due to this, the drag coefficient records its local minima at theat instance,
as evident in Fig. 3. The lens effect slowly diminishes and the drag value starts rising
further. A representative schlieren image of this phase (t = 50 μs) is given in Fig. 4.
Subsequent to low-density core, the rear high-pressure region of the blast wave hits
the blunt body, which in turn leads to further increase in stagnation pressure. As a
result of interaction of blast wave with shock layer a vortex sheet gets formed. Thus,
a formed vortex sheet observed to be wetting the blunt body surface and reducing
the surface pressure values further. That is the reason for the observed second stage
drag reduction in Fig. 3.
t=9μs t=15μs
t=25μs t=50μs
t=75μs t=100μs
Fig. 4 Large view of the flowfield at different times
As the vortex passes the outlet of the domain, the flow will regain to its initial
steady-state condition. Hence, it is important to investigate the variation of drag
during this entire process. Therefore, the drag variation from the point of initiation
of blast wave to the point of vortex disappear has been recorded. Thus, obtained
drag coefficient variation for two different blunt bodies is compared in Fig. 5. The
sphere model has higher steady-state drag as compared to blunt cone model. This
observation is very obvious, since the sphere has a larger frontal area normal to the
flow. The drag rise corresponding to the hitting of blast wavefront on the stagnation
Fig. 5 Comparison of drag variation for two different blunt bodies
region is noted as almost same in both the cases. The drag histories of two blunt
bodies have notable disparity during vortex interaction period.
Parametric study revealing the influence of BW origin location analysed in terms
of drag variation is presented in Fig. 6. It is evident from Fig. 6 that the farther the
energy addition location larger the drag reduction. This is due to the fact that the blast
wave originated at a location 50 mm ahead of the stagnation point will be bigger
in size than that of 30 mm case at the time of interaction. So, the interaction of a
bigger low-density core with the body can induce much-improved drag reduction as
observed in Fig. 6. However, the time instances of peak drag and minimum drag are
different for different energy addition locations. Further, it has been revealed that
too much upstream blast wave origination is not favourable as far as drag reduction
is concerned. In such a case, the blast wave may grow bigger than the blunt body
before its reach to the blunt body location.
4 Conclusions
Computational study has been carried out to explore the possibility of time-dependant
drag reduction through pulsed energy addition. Drag reduction obtained by adding
single energy pulse added upstream of the stagnation point was studied. It has been
observed that the pulsed energy addition creates a blast wave that propagates and
interacts with the shock layer of the blunt body. The nature of interaction found to
be similar for both the blunt bodies. However, the amount of drag reduction differs.
Fig. 6 Comparison of drag variation for different BW origins
The relative sizes and configurations of blast wave and shock layers are noted as
the deciding factors of instantaneous drag reduction. A change in initial location
of the BW leads to change in the size of blast wave involved in the shock layer
interaction. Higher drag reduction has been recorded when the origin of the blast
wave shifted from 30 mm upstream to 50 upstream. Finally, it has been revealed
that the location of energy addition/blast wave origination must be chosen to ensure
complete swallowing of low-density core of the blast wave by the shock layer.
Acknowledgements This research was supported by SERB, a statutory body of the Depart-
ment of Science and Technology, Government of India through the early career research grant
(ECR/2016/001873) issued to the second author.
References
1. Anderson JD (2000) Hypersonic and high temperature gas dynamics. AIAA Press, USA
2. Crawford DH (959) Investigation of the flow over a spiked-nose hemisphere-cylinder. NASA
TN-D-118
3. Kuo SP (2005) Shock wave modification by a plasma spike: experiment and theory. Phys Scripta
71:535–539
4. Zukovski, Spaid F (1964) Secondary injection of gases into a supersonic flow. AIAA J.
2(10):1689–1696
5. Satheesh K, Jagadeesh G, An experimental study of the effect of plasma injection on the wave
drag on a blunt body in a hypersonic flow. In: Jagadeesh G, Arunan E, Reddy KPJ (eds) Shock
waves–Proceedings 25th international symposium on shock waves, Bangalore , pp 89–93
6. Versteeg HK, Malalasekera W (2007) An introduction to computational fluid dynamics: the

finite volume method
7. Sahoo N, Saravanan S, Jagadeesh G, Reddy KPJ (2006) Simultaneous measurement of aerody-
namic and heat transfer data for large angle blunt cones in hypersonic shock tunnel. Sadhana
31(part 5):557–581
8. Golbabaei-Asl M, Knight DD (2014) Numerical characterization of high-temperature filament
interaction with blunt cylinder at Mach 3. Shock Waves 24(2):123–138
Methodologies to Achieve High
Performance in Cast Aluminum Diesel
Pistons to Meet BS6 and CAFE Norms
Hariprasath Arivukkarasu, Rajkumar Mani, and Selvaraji Muthu
1 Introduction
Over the years, numerous aspects of the diesel engine have been subjected to opti-
mization in attempts to make the engine more efficient. So continuous developments
in diesel engine technology have progressively resulted in continuous advances in
power density and peak cylinder pressure, presently reaching (100 kW/L and 21 MPa)
and beyond in aluminum diesel applications. In order to meet increasingly demanding
regulatory emission, fuel economy and commercial requirements, engine designers
are constantly seeking to improve and optimize present combustion strategies.
PCP in excess of 21 MPa and specific power approaching 100KW/l are expected
for highly rated development projects (Fig. 1).
As temperatures and pressures in the piston combustion bowl rapidly approaching
and exceeding 420 °C and 21 MPa, it is largely increasing the thermal and mechanical
loads on the piston that leads to premature fatigue cracking in the combustion bowl
rim and/or base regions, Fig. 2.
H. Arivukkarasu (B) · S. Muthu

Design and Development, MAHLE Engine Components India Pvt Ltd, Chennai, India
e-mail: hariprasath.arivukkarasu@mahle.com
S. Muthu
e-mail: Selvaraji.muthu@mahle.com
R. Mani
Numerical Simulation, Design and Development, MAHLE Engine Components India Pvt Ltd,
Chennai, India
e-mail: rajkumar.mani@mahle.com
184 H. Arivukkarasu et al.
Fig. 1 Market trend and development target for HD diesel segment
Fig. 2 Bowl failure crack at combustion bowl rim
2 Loads on Diesel Piston Combustion Bowl
During combustion, the piston is subjected to some of the highest thermal and
mechanical stresses occurring in the engine. The lifetime of a diesel piston is typi-
cally determined by the rim of the combustion bowl, which is subjected to particularly
heavy stresses resulting from the peak cylinder pressure in conjunction with the high
temperatures occurring in that area. Under the highest stresses, temperatures there
can reach 400–450 °C, depending on the material, as explained in detail by Span-
genberg et al. [1]. The critical mechanical stress at the bowl rim results from the
alternating bending of the piston around the piston pin under gas pressure (Fig. 3).
The thermal load from the combustion process is also a cyclic load on the piston.
It acts primarily during the expansion stroke on the combustion chamber side of the
piston.
In the other strokes, depending on the operating principle, the thermal load on the
piston is reduced, interrupted, or even has a cooling effect during gas exchange. In
Methodologies to Achieve High Performance … 185
Fig. 3 Superposition of mechanical and thermal stresses at the bowl rim of a diesel piston
general, heat transfer from the hot combustion gases to the piston occurs primarily
by convection and only a slight portion results from radiation.
In relation to the expansion stroke, the duration over which the thermal load
from combustion acts is very short. Therefore, only a very small portion of the
component mass of the piston, near the surface on the combustion side, follows
the cyclic temperature fluctuations. Nearly, the entire mass of the piston reaches a
quasi-static temperature, which can have significant local variations [1].
3 Combustion Bowl Failure Mechanism
Generally, two different types of combustion bowl failure mechanisms are known to
dominate and interact in diesel pistons:
(i) Failure due to high cycle fatigue (HCF)
(ii) Failure due to thermo-mechanical fatigue
3.1 Failure Due to High Cycle Fatigue (HCF)
The high cycle stress is due to the maximum gas force acting on the piston for a large
number of load cycles. During maximum combustion pressure loading, the piston
bends around the piston pin, resulting in a complex distribution of stresses through
the piston. Overloading a diesel piston in this mode typically results in bowl rim
crack. Failure of this type can be fast propagating and may result in catastrophic
engine failure (further information can be found at [2]).
3.2 Failure Due to Thermo-Mechanical Fatigue (TMF)
The second typical case of piston combustion bowl failure is thermo-mechanical

fatigue (TMF). Various sources and drains of heat transfer from the engine
to the piston resulting in non-homogeneous temperature distribution. The non-
homogeneous thermal expansion driven by that temperature field induces high
compressive stresses during heating and tensile stress during cooling at the
combustion bowl. These stresses can result in initiation of slow-growing cracks [2].
4 Technological Solutions
In order to overcome these failures, Mahle has developed advanced product solutions
as given below.
• Topcast piston
• Elevated salt core cooling gallery (eSC1/2/3)
Depending on the individual engine application, the criterion for improving
aluminum piston properties can either be used individually or in combination.
5 Topcast Piston
To increase the piston crown durability by material modification, the introduction

of a more refined aluminum microstructure is the proven technology for the thermal
fatigue in aluminum pistons.
In piston production for diesel engines, Mahle utilizes the extremely heat-resistant
aluminum alloy. In addition, Mahle improved the casting process with its newly
developed Topcast method. With TOPCAST, a fine microstructure can be achieved
in the high-stress zone of the bowl rim, which improves fatigue resistance and the
resistance to temperature fluctuations, as explored in advanced piston technologies

by Selvaraji et. al. [3] and [4].
TOPCAST technology is composed by acting in the casting process using special
Al-Si alloy, special mold design, combined with a controlled cooling process for a
faster solidification of the melted aluminum inside the mold (Refer Fig. 4).
The controlled aluminum solidification leads to a very refined and homogeneous
microstructure, finer intermetallic phases, finer primary silicon and reduction of
micro-porosity in the region of the piston bowl rim. This is the region in the piston
most affected by operational thermal fatigue strength where generally might initiate
cracks. The crack usually propagates during the operation time of the engine and
could lead to engine failure, as explained by Rafael Bruno et. al. [5].
Fig. 4 Piston cooling rates in the casting process [1]
Fig. 5 Piston microstructure for Al-Si alloy

Fig. 6 FEA relative lifetime results of standard casting and topcast piston
An example of microstructure refinement is shown in Fig. 5 and is part of the

technology.
Go-Motoring testing is simulated for 1600 h for standard casting piston and
TOPCAST piston. Results are as follows (Refer Fig. 6).
In the simulated piston, The TOPCAST piston has almost five times greater
relative lifetime in the highly critical than standard casting piston.
The TopCast pistons were tested more than 10,000 hours in the test bench using
thermal shock accelerated failure test protocol, in various engines (Fig. 7). In all
validations, the TopCast pistons achieved the validation targets.
Result Summary
(i) TOPCAST piston has 5 × higher relative lifetimes than standard casting
piston.
(ii) Up to 2.5 times higher lifetime than remelted pistons referred to thermal cycle
engine tests.
(iii) Superior lifetime of bowl rim area compared to standard casting pistons.
6 Piston Cooling Gallery
In order to improve aluminum pistons structural strength, the aluminum piston

cooling gallery is the main area of studies. The temperature reduction by increasing
the heat transfer from the piston material to the lube oil in the gallery will increase
the fatigue strength level of Al-Si alloys.
Fig. 7 Engine test results comparing aluminum piston technologies
Significance of cooling gallery position

For positioning of the cooling channel, the causes which lead to HCF and LCF have
to be considered. If a cooling channel is positioned too high, It can lead to too high-
temperature gradients in the bowl rim area. Based on restraint to thermal expansion,
it leads to compressive stress at the bowl rim. Based on the relaxation characteristics
of the material, this may result in higher tension stress under low load/zero load
condition which leads to increased TMF loading and temperature cycling cracks in
thrust—anti-thrust direction [2, 3].
If a cooling channel is positioned too low, it is responsible for a too high-
temperature level in bowl rim and 1st ring groove. This leads to poor material prop-
erties, which can be exceeded while the piston is loaded, and edge cracks will be
formed (HCF).
A low-cooling gallery position is unfavorable because of stress concentration
between the cooling channel and boss. Stress concentrations exist at the upper inner
cooling channel radius (distance to bowl) and at the outer lower area (distance to
groove) as well as at the intersections with inlet bore and outlet bore.
7 Elevated Cooling Gallery (eSC)
With its various versions of elevated salt core cooling gallery (eSC), Mahle has
developed a solution for achieving a huge improvement in piston cooling in the
critical areas of the bowl rim and first ring groove and which is already used in new
engine concepts.
Working on cooling gallery shape design predefined parametric models and

calculating their influences in piston durability using finite elements analysis, it
was possible to optimize for the geometry and position capable to bring greater
temperature reduction aligned with piston lifetime increase (Fig. 8) [5, 6].
Analysis and Experiment
After definition of cooling gallery design and positioning, temperature real-time
measurements were performed to prove the concept that showed the reduction of
piston temperature from 402 to 363 °C in the piston higher temperature with extended
benefits to the other piston regions [1]. The temperature gradient is shown in Fig. 9
for the piston crown, which is the most critical area, the bowl rim.
Once the temperature was measured in the engine, the finite element analysis
calculation was performed simulating a thermal cycle validation test, with the
improvement of two times in total piston lifetime for the thermal cycle engine test
procedure (Fig. 10).
Fig. 8 Cooling gallery

design features
Fig. 9 Temperature reduction comparison

Fig. 10 Piston lifetime prediction by numerical simulation
The simulated cycle results were confirmed in the bench test, with all pistons free
of cracks after 2000 hours engine test.
8 Conclusion
• For diesel pistons, excellent cooling efficiency can be achieved by introducing

elevated cooling gallery instead of baseline cooling galley.
• The relative lifetime of diesel aluminum alloy piston is hugely enhanced by
TOPCAST bowl piston up to five times compared to a standard casting diesel
piston and 2.5 times compared to remelt bowl rim piston.
So these product technologies are exceptional solutions to dominant causes of

piston combustion bowl rim/base failure.
References
1. Spangenberg S, Hettich T, Lazzara M, Schreer K (2014) Pistons for passenger car diesel
engines—aluminum or steel. In: Internationales Wiener Motorensymposium 2014, vol 35
2. Mahle GmbH (2020) Pistons and engine testing-second edition. ISBN 978–3–658–10033–9
3. Selvaraji M, Pramod K, Sathishbabu S, Rajkumar M, Li Q, Zeng S, Chai H (2017) Advanced
Piston technologies for gasoline and diesel engine applications to meet EU6 emission norms. ©
Springer Nature Singapore Pte Ltd. 2017 In: Chakrabarti A, Chakrabarti D (eds) Research into
design for communities. Smart innovation, systems and technologies vol 1. 65. https://link.spr
inger.com/chapter/ https://doi.org/10.1007/978-981-10-3518-0_68
4. Selvaraji M, Pramod K (2019) Methodologies to achieve highest CO2 emission reduction in

automotive systems to meet global CAFE/CAFC norms. © Springer Nature Singapore Pte Ltd.
2019 In: Chakrabarti A (ed) Research into design for a connected world, smart innovation,
systems and technologies, vol 134. https://link.springer.com/chapter/10.1007%2F978-981-13-
5974-3_76
5. Rafael B, Rafael D, Paulo Mordente-Mahle Metal Leve SA (2017) Power cell unit (PCU) for
modern HDD engines. SAE Technical paper 2017–36–0301
6. Eike S, Matthias G, Reiner K (2019) Influence of cooling channel design on piston temperature
of HSD engines. In: 11th Stuttgart international symposium automobile and engine technology,
vol 2
Effect of the Pivot Point Locations
on the Wake Dynamics and Thrust
Performance of a Flapping Elliptic
Airfoil: A Numerical Study
Jit Sinha, Nishanth Sreedharan, and Sunil Manohar Dash
1 Introduction
The understanding of the flapping aerodynamics is seemingly important for the devel-
opment of energy-efficient bio-inspired micro air vehicle (MAV) and unmanned
underwater vehicle (UUWV) (Dash et al. [1]; Lua et al. [2]) that can be used in
traffic monitoring, surveillance, and other combat operations. Previous numerical
and experimental studies (Dash et al. [3]; Triantafyllou et al. [4]) have revealed that
aerodynamics performance by the flapping airfoil is associated with the generation of
the leading edge vortex (LEV) and trailing edge vortex (TEV). In the thrust production
regime, these vortices shed and arranged themselves in a reverse von Karman vortex
street formation in the wake of the flapping airfoil. Ol [5] has performed water tunnel
experiments on pitching and plunging SD7003 airfoil at Reynolds number greater
than 10,000 and identified how the far wake regime gets affected with the change in
the pivot point location. Webb et al. [6] have conducted water tunnel experiments and
numerical simulations to investigate the effects of the pivot point location on SD7003
airfoil with unequal pitching and plunging frequencies. Tian et al. [7] showed that
how propulsive performance of pitching NACA0012 airfoil is modified by changing
its pitch pivot point location. In these previous studies, the significance of pivot point
locations on the flow physics of the near wake region and transient and time-averaged
thrust performance of flapping airfoil are unexplained. To bridge the knowledge gap,
in this study, we have performed a systematic numerical investigation on the near
J. Sinha (B) · N. Sreedharan · S. M. Dash

Department of Aerospace Engineering, IIT Kharagpur, Kharagpur, India
e-mail: jit.snh@gmail.com
N. Sreedharan
e-mail: nishanths1234@gmail.com
S. M. Dash
e-mail: smdash@aero.iitkgp.ac.in
194 J. Sinha et al.
wake vorticity pattern and the propulsive force generated by the sinusoidal flapping
elliptic airfoil when its pivot point location is varied.
2 Methodology
2.1 Problem Statement
The schematic drawing in Fig. 1 depicts the upstroke and downstroke of the flapping
cycle of the elliptic airfoil. Here, the airfoil follows a sinusoidal heaving and pitching
motion profile as shown in the Eqs. (1–2). The phase angle (φ) difference between
heaving and pitching profile is considered as 90° since at this phase angle the flapping
airfoil generates highest propulsive efficiency [8].
h(t) = h 0 cos(ωt) (1)
θ (t) = θ0 cos(ωt − φ) (2)
Here, h 0 = 0.75c and θ0 = 300 are the heaving and pitching amplitude, respec-
tively. c = 0.04 m is the chord length of the flapping airfoil. The incoming free-stream
flow speed is U ∞ = 0.125 m/s. f and ω are the flapping frequency and angular flap-
ping frequency, respectively. In this study, to investigate the thrust performance of the
flapping airfoil, three pivot points A, B, and C at locations 0.25c, 0.5c, and 0.75c are
considered, respectively, as shown in Fig. 1. For the analysis, two non-dimensional
numbers, Reynolds number, Re = U ∞ c/ν, and Strouhal number St = fA/U ∞ , are
defined. Here, ν is the kinematic viscosity of the fluid, and A is the width of the wake
(= maximum excursion of the trailing edge of the airfoil, i.e., 2 h 0 ). The transient
Fig. 1 Shows the flapping

cycle of the elliptic airfoil.
The pivot points A, B, and C
on the airfoil are located at
distance 0.25c, 0.5c, and
0.75c from the leading edge
respectively
Effect of the Pivot Point Locations on the Wake Dynamics … 195

thrust coefficient is defined as Ct = 0.5 ρU
T
2 S , where T is the thrust force, ρ is the
∞
density of the fluid, and S = 1.0 is the plan form area of the airfoil.
In the present study, a finite volume method (FVM)-based commercial software

package, ANSYS FLUENT, is utilized to simulate the transient incompressible flow
over a flapping elliptic airfoil at Re = 5000 and St = 0.32 with a laminar viscous
flow model. The effectiveness of the ANSYS FLUENT solver in simulating flapping
airfoil problems has been thoroughly verified and documented in the literature [2, 9].
Using the dynamic mesh principle, the conservative form of the governing equations
shown in Eq. (3) is solved in an inertial frame of reference on an arbitrary control
volume cν with moving control surface cs [10].

∂
ρψd V + ρn · ν − νg ψdS − ρ ψ n · ∇ψdS
∂t
cν cs cs

= Sψ (ψ)dV (3)
cν
Here, the variables ν, ν g , ψ , S ψ , and ρ are the flow velocity, moving mesh
velocity, diffusion coefficient, source term of ψ per unit volume, and fluid density,
respectively. n is the unit vector normal to the control surface. Pressure and velocity
coupling are done using the pressure-implicit with splitting operators [10] (PISO)
algorithm. We have used the PISO skewness correction for highly distorted meshes,
and the under-relaxation factors for the momentum and pressure are defined as 0.3
and 0.7, respectively. Spatial and temporal discretization of the momentum equa-
tion is carried out using the second-order upwind scheme and first-order implicit
scheme [10], respectively. A double precision solver setting is followed to approx-
imate the computer-generated round off errors. The convergence criterion for the
iterative method for both momentum and mass residuals is kept O(10–5 ).
2.3 Computational Domain, Mesh, and Validation
An O-type computational domain as shown in Fig. 2a is adopted in the present

simulations. The domain is subdivided into two zones, namely a moving inner zone
1 and a stationary outer zone

2 . Between the two zones, an interface boundary
condition is applied. The inner ( 1 ) and outer ( 3 ) boundaries are located at a radius
of 15c and 30c, respectively, so that the far-field boundary effects on the airfoil ( 0 )
are negligible. At the start of the flapping stroke, the center of the elliptic airfoil is
196 J. Sinha et al.
Fig. 2 a Represents O-type computational domain and b close-up view of the structured mesh near
the elliptic airfoil
located at 0.75c from the origin (O) of the domain. On the inlet section, a uniform
free-stream velocity (U ∞ ) in the X-direction is assigned. No-slip velocity condition
on the airfoil surface and atmospheric pressure condition on the outlet boundary
is applied, respectively. Standard initialization is adopted for computing the result
from the inlet. To obtain the flapping motion profile shown in Eqs. (1) and (2),
a C-programming-based user-defined function (UDF) is written and compiled in
FLUENT.
Grid independency test and validation study.
A structured mesh with quadrilateral elements has been generated in the compu-
tational domain using face meshing. A close-up view of mesh around the airfoil is
shown in Fig. 2b. Here, keeping the outer domain stationary, diffusion smoothing-
based dynamic mesh technique was incorporated only in the inner zone to mini-
mize computational time. To perform the grid independence study, four different
mesh sizes as shown in Table 1 are selected. Since for the mesh size M2, the error
percentage of time-averaged C T is below 0.5%, and therefore, it is chosen for the
rest of the simulations.
Table 1 Grid independency

Grid name No of nodes C T value % Error
test
M4 50,960 0.367 0.19
M3 39,520 0.366 0.25
M2 29,280 0.365 0.41
M1 19,000 0.364 –
Grid (M2) is selected for all the simulations in the present study
In the following section, the effect of the pivot point locations A, B, and C on the
thrust performance of the flapping airfoil is investigated. It is noticed that the transient
C T for three pivot points has distinct profiles as shown in Fig. 3.
To find out the associated near wake flow features for different transient C T
profiles, we have selected a few salient points as shown with the markers in Fig. 3.
Since in Fig. 3, the periodic transient C T profile for upstroke and downstroke is
symmetric, and therefore, we have limited our discussions to downstroke flapping
cycle only.
3.1 Flow Regime at Pivot Point A
At this pivot point location, up to t* < 0.14, transient C T profile has negative values
(see in Fig. 3). The associated vorticity plot in Fig. 4 shows that the residue leading
edge vortex (RLEV), i.e., the leading edge vortex (LEV) generated in the previous
flapping stroke remains attached to the airfoil bottom surface up to t* < 0.14.
This RLEV creates a suction effect on the airfoil along the drag direction that
results in the negative transient C T value seen in Fig. 3. After, RLEV has shed
the airfoil generates positive thrust due to suction effects from the shear layers on
the upper airfoil surface. An LEV formation is noticed at t* = 0.30. As the airfoil
Fig. 3 Shows the transient

C T comparison for different
pivot points. Here, t* is the
non-dimensional flapping
cycle time
198 J. Sinha et al.
Fig. 4 Vorticity contours for pivot location A at the selected instances shown in Fig. 3
decelerates and reduces its pitching angle in preparation of the next flapping stroke,
a reduction in the transient C T profile is noticed for t* > 0.30.
3.2 Flow Regime at Pivot Point B
When the pivot point location is changed to point B, the duration of the negative
transient C T value compared to pivot point A case is reduced (see in Fig. 3). This is
because RLEV, in the point B case, has shed t* > 0.06 that is much earlier than point
A case (compare Figs. 4 and 5). The vorticity plots in Fig. 5 for selective instances
t* > 0.06 are similar to point A cases (see in Fig. 4), and therefore, the discussions
in the previous section can be suitably followed to explain the transient C T profile
of point B case. Note that LEV in point B case is formed before point A case that
leads to an early increasing trend of the transient C T (see in Fig. 3).
3.3 Flow Regime at Pivot Point C
For pivot location C, the vorticity contours near the trailing edge are different
compared to pivot A and B cases. As shown in Fig. 6, the RLEV, in this case,
shed by forming a counter-rotating vortex pair (see at t* = 0.10) which also induce a
RLEV
LEV
Fig. 5 Vorticity contours for pivot location B at the selected instances shown in Fig. 3
Counter rotating
vortex pair
Fig. 6 Vorticity contours for pivot location C at the selected instances shown in Fig. 3
200 J. Sinha et al.
Fig. 7 Time-averaged C T
comparison for different
pivot points, where x is the
distance measured from the
leading edge of the flapping
airfoil
force on the airfoil along the thrust direction. In the presence of LEV suction effect
and the vortex-induced force, the transient C T for point C does not show a negative
value at the start of the downstroke. As the airfoil continues in the flapping stroke,
the LEV grows, and an increasing thrust profile is observed (see in Fig. 3) which
attains a maximum at t* = 0.23. In preparation of the next flapping stroke, as the
airfoil decelerates after t* = 0.23, a reduction in the transient C T is observed.
We have determined the time-averaged C T for different pivot point cases by taking
the cyclic average of the transient C T when the simulation results become periodic.
It can be seen from Fig. 7 that the time-averaged C T is higher for the pivot point A
and C cases, i.e., when the pivot point is located away from the center of the airfoil.
The maximum time-averaged C T is observed for pivot point C, when the pivot point
is located at 0.75c.
4 Conclusions
In the present study, how the thrust performance of a two-dimensional flapping airfoil
is affected by the pivot point location is investigated using numerical simulation. The
elliptic airfoil flaps follow a sinusoidal motion profile at Re = 5000 and St = 0.32. It
noticed that the transient C T for all pivot point cases is symmetric for both upstroke
and downstroke flapping cycle. When the pivot point is located near the leading edge
of the airfoil, a negative transient C T is observed at the beginning of the flapping
stroke. The negative region shortens as the pivot point shifts toward the trailing edge
of the flapping airfoil. By comparing the vorticity plots for different pivot point
locations, we identified that shedding of RLEV and formation of LEV mainly affects
the transient C T behavior. Here, a maximum in the time-averaged C T is observed
when the pivot point is located at 0.75c.
Acknowledgements First author is immensely grateful to IIT Kharagpur for computational support
and Science and Engineering Research Board (SERB) for financial support.
References
1. Dash SM, Lua KB, Lim TT, Yeo KS (2018) Enhanced thrust performance of a two dimensional
elliptic airfoil at high flapping frequency in a forward flight. J Fluids Struct 76:37–59
2. Lua KB, Dash SM, Lim TT, Yeo KS (2016) On the thrust performance of a flapping two-
dimensional elliptic airfoil in a forward flight. J Fluids Struct 66:91–109
3. Dash SM, Lua KB (2016) Thrust enhancement on a two dimensional elliptic airfoil in a forward
flight. Int J Mech Aerospace Indus Mechatron Manuf Eng 10(2)
4. Triantafyllou MS, Triantafyllou GS, Gopalkrishnan R (1991) Wake mechanics for thrust
generation in oscillating foils. Phys. Fluids, A 3:2835
5. Ol M (2007) Vortical structures in high frequency pitch and plunge at low reynolds number.
In: AIAA Fluid dynamics conference and exhibit, Miami, FL, pp 4233
6. Webb C, Dong H, Ol M (2008) Effects of unequal pitch and plunge airfoil motion frequency
on aerodynamic response. In: AIAA Aerospace sciences meeting and exhibition, Reno, NV,
pp 0582
7. Tian W, Bodling A, Liu H, Wu JC, He G, Hu H (2016) An experimental study of the effects of
pitch-pivot-point location on the propulsion performance of a pitching airfoil. J Fluids Struct
60:130–142
8. Anderson J, Streitlien K, Barrett D, Triantafyllou MS (1998) Oscillating foils of high propulsive
efficiency. J Fluid Mech 360:41
9. Anderson JD, Wendt J (1995) In: Computational fluid dynamics, Springer
10. Ferziger JH, Perić M (2002) Computational methods for fluid dynamics. Springer, Berlin
Lattice Boltzmann Simulations
of a Lid-Driven Cavity at Different
Moving Lengths of the Top Lid
Prabir Sikdar, Sunil Manohar Dash, and Kalyan Prasad Sinhamahapatra
1 Introduction
The fluid dynamics of a lid-driven square cavity is a common problem in the field
of numerical investigation, which is often used for validating the new computa-
tional algorithms. The cavity flow regime mainly contains counter-rotating corner
vortices, primary vortices which also has several engineering applications including
the solar collectors [1], short-dwell and flexible blade coaters [2], etc. The factors
like Reynolds number (Re) [3, 4], the number of the moving lids/sides [5], and orien-
tation [6] of the cavity govern the cavity flow regimes. Burggraf [7] has proposed the
analytical and numerical method to find the steady solution of the lid-driven cavity
flow for Reynolds number up to 400. The result of this study shows that a large
primary vortex and two small corner vortices exist inside the cavity. Ghia et al. [3]
used the stream function-vorticity formulation of the Navier–Stokes (NS) equation
and showed that an additional counter-rotating vortex is formed close to the cavity
corner at higher Re.
In recent years, the lattice Boltzmann method (LBM) with Bhatnagar–Gross–
Krook (BGK) collision model has gained considerable success in simulating fluid
flow problem. As compared to the conventional NS solvers, LBM is simple and easy
to implement. Also, it is suitable for parallel computation [8] and complex boundary
value problems. Hou et al. [9] studied the lid-driven cavity flow using LBM with
a single relaxation time (SRT) model. Since SRT model is numerically unstable at
P. Sikdar (B) · S. M. Dash · K. P. Sinhamahapatra

e-mail: iamprabirsikdar@gmail.com
S. M. Dash
K. P. Sinhamahapatra
e-mail: kalyanps@aero.iitkgp.ac.in
204 P. Sikdar et al.
higher Re, Lin et al. [4] have proposed a multi relaxation time (MRT) model with a
different collision term formulation to simulate the cavity flow at high Re.
In order to investigate the effect of the geometric parameter of the cavity on the
vortex structure, researchers have focused on the cavity with different aspect ratios,
including deep and shallow cavities. Lin et al. [4] have investigated the flow regime
in the cavity by varying the aspect ratio from 1 to 4 and Re ranges from 100 to
7500. Mussa et al. [10] have studied the flow features inside a rectangular cavity for
Reynolds number 100 to 1000 and aspect ratio 0.25 to 4. Cheng and Hung [11] have
characterized the flow domain in the cavity of aspect ratio 0.1 to 7 starting from the
Stokes flow regime (Re = 0.01) to Re = 5000. The above studies show that the shape
and size of the vortex, vortex center, and the number of vortices in the cavity depend
on the cavity aspect ratio and Re.
Even though the flow physics of the square lid-driven cavity is well established
in the previous literature, most of these studies are only limited to the square cavity
with the moving lid length equals to side of the square cavity. Since the inertia effect
imparted by the moving lid governs the flow regime development inside the square
cavity, in this study to thoroughly characterize the inertia effect, we have varied the
moving lid length and investigated the associated physics of the flow regime using
LBM.
In two-dimensional space, the governing equation for a steady, viscous, laminar, and
incompressible flow inside a rectangular cavity is given by
∇ · (ρu) = 0 (1)
∇ · (ρuu) = −∇ P + ν∇ · [ρ∇u + (∇u)T ] (2)
where the flow velocity, density, kinematic viscosity, and pressure of the fluid are
denoted by the variables u, ρ, ν, and P, respectively. Here, the effect of gravity
is neglected. In the present study, the above governing equations are solved in the
LBM framework using a D2Q9 lattice model (D2: two dimensions; Q9: nine lattice
directions). The governing equation of LBM is shown in Eq. (3), where along the α
direction, f α is the particle density distribution function, cα is the discrete velocity of
the particle, and f α eq is the equilibrium distribution function. r is the position vector
of the lattice node point and τ is the relaxation time factor.
f α (r + cα t, t + t) − f α (r, t)

eq
f α (r, t) − f α (r, t)
=− (3)
τ
Lattice Boltzmann Simulations of a Lid-Driven Cavity … 205
The following equations are used to calculate the macroscopic density and
velocity.

8
8
ρ= f α and ρu = f α cα (4)
α=0 α=0
For brevity of the manuscript, additional details of LBM formulation and its
implementation are not discussed here which can be referred in Cheng and Hung
[11].
2.1 Problem Definition and Computational Domain
The schematic diagram with appropriate boundary conditions of the present lid-
driven square cavity problem is shown in Fig. 1. At t = 0, velocity (u, v) at all
grid points except on the top wall is taken as zero, as the initial condition. The non-
dimensional velocity of the moving part of the top wall is U = 0.1. Here, the velocity
of the top wall is suitably selected so that the Mach number (Ma < 0.2) of the flow
lies in the incompressible limit. The Reynolds number is defined with a length scale
of the square cavity ‘L’ as Re = UL/ν.
The top wall is divided into two parts in which one part is stationary and the
other part is moving. The length of the stationary part is denoted by L x . Here, L x is
varied from 0.1L to 0.9L with an interval of 0.1L. Using the bounce-back technique,
no-slip condition in the LBM framework is applied on the stationary solid walls. To
apply the moving wall boundary condition, the equilibrium density distribution, f α eq
is suitably modified as discussed by Hou et al. [9].
Fig. 1 Schematic diagram

of the present lid-driven
square cavity problem
2.2 Grid Independence and Validation
To ensure that the results obtained from the present LBM code are independent of
grid size, we have performed the simulations of the square lid-driven cavity with L x
= 0.0L, at Re = 400, for mesh sizes 65 × 65, 129 × 129, 257 × 257 and 513 ×
513. It is clear from Fig. 2 that the predicted v-velocity profile along the horizontal
centerline of the cavity has a marginal error when the grid size is greater than 65 ×
65. Therefore, in the remaining simulations, for Re = 100 and 400, we have followed
the grid size of 129 × 129 and for Re = 1000, the grid size is increased to 257 ×
257 to keep the value of τ within the stability range, i.e., τ < 0.5.
To validate the present LBM code, the computed u and v-velocity along the vertical
and horizontal centerline of the cavity, respectively, are compared with Ghia et al.
[3] and Lin et al. [4] at Re = 400 and 1000. As shown in Fig. 3 the results obtained
from the present solver are in excellent agreement with Ghia et al. [3] and Lin et al.
[4] which confirms the accuracy of the present solver.
The typical flow regime of a two-dimensional full (L x = 0) top lid-driven cavity

condition contains a large primary vortex and two corner vortices as shown in the
streamline plots in Fig. 4 which is also seen in ref. [4, 11]. Notice that when Re of the
flow is increased, the center of the primary vortex gradually moves in the direction of
the geometric center of the cavity and corner vortices are enlarged (Compare Fig. 4
a, b). This shift in the primary vortex center is because of the increased inertia effect
at higher Re, which is also reported by Cheng and Hung [11].
Fig. 2 Comparison of the

computed vertical (v)
velocity for different grid
sizes at Re = 400
Fig. 3 Comparison of a u-velocity and b v-velocity along the vertical and horizontal centerline at
different Re
3.1 Flows for Lx > 0
To investigate how the flow regime in the cavity is affected when a part of the top
lid (i.e., L x > 0) moves, we have carried out simulations by varying L x in the range
0.1L–0.9L, at Re = 100, 400 and 1000.
It is noticed that the number of the primary vortex (PV) within the cavity gets
modified by changing the L x and Re. We have categorized the flow type as Type-I if
the cavity contains one PV and Type-II if the cavity contains two PV. Figure 5 shows
the flow types observed at different Re and L x . Figure 6 depicts the change of the
center location (R) of the primary vortex from the geometric origin as functions of
Re and L x .
Fig. 4 Streamline distributions of a lid-driven cavity (L x = 0) at a Re = 100 and b Re = 1000
Fig. 5 Flow types observed at different L x and Re
Streamlines in Figs. 7, 8, and 9, show Type-I and Type-II flow types, respectively,
at Re = 400 and 1000. Notice that when L x (or the stationary part of the top lid) is
increased, the inertia effect is dominated near the moving part of the top lid. As a
result, the center of the PV shifts upward and toward the moving part of the top lid (see
Fig. 6) and the bottom corner vortices are enlarged (compare Fig. 7 a and 8 a). When
L x is further increased, the inertia effect is significantly reduced. In consequence,
the size of the top primary vortex is shrunk and the enlarged bottom corner vortices
merge to form a second PV as shown in Fig. 9.
Note that the transition from Type-I to Type-II flow types depends on both L x and
Re (see Fig. 5).
Fig. 6 Comparison of the center of the primary vortex (PV) at different L x and Re
Fig. 7 Streamline distributions for a Re = 400 and b Re = 1000 at L x = 0.6 L
4 Conclusions
In the present study, steady-state numerical simulations on a 2D lid-driven square

cavity are performed to investigate the effects of the length of the moving lid and Re
on the flow regime inside the cavity. It is noticed that for a larger moving lid length,
the flow regime is of Type-I type that contains single primary vortex. However, when
the moving lid length is reduced below, a critical value the flow regime changes
to Type-II, containing two primary vortices. This critical length of the moving lid
depends on the Reynolds number of the flow. Moreover, a change in the moving lid
length modifies the inertia effect, which affects the shape and location of the primary
and corner vortices in the cavity.
Acknowledgements The first author is grateful to Indian Institute of Technology Kharagpur for
providing computational support.
References
1. Zdanski PSB, Ortega MA, Fico NGCR Jr (2003) Numerical study of the flow over shallow
cavities. Comput Fluids 32:953–974
2. Aidun CK, Triantafillopoulos NG, Benson JD (1991) Global stability of a lid-driven cavity
with throughflow: Flow visualization studies. Phys Fluids A 3:2081–2091
3. Ghia U, Ghia KN, Shin CT (1982) High-Re solutions for incompressible flow using the Navier-
Stokes equations and a multigrid method. J Comput Phys 48:387–411
4. Lin LS, Chen YC, Lin CA (2011) Multi relaxation time lattice Boltzmann simulations of deep
lid driven cavity flows at different aspect ratios. Comput Fluids 45:233–240
5. Arun S, Satheesh SA (2015) Analysis of flow behaviour in a two sided lid driven cavity using
lattice boltzmann technique. Alexandria Engineering Journal 54:795–806
6. Cheng TS, Liu WH (2014) Effects of cavity inclination on mixed convection heat transfer in
lid-driven cavity flows. Comput Fluids 100:108–122
7. Burggraf OR (1966) Analytical and numerical studies of the structure of steady separated flows.
J Fluid Mech 24:113–151
8. Succi S (2001) In: The lattice boltzmann equation: for fluid dynamics and beyond. Oxford
University Press
9. Hou S, Zou Q, Chen S, Doolen G, Cogley AC (1995) Simulation of cavity flow by the lattice
Boltzmann method. J Comput Phys 118:329–347
10. Mussa MA, Abdullah S, Azwadi CSN, Muhamad N, Sopian K, Kartalopoulos S, Buikis A,
Mastorakis N, Vladareanu L (2008) Numerical simulation of lid-driven cavity flow using the
lattice Boltzmann method. In: WSEAS International conference proceedings. mathematics and
computers in science and engineering, WSEAS
11. Cheng M, Hung KC (2006) Vortex structure of steady flow in a rectangular cavity. Comput
Fluids 35:1046–1062
Shape Modification of Circular Cam
Cylinder for Enhanced Thermal
Performance
Refiya Hussain, B. Ganesh Kumar, and K. Arul Prakash
Nomenclature
Cp Coefficient of pressure
Cd Coefficient of drag
Re Reynolds number
Nu Nusselt number
D1 Diameter of front cylinder
D2 Diameter of rear cylinder
M Major axis length of the ellipse
N1 Minor axis length of the front ellipse
N2 Minor axis length of the rear ellipse
l Length of joining tangent
Deq Equivalent diameter
1 Introduction
A body immersed in fluid flow with separated flow over a substantial part of its
surface is termed as a bluff body. Pressure drag is dominant on the surface of a bluff
body due to the separated flow and has a larger wake region. As the fluid boundary
layer separates from the bluff body due to the adverse pressure gradient, vortex
R. Hussain (B) · K. A. Prakash

Dept. of Applied Mechanics, IIT Madras, Chennai, India
K. A. Prakash
e-mail: arulk@iitm.ac.in
B. G. Kumar
Dept. of Mechanical Engineering, PSGCT Coimbatore, Coimbatore, India
214 R. Hussain et al.
sheets are formed which are shed in a periodic manner from the top and bottom of
the body, and this phenomenon known as vortex shedding has many implications on
the performance and life of the applications.
Researchers are continuously developing new methodologies to suppress vortex
shedding and they are classified as active and passive methodology. Choi et al. [1]
reviewed both the active and passive methodology of vortex shedding and claims that
the passive methodology is easy to implement when compared to active methodology
as they do not require the use of external energy. Also, Chamoli et al. [2] from their
study suggest that to improve the thermal performance of thermal exchangers and
industrial exhaust chimneys, the passive methodology is the predominantly used
technique.
The study so far shows that unlike a circular cylinder with severe flow separation
and large wake region, the non-circular tubes offer less drag and enhanced thermal
performance.
Cam-shaped cylinder is one such shape that is of fundamental importance in
the engineering applications such as crossflow heat exchangers and industrial plant
exhaust chimneys. Borujerdi and Lavasani [3] and Tang et al. [4] emphasized their
study on cam-shaped tubes and suggested that they have better thermal performance.
The literature review indicates that the majority of the work is focused on
suppressing the vortex shedding from circular, triangular elliptical and square cylin-
ders, and only limited study is available on cam-shaped cylinders. The objective
of this paper is to attempt to bridge this gap by numerically investigating the flow
and heat transfer characteristics past a cam-shaped cylinder and modify the shape to
achieve a low drag and a high heat transfer rate for better thermal performance in
crossflow heat exchangers and industrial exhaust chimneys.
2 Methodology
To study the flow and heat transfer characteristics, numerical simulations are carried
out using ANSYS Fluent 15.0 Software. The geometry of the existing circular cam-
shaped cylinder consists of two cylinders joined by their common tangent as shown
in Fig. 1a.
Tang et al. [4] carried out a numerical study and suggested that a cam-shaped
cylinder offers a low drag and a high heat transfer when compared to the circular
cylinder; however, there is no common domain where both a low drag and high heat
transfer is achieved. In this study, the cam-shaped cylinder shape is modified keeping
Deq constant by adjusting the diameter of the front and rear cylinder and the length
of joining tangent, to attain a geometric domain where both a low drag and a high
heat transfer can be achieved. The modified circular cam cylinder and the modified
elliptic cam cylinder are shown in Fig. 1b and c, respectively.
Rectangular flow domain as shown in Fig. 2 is considered for the simulation.
Structured mesh is obtained by employing quadrilateral elements in the flow domain.
The circumference of the cylinder is discretized into 400 grid points and a total of
Shape Modification of Circular Cam Cylinder … 215
Fig. 1 Geometry: a) existing circular cam, b) modified circular cam, c) modified elliptic cam
Fig. 2 Flow domain
200 grid layers are used around the cylinder. Biasing is done to obtain a coarse mesh
away from the cylinder and a finer mesh around it, to capture the wake properties
accurately. The present study is done for Re = 100 and is purely a scientific study
which can be further extended for higher Reynolds number.
The following boundary conditions are considered:
Inlet (velocity inlet): u = U ∞ ; v = 0; T = T ∞ (300 K) cylinder wall (no slip): u
= 0; v = 0; T = 310 K.
Small temperature difference is imposed to neglect the effect of natural convection.
Boundary wall (slip): ∂u ∂y
= 0; v = 0; T = T ∞ (300 K).
Fig. 3 Validation with numerical results
Outlet (pressure outlet): P = Patm .

Pressure-based solver is used with absolute velocity formulation for the tran-
sient 2D simulation. The viscous laminar model with the energy equation enabled is
utilized for the simulation.
The solver setup for the simulations is as follows.
1. Pressure velocity coupling—SIMPLE.
2. Pressure—Second-Order Upwind.
3. Momentum—QUICK.
4. Energy—Second order.
5. Transient—Second-Order Upwind.
The tolerance convergence is set to 10–5 . The time step is given as 0.002 s and
simulated for 2000 iterations.
Validation
Mesh independent and domain independent study had been done and an optimized
mesh with the appropriate domain has been chosen for the study. The solver setup
is validated with the existing literature, as shown in Figs. 3 and 4, respectively.
Case 1: Modified circular cam cylinder

The simulation is carried out for the following dimensions (in meter):
D1 = 0.9, D2 = 0.203; D1 = 0.9, D2 = 0.372

D1 = 0.9, D2 = 0.617; D1 = 0.9, D2 = 0.842
Fig. 4 Validation with experimental results
From the results obtained, it is evident that the modified circular cam has a lower
drag and higher heat transfer when compared to that of an existing cam cylinder.
Also, as D2 increases, the drag decreases and heat transfer increases (except for the
case where D1 and D2 are comparable, i.e., D1 = 0.9 m and D2 = 0.842 m). As
D2 increases the flow separation, denoted as a‘ in Cp versus theta plot in Fig. 5, is
delayed and hence a narrower wake region contributes to a lower drag.
The Nusselt Number is determined by the local temperature gradient which is
inversely proportional to thermal boundary layer thickness. The isotherms as shown
in Fig. 6 show that as D2 increase, the thermal boundary layer thickness decreases
which indicates an increase in heat transfer.
Hence, we have obtained a common domain where both the drag is lower and
heat transfer is higher; however, the change is only marginal when compared to the
Fig. 5 Cp versus theta plot for modified circular cam

Fig. 6 Isotherms for modified circular cam
existing circular cam, and thus the shape is further modified by changing the circular
cylinder to ellipse.
Case 2: Modified elliptic cam cylinder
The simulations are carried out for the following dimensions (in meter):
M = 0.9, N 1 = 0.203, N 2 = 0.2

M = 0.9, N 1 = 0.372, N 2 = 0.3
M = 0.9, N 1 = 0.617, N 2 = 0.6
M = 0.9, N 1 = 0.842, N 2 = 0.8
The results obtained infers that the modified elliptic cam offers lowest drag and
enhanced heat transfer when compared to existing circular cam and modified circular
cam. Also, as N1 and N2 decreases the bluff body transforms into a streamlined body
with complete suppression of vortex shedding as is evident in the streamline plots as
shown in Fig. 7.
Also, the isotherms in Fig. 8 show that the modified elliptic cylinder offers higher
heat transfer because the thermal boundary layer thins as we modify the shape from
circular cam cylinder to elliptic cam cylinder.
Fig. 7 Streamline plots for modified elliptic cam
Having achieved the objective of low drag and high heat transfer simultaneously
at lower values of N1 and N2, further the effect of change in major axis (M) length is
investigated numerically. The simulations are carried out for a major axis length of
0.8, 0.7 and 0.6 m with an increase in length of tangent line to maintain the equivalent
diameter of 1 m. Also, the case of N1 = 0.203m and N2 = 0.2 m is neglected for
further study as they do not confer to the requirement of bluff body applications.
Figure 9 shows that for all the cases (M = 0.9, 0.8, 0.7, 0.6 m) the Cd decreases
with the decrease in N1 (N1 = 0.842, 0.617, 0.372). Also, Cd decreases with the
decrease in M. So, smaller dimensions of M, N1 and N2 can be considered optimum
for low drag.
Figure 10 shows that for all the cases (M = 0.9, 0.8, 0.7, 0.6 m) the Nu increases
with the decrease in N1 (N1 = 0.842, 0.617, 0.372 m). As M decreases from 0.9
to 0.8 m Nu increases; however, for further decrease in M, Nu decreases. Hence,
moderate value of M can be considered as an optimum major axis for high heat
transfer.
Fig. 8 Isotherms of the existing circular, modified circular and modified elliptic cam
Fig. 9 Cd versus N1 plot for varying major axis (M)
4 Conclusion
1. Modified circular cam (Deq = 1 m) has a lower drag and higher heat transfer
when compared to the existing circular cam (Deq = 1 m).
Fig. 10 Nu versus N1 plot for varying major axis (M)
2. As D2 increases (except for when D1 and D2 are comparable) for a constant

D1 = 0.9 m and Deq = 1 m, drag decreases and heat transfer increases serving
as an optimum domain satisfying the objective.
3. Modified elliptic cam (Deq = 1 m) offers the lowest drag and highest heat transfer
when compared to the existing circular cam (Deq = 1 m) and modified circular
cam (Deq = 1 m) and is an optimum shape for the application of heat exchangers
and industrial exhaust chimneys.
4. As N1 (and N2) decreases for a constant M (=0.9 m, 0.8 m, 0.7 m, 0.6 m) and
Deq = 1 m, drag decreases and heat transfer increases transforming the bluff
body to a streamlined body with complete suppression of vortex shedding.
5. A moderate dimension of major axis with low dimensions of minor axis is
considered an optimum geometric domain for a low drag and high heat transfer.
References
1. Choi H, Jeon WP, Kim J (2008) Control of flow over a bluff body. Annual Rev Fluid Mechanics
40:13–39
2. Chamoli S (2017) Thermal characteristic of a turbulent flow through a circular tube fitted with
perforated vortex generator inserts. Appl Therm Eng 121:11171134
3. Borujerdi AN, Lavasani AM (2008) Pressure loss and heat transfer characterization of a cam-
shaped cylinder at different orientations. ASME 130
4. Chamoli S, Tang T, Yu P (2012) Effect of shape modification on heat transfer and drag for fluid
flow past a cam-shaped cylinder. Heat Mass Transf 12
A Parametric Study on the Taylor
Analogy Breakup Model
Sumit Joshi, Yashodhan Kadam, and Sushrut Ranade
1 Introduction
A liquid droplet, after encountering a flow field at a different velocity, may deform and
break up. This breakup is known as secondary breakup. In a spray, secondary breakup
of a droplet plays an important role in deciding the final droplet size distribution [1].
This breakup phenomenon can be divided into two stages: (1) deformation and (2)
breakup. In the first stage, the droplet deforms into a disk-like shape, and then,
structures such as bag, stamen and liquid sheets appear depending on the breakup
mode, while in the second stage, the breakup starts and happens over a period of
time [1]. The initial deformation into a disk-like structure is governed by relatively
simple physics, and hence, can be modeled analytically [1]. The Taylor Analogy
Breakup (TAB) model is one such analytical model where the initial deformation
stage of the droplet is modeled using an analogy between an oscillating and distorting
droplet and a forced spring-mass-damper system [2]. The TAB model is used widely
to simulate secondary atomization and sprays [2–4], and a number of shortcomings
have been pointed out: (1) breakup time is underpredicted most of the times [3, 4]
as the breakup criterion is arbitrary, (2) temporal evolution of droplet deformation
is not predicted accurately [5], (3) droplet drag model based on the TAB model
underestimates the drag force acting on the droplet at high relative velocities [6], (4)
final droplet sizes in a spray (Sauter mean diameter) are predicted inaccurately [7].
Some of these shortcomings are overcome by changing the constants in the TAB
model. However, if the constants are not changed according to the constraints of the
model and experimental results (which is usually done in the literature), it may lead
to unrealistic results. In the present study, we varied these constants (considering the
TAB model and experimental constraints) and predicted the effect of this variation on
the following TAB model outcomes: droplet deformation (y(t)), breakup time (tbu ),
velocity of the product droplets normal to the path of the parent droplet (V⊥ ) and,
S. Joshi (B) · Y. Kadam · S. Ranade

Department of Mechanical Engineering, IIT Madras, Chennai 600036, India
224 S. Joshi et al.
Sauter mean radius (r32 ). The main objective of this study is to provide researchers
with a ready-made guide for choosing the appropriate TAB model constants since
it is difficult to gauge the effect of variation of these constants (which are coupled)
on the TAB model outcomes from the model equations. We also suggest a change
in the energy conservation equation of the TAB model (which equates the energy of
the parent droplet with that of the product droplets after breakup to calculate r32 ) and
quantify its effect on r32 .
2 Tab Model
The TAB model assumes equivalence between an oscillating and distorting droplet
in an external uniform flow field and a spring-mass system undergoing forced and
damped oscillations [2]. The restoring force of the spring, the external force on the
mass and the dampening are analogous to the surface tension, aerodynamic force,
and the droplet viscosity, respectively [2]. The dynamics of a forced and damped
spring-mass system are given by,
m ẍ = F − kx − dẋ, (1)
where x represents the displacement of the droplet equator from its equilibrium
position (Fig. 1). After non-dimensionalizing x in Eq. (1) as y = Cxb r (where Cb is a
dimensionless constant) and using the Taylor analogy, Eq. (1) becomes,
C F ρg u 2 Ck σ Cd μl
ÿ = − y− ẏ, (2)
Cb ρl r 2 ρl r 3 ρl r 2
Fig. 1 Deformation of the droplet under a uniform flow field

A Parametric Study on the Taylor Analogy Breakup Model 225
where r is the initial droplet radius, u is the gas velocity, ρg and ρl are the densities
of gas and liquid, μl is the dynamic viscosity of the liquid, σ is the surface tension
coefficient, and C F , Ck , Cd and Cb are dimensionless constants whose values are
1/3, 8, 5, and 0.5, respectively [2].
Equation (2) can be solved analytically. The solution is,

CF −t/td CF
y(t) = We + e y0 − W e cos(ωt)
Ck Cb Ck Cb

1 . y0 − CCk CF b W e
+ y0 + sin(ωt) , (3)
ω td
where y(t) represents non-dimensional

droplet 2deformation
in the cross-stream

direc-
ρg u r 2ρl r 2
tion, W e represents Weber number W e = σ , td is relaxation time td = Cd μl ,

ω represents droplet oscillation frequency ω2 = Cρlkrσ3 − t12 , y0 = y(t = 0) and
d
.
y0 = dydt (t
= 0) [2]. The underlying assumption for obtaining the above solution
(Eq. (3)) is that the system is underdamped [2]. Hence, Eq. (2) implies,
2
Cd μl Ck σ
−4 < 0. (4)
ρl r 2 ρl r 3
This is the TAB model constraint that the constants Cd and Ck have to satisfy. Another
constraint on the constants comes from the experimental observation.
A widely accepted value of the critical Weber number (W ecrit ), which marks the
beginning of the bag breakup mode, is 5.5 ± 1 [1]. In the present study, we take
W ecrit = 6 as in the original TAB model [2]. Since μl is very small for the most
common liquids and y0 = ẏ0 = 0 [2], Eq. (3) reduces to
CF
y(t) = W e{1 − cos(ωt)}. (5)
Ck Cb
According to the TAB model, the droplet should break for y(t) > 1 [2]. Hence, for
breakup to happen, the maximum value of y (t) should be greater than 1,
2C F
max{y(t)} = W e > 1. (6)
Ck Cb
At W ecrit , the TAB model predicts breakup if,
Ck Cb
2W ecrit = . (7)
CF
226 S. Joshi et al.
Equation (7) is the experimental constraint which relates Ck , Cb , and C F [2].

The TAB model is also used to calculate the velocity of the product droplets
(formed after breakup of the parent droplet) in a direction perpendicular to the path
of the parent droplet [2]. This velocity can be used to calculate the spray cone angle
if the liquid jet breakup is simulated by injecting droplets of radius equal to injector
radius [2]. This velocity of product droplets is given by,
V⊥ = Cv Cb r ẏ(t), (8)
where Cv is approximately unity [2]. The TAB model also predicts r32 by equating
the energy of the droplet before breakup to combined energies of the product droplets
after breakup. The energy conservation equation for the parent droplet (in its own
frame of reference) is,
4π Kρl r 5 Cb2 2
(T S A)σ + ẏ + ω2 y 2 |t=tbu
5
5 2
r 2πρ l r Cb 2
= 4πr 2 σ + ẏ |t=tbu , (9)
r32 3
where TSA is the total surface area of the parent droplet (T S A = 4πr 2 ), tbu is the
breakup time (time corresponding to y(t) = 1) and K = 10 3
[2]. The terms on
the left-hand side of Eq. (9) represent the total energy of the parent droplet, while
the terms on the right-hand side represent the total energy of the product droplets.
The total energy of the parent droplet consists of surface energy (first term on the
left-hand side of Eq. (9)) and energy due to distortion and oscillation (second term
on the left-hand side of Eq. (9)). Similarly, the first term on the right-hand side of
Eq. (9) represents the total surface energy of the product droplets, and the second
term represents their kinetic energy.
3 Objectives
1. Vary the constants of the TAB model (considering the constraints, Eqs. (4) and
(7)) and quantify the effect of this variation on the model outcomes—breakup
time, droplet deformation, the velocity of the product droplets normal to the
path of the parent droplet, and Sauter mean radius.
2. Use the expression for surface energy of an oblate spheroid in the place of
that for a sphere on the left-hand side of Eq. (9), and quantify the effect of
this change on r32 . The reason for doing this is as follows: According to [3],
the breakup time predicted by the TAB model matches well with the initiation
time. At the initiation time, the droplet attains a shape like an oblate spheroid
[1]. Hence, considering the surface area of an oblate spheroid instead of a
sphere for calculating surface energy of the parent droplet before breakup seems
appropriate. Thus, T S A in Eq. (9) equals,

πr 2 1+e
T S A = 2πr (1 + Cb ) +
2 2
ln , (10)
e(1 + Cb )4 1−e
where e is the eccentricity of the oblate spheroid which is

1
e= 1− . (11)
(1 + Cb )6
4 Methodology
We varied the constants in the TAB model (without violating the constraints, Eqs. (4)
and (7)) in the following ranges: (1) Ck = 8−50, (2) Cd = 5−105, (3) Cb = 0.5−2.2.
For a given value of Cb , C F changes proportionately with Ck through Eq. (7). Hence,
C F is expected to have a similar effect on the TAB model outcomes as Ck . The
range of Cd was obtained for every Ck from Eq. (4). The parameter Cb is associated
with the droplet deformation. Keeping y(tbu ) = 1 and Cb = 0.5 as in the original
TAB model, b in Fig. 1 equals 0.44r , assuming the deformed droplet to be an oblate
spheroid [8]. If Cb is increased, b decreases, while r + x (cross-stream length of the
droplet) increases. b = 0.44r corresponds to r + x = 1.5r in the TAB model [8].
However, in experiments, it has been observed that r + x > 1.5r at the breakup time
[5]. Hence, in this study, Cb was varied in the range 0.5−2.2 (which corresponds
to b = 0.44r −0.1r and r + x = 1.5r −3.2r ). The Weber number was kept at
W e = 6.4 (where bag breakup is observed experimentally) for a water droplet of
diameter 1.76 mm. The properties of the fluids (air and water) at NTP were used.
Time (t) was made non-dimensional by t = u ρρgl [9].
∗ 2r
Figure 2 shows the effect of a change in Ck and Cd on the temporal evolution of

droplet deformation (y(t)) for Cb = 0.5 (variation in Cb does not affect y(t)). The
rate of droplet deformation ( ẏ(t)) increases significantly with Ck , i.e., the frequency
of droplet oscillations (ω) increases with Ck , whereas increasing Cd reduces ω.
However, the effect of variation in Cd on ω
is negligibly small compared to that of
Ck σ
Ck . The first term in the expression for ω ω = ρl r 3 − t 2 is at least three orders
2 2 1
d
of magnitude higher than the second term for all the values of Ck and Cd considered
228 S. Joshi et al.
Fig. 2 Effect of variation in

Ck and Cd on the temporal
evolution of droplet
deformation y(t))
in this study. Hence, Ck affects ω significantly compared to Cd . The amplitude of

oscillations reduces as Cd is increased, whereas a change in Ck does not affect it
significantly. At larger values of Cd and lower values of Ck (e.g., Cd = 105, Ck=8 ),
y(t) does not attain y(t) = 1 (Fig. 2), and hence, the breakup is not predicted
even though it may be observed experimentally. The magnitude of reduction in the
amplitude as Cd increases is larger at smaller values of Ck (Fig. 2).
Figure 3 shows the effect of a change in Ck and Cd on the breakup time for
Cb = 0.5 (variation in Cb does not affect breakup time). As ẏ(t) increases with
increase in Ck (Fig. 2), the breakup time decreases (Fig. 3). At low values of Ck ,
breakup time increases with an increase in Cd . At Cd = 85 and Cd = 105 for low
Ck , breakup is not predicted as y(t) does not attain y(t) = 1. At larger values of Ck ,

Ck and Cd on the breakup
time (tbu )

Ck , Cd , and Cb on the
product droplet velocity
normal to the path of the
parent droplet (V ⊥ )
the effect of Cd on breakup time diminishes (Fig. 3), because, as mentioned before,
Ck is dominant in affecting ω, and hence, the breakup time.
Figure 4 shows the influence of Ck , Cd , and Cb on (V ⊥ ). As Cb increases, (V ⊥ )
increases (Eq. (8)). For a fixed value of Cbr , (V ⊥ ) increases with Ck (Fig. (4)) because
ẏ(t) increases with Ck (Eq. (8)). For a given Ck and Cbr , (V ⊥ ) decreases with an
increase in Cd . This reduction is larger at larger Cb . Also, the magnitude of this
reduction appears to be the same for different Ck values (Fig. 4).
Figure 5 shows the effect of Ck , Cb , and the change proposed in the energy equation
on r 32 (the effect of a change in C D on r 32 is not shown as the effect is insignificant).
r 32 decreases as Ck and Cb are increased. However, at higher values of Cbr , the effect
Ck on r 32 diminishes. Using the expression for surface energy of an oblate spheroid

Ck and Cb , and the proposed
change in the energy
equation on r 32 (‘sphere’ and
‘oblate spheroid’ in the
legend respectively
correspond to r 32 calculated
using surface energy of the
sphere and oblate spheroid in
the energy equation)
230 S. Joshi et al.
instead of a sphere before breakup reduces r 32 for all Cbr and Ck , and the magnitude
of this reduction decreases as Ck increases.
6 Conclusions
The effects of variation in the constants and the proposed change in the energy
equation of the TAB model on its outcomes are discussed.
Ck (and C F ) seems to affect all the model outcomes (droplet deformation y(t)),
breakup time (t bu ), product droplet velocity normal to the path of the parent droplet
(V ⊥ ), and Sauter mean radius (r 32 )) significantly except for the amplitude of y(t). Cd
also affects all the model outcomes except r 32 . Cb affects (V ⊥ ) and r 32 significantly.
However, the effects of variation in Ck , Cd , and Cb on the TAB model outcomes
are not independent of each other. Considering the surface energy of the oblate
spheroid instead of a sphere before breakup in the energy equation reduces r 32 , and
this reduction depends on Ck . The effect of variation in the Weber number and the
critical Weber number on the TAB model outcomes could be investigated in of the
future work.
References
1. Guildenbecher DR, López-Rivera C, Sojka PE (2009) Secondary atomization. Exp Fluids

46:371–402. https://doi.org/10.1007/s00348-008-0593-2
2. O’Rourke PJ, Amsden AA (2017) The TAB method for numerical calculation of spray droplet
breakup. SAE Technical Paper (1987). Accessed July 17 2017. https://papers.sae.org/872089/
3. Hwang SS, Liu Z, Reitz RD (1996) Breakup mechanisms and drag coefficients of high-speed
vaporizing liquid drops. Atomization Sprays 6:353–376
4. Park JH, Yoon Y, Hwang SS (2002) Improved TAB model for prediction of spray droplet
deformation andbreakup. Atomization Sprays 12:387–401
5. Kim S, Hwang JW, Lee CS (2010) Experiments and modeling on droplet motion and atomization
of diesel and bio-diesel fuels in a cross-flowed air stream. Int J Heat Fluid Flow 31:667–679.
https://doi.org/10.1016/j.ijheatfluidflow.2010.02.001
6. Liu Z, Reitz RD (1997) An analysis of the distortion and breakup mechanisms of high speed
liquid drops. Int J Multiph Flow 23:631–650. https://doi.org/10.1016/S0301-9322(96)00086-9
7. Park SW, Sik Lee C (2004) Investigation of atomization and evaporation characteristics of
high-pressure injection diesel spray using Kelvin-Helmholz instability/droplet deformation and
break-up competition model. Proc Inst Mech Eng Part J Automob Eng 218:767–777. https://
doi.org/10.1243/0954407041580067
8. Lee MW, Park JJ, Farid MM, Yoon SS (2012) Comparison and correction of the drop breakup
models for stochastic dilute spray flow. Appl Math Model 36:4512–4520. https://doi.org/10.
1016/j.apm.2012.02.015
9. Nicholls JA, Ranger AA (1969) Aerodynamic shattering of liquid drops. AIAA J 7:285–290.
https://doi.org/10.2514/3.5087
Effect of Slits in Diffuser Casing
of Diffuser Augmented Wind
Turbines(DAWTs)
S. Surya, Jayakrishnan Radhakrishnan, and Amit Kumar
1 Introduction
The development of renewable energy systems has become a serious issue due to the
effects of global warming and climate change. In order to address the current energy
crisis, alternative methods of energy production are being evaluated [1]. Wind energy
is one of the most common alternatives which are looked upon to due to its use of
wind for power generation, which is virtually considered to be an endless resource.
In conventional wind turbines, the airflow speed is reduced and widened. This
effect reduces the efficiency of the wind turbine. Other than this, the tip losses and
losses due to vibrations and noise at the tips further exploit the efficiency of the
turbine. This loss can be countered by the addition of a region of low pressure
behind the wind turbine [2].
In conventional wind turbines, the efficiency of the rotor is limited by the Betz–
Joukowsky limit (59.3%) [3]. This limit has been an interest of study, and various
researches have been done so as to obtain a method to cross this limit. Studies show
that the addition of shroud or a ducted shape around a turbine was found to be
beneficial [4]. The addition of shrouds enhances the wind velocity at the rotor plane,
thus increasing the power output of the turbine, since wind power is proportional to
the cube of the freestream velocity. This property of shrouding the turbines helps in
the setup of DAWTs in areas with low wind speeds also [4].
A study conducted by Takahashi et al. [5] shows a significant reduction in the wind
noise due to the damping of vortices produced by blade tips through the interaction
with the boundary layer produced by the shroud. DAWTs have been explored in the
S. Surya · A. Kumar
Aeronautical Engineering, Department of Aeronautical & Automobile Engineering, Manipal
Institute of Technology, Manipal, India
J. Radhakrishnan (B)
Assistant Professor (Senior Scale), Department of Aeronautical and Automobile Engineering,
Manipal Institute of Technology, Manipal, India
232 S. Surya et al.
published literature by Rainbird et al. [6]. Further research on the same was done by
Gilbert et al. [7–9] by studying the performance of a family of compact diffusers using
small-scale wind tunnels. Porous membranes were used to approximate the turbine
rotors. Based on these results, a simple cost model was devised which showed that
specific power cost ($/kW) for DAWTs was found to be comparatively less than
that of conventional wind turbines. Further research on the shape of the diffuser was
carried out by Igra et al. [10, 11] and focused on the increase of power coefficient(CP)
by using bleed air through the shrouds.
Nevertheless, none of the above experiments crossed or came near to Betz’s limit,
even though the power output was normalized to the diffuser exit diameter rather
than the rotor diameter. But recent experiments conducted by Ohya et al. [12–15]
showed that the Betz limit can be surpassed by the addition of flanges at the outlet of
the diffuser. This is because of the low pressure region formed due to the formation
of vortices in the presence of flanges.
In summary, DAWTs are found to be more efficient in various ways and have the
capability to surpass the Betz limit [12, 13]. To our knowledge, till date, no one has
done an extensive study on the addition of slits to a diffuser to increase the mass flow
rate of air through the rotor plane.
2 Methodology
2.1 Optimum Design of the Diffuser
The converging section of the diffuser consists of the top edge curve of a NACA
2412 airfoil. The throat of the shroud was optimized for maximum velocity, since
the throat will consist of the wind turbine. Based on these two parameters, the outlet
area was chosen so as to avoid flow separation at the exit of the diffuser. In both
cases, the
length of the diffuser was 66.2 cm with an L/D ratio of the diffuser was 0.331
where D is the minimum diameter at the throat.
The diffuser is modelled with the varying parameter as the number of slits. Other
parameters such as the length, inlet and outlet areas are kept constant (Fig. 1).
2.2 Computational Fluid Dynamics
ANSYS Fluent was used to run the simulations to obtain flow velocity and other
parameters. The RANS equations are in conjunction with the k- SST turbulence
model. The governing equations are as follows
Effect of Slits in Diffuser Casing of Diffuser … 233
Fig. 1 Diffuser with slits and bleed air slots and diffuser without slits

∂k ∂k ∂ vt ∂k
+ Uj = Pk − β ∗ kω + v+ (1)
∂t ∂x j ∂x j σj ∂x j

∂ω ∂ω ω ∂ vt ∂ω
+ Uj = α Pk − βω +
2
v+
∂t ∂x j k ∂x j σj ∂x j
(1 = F1 ) 1 ∂k ∂ω
+2 (2)
σω2 ω ∂ x j ∂ x j
∂ ū
=0 (3)
∂x j
A rectangular domain was chosen with the rotor plane as

a porous membrane and the diffuser as a no-slip stationary wall. The freestream
velocity at the inlet was given as 6 ms-1 as shown in Fig. 2. A structured mesh was
used with high refinement inside the diffuser and at the rotor plane. Grid independence
study was conducted for the verification of the results. The mesh around the diffuser
section is shown in Fig. 3.
Grid Independent Study
The number of elements involved in the grid study ranged from 3 × 105 to 6 × 106
and the results were found to remain almost constant with a minimal variation. By
increase in the number of elements, the results remained constant. Beyond 6 × 106
elements no further refinement in the mesh was required. The details of the grid study
are given in Table 1.
234 S. Surya et al.
Fig. 2 CFD domain
Fig. 3 Refined mesh around

the diffuser
Table 1 Grid independent

Grid type With slits Without slits
study
Coarse 3.7 × 105 3.5 × 105
Medium 2.3 × 106 2.9 × 106
Fine 5.45 × 106 5.9× 106
From simulations, it was found that the velocity at the throat was maximum in case
of diffusers with slits and without slits. Also with the addition of slits, there was an
increase in the velocity at the rotor plane, thus resulting in an increased mass flow
rate through the rotor plane. Based on the velocities obtained, there is approximately
an increase of mass flow by 22% with the presence of the slits at the rotor plane.
For the given diffuser, a maximum of 22% increase in the mass flow rate was seen
with 20 slits at the converging section. Theoretically, this number must increase with
increase in the number of slits. But beyond 20 slits, there was a decrease in the mass
flow rate through the diffuser, because most of the flow started moving out of the
diffuser even before reaching the rotor plane. Thus, it can be shown that the number
of slits also can affect the mass flow rate of air through the rotor plane. The velocity
contours and streamlines through the diffuser are shown in Figs. 4 and 5. In the same
manner, the simulation was run for various freestream velocities and it was found
that, the velocity at the rotor plane linearly increased with increase in the freestream
velocity, in case of diffusers with slits and without slits. The plot between velocity
at the rotor plane and freestream velocity is shown in Fig. 6. The values of velocities
at the rotor plane for different freestream velocities are given in Table 2.
Fig. 4 Velocity contours for diffuser without slits
Fig. 5 Velocity contours for diffuser with slits

236 S. Surya et al.
Fig. 6 Variation of velocity at the rotor plane with freestream velocity, for both diffusers with slits
and without slits
Table 2 Velocity at the rotor

Freestream Velocity Velocity at the rotor Velocity at the rotor
plane for various freestream
plane without slits plane with slits
velocities (in ms−1 )
2 2.99 3.87
4 5.87 7.59
6 8.91 11.56
8 11.27 15.18
10 14.79 19.27
In the case of diffuser with slits, at the exit of the diffuser, there is some flow
separation occurring. This was rectified using some bleed air holes, which reduces
the flow separation by re-energizing the flow at the point of separation. But this
addition of slits at the divergent/rear section of the diffuser caused the reduction of
velocity at the rotor plane by 0.1 ms−1 which may not affect the out-stream and
upstream properties of the flow. The bleed air slits were added at the divergent/rear
end of the diffuser based on the following assumption: Flow separation at the rear
end of the diffuser causes a minimal change in the velocity at the rotor plane. But
this causes an increase in aerodynamic drag on the diffuser. This increase in drag can
reduce the efficiency of the turbine in long-term running (Fig. 5 ).
4 Conclusions and Future Scope
The effect of slits on a diffuser for DAWT has been shown in this article. We can
conclude the following things from these results
1. The mass flow rate through the rotor plane increases with the addition of slits
to the diffuser
2. Theoretically, with increase in number of slits, there must be an increase in mass
flow rate of air through the rotor plane, but from the results, it was found that
this was true up to a specific no of slits and beyond that specific number the
results are deteriorating. Therefore, for a specific diffuser there is an optimum
number of slits which gives ideal mass flow through the rotor.
3. Addition of bleed air slits is beneficial in the form of reduction of aerodynamic
drag and reduction of the shell weight, thus increasing the efficiency of the entire
system.
Further, these results will be used along with a wind turbine for obtaining the power
and thrust coefficients. These experiments will be compared with past experiments
and also be validated via wind tunnel experiments
Acknowledgements The authors would like to thank Department of Aeronautical and Automobile
Engineering of Manipal Institute of Technology for their continuous support
References
1. Ackermann T, Soder L (2002) An overview of wind energy-status. Renew Sustain Energy Rev
6(1–2):67–127
2. Ueno H, Serada T (2002) Wind turbine generation system and photovoltaic power generation
system. NTT Power Build Facil J 39(231):42–45
3. https://en.wikipedia.org/wiki/Betz%27s_law
4. Khamlaj TA, Rumpfkeil MP (2018) Analysis and optimization of ducted wind turbines. Energy
Elsevier 162(C):1234–1252
5. Takahashi S, Hata Y, Ohya Y, Karasudani T, Uchida T (2012) Behavior of the blade tip vortices
of a wind turbine equipped with a brimmed diffuser shroud. Energies 5(12):5229–5242
6. Lilley G, Rainbird W (1956) A preliminary report on the design and performance of ducted
windmills. Tech. rep., College of Aeronautics Cranfield
7. Foreman K, Gilbert B, Oman R (1978) Diffuser augmentation of wind turbines. Sol Energy
20(4):305–311
8. Oman R, Foreman K, Gilbert B (1975) A progress report on the diffuser aurgmented wind
turbine In: Proceedings 3rd Biennal conference and workshop on wind energy conversion
systems, Washington, DC, USA, pp 819–826
9. Foreman K, Gilbert B, Oman R (1976) Di user augmentation of wind turbines. Tech. rep.,
Grumman Aerospace Corp., Bethpage, NY (USA). Research Dept
10. Igra O (1977) Compact shrouds for wind turbines. Energy Convers 16(4):149–157
11. Igra O (1981) Research and development for shrouded wind turbines. Energy Convers Manage
21(1):13–48
238 S. Surya et al.
12. Abe K, Ohya Y (2004) An investigation of flow fields around flanged diffusers using cfd. J
Wind Eng Indus Aerodyn 92(3):315–330
13. Abe K, Nishida M, Sakurai A, Ohya Y, Kihara H, Wada E, Sato K (2005) Experimental and
numerical investigations of flow fields behind a small wind turbine with a flanged diffuser. J
Wind Eng Indus Aerodyn 93(12):951–970
14. Ohya Y, Karasudani T, Sakurai A, Abe K-I, Inoue M (2008) Development of a shrouded wind
turbine with a flanged diffuser. J Wind Eng Indus Aerodyn 96(5):524–539
15. Ohya Y, Karasudani T (2010) A shrouded wind turbine generating high output power with
wind-lens technology. Energies 3(4):634–649
Transition of Stratified to Intermittent
Flow Pattern
Sunny Saini, Nihar Shah, and Jyotirmay Banerjee
1 Introduction
Gas-liquid flow exhibits several flow patterns which are identified as separated flow,
intermittent flow and dispersed flow. A small instability at the interface due to change
in flow rate and properties of the fluids leads to intermittent flow pattern. Intermittent
flow patterns include plug and slug flow. Slug flow has intermittency [1–3] in flow
which causes sudden pressure surge and mechanical impact. Moreover, the gas and
liquid slugs are alternatively in contact with the wall of the pipe, therefore, resulting
in significant variations in wall temperature, pressure gradient and mechanical impact
on the pipeline. The momentum transfer occurring during the appearance of liquid
and elongated bubbles produces a mechanical force on the pipeline which introduces
structural issues that lead to cracks, accelerate the erosion-corrosion [4–6] and fatigue
at the bends, T- and I-sections. Hence, it is required to predict the formation of
intermittent flow pattern in gas-liquid transportation pipeline to reduce the economic
loss and hazards in the industries.
Slug is produced when the roll waves on the liquid surface undergo constructive
interference and the magnitude of the resulting wave increases. Thus, the proximity
of the wave crest to the upper surface of the pipe decreases, decreasing the area of the
gas flow. Venturi effect created produces pressure drop that pulls the liquid surface
of the wave upwards. The wave also experiences a downward force due to gravity
and surface tension. When the wave amplitude is high enough to create suction pull
such that it beats the downward forces, the liquid instantly sticks to the pipe wall
forming a slug. This is observed in the present research through flow visualization
using Photron FASTCAM high speed camera.
S. Saini (B) · N. Shah · J. Banerjee

Department of Mechanical Engineering, S. V. National Institute of Technology, Surat, India
240 S. Saini et al.
2 Methodology
The experiments are carried out on a two-phase flow test rig described earlier by
Thaker and Banerjee [6]. Intermittent flow patterns are visualized 25 ± 0.15 mm
diameter and 14 m long transparent pipe. Photron FASTCAM high speed camera is
used to capture the images at the different ReSL and ReSG .
Separated flow is desirable in most of the industrial applications because it has less
influence on the pipe wall. On the other hand, intermittent type flow and dispersed
type of flow patterns induce significant pressure pulsations which may lead to pipe
failure. Pressure pulsation associated with intermittent flow (slug/plug) produces
cyclic loading with sudden impact at bend and T-sections. In this article, we visualize
the flow to analyze the dynamics of flow patterns and their transitions by varying
the ReSL and ReSG . Intermittent flow pattern is developed due to instability in the
stratified and wavy-stratified flow pattern which grows and bridges the gas flow path.
3.1 Development of Intermittent Flow Pattern
Intermittent flow includes plug and slug flow. The plug flow is almost free of gas
bubbles inside the liquid body, whereas the aeration of gas bubble is observed in the
slug flow. As reported in Fig. 1 at low ReSG and ReSL 400 and 1800, respectively,
stratified flow pattern is observed. As the ReSL is increased keeping fixed low ReSG
small disturbance is observed on the interface which however does not grow to restrict
the flow path of the gas or touch the upper pipe wall surface. At low ReSG of 400 and
with gradual increase the ReSL , the plug is observed at the upstream of the pipe which
is, however, not capable of flowing entire length of pipe and diminishes along the
length of pipe as shown in Fig. 2a. Zukoski [7] suggested that the rate of propagation
of the plug flow is influenced by surface tension and viscosity of the fluid at low
ReSL and ReSG . The disturbance at the inlet section or at high liquid holdup at the
upstream side produces the plug flow. However, the inertia is lower than the surface
Fig. 1 Stratified flow at ReSG = 400 and ReSL = 1800

Transition of Stratified to Intermittent Flow Pattern 241
Fig. 2 a Influence of inertia over viscous and surface tension and b oscillation of gas and liquid
plug
242 S. Saini et al.
Fig. 3 Influence of the

inertia on the surface tension
and viscosity
tension and viscous forces (as Fig. 3) and the plug is not able to flow in the pipe for
longer distance. With further increase in the ReSL inertia increases but viscous and
surface tension forces are not negligible. As shown in Fig. 2b, oscillation of the plug
flow is capable to flow till end. Due to gas bubble oscillations energy is transferred
to the liquid plugs. At high ReSL oscillations of the gas bubble produces energy due
to which the plug length increases.
The necessary condition for development of intermittent flow is that the liquid
hold up must be above the value of 0.5 [8]. Barnea and Taitel [8] suggested the
phenomena of development of plug/slug flow pattern in viscous and inviscid flow
using linear stability analysis. This phenomenon is known as Kelvin-Helmholtz (K-
H) stability. Barnea and Taitel [8] described that from the Bernoulli’s equation, as the
wave increases in amplitude the flow area of the gas reduces as in Fig. 4. The reduction
in gas flow path increases the velocity of gas but the pressure at upper surface of the
wave decreases. According to Jeffery [9] through wave phenomena, pressure at the
wave slope is higher due to contact of air flow to wave slope. The air flow motion
is separated at the crest surface where pressure is reduced and develops eddies at
the leading trough region where again the pressure is increased. This phenomenon
creates suction at the upper surface of the wave. Hence, the liquid elevates in the
upward direction and block the flow path of the gas. As shown in Fig. 2, it forms
liquid pockets in between the elongated gas bubble which is the plug flow pattern.
A similar phenomenon is observed for the initiation of the slug flow at high gas
flow rate where the roll wave coalescence is observed and leads to K-H instability.
As the ReSG increases disturbances are developed at the liquid surface in the form
of waves. Spending and Spence [10] and Andritsos and Hanratty [11] suggested that
these waves are large amplitude waves also known as inertia or K-H waves. The wave
velocity is high at higher ReSG . As shown in Fig. 5, according to Jeffery theory [9],
the leading waves are at higher velocity which merges to the trailing waves (lesser
velocity compared to leading waves) because high amplitude wave is propagated in
the direction of air flow where momentum is lost to waves and shedding side wave
having low momentum force. The merging of the roll waves increases the critical
liquid height above the value of 0.5. Hence, due to the phenomena of roll wave
Fig. 4 Initiation of the plug

flow at ReSG = 750 and
ReSL = 2200
Fig. 5 Initiation of the slug

flow at ReSG = 1875 and
ReSL = 1800
244 S. Saini et al.
Fig. 6 Initiation of
intermittent flow from
stratified flow
coalescence and K-H stability, liquid pockets are observed with the aeration at slug
front as shown in Fig. 5. This flow pattern is identified as the slug flow pattern.
A careful observation depicts that as the plug/slug passes a section of the pipe
the liquid film level at that section drop which takes time to rise for the initiation
of the next plug/slug. At low ReSL , the rate of the plug/slug formation is not higher.
As shown in Fig. 6, initially, surface tension and viscous effect are dominant on
the inertia due to which at high U SL plug flow is initiated. With increase in U SG
inertia dominant over the surface tension and viscous effect, hence, intermittent flow
initiated at low U SL . With further increase in U SG , liquid height dampens and roll
waves merges. Pseudo-slug develops due to less liquid height. Hence, beyond a U SG
value, the intermittent flow pattern is initiated at higher value of U SL . The obtained
results based on the flow visualization with change in U SL and U SG are used to
develop the transition boundary between the stratified-wavy and intermittent flow
pattern. The obtained transition boundary is in good agreement with Andritsos and
Hanratty [11] and Lin and Hanratty [12].
Similar, results were reported by Bendiksen [13] for different inlet flow section.
He suggested that above a gas superficial velocity 4.5 m/s the inlet section does not
influence the onset of slug formation. At the low ReSL , the stability of the plug/slug
is influenced by the accumulation and shedding of the water at the front side and tail
of the plug/slug, respectively.
The surface tension and viscous forces also influenced the rate of the initiation of
the plug/slug flow as shown in Fig. 3. Ruder et al. [14] described the stability of the
plug/slug as:
• If accumulation of the water from the carpet of the liquid film is higher than
the shedding of the water at the plug/slug tail, then the length of the plug/slug
increases as it flows downstream shown from our visualization images as shown
in Figs. 2 and 5.
• If accumulation of the water from the carpet of the liquid film is lesser than
the shedding of the water at the plug/slug tail, then the length of the plug/slug
decreases as it flows downstream shown from our visualization images as shown

in Fig. 3.
• If accumulation of the water from the carpet of the liquid film is equal to the
shedding of the water at the plug/slug tail, then the length of the slug is constant
as the flow moves downstream side. This flow is called stable plug/slug flow.
4 Conclusions
The following conclusions are derived from the flow visualization analysis:
• At low ReSL and ReSG , viscous and surface tension are dominant on the inertia of
the plug which destabilize the plug length along the pipe.
• With increase in ReSG inertia of plug is increased and the plugs are capable to
flow along the pipe length.
• At high ReSG mean liquid height of the flow decreases and high amplitude roll
waves are observed at the interface.
• Roll waves have high steepness front which forms curvature at the wave front (as
Fig. 5), when it comes in contact with high accelerated air and suck the air inside
the liquid region.
• At high ReSG , high amplitude wave merges to small amplitude wave and leads to
slug flow formation.
• After the formation of the slug flow, aeration inside the liquid front is developed
through the small surface waves which are flow in front of the slug and captured
by it.
• Development of intermittent flow initiated at low USL with increase in USG .
Beyond a USG value roll waves developed pseudo-slug due to which initiation
of intermittent flow shifted to high USL value.
References
1. Crowe CT (2006) In: Multiphase flow handbook. Taylor and Francis group
2. Mandhane JM, Gregory GA, Aziz KA (1954) A flow pattern map for gas-liquid in horizontal
pipeline. Int J Multiphase Flow 1:537–554
3. Vaze MJ, Banerjee J (2011) Experimental visualization of two-phase flow patterns and
transition from stratified to slug flow. Proc ImechE, Part C, J Mech Eng Sci 225:382–389
4. Sun JY, Jepson W (1992) Slug flow characteristics and their effect on corrosion rates in hori-
zontal oil and gas pipelines. In: SPE Annual technical conference and exhibition, society of
petroleum engineers
5. Ahmed WH, Bello MM, El Nakla M, AlSarkhi A, Badr HM (2014) Experimental investigation
of flow accelerated corrosion under two-phase flow conditions. Nuclear Eng Design 267:34–43
6. Thaker J, Banerjee J (2016) Influence of intermittent flow sub-patterns on erosion-corrosion in
horizontal pipe. J Petrol Sci Eng 145:298–320
7. Zukoski E (1966) Influence of viscosity, surface tension, and inclination angle on motion of
long bubbles in closed tubes. J Fluid Mech 25(4):821–837
246 S. Saini et al.
8. Barnea D, Taitel Y (1993) Kelvin-Helmholtz stability criteria for stratified flow: viscous versus
non-viscous (inviscid) approaches. Int J Multiphase Flow 19(4):639–649
9. Jeffreys H (1925) On the formation of water waves by wind. In: Proceedings of the royal society
of London. Series A, containing papers of a mathematical and physical character. vol 107(742)
pp 189–206
10. Spedding P, Spence D (1993) Flow regimes in two- phase gas-liquid flow. Int J Multiphase
Flow 19(2):245–280
11. Andritsos N, Hanratty T (1987) Interfacial instabilities for horizontal gas-liquid flows in
pipelines. Int J Multiphase Flow 13(5):583–603
12. Lin PY, Hanratty TJ (1986) Prediction of the initiation of slugs with linear stability theory. Int
J Multiphase Flow 12:79–98
13. Bendiksen K, Malnes D (1987) Experimental data on inlet and outlet effects on the transition
from stratified to slug flow in horizontal tubes. Int J Multiphase Flow 13(1):131–135
14. Ruder Z, Hanratty P, Hanratty T (1989) Necessary conditions for the existence of stable slugs.
Int J Multiphase Flow 15(2):209–226
Effects of Partial Slip and Stretching
on the Oscillations of the Velocity Profiles
of the Bödewadt Flow
Dip Mukherjee and Bikash Sahoo
1 Introduction
The study of steady laminar flow of a Newtonian viscous, incompressible fluid near
a rotating disk was first presented by von Kármán [1]. The twin problem, where
the fluid rotates like rigid body over a stationary disk, was considered by Bödewadt
[2]. Hence, the flow is known as the Bödewadt flow. It is observed that the flow
is driven by pressure gradient being balanced by centrifugal force, and as a result,
the flow is drawn toward the axis of rotation and is swept in the upward direction.
Bödewadt [2] found an exact solution of the Navier–Stokes equations with no-slip
boundary conditions. The seminal books by Batchelor [3], Schlichting [4] and Owen
and Rogers [5] give vast idea on rotating disk boundary-layer flows.
Due to its wide industrial and practical applications, the Bödewadt flow problem
has been revived with diverse physical effects. Recently, Sahoo et al. [6], Turkyil-
mazoglu [7] have considered the classical Bödewadt flow problem subject to partial
slip boundary conditions and disk surface stretching, respectively. The flow between
two stretchable rotating disks has been considered by Fang [8].
Unlike Kármán flow, the velocity profiles of the Bödewadt flow problem exhibit
oscillations. In this conference paper, we will see the combined effects of surface
roughness and surface stretching on the oscillations of the velocity profiles of the
Bödewadt flow.
D. Mukherjee (B) · B. Sahoo

National Institute of Technology Rourkela, Rourkela, Odisha 769008, India
B. Sahoo
e-mail: bikashsahoo@nitrkl.ac.in
248 D. Mukherjee and B. Sahoo
2 Formulation of the Governing Equations
We have considered the cylindrical coordinates system (r, θ, z). The infinite
stationary disk coincides with the plane z = 0. The viscous, incompressible fluid
occupies the space z > 0. Due to rotational symmetry, it is assumed that ∂θ∂ ≡ 0. The
velocity field is given by u= (u, v, w). The continuity equation and the Navier–Stokes
equations are given by:
∇ ·u =0 (1)
ρ(u · ∇)u = −∇ p + μ∇ 2 u (2)
Here ρ is the density, p is the pressure, and μ is fluid viscosity.

By considering the similarity transformations [1]:

ω
ζ = , u = r ωU (ζ ),
ν
√
v = r ωV (ζ ), w = ωνW (ζ ),
p − p∞ = −ρνω P (3)
Eqs. (1) and (2) take the form [2, 8]:
dW
+ 2U = 0 (4)
dζ
d2 U dU
2
−W − U2 + V 2 = 1 (5)
dζ dζ
d2 V dV
−W − 2U V = 0 (6)
dζ 2 dζ
dP dW d2 W
−W + =0 (7)
dζ dζ dζ 2
The boundary conditions at z → ∞ are u = 0, v = ωr.

By using the transformations, the boundary conditions become:
U (∞) → 0, V (∞) → 1 (8)
Case-I:Partial Slip Boundary Conditions:

The partial slip boundary conditions in the radial and azimuthal directions are
given by:
Effects of Partial Slip and Stretching on the Oscillations … 249
u(0) = λ1 τr z |z=0 , v(0) = λ2 τθ z |z=0 , w(0) = 0 (9)
where λ1 , λ2 are slip coefficients and τr z , τθ z are stress tensors. The dimensionless
slip coefficients are

ω ω
λ = λ1 μ, η = λ2 μ. (10)
ν ν
Using the transformations (3), the slip conditions become:
U (0) = λU (0), V (0) = ηV (0), W (0) = 0 (11)
Case-II: Stretch Boundary Conditions:

The stretch boundary conditions are given by:
u(0) = sr, v(0) = 0, w(0) = 0 (12)
where s is the radial stretch. Using the transformations (3), the boundary conditions
become:
U (0) = c, V (0) = 0, W (0) = 0 (13)
where c = s
ω
is the stretching strength parameter.
We have solved the resulting system of coupled and highly nonlinear similarity
Eqs. (4–6) subject to boundary conditions (8), (11), and (13) by the bvp4c func-
tion of MATLAB software. The effect of radially anisotropic surface roughness on
the oscillations of the components of the velocity profiles is shown in Figs. 1, 2, 3.
If we take a look insight of these figures, it can be observed that as the value of η is
increased, the magnitude of the oscillations decreases; i.e., the slip parameter has a
damping effect on the oscillations of the mean flow profiles. The effect of azimuthally
anisotropic surface roughness on the oscillations of the velocity profiles is shown
in Figs. 4, 5, 6. Interestingly it can be seen that λ and η have opposite effects on
oscillations. The isotropic surface roughness has the same effect as for the case of
radially anisotropic roughness (Figs. 7, 8, 9). For the case, when the disk is radially
stretched a damping effect on the oscillations of the velocity profiles is observed,
which are shown in Figs. 10, 11, 12.
To validate the results obtained in this paper the, values of U for different values
of ζ are compared to those values given by Sahoo et al. which is shown in Table 1.
Fig. 1 Oscillations of U (ζ )
for radially anisotropic
surface roughness
Fig. 2 Oscillations of V (ζ ) 1.4

for radially anisotropic =0, =0
1.2 =1, =0
surface roughness =2, =0
1
0.8
V( )
0.6
0.4
0.2
0
0 2.5 5 7.5 10 12.5 15
Fig. 3 Oscillations of W (ζ ) 2
for radially anisotropic =0, =0
=1, =0
surface roughness
=2, =0
1.5
W( )
0.5
0
0 2.5 5 7.5 10 12.5 15
Fig. 4 Oscillations of U (ζ ) 1
for azimuthally anisotropic =0, =0
=0, =0.25
surface roughness =0, =0.75
0.5
U( )
-0.5
-1
-1.5
0 2 4 6 8 10 12
Fig. 5 Oscillations of V (ζ ) 3
=0, =0.25
surface roughness 2.5 =0, =0.75
2
V( )
1.5
0.5
0
0 2 4 6 8 10 12
=0, =0.25
surface roughness =0, =0.75
4
3
W( )
0
0 2 4 6 8 10 12
Fig. 7 Oscillations of U (ζ ) 0.2

for isotropic roughness =0, =0
0.1 =1, =1
=2, =2
0
-0.1
U( )
-0.2
-0.3
-0.4
-0.5
0 2 4 6 8 10 12
Fig. 8 Oscillations of V (ζ ) 1.4

1.2 =1, =1
=2, =2
1
0.8
V( )
0.6
0.4
0.2
0
0 2 4 6 8 10 12
=1, =1
=2, =2
1.5
W( )
0.5
0
0 2 4 6 8 10 12
Fig. 10 Oscillations of 0.2

c=0
U (ζ ) for radial stretching
0.1 c=0.01
c=0.1
-0.1
U( )
-0.2
-0.3
-0.4
-0.5
0 2 4 6 8 10 12
Fig. 11 Oscillations of 1.4

c=0
V (ζ ) for radial stretching
1.2 c=0.01
c=0.1
0.8
V( )
0.6
0.4
0.2
0
0 2 4 6 8 10 12
Fig. 12 Oscillations of 2
W (ζ ) for radial stretching c=0
c=0.01
c=0.1
1.5
W( )
0.5
0
0 3 6 9 12
Table 1 Comparison of
ζ U (ζ )
numerical values of U for
different values of ζ Obtained value Sahoo et al.
0 0 0
0.5 −0.3484 − -0.3486
1 −0.4786 −0.4788
1.5 −0.4498 −0.4496
2 −0.3287 −0.3287
2.5 −0.1763 −0.1762
3 −0.0360 −0.0361
4 Conclusions
In this paper, the combined effects of surface roughness and surface stretching on
the oscillations of the velocity profiles of classical Bödewadt flow have been studied
numerically. It is observed that the slip parameters enhance (radial slip) oscillations
as well as damps (azimuthal slip) oscillations; on the other hand, stretching has only
damping effects on the oscillations. Further qualitative oscillations of the Bödewadt
flow will be carried out in the near future.
Acknowledgements The first author would like to thank MHRD and the institute for financial
support.
References
1. von Kármán T (1921) Uber laminare und turbulente Reibung. Z. Angew Math Mech 1:233–252
2. Bödewadt UT (1940) Die Drehströmung über festem Grunde. Z. angew. Math Mech Bd 20S:241–
253
3. Batchelor GK (2000) In: An introduction to fluid dynamics. Cambridge University Press
4. Schlichting H (1960) In: Boundary layer theory. New York, McGraw-Hill
5. Owen JM, Rogers RH (1992) Flow and heat transfer in rotating-disc systems
6. Sahoo B, Abbasbandy S, Poncet S (2014) A brief note on the computation of the Bödewadt flow
with Navier slip boundary conditions. Comput Fluids 90:133–137
7. Turkyilmazoglu M (2015) Bödewadt flow and heat transfer over a stretching stationary disk. Int
J Mech Sci 90:246–250
8. Fang T, Zhang J (2008) Flow between two stretchable disks—An exact solution of the Navier–
Stokes equations. Int Commun Heat Mass Transf 35(8):892–895
Flow and Heat Transfer Characteristics
of Oscillatory Thermo-Capillary
Convection in Half Floating Zones
R. Jayakrishnan and Shaligram Tiwari
1 Introduction
The relevance of the semiconductor crystal growth using the floating zone technique
is motivating the researchers for the past few decades. Steered by the recent develop-
ments in the area of micro and nano technologies, thermo-capillary convection (TC)
in liquid bridges (LB) has achieved increased attention. The generation of surface
tension gradients due to the temperature gradients at the liquid-liquid interface results
in a surface flow called as TC from high-temperature region to the low-temperature
region [1]. It is well established that beyond a particular Marangoni number (Macr )
corresponding to the applied temperature difference (Tcr ), the convection under-
goes a transition to a time-dependent and 3D state which is commonly referred to as
oscillatory TC. This onset of super criticality in the liquid zones is distinguished by
the spatiotemporal convective structures such as standing or traveling waves. Exper-
iments conducted under microgravity conditions demonstrated that this spatiotem-
poral elements of the convective flow can be analyzed in detail in the framework of
hydrothermal wave (HTW) instability [2]. The toroidal convection roll inside the LB
is reported to be disturbed and displaced circumferentially along the free surface like
a sinusoidal wave.
R. Jayakrishnan (B) · S. Tiwari

S. Tiwari
e-mail: shaligt@iitm.ac.in
256 R. Jayakrishnan and S. Tiwari
2 Problem Description and Numerical Methodology
A LB consisting of a high Pr liquid (Pr = 39.3) of 5 cSt silicone oil suspended

in between two cylindrical rods which are heated differentially is considered for the
present investigation as illustrated in Fig. 1.
The dimensions of the present LB geometry and the properties of the working
liquid are taken from the published experimental results of [3]. The modeling and
the analysis of the computed oscillatory flow regimes of the present LB problem are
performed with the help of a commercially available CFD package ANSYS Fluent
18.1 which is based on finite volume method (FVM). The pressure-velocity coupling
is handled using semi-implicit method for pressure-linked equations (SIMPLE)
scheme. The dynamics of the LB problem is characterized by thermal convection
equations given by continuity, Navier–Stokes, and energy equations as mentioned
below.
1 ∂ 1 ∂ ∂
(r u r ) + u ϕ + (u z ) = 0 (1)
r ∂r r ∂ϕ ∂z

∂u r ∂u r 1 ∂u r ∂u r 1 ∂P ur 2 ∂u ϕ
+ ur + uϕ + uz =− + υ ∇ 2 ur − 2 − 2 (2)
∂t ∂r r ∂ϕ ∂z ρ ∂r r r ∂ϕ

∂u ϕ ∂u ϕ 1 ∂u ϕ ∂u ϕ 1 ∂P uϕ 2 ∂u r
+ ur + uϕ + uz =− + υ ∇ 2uϕ − 2 + 2
∂t ∂r r ∂ϕ ∂z ρr ∂ϕ r r ∂ϕ
(3)
∂u z ∂u z 1 ∂u z ∂u z 1 ∂P
+ ur + uϕ + uz =− + υ∇ 2 u z (4)
∂t ∂r r ∂ϕ ∂z ρ ∂z
∂T ∂T 1 ∂T ∂T
+ ur + uϕ + uz = α∇ 2 T (5)
∂t ∂r r ∂ϕ ∂z
The hot and cold ends of the rods in contact with the working liquid have no-
slip, impermeable, and isothermal boundary conditions. Meanwhile, tangential shear
stress balance [4] and the heat flux balance at the free surface of the LB are represented
as shown below.
Fig. 1 Configuration of half

floating zone (HFZ)
Flow and Heat Transfer Characteristics of Oscillatory … 257
Fig. 2 3D grid mesh near

the free surface of LB
2
dR ∂u r ∂u z dR ∂u z ∂u r
2μ − +μ 1− +
dz ∂r ∂z dz ∂r ∂z
2 1/2
dR dR ∂ T ∂T
= σT 1 + + (6)
dz dz ∂r ∂z

∂T dR ∂ T
−k − = h(T − Tamb ) (7)
∂r dz ∂z
A non-uniform grid mesh which is well refined near the free surface and the
solid walls is adopted to resolve the velocity and thermal boundary layers. Grid
and time independence study is performed by computing the value of Tcr and the
corresponding amplitude of liquid temperature fluctuations. Eventually, a grid of size
60 × 140 × 60) and a time step of dt = 0.0025 are selected as the optimum one.
Figure 2 shows the grid mesh employed for the present numerical investigation.
The critical values deciding the onset of instability Tcr and Macr have been numer-
ically validated with the experimental results of [3] and were found to be 50 K and
27136, respectively, which is corresponding to a specific ambient temperature of
300 K. The emerging spatiotemporal flow structures observed in the present case are
due to the occurrence of a pulsating or standing waveform.
The pulsation of the disturbances is observed to form at particular azimuthal
locations of the LB free surface. The critical azimuthal wave number of the present
spatiotemporal mode of convection is found to be m = 2 such that 2 m number of
thermal spots can be noticed along the free surface and along the generic sections
Fig. 3 Disturbance temperature field in the cross section at z = 0.5 at different time intervals a t =
τ b t = τ /4 c t = τ /2 d t = 3τ /4
perpendicular to z-axis [5] as shown in Fig. 3. The instantaneous disturbance temper-

ature field shows that the disturbances are pulsating at fixed azimuthal positions
instead of rotating. The hot and cold thermal spots are observed to grow or shrink in
the axial direction of the LB. The intensity of thermal spots is also found to change
with respect to time.
The azimuthal convective cells in the generic section orthogonal to the z-axis show
a periodic variation of their direction of rotation which is a characteristic of the SW,
i.e., a particular convective cell which is oriented in the clockwise direction in the first
half period of oscillation will change to the anti-clockwise direction in the second
half period as presented in Fig. 4. The variation in the oscillatory flow structure at
various instants of time is more apparent from the above-shown distribution of the
streamlines.
The hydrothermal mechanism for the onset of instability can get initiated due to
the random occurrence of a hot or cold temperature disturbance on the free surface
of the LB. Generation of a hot disturbance results in a thermo-capillary flow in the
azimuthal direction, thereby creating two counter-propagating hydrothermal waves
away from the disturbance. A radial flow from the cold bulk to the free surface occurs
to satisfy the continuity which consequently causes the initial hot disturbance to get
cooled. Thus, the disturbances convect along the free surface and replace each other
and finally resulting in a pulsating structure as shown in the instantaneous velocity
and temperature fields. Moreover, the observed transition to the supercritical state
is due to the HTW instability due to the intercoupling between the velocity and
Fig. 4 Disturbance flow field in the cross section at a location z = 0.5 at different time intervals
a t = τ b t = τ /4 c t = τ /2 d t = 3τ /4
temperature fields. It is reported to be Hopf bifurcation which is responsible for the

consequent dynamic evolution of the system. The peculiarity of this pulsating mode
m = 2 is the divisibility of the oscillatory convective flow structure into four regions
about a plane of symmetry. The free surface flows from hot rod to the cold rod created
due to the surface temperature gradients, and the corresponding return flow of the
bulk fluid inside the liquid zone can be clearly understood from Fig. 5a.
As the present class of critical mode is observed to be the symmetric mode (m =
2), the flow and temperature fields in the meridional section of the LB show a mirror
symmetry with respect to the z-axis as shown. Figure 5b illustrates the corresponding
disturbance temperature and velocity fields in the same section which is also found
to be symmetric about the axis of the LB. The presence of two hot thermal spots on
the free surface results in two HTW to generate in the azimuthal direction, each gets
split into two counter-propagating partial HTWs. The distribution of the disturbance
Fig. 5 a Flow and temperature fields and b disturbance flow and temperature fields in the meridional
plane of the liquid bridge
velocity field shows that the crests of the generated HTWs are oriented in the axial
direction of the LB as presented in Fig. 6. A close observation of the orientation
of the core region of the azimuthal convective cells in the generic cross section at
different axial positions reveals that the generated waveform is an oblique HTW.
Hence, these waves exhibit an angle of propagation with respect to the direction of
basic free surface flow, having two distinct components along both the directions.
Fig. 6 Disturbance flow

fields in the cross section at
different axial locations
Fig. 7 Temperature monitor

measured at a point on the
free surface
The critical azimuthal wave number decides the extension of thermal spots
appearing along the azimuthal direction. Generally, the azimuthal extension of
temperature spots can be calculated as (360/2 m) [5]. Hence, for m = 2, eight temper-
ature probes are supposed to be necessary to analyze the spatiotemporal flow struc-
ture in the liquid. Stabilized liquid temperature fluctuations show that the phase shift
between any of the observed temperature fluctuations located in azimuthal direction
measure the values 0 and 180 only which is a characteristic of the pulsating behavior
as shown in Fig. 7. Furthermore, the amplitudes of the fluctuations are also found to
be changing with respect to the spatial locations unlike that of a TW.
The analysis of the oscillation frequency of liquid temperature fluctuations on the
free surface is carried out using fast Fourier transform (FFT) and presented in Fig. 8.
The frequency and amplitude of liquid temperature oscillations in the present case
are observed to be 1.6 Hz and 0.21 K, respectively.
4 Conclusions
A numerical investigation of spatiotemporal flow structures of oscillatory thermo-

capillary convection in LBs of a high Pr liquid is carried out under microgravity
conditions. The spatiotemporal mode of convection in the present case is observed
to be a pulsating pattern which changes the flow and thermal fields in the liquid zone
periodically. The critical azimuthal wave number is found out be equal to 2 which
causes four thermal spots to generate on the free surface and on the generic section
orthogonal to the z-axis. The dynamic evolution of the flow system has been studied
with the help of an analysis of the amplitude and phase shift between the liquid
temperature fluctuations.
Fig. 8 Fourier spectrum for

liquid temperature
oscillations
References
1. Chang E, Wilcox WR (1976) Analysis of surface tension driven flow in floating zone melting.
Int J Heat Mass Transf 19:355–356
2. Kawamura H, Nishino K, Matsumoto S, Ueno I (2012) Report on microgravity experiments of
marangoni convection aboard international space station. J Heat Transf 134:031005-1–031005-
13
3. Tiwari S, Nishino K (2009) Effect of confined and heated ambient air on onset of instability in
liquid bridges of high Pr fluids. Fluid Dyn Mater Process 182:1–28
4. Lai CL (1984) Studies of thermocapillary oscillation phenomena, Ph.D. Dissertation, Depart-
ment of Mechanical and Aerospace Engineering, Case Western Reserve University
5. Lappa M (2009) Thermal convection: patterns, evolution and stability, Wiley, Chichester,
England
Dynamic Contact Angle Formulation
for Numerical Analysis of Taylor Bubble
Flow Over Obstacle
Kathankumar N. Khalasi, Nishant M. Shah, and Jyotirmay Banerjee
1 Introduction
Microfluidics is the miniaturization of procedures concerning the efficient and

controlled manipulation of very tiny quantities of liquids (pico—to nano-liter).
Consequently, it has been proven to be a promising technology for high-performance
systems. With the trend toward increasing levels of integration for high power
density within a confined region, high-performance heat sink technologies are
intrinsic to meet the elevated heat dissipation requirement in electronic devices.
Microchannels enable incredibly large surface-to-volume ratio ranging from 10,000
to 50,000 m2 ·m−3 which results in high heat transfer coefficient for convective heat
transfer, while in conventional devices, this ratio is approximately 100 m2 ·m−3 and
rarely reaches 1000 m2 ·m−3 [2]. Microscale effects and increase in heat sink area
enhance the heat transfer rate but also increase the pressure drops.
Several researchers considered combining pin fin within two-phase flows in the
microchannel considering the heat transfer enhancement [3]. With this background,
the present work is motivated to consider hydrodynamics of microscale two-phase
flows over micro-obstacle.
K. N. Khalasi (B) · N. M. Shah · J. Banerjee

Mechanical Engineering Department, National Institute of Technology, Surat, India
J. Banerjee
e-mail: jbaner@med.svnit.ac.in
264 K. N. Khalasi et al.
2.1 Governing Equations
Continuity and momentum equations in integral form with surface tension force
using CSF methodology are written as
¨
−
→
V · dA = 0 (1)
CS
˚ ¨ − ¨
∂ V → 1 →
−
d∀+ V V · dA = μ ∇ V · dA
∂t ρ
C∀ CS CS
˚ ˚ ˚
1 1
− ∇ pd∀− gd∀ + σ k∇Cd∀
ρ ρ
C∀ C∀ C∀
(2)
Here, κ is curvature, which is obtained by

1 n
κ = −∇ · n̂ = · ∇ |n| − ∇ · n (3)
|n| |n|
2.2 Contact Angle Formulation
The contact angle can be defined as an angle between the unit normal to the wall and
unit normal to the interface or unit tangent to the wall and tangent to the interface as
shown in Fig. 1. Using this definition,
n̂ = n̂ w cos θ + n̂ t sin θ (4)
where n̂ w is the normal pointing inside the wall, n̂ t is tangent to the wall, and n̂ is the
normal to the interface. The values of the unit normal n x and n y of the cell lying on
the wall are replaced by − cos θ and ± sin θ according to the volume fraction field
around the cell (horizontal wall).
When the contact line is moving with some velocity, the contact angle does not
remain same as static contact angle. This phenomenon was first observed by Hoffman
[4] in his experiments. It was concluded that the apparent contact angle is dependent
on the capillary number and shift factor (which is the function of θs ). In the present
work, the model proposed by Cox [1] is used for the implementation of dynamic
contact angle.
Dynamic Contact Angle Formulation … 265
Fig. 1 Wall adhesion

boundary condition
θd3 = θs3 + 144 · Ca (5)
μ U
Here, Ca is capillary number which can be given by liq contact−line
σ
. To allow free
movement of the interface on the wall, slip boundary is introduced. The velocity is
taken zero at some distance (slip length λ N ) inside the wall rather than at the actual
wall.
∂U
Uw = λ N (6)
∂n w
3 Validation of Contact Angle Formation
The drop of radius (Ro )0.01 m is initially placed at (0, 0). Mesh size is selected
such that dx = dy = 40 Ro
. Properties of fluids are ρ1 = 1000 kg/m3 , and ρ2 =
1 kg/m , μ1 = 10 Pa · s, μ2 = 10−5 Pa · s σ = 0.072N/m (Fig. 2).
3 −2
The drop shape (0.5 volume fraction contour) obtained numerically (red line) is
compared with the analytical shape (black-dotted line) as seen in Fig. 3.
Fig. 2 Initial position of the

drop and domain size
Fig. 3 Comparison of numerical results with analytical shape
Fig. 4 Initial position of

droplet on a flat surface
For the comparison of static contact angle (SCA) and dynamic contact angle
(DCA), droplet spreading on a flat surface is analyzed. The initial condition of the
droplet on surface is shown in Fig. 4.
The density and viscosity of liquid droplet are 809 kg/m3 and 34 × 10−3 Pa · s.
σ = 0.032N/m, and equilibrium contact angle is 41.5o . λ N = 0.01 · Ro and dx =
dy = Ro /20. Figure 5 shows the evolution of the shape of the droplet for DCA.
Fig. 5 Evolution of the

shape of droplet released on
a wall
Fig. 6 Comparison of SCA

and DCA with experimental
data

The dimensionless area A∗ = A−Ao
A f −Ao
is plotted against the dimensionless time
τ= tσ
1 for SCA and DCA to compare with experiments [5], where A f and Ao are
μV 3
final and initial droplet area, respectively.
It is observed from Fig. 6 that the drop spreads faster with SCA compared to DCA
formulation. The results obtained by DCA are compatible with the experiments.
4 Numerical Investigation of Taylor Bubble Flow Over

an Obstacle
The results of the present numerical study are compared with the experimental results
of Guo and Chen [6]. Guo and Chen carried out experiments in a microchannel with
T-junction using air–water flow.
The numerical results are validated for gas slug lengths with the experiments [6].
The variation of gas slug length by varying the liquid superficial velocity and keeping
gas superficial velocity constant was analyzed. The bubble length decreases with
increase in liquid superficial velocity. The comparison is shown in Fig. 7. (Table 1).
The results suggest that the bubble length observed in the numerical modeling
nearly fall within the range of gas slug measured experimentally. Certain deviation
is observed for the case when the liquid inlet velocity is low.
Fig. 7 Comparison of the

numerical gas slug length
with the experimental data
[6]
Table 1 Gas slug length comparison with experimental data [6]

QG QL UG UL L G (mm) L G (mm)
(ml/h) (ml/h) (m/s) (m/s) (Guo-Chen) (Present)
1.0 0.5 0.0347 0.0174 1.10–1.17 1.07
1.0 1.0 0.0347 0.0347 0.66–0.72 0.662
1.0 2.0 0.0347 0.0694 0.47–0.49 0.487
1.0 3.0 0.0347 0.1042 0.38–0.40 0.392
1.0 5.0 0.0347 0.1736 0.35–0.39 0.368
1.0 10.0 0.0347 0.3472 0.20–0.23 0.212
5 Taylor Bubble Flow Over an Obstacle
Taylor bubble is allowed to flow past a micro-obstacle placed well downstream of the
junction. Fully developed laminar flow is considered at both air and water inlet. The
water inlet velocity is kept such that Reynolds number is 50, and air inlet velocity is
set such that Reynolds number is 1.
The width of both the channels is 0.2 mm. The main channel length is 4.5 mm
while that of the upper arm is 0.5 mm. The square obstacle of length 25 μm is placed
2.0 mm downstream of the junction. The position of the obstacle is varied vertically
from 0.1 mm (at the center of the channel) to 0.08 mm. The schematic computational
domain is shown in Fig. 8.
Initially, the main channel is filled with water. As the water and air start entering
from inlet ports, the water flows over the obstacle placed downstream of the channel.
The result suggests the formation of a wake behind the obstacle, as seen in Fig. 9.
Fig. 8 Schematic of T-junction with a downstream obstacle
Fig. 9 Streamlines of water

flow past a square cylinder
When the obstacle is placed exactly at the center of the main channel, that leaves an
equal gap of 0.875 mm on either side of an obstacle. The bubble is symmetric when
it approaches the obstacle as it has attained a steady state during its journey from the
junction, as seen in Fig. 10a. As soon as the bubble touches the obstacle (Fig. 10b), it
splits in half and moves through both the passages, as seen in Fig. 10c–d. These two
parts of the bubble expand toward the center of the channel and touch each other at
some distance downstream of an obstacle Fig. 10e. They merge to form an original
bubble and make the obstacle completely entrapped by the bubble. As the merging
takes place at some distance downstream of the obstacle, this process captures some
amount of water downstream of an obstacle Fig. 10 f–k.
As the bubble advances ahead, the rear interface comes in contact with an obstacle,
and it again splits in half. This makes the rear interface to join the interface of the
entrapped water to create a single meniscus. As the surface tension tries to minimize
the interface area, this pushes the rear end of the bubble to attach with the obstacle.
While moving further downstream, the bubble elongates and creates a V-shaped tail,
as seen in Fig. 10m. Once detached from the obstacle, the bubble regains its original
shape while moving downstream Fig. 10n–p.
Next numerical simulation performed with an obstacle is shifted from the center-
line of the channel to a distance of 0.08 mm from the bottom wall as shown in Fig. 11.
This initial condition leaves a gap of 0.1075 mm above an obstacle and the gap of
0.0675 mm below it. The obstacle touches the bubble interface just below the leading
tip of the bubble, as seen in Fig. 11b. The bubble moves through both the passages
Fig. 10 Taylor bubble flow over a square obstacle at Y = 0.10 mm
Fig. 11 Taylor bubble flow over a square obstacle at Y = 0.08 mm

initially. This can be observed in Fig. 11c. The part of the bubble moving through
the larger passage continuously moves ahead, while the lower bubble moves to some
extent only, as seen in Fig. 11d–f. The upper interface grows toward the center to
attain a circular shape. This process forces the water in the vicinity below an obstacle
to move in the direction opposite to the actual flow. The contact line of the bubble on
the lower wall shrinks as the flow keeps coming. This process forms a V-shaped tail
that connects the bubble with a wall, as seen in Fig. 11i. When the interface detaches
from the wall, it tries to attain the circular shape while moving toward the upper
passage. This can be observed in Fig. 11j–m.
The findings indicate that when the obstacle is shifted from the centerline, the
bubble does not split as in the event of placing the obstacle at the centerline. The
bubble moves through the passage with a large cross-sectional area only.
The vorticity contours for the Taylor bubble flow over an obstacle placed at Y =
0.1 mm and Y = 0.08 mm are shown in Figs. 12 and 13, respectively.
The results show change in vorticity as the Taylor bubble flows over the obstacle.
As the interface changes its shape throughout the process, it creates local turbulence,
resulting in improved heat transfer. As the bubble passes over the obstacle placed at
Fig. 12 Vorticity contours for Taylor bubble with obstacle (Y = 0.1 mm)
Fig. 13 Vorticity contours for Taylor bubble with obstacle (Y = 0.08 mm)
Y = 0.08 mm, the change in liquid side vorticity is more compared to the case when
the obstacle is placed exactly at the centerline of the channel.
6 Conclusions
Contact angle is applied as a boundary condition in an in-house two-phase flow solver

using continuum surface force (CSF) approach.
• The results of droplet spread using SCA and DCA are compared with experimental
data. In the case when SCA is used, spreading of the droplet is much faster as
compared to the case when the DCA model is implemented.
• The obstacle in the microchannels can be used as heat transfer augmenter. As
it creates local turbulence in the channel, it is possible to dissipate heat more
effectively.
• The change in liquid side vorticity is more in the case when the obstacle is placed
slightly offset from the channel centerline.
References
1. Cox RG (1986) The dynamics of the spreading of liquids on a solid surface. part 1. viscous flow.
J Fluid Mech 168:169–194
2. Jähnisch K, Hessel V, Löwe H, Baerns M (2004) Chemistry in microstructured reactors. Angew
Chem Int Ed 43(4):406–446
3. Kishimoto T, Saski S (1987) Cooling characteristics of diamond-shaped interrupted cooling fin
for high-power LSI devices. Electron Lett 23(9):456
4. Hoffman RL (1975) A study of the advancing interface I. Interface shape in liquid—gas systems.
J Colloid Interface Sci 50(2):228–241
5. Lavi B, Marmur A (2004) The exponential power law: partial wetting kinetics and dynamic
contact angles. Colloids and Surfaces A: Physicochem Eng Aspects 250(1):409–414
6. Guo F, Chen B (2009) Numerical study on taylor bubble formation in a micro-channel T-junction
using VOF method. Microgravity Sci Technol 21(S1):51–58
Effect of Angle of Incidence on Unsteady
Wake Characteristics for Flow Past
Surface Mounted Square Cylinder
Prashant Kumar and Shaligram Tiwari
1 Introduction
Flow past bluff bodies mounted on a plate surface has attracted interests of researchers
due to its vast engineering applications. Such applications include cooling towers,
offshore structures, high-rise buildings, chimneys and fuel and gas storage tanks.
Experimental and numerical investigations have been carried out to study the
flow structure and dynamic loading for different values of Reynolds number (Re),
boundary layer thickness and aspect ratio (AR, ratio of height to side length of
the cylinder). Horseshoe vortices are formed due to flow separation in front of the
cylinder [1]. These vortices are observed to be steady for Re < 1500 [2]. A vortex
pair is formed of opposite sense of rotation near free-end named as ‘trailing vortex’
[3]. Liu et al. [4] studied effects of horseshoe and trailing vortices on dynamic forces
and Strouhal number for AR = 10 and Re changing from 100 to 200. Another vortex
pair named as ‘base vortex’ is formed near plate surface due to local flow field in the
upward direction [3].
Wake transition from steady to turbulent behind a square cylinder (AR = 4) has
been investigated by Zhang et al. [5] for Re varying from 50 to 1000 using direct
numerical simulations (DNS). Long streamwise slender structures and formation of
hairpin vortices for different values of AR = 2, 3 and 4 have been studied by Saha
[6] using DNS in flow past wall mounted square cylinder at Re = 250. Effect of
impinging shear on modes of vortex shedding in flow past surface mounted circular
cylinder of different heights at Re = 300 has been studied by Kumar et al. [7].
Effect of angle of attack on unsteady wake behaviour has not been investigated
to the authors’ best knowledge for surface mounted square cylinder. This forms the
P. Kumar (B) · S. Tiwari

Department of Mechanical Engineering, Indian Institute of Technology Madras, Chennai 600036,
India
S. Tiwari
276 P. Kumar and S. Tiwari
Fig. 1 Schematic of
computational domain
motivation behind the present study. Temporal wake oscillations have been illustrated
using Hilbert Huang transformation of time signals in the wake. Effect of incoming
shear on unsteady wake characteristics has also been illustrated for different values
of angle of incidence.
2 Problem Statement
Schematic of three-dimensional computational domain of finite-height square

cylinder of side length ‘d’ mounted on a plate surface is shown in Fig. 1. Cylinder
is exposed at different angle of incidence (α), measured with respect to streamwise
direction as depicted in Fig. 1. Projected length D = d × cos α + d × sin α, is
considered as characteristic dimension. Ratio of height (H) to the projected length
(D) is defined as aspect ratio (AR) of the cylinder and is fixed equal to 4. Cylinder is
placed at a distance of 5D from inlet. Length, width and height of the computational
domain are L 1 = 25D, L 2 = 10D and L 3 = 8D. Origin is fixed at geometric centre of
the square cylinder as shown in Fig. 1. Uniform and non-uniform velocity profiles
with constant shear rate (K = 0 and 0.025) have been considered up to height of 4D
from plate surface and above that uniform velocity is assumed at inlet corresponding
to Re equal to 150.
3 Governing Equations and Boundary Conditions
Flow is considered to be laminar and incompressible. Three-dimensional unsteady

continuity and momentum equations have been solved. Non-dimensionalized form
of governing equations in indicial form can be represented as
Effect of Angle of Incidence on Unsteady Wake Characteristics … 277
∂Ui
=0 (1)
∂ Xi
2
∂Ui ∂Ui U j ∂P 1 ∂ Ui
+ =− + (2)
∂t ∂Xj ∂ Xi Re ∂ X j ∂ X j
where Ui is non-dimensionalized Cartesian velocity component along X i -direction

and P is non-dimensionalized pressure. Here Ui represents U, V and W compo-
nents of velocity along X, Y and Z-directions respectively. Velocity components and
coordinate directions have been non-dimensionalized using characteristic velocity
(free-stream velocity, U∞ ) and characteristic dimension (D). Pressure (p) is non-
dimensionalized using ρU∞ 2
, where ρ is density of fluid. Flow Reynolds number
(Re) is defined as
U∞ D
Re = (3)
ν
where ν is kinematic viscosity of the fluid.
Boundary conditions have been employed on inlet, outlet, side, top, bottom
and cylinder surfaces. Uniform as well as non-uniform velocity profiles have
been
considered at inlet. Shear impinging flow with constant shear rate
K = G×D
U∞
, where G is velocity gradient up to height of 4D from bottom surface
is considered at inlet. Flow above the height of 4D is assumed to be uniform. Pres-
sure outlet boundary condition has been applied at outlet. Free slip and impermeable
boundary conditions have been imposed on top and side surfaces of the computa-
tional domain. No-slip and impermeable boundary conditions have been considered
on cylinder and bottom surfaces.
4 Grid and Numerical Technique
4.1 Grid Mesh and Grid Independence Study
Commercial software ANSYS ICEM CFD 18.1 has been used to generate hexahedral
grid mesh as shown in Fig. 2. Grids are refined near the cylinder surface and are
stretched uniformly in all the direction with constant expansion ratio less than 1.1.
Grid independence study has been carried out for different grid distributions around
the cylinder with zero angle of incidence at Re equal to 150. Total number of nodes
Fig. 2 Grid mesh around

cylinder
Table 1 Grid independence

Total no of No of cells Near wall grid Cdmean
study
nodes around cylinder size
210,325 80 0.013 1.2481
290,560 100 0.010 1.2593
376,732 120 0.007 1.2678
420,846 140 0.004 1.2691
and corresponding mean drag coefficient (Cdmean ) have been reported in Table 1. It
can be observed that for a grid divisions of 120 and near wall grid size of 0.007,
the mean drag coefficient (Cdmean ) is 1.2678 and does not change significantly with
further grid refinement. Hence, all the computations have been carried out considering
120 grid divisions around cylinder and 0.007 near wall grid size.
4.2 Numerical Techniques
Finite volume based Open Source Field Operation and Manipulation (OpenFOAM)
has been used to solve governing equations on collocated grids. Pressure and velocity
coupling has been resolved using Pressure Implicit with Splitting of Operators (PISO)
algorithm with two PISO corrector loops. Convective, pressure and diffusive terms
are discretized using Gauss linear scheme. Second order implicit scheme is employed
for time discretization in the momentum equation. For accuracy and numerical
stability of time derivative, time step has been adopted such that maximum Courant
number remains below 0.5 for all the computations. Absolute convergence criteria is
set as 10−6 for pressure and momentum convergence. A generalized code has been
Fig. 3 Variation of umean along transverse direction (Y-direction) for AR = 3 at X = 3 and Z = 2.5
written to incorporate the impinging shear flow at inlet using GroovyBC function
and integrated with the existing solver.
5.1 Validation of Computations
Numerical scheme has been validated with direct numerical simulations (DNS) for
flow past finite-height square cylinder (AR = 3) mounted on a plate surface at Re
= 250. Variation of time-averaged streamwise velocity component (U mean ) along
transverse direction (Y-direction) has been compared from the present computation
with the variation reported by Saha [6] at X = 3 and Z = 2.5 as shown in Fig. 3.
Velocity variation is found to be in good agreement with the variation observed using
DNS [6].
5.2 Hilbert Huang Transformation (HHT)
Wake oscillations might be nonlinear and nonstationary for flow past surface mounted
square cylinder due to combined effect of free-end and boundary layer thickness.
Such fluctuations can be analysed by decomposing transverse velocity signals (Y-
component) in the wake. HHT presents frequency-energy-time distribution of signal
known as Hilbert spectra, by decomposing any nonlinear, nonstationary time signal

into its components. An iterative technique known as “Empirical Mode Decompo-
sition (EMD)” decomposes signal into basis functions known as “intrinsic mode
functions (IMFs)”. Frequency of these IMFs may vary with time and such varia-
tion is presented in terms of “instantaneous frequency (IF)”. Explanation of HHT
algorithm can be found in the earlier study [7].
Hilbert spectra
Figure 4 shows variation of instantaneous frequency with time for each IMF.
Frequency and time have been labelled on Y and X-axes respectively. Colour in the
spectra presents square of amplitude or energy density corresponding to each IMF.
Red and blue contours present maximum and minimum energy density respectively.
For K = 0 and α = 0, single dominant IMF appears in the Hilbert spectra, indicating
wake oscillation is periodic as shown in Fig. 4a. However, with increase in α up to 30º,
intra-wave frequency modulations increase signifying growth in wake nonlinearity. In
addition, low frequency IMFs appear for α = 30º and 45º. Presence of low frequency
IMFs, may disturb the periodic of the wake for K = 0. For a given value of α, with
increase in K, bandwidth of the dominant IMF decreases revealing decrease in wake
unsteadiness as shown in Fig. 4b. Furthermore, low frequency IMFs are almost absent
for K equal to 0.025 corresponding to each value of α.
Marginal spectra
Hilbert spectra is integrated over time that gives marginal spectra. Figure 5 illus-
trates variation of energy density with Strouhal frequency (st = Uf ∞D ) for different
values of α and K. For K = 0 and α = 0, single dominant frequency appears in
the marginal spectra revealing periodic wake as shown in Fig. 5a. With increase in
values of α, low frequency peaks are observed other than dominant frequency that
may be caused by growth of nonlinearity in the wake. At higher values of α corre-
sponding to each value of K, spectra is spread over a range of frequency rather than
presence of sharp peaks. This indicates that multiple frequencies are responsible in
wake unsteady fluctuations.
Degree of stationarity
Presence of wake nonlinearity can be quantified using an index known as “degree
of stationarity DS(f )”. Zero index value corresponds to linear and stationary data.
Increase in values of DS(f ) corresponds to deviation from stationary behaviour
caused by nonlinear evolutions in the wake. With increase in values of α, range
of frequencies contribute to wake fluctuations as shown in Fig. 6. Index values are
higher at low frequencies, indicating these frequency components add more to wake
unsteadiness.
6 Dynamical Characteristics
Variation in mean drag coefficient (Cdmean ) with α for different values of K is shown
in Fig. 7. Mean drag coefficient is found to decrease with increase in α from 0º to
15º. However, no significant change in the value of Cdmean is observed as α increases
Fig. 4 Hilbert spectra of transverse velocity signals for different values of α a K = 0 and b K =
0.025
Fig. 5 Marginal spectra of transverse velocity signals for different values of α a K = 0 and b K =
0.025
Fig. 6 Degree of stationarity of transverse velocity signals for different values of α a K = 0 and
b K = 0.025
Fig. 7 Variation of Cdmean with α for different values of K
from 15º to 30º. With further increase in α, mead drag coefficient is found to increase
as shown in Fig. 7. For a given value of α, Cdmean is larger at lower value of K.
7 Conclusion
Effect of angle of attack on unsteady wake characteristics for flow past surface
mounted square cylinder has been investigated numerically using OpenFOAM. Intra-
wave frequency modulations increases with increase in values of α for uniform flow.
Appearance of IMFs in different frequency bands reveal that wake oscillation is
periodic. On the other hand, increase in bandwidth of the dominant IMF signifies
growth of wake nonlinearity. Frequency of the dominant IMF decreases with increase
in values of α corresponding to each value of K. Low frequency components appear
in the marginal spectra resulting into growth of wake unsteadiness with increase in
values of α. Wake nonlinear fluctuations have been quantified in terms of degree of
stationarity and found to be higher at larger angle of incidence. Mean drag coefficient
is found to decrease and then increase with α corresponding to each value of K.
References
1. Kawamura T, Hiwada M, Hibino T, Mabuchi T, Kumada M (1984) Flow around a finite circular
cylinder on a flat plate. Bull JSME 27:2142–2151
2. Lin C, Ho TC, Dey S (2008) Characteristics of steady horseshoe vortex system near junction of
square cylinder and base plate. J. Eng Mec 134:184–197
3. Sumner D, Heseltine JL, Dansereau OJP (2004) Wake structure of a finite circular cylinder of
small aspect ratio. Exp Fluids 37:720–730
4. Liu Y, So RMC, Cui ZX (2005) A finite cantilevered cylinder in a cross-flow. J Fluids Struct
20:589–609
5. Zhang D, Cheng L, An H, Zhao M (2017) Direct numerical simulation of flow around a surface-
mounted finite square cylinder at low reynolds numbers. Phys Fluids 29:
6. Saha AK (2013) Unsteady flow past a finite square cylinder mounted on a wall at low Reynolds
number. Comput Fluids 88:599–615
7. Kumar P, Manelil NP, Tiwari S (2019) Effects of shear intensity and aspect ratio on three-
dimensional wake characteristics of flow past surface mounted circular cylinder. Phys Fluids
31:043602(1–23)
Investigations on Thermal Performance
of Spiral Finned Latent Heat Storage
Unit
Bhavesh V. Vaghela, Digant S. Mehta, Manish K. Rathod,

and Jyotirmay Banerjee
1 Introduction
Solar thermal energy is one of the most prospective clean and renewable sources
of energy. However, the absorption of solar energy is intermittent in the absence of
sunny days. Therefore, there is a need to store solar energy as much as possible in the
presence of sunshine hours. Thus, there is strong need of to develop the energy storage
system which stores the solar energy during sunshine hours, and same energy can be
retrieved during non-sunshine hours. Thermal energy storage can be accomplished
using phase change materials (PCMs) which is an attractive way to store the energy.
This stored energy can be retrieved back to the system when solar radiation is not
available.
However, due to lower thermal conductivity (0.15 to 0.3 W/mK) of PCM, the
widespread utilization of PCM is restricted in thermal energy storage. Therefore,
in addition to energy storage, energy recovery time is much affected by such lower
thermal conductivity. In order to meet the expected level of melting/solidification
characteristics for large-scale utilization, thermal performance improvements of such
LHSU containers are necessary. To achieve the desired outcome, the energy must
be stored and retrieved within a minimum time for solar thermal applications. Thus,
the performance of any latent heat storage unit (LHSU) largely depends on heat
transfer mechanism. The heat transfer mechanism can be altered by active and
passive heat transfer enhancement techniques. The active heat transfer enhance-
ment includes thermal conductivity enhancement [3], extended surfaces (fins) [6, 7],
multiple PCM method [4] and micro-encapsulation [5]. Passive techniques includes
use of eccentricity of HTF tubes, rotation of the LHSU and multitube LHSU [8].
B. V. Vaghela (B) · D. S. Mehta · M. K. Rathod · J. Banerjee

Department of Mechanical Engineering, SVNIT, Surat, India
M. K. Rathod
e-mail: mkr@med.svnit.ac.in
286 B. V. Vaghela et al.
The motivation for the present research is to establish the feasibility of fins to
augment the heat transfer rate in LHSUs using PCMs. In this regard, analysis is
carried out to establish the heat transfer augmentation of LHSU using spiral fins.
2 Methodology
2.1 Experimental Set-up
A schematic of the shell-and-tube spiral finned latent heat storage unit is shown in
Figs. 1 and 2. The shell- and tube-type spiral finned heat exchanger is made of two
concentric tubes which are having a length of 0.6 m. The inner tube is manufactured
from brass, having an internal diameter of 0.028 m and an external diameter of
0.030 m. The outer tube, with an internal diameter of 0.088 m and an external
diameter of 0.092 m, is manufactured from stainless steel. A spiral fin is brazed on
inside tube. All details of spiral finned latent heat storage unit features are listed in
Table 1.
Fig. 1 Experimental test set-up for horizontal positioned LHSU

Investigations on Thermal Performance of Spiral Finned Latent Heat Storage Unit 287
Fig. 2 Schematic view of the spiral finned LHSU
Table 1 Details of spiral

Length L (mm) 600
finned LHSU features
Thickness t (mm) 2.5
Pitch P (mm) 37
Height h (mm) 20
Material Aluminium
2.2 Measurement of PCM and HTF Temperature
45 K-type thermocouples are located in the annular space between the shell and the
tube in order to record the temperature variation in the PCM as shown in Fig. 3. These
thermocouples are arranged in such a way that PCM temperature measurement is
feasible in all three directions (radial, axial and angular). Three thermocouples are
positioned radially to record temperature in the radial direction, one near the HTF
tube (1 mm radial distance), one near the middle (15 mm radial distance) and one
near the outer shell (28.5 mm radial distance) on the same axial plane as shown in
Fig. 3. These three temperature sensors are also arranged in the same radial plane at
120-degree angular interval. Thus, nine thermocouple sensors are used in one radial
Fig. 3 Schematic arrangement for different location of the thermocouple

plane. These nine sensors are mounted at the same axial distance on five distinct
axial planes (A, B, C, D and E).
2.3 Numerical Analysis
Phase change is studied in the present research numerically using the enthalpy-
porosity technique in the shell- and tube-type LHSU [1]. Melting and solidification
phenomena are simulated with ANSYS-FLUENT 16.2. The formulation of enthalpy-
porosity is used to monitor the interface between solid and liquid. In this method, the
liquid fraction shows the liquid state fraction of the cell volume. At each iteration, the
liquid fraction is calculated on the basis of an enthalpy equilibrium. The mushy zone
in which the liquid fraction lies between 0 and 1 is modelled as a “pseudo”-porous
medium in which the porosity decreases from 1 to 0 as the material solidifies. When
the material has fully solidified in a cell, the porosity becomes zero and hence the
velocities also drop to zero. The momentum source term in the mushy zone takes the
following from.
(1 − β)2
S= Amush −
→
v (1)
(β 3 + ε)
Liquid fraction β can be computed as
β = 0 if T < Tsolidus
β = 1 if T > Tliquids
T − Tsolidus
β= if Tsolidus < T < Tliquids (2)
Tliquids − Tsolidus
∂
(ρ H ) + ∇ · (ρ −
→
v H ) = ∇ · (k ∇T ) + S (3)
∂t
2.4 Initial and Boundary Conditions
In this numerical work, initially in the melting process at time t = 0, PCM inside
the annulus having a temperature of 28 °C as it is in the solid state. After time t > 0,
hot water flows through the tube at the temperature and mass flow rate of 85 °C and
1 kg/min, respectively. The outer wall of LHSU is kept insulated.
2.5 Validation with Experiment
The numerical result obtained is compared with present experimental results for
horizontal shell-and-tube latent heat storage unit as shown in Fig. 4. For validation, the
temperature recorded by thermocouple TD4 and TB6 in the experiment agrees well
with the numerical simulation as shown in Fig. 4. Table 2 presents thermophysical
properties of stearic acid (PCM).
Fig. 4 Validation of 100

numerical result with the
present experimental set-up
80
60
40
20
20 40 60 80 100
Table 2 Thermophysical
Property Stearic acid (PCM) Aluminium (Fin)
property of stearic acid
Melting point (°C) 55.7–56.6 –
Latent heat (kJ/kg) 196.1 –
Density (kg/m3 ) Solid-960 2719
Liquid-840
Specific heat (J/kg Solid-3000 871
°C) Liquid-2100
Thermal conductivity Solid-0.3 202.4
(W/m °C) Liquid-0.172
Fig. 5 Melting fraction

contours for horizontal latent
heat storage unit 5 min
10 min
30 min
60 min
3.1 Heat Transfer Intensification in Horizontal LHSU
The influence of a spiral fin in a horizontal latent heat storage unit on the melting
process of a phase change material is analysed in this paper. The heat generated by
the HTF is transmitted by conduction mode of heat transfer to the spiral fin and then
to the PCM around the tube at the starting of the melting process. The heat leads to a
gradual rise in the temperature of the spiral fin and the PCM. After the time progress,
the quantity of liquid PCM increases owing to a mixed convection and conduction
process in such a manner that as the melt region grows, the conduction dominance
gradually weakens and convection as the dominant mode of heat transfer leads to
an enhancement in the melting rate due to buoyancy-driven convection leads to the
circulation of liquid PCM in the upper half of the horizontal plane. So, the melted
PCM accumulates at the top of the shell, and solid PCM remains at the bottom as
shown in Fig. 5. The lower half melts mainly due to conduction mode of heat transfer.
Hence, the melting rate is slow in the region below the HTF tube.
3.2 Heat Transfer Intensification in Vertical LHSU
When hot HTF is passed from inner tube, heat is diffused from HTF to the fin and
then to the PCM around the inner tube. Initially (up to 5 min), a small layer of evenly
distributed molten PCM is formed around the HTF wall, which mainly occurs due
to the conduction mode of the heat transfer as shown in Fig. 6. As time progresses,
more amount of heat is diffused to the PCM which leads to decreasing in total
melting time. In vertical LHSU, heat transfer is uniformly distributed in axial and
5 min 10 min 30 min 40 min
Fig. 6 Melting fraction contours for vertical latent heat storage unit
radial direction throughout length of LHSU only conduction mode of heat transfer is
uniformly distributed in axial and radial directions throughout the length of LHSU
because only the conduction mode of heat transfer prevails. The upward melt flow
is restricted between two consecutive spiral passages which reduces the strength of
natural convection during melting.
Fig. 7 Liquid fraction 1

versus time for horizontal
LHSU
0.8
0.6
0.4
0.2
0
0 50 100 150 200 250 300
3.3 A Comparative Assessment of Horizontal LHSU Without

and with Spiral Fins During Melting
From Fig. 7, it is noted that around 310 min is required to melt the complete mass
of PCM in horizontal LHSU without fins. The upper half (till melt interface reaches
below the HTF tube) takes only 120 min to melt while a longer time of 190 min is
required for melting the lower half. This is due to the lower half melts mainly by
conduction mode heat transfer. By installation of spiral fins more amount of heat is
penetrated in to PCM via fins. So, time required to melt complete mass of PCM is
115 min with fins. Hence, total 63.37% reduction in overall melting time.
3.4 A Comparative Assessment of Vertical LHSU Without

and with Spiral Fins During Melting
From Fig. 8, it is observed that time required to reach liquid fraction one for vertical
LHSU is almost similar to horizontal LHSU without fins. By provision of spiral fins,
heat is diffused very fast from HTF to fins and then to PCM around the tube. Time
required to reach liquid fraction one is around 65 min with fins. Hence, total 80.30%
reduction is in overall melting time.
4 Conclusion
Following points can be concluded:

• In horizontal LHSU, the upper half region of PCM melts at a quicker rate due
to natural convection. Horizontal LHSU is thus better for part load operation as
Fig. 8 Liquid fraction 1

versus time for vertical
LHSU
0.8
0.6
0.4
0.2
0
0 50 100 150 200 250 300
compared to vertical LHSU. The total melting time for horizontal and vertical
LHSU is 314 min and 330 min, respectively without fins.
• The total melting rate is enhanced by 63.37% and 80.30% with spiral fins for
horizontal and vertical LHSU, respectively.
• By tilting LHSU from horizontal to vertical position, the total melting time reduces
from 115 to 65 min which means 43.47% reduction.
• The amount of total stored energy is 961.74 kJ and 983.57 kJ for horizontal and
vertical LHSU without fins, respectively. By provision of fins, total mass of PCM
in annulus reduces which leads to decrease in total stored energy. While using
spiral fins, amount of stored energy is 773.17 kJ and 732.77 kJ for horizontal and
vertical LHSU.
References
1. Brent AD, Voller VR, Reid KJ (1988) Enthalpy-porosity technique for modeling convection-
diffusion phase change: application to the melting of a pure metal. Number Heat Transf 13(3):
297–318
2. Sharma A, Tyagi VV, Chen CR, Buddhi D (2009) Review on thermal energy storage with phase
change materials and applications. Renew Sustain Energy Rev 13(2): 318–345
3. Esapour M, Hosseini MJ, Ranjbar AA, Pahamli Y, Bahrampoury R (2016) Phase change in
multi-tube heat exchangers. Renew Energy 85:1017–1025
4. Fang M, Chen G (2007) Effects of different multiple PCMs on the performance of a latent
thermal energy storage system. Appl Therm Eng 27:994–1000
5. Zheng Y, Barton JL, Tuzla K, Chen JC, Neti S, Oztekin A, Misiolek WZ (2015) Experimental and
computational study of thermal energy storage with encapsulated NaNO3 for high temperature
applications. Sol Energy 115:180–194
6. Rathod MK, Banerjee J (2015) Thermal performance enhancement of shell and tube Latent Heat
Storage Unit using longitudinal fins. Appl Therm Eng 75:1084–1092
7. Borhani SM, Hosseini MJ, Ranjbar AA, Bahrampoury R (2019) Investigation of phase change
in a spiral-fin heat exchanger. Appl Math Model 67:297–314
8. Mehta DS, Vaghela B, Rathod MK, Banerjee J (2019) Heat transfer intensification in horizontal
shell and tube latent heat storage unit. Numer Heat Transf Part Appl 75(7):489–508
Iterative Three-Dimensional
Tomographic Reconstruction of Flame
Using MENT
Jay Patel, Nandini Rajguru, Bhumi Dhameliya, Preet Tejwani,

and Mitesh Shah
1 Introduction
Having in-depth understanding of combustion process is necessary for controlling

and optimizing it. For doing so, it is important to analyse various phenomena that
occur in flame, such as temperature distribution, flame propagation, and physical
parameters of a flame such as size, location, brightness, uniformity, and oscillation
frequency. However, to fully reveal these characteristics, predicting the 3-D shape
of the flame is desirable.
Earlier, mechanical probes were used, but due to the obtrusive behaviour of these
probes, the results obtained were inaccurate. This led to the development of non-
intrusive experimental methods, like Laser Doppler Velocimetry (LDV), Particle
Image Velocimetry (PIV), and Raman–Rayleigh Method [1–3]. Such methods focus
only on a section of the flame at one point of time and require powerful lasers;
hence, the method becomes quite costly and unsuitable for hostile industrial envi-
ronments [4–6]. Then, visualization and quantitative characterization of flames using
3-D deconvolution and tomographic reconstruction techniques was done by a multi-
camera and single-camera system [7–10]. With the advent of computed tomography
(CT), reconstruction of flames became easier due to the ease of data collection,
robustness, and large field of view. Gilabert et al. [11] proposed a tomographic
approach that combines the logical filtered back-projection and Algebraic Recon-
struction Technique (ART) to reconstruct the 3-D flame shape. All the CT techniques
form a system of equations from the projection data, and then use the appropriate
computational method to solve it. Among the analytical and iterative methods, latter
one provides a good quality of reconstruction with sparse projection data; therefore,
it is most suitable for industrial applications. Subbarao et al. [12] compared several
iterative techniques: ART, MENT, Multiplicative Algebraic Reconstruction Tech-
nique (MART) and Minimization of Energy Technique (MEM). The conclusion is
J. Patel · N. Rajguru · B. Dhameliya · P. Tejwani · M. Shah (B)

Department of Mechanical Engineering, A. D. Patel Institute of Technology, Anand 388001, India
296 J. Patel et al.
drawn that MENT works very efficiently with sparse data, gives lesser errors, and
takes lower computing time. It was claimed by Minerbo [13] and Goyal et al. [14]
that MENT operates with a faster rate of convergence and greater accuracy even
with noisy data. To evaluate the crucial importance of relaxation factor in the quality
of the reconstruction image, Jiang and Zhang [15] examined the optimal relaxation
factors of the ART, MART, and MENT algorithms for bimodal asymmetrical and
three-peak asymmetrical-tested images.
Due to the above mentioned reasons, MENT has been chosen in this work for
reconstructing the 3-D flame shape. First, the algorithm is verified with a standard
benchmark problem. Thereafter, to evaluate anti-noise ability of the MENT, various
noise levels are intentionally induced in the reconstructed object. Furthermore, effect
of noise on the optimum relaxation factor has been analysed, and lastly, the 3-D flame
shape at various sections has been reconstructed.
2 Mathematical Modelling
2.1 Methodology for Generating Projection of an Object
A flame can be assumed as the 3-D distribution of luminous intensity. As flame

images represent the 2-D distribution of intensity at various angles, the aim here is to
utilize it and reconstruct once again the 3-D distribution of intensity to obtain the 3-D
flame shape. Inside a square region a luminosity distribution is considered, through
which the light beam having a constant cross section is passing. The square region
and light beam both are discretized into number of pixels and number of beams
respectively as shown in Fig. 1.
Fig. 1 Visualization of a
light beam passing through a
square grid
Iterative Three-Dimensional Tomographic Reconstruction of Flame Using MENT 297
One such discretized jth beam is shown, which passes at a projection angle (θ ) and
generates projection on a projection plane. Every pixel has a different intensity field
( f i ), which can either be zero or non-zero. The pixels which have a non-zero value
of f i represent the portion of the flame region. It can be seen from the Fig. 1, that the
jth beam will not intersect with each pixel and may intercept only a fraction of area in
some of the pixels. Henceforth, for computing contribution of a specific pixel in total
projection value of one particular discretized beam, only that portion of intercepted
area requires to be considered. Summation of all such projection value of pixels,
from which beam passes represents the total projection value of one discretized
beam. Similarly, projection value for all the beams can be calculated, and together
they all represent the projection of an object at the projection angle θ. For a specific
pixel, the fraction of area intercepted by a discretized light beam in a considered
projection is called weight function. Mathematically, it can be represented as
A j,k,i
w j,k,i = (1)
dx · dy
where, A j,k,i is the area of intercept for the ith cell of the jth beam in the k th projection,
dx and dy are the dimensions of cell along the x and y directions. If f i is the field
value in the ith cell and M is the total number of cells, then projection value of the
jth beam in the k th projection is,

M
{φ jk } = w j,k,i f i (2)
i=1
From the discretization, matrix equation can be formed:
{φ jk } = w j,k,i { f i } (3)
Thus, the problem of tomographic reconstruction reduces to an inversion of the

matrix. In this work, MENT is used to obtain field value for the reconstruction from
the projection data received out of Eq. (3).
2.2 Solution Technique—MENT
Consider a continuous function, g(x, y, z) ≥ 0. Entropy maximization refers to

maximizing following function.
˚
G(x, y, z) = − g(x, y, z) ln g(x, y, z)dzdydx (4)
298 J. Patel et al.
where g ln g is the entropy of function g. In image reconstruction, the collected

data and any other prior information comprise the constraint, over which entropy is
maximized. The aim is to maximize Eq. (4) subjected to the constraint represented in
Eq. (2) of projection data. In the mathematics, there is an approach for finding the local
maxima and minima of a function, which is subjected to equality constraints, known
as Lagrange multiplier method. Hence, using that method following expression for
optimum value of f i is obtained:
⎛ ⎞

Z
z(k)
f i = exp⎝−1 − λ jk w j,k,i ⎠ (5)
k=1 j=1
where, λjk is the Lagrangian multiplier, Z is the total no. of projections, and z(k) is
the no. of beams in k th projection. Substituting this f i in Eq. (2) of projection data,
⎛ ⎞

M
Z
z(k)
φ jk = w j,k,i exp⎝−1 − λ jk w j,k,i ⎠ (6)
i=1 k=1 j=1
For all k’s and all j’s Eq. (6) produces a square matrix with an equal no. of unknown
Lagrangian multipliers. Initially, they are initialized to zero, then subsequent iteration
values are calculated using:

G λkst
λk+1
st = λkst − X relaxation factor (7)
G λkst
These iterations are carried till all the Lagrangian multipliers have converged and
then using them in Eq. (5), the original field can be reconstructed.
3 Reconstruction of a Benchmark Problem
3.1 Cosine Phantom Object
In the present study, cosine phantom object is treated as a benchmark problem. Hence-
forth, the 3-D intensity distribution of a cosine phantom object has been generated,
whose shape is like ideal flame shape and can be formulated as follows:

f (x, y) = 0.082 1 − cos 2π (x + 0.5)0.95 1 − cos 2π (y + 0.5)0.98
where, −0.5 < x < 0.5, −0.5 < y < 0.5. The intensity field is discretized into 50 ×
50 grids for reconstruction purpose. The plot of the cosine phantom object is carried
out through its mathematical equation using MATLAB R2017a as shown in Fig. 3a.
3.2 Implementation of MENT on a Benchmark Problem
For verification of MENT algorithm, first it has been implemented on a standard

cosine phantom object. While reconstructing the flame shape, the images captured
by the camera can be directly used as projections; but for standard cosine phantom
object, it is necessary to generate its theoretical images at different projection angles
using its mathematical equation. In this work, theoretical projections are generated
at angles 0°, 45°, 90°, and 135° by method as described in Sect 2.1. To verify
these projections, the graphs of projection value versus beam no. are plotted, which
signifies that the image of cosine phantom at that angle will be the same as the graph
and Fig. 2 vouches for that.
From the theoretical generated projections shown in Fig. 2, reconstruction of the
cosine phantom object has been performed for only two projections, i.e. at 0° and
90° using the MENT algorithm. The obtained 3-D reconstructed shape of the cosine
phantom object is shown in Fig. 3b. From the comparison of Fig. 3a and 3b, it
is concluded that MENT accurately reconstructs the 3-D shape of cosine phantom
object.
Concept of root mean squared (RMS) error is used here for checking accuracy of
the reconstructed field with the original known intensity field. The formula for root
mean square error is as follows:

f (x, y, z)− f(x, y, z) 2
E = 100 ∗ / f max
n
Fig. 2 Cosine phantom object and its theoretical projections at different projection angles
300 J. Patel et al.
Fig. 3 Plot of cosine phantom object: (a) By using its mathematical equation and (b) By
reconstructing it using MENT
where, f (x, y, z) is the reconstructed field, f (x, y, z) is a known intensity field,
f max is the maximum intensity in the whole domain of f (x, y, z) , and n is the total
points at which the field is defined, i.e. 125,000 in the present study. For the above
reconstruction, E is found to be 0.6% by considering relaxation factor as 0.035.
Hence, this proves that the accounted MENT algorithm gives accurate reconstructed
shape even with a sparse data set. The relaxation factor can be taken as any positive
value but in the present case, it is found from the experiments that at 0.035 value,
minimum reconstruction error E is obtained.
3.3 Effect of Noise on the Performance of MENT
While capturing flame images, it may possible that its ideal images cannot be obtained
due to noise, attenuation, and other surrounding conditions. Therefore, the MENT
should be able to reconstruct the accurate shape even with limited noisy projection
data. To analyse the effect of noise and to check the anti-noise ability of the MENT,
various noise levels were induced intentionally in the above reconstructed shape,
which is shown in Fig. 4.
As the induced noise level increases, it is observed from Fig. 4 that the distortion
of reconstructed shape also increases gradually. However, up to 10% noise level, the
reconstructed shape of an object is still predictable; but after that, dramatic change
in reconstructed shape is observed. Hence, it is crystal clear that MENT algorithm
used in this work can reconstruct the accurate shape of an object up to 10% noisy
and sparse projection data set.
(a) 2% noise level (b) 5% noise level
(c) 10% noise level (d) 15% noise level
Fig. 4 Effect of noise on reconstructed shape of cosine phantom object
3.4 Effect of Noise on Optimum Relaxation Factor
Relaxation factor makes Lagrange multipliers to converge faster and reduces compu-
tation time, but it also changes RMS error. The value of relaxation factor, which gives
almost the same value of RMS error as the intentionally induced noise level, is consid-
ered as the optimum relaxation factor for that specific noise level. Table 1 shows the
obtained value of optimum relaxation factor and RMS error for the various noise
levels at 90° and 180° angle of views. It is clear from the Table 1 that, with the rise
in noise level, the optimum relaxation factor tends to alleviate to reduce the RMS
error. Furthermore, as the angle of view widens, the more projections are available,
thus more accurate results can be obtained. Such trend can be clearly witnessed in
Table 1.
302 J. Patel et al.
Table 1 RMS error and optimum relaxation factor for various noise levels
Noise level 90° angle of view with 2 projections 180° angle of view with 3 projections
RMS error Optimum relaxation RMS error (%) Optimum relaxation
(%) Factor Factor
1% 1.49 0.02 1.02 0.019
2% 2.26 0.017 2.11 0.011
5% 5.64 0.0065 5.16 0.005
7.50% 7.93 0.0045 7.62 0.0025
4 Reconstruction of Flame
The reconstruction of flame is performed using projections at 0° and 90°, which are
shown in Fig. 5. The original size of images was 650 × 200 pixels, which is resized
into 163 × 50 pixels, and then used to reconstruct the flame.
As thickness of flame at the bottom is an important parameter, to obtain it, the
projection data of 161th pixel row was taken from the top of images and by its
reconstruction, flame thickness at the bottom is found as 5.5 cm as shown in Fig. 6a.
Furthermore, as we go from bottom to top, the reconstructed shape becomes narrow
gradually as shown in Fig. 6. It is observed that the flame shape is rectangular
at bottom and has abrupt changes at corners, which might happened due to the
lack of projection data, as only two projection angles were taken in this work. If
more projection angles are considered between 0° and 90°, then perhaps oval type
cross-sectional shape of flame can be obtained.
Fig. 5 Flame image at (a) 0° angle and (b) 90° angle

Fig. 6 Reconstructed flame shape: (a) For 161st pixel row, (b) For 100th pixel row, (c) For 50th
pixel row, and (d) For 5th pixel row from the top of images
5 Conclusion
In the present work, focus had been made on the MENT for reconstructing 3-D lumi-
nosity distribution of flame from limited projections, and to check its correctness,
first it was implemented on the standard cosine phantom object. From the results, it
was observed that the reconstructed object shape came out as similar as the standard
object even with sparse projection data. Later, intentionally noise had been intro-
duced in the projection data to check the anti-noise ability of MENT and from the
observations, it can be stated that MENT can reconstruct the shape of an object accu-
rately up to 10% noisy sparse projection data. Further, effect of induced noise level
on optimum relaxation factor had been analysed; it followed decreasing trend. Lastly,
3-D reconstruction of flame had been carried out for 0° and 90° projection angles,
and flame thickness at the bottom was obtained as 5.5 cm. Moreover, it can be stated
that taking more projections between 0° and 90° might give oval type cross-sectional
flame shape. Hence, future work may include taking more projections for better
results of flame reconstruction and developing Bunsen burner set-up with multiple
cameras to analyse effects of turbulence in flame.
304 J. Patel et al.
References
1. Lee T, Bessler WG, Kronemayer H, Schulz C, Jeffries JB (2005) Appl Opt 44(31): 6718–6728
2. Doi J, Sato S (2007) Opt Eng 46(1):015601
3. Yang HN, Yang B, Cai XS, Hecht C, Dreier T, Schulz C (2015) Laser Eng 31:285–297
4. Sun J, Xu C, Zhang B, Moinul Hossain M, Wang S, Qi H, Tan H (2016) Three-dimensional
temperature field measurement of flame using single light field camera 24(2)
5. Floyd J, Geipel P, Kempf AM (2011) Comb Flame 158:376–391
6. Sislian JP, Jiang L-Y, Cusworth RA (1988) Prog Energy Combust Sci 14(2):99–146
7. Zhou HC, Han SD, Sheng F (2002) J Quant Spectrosc Radia Transf 72(4):361–383
8. Zhou HC, Lou C, Cheng Q, Jiang Z, He J, Huang B, Pei Z, Lu C (2005) 30:1699–1706
9. Bheemul HC, Lu G, Yan Y (2002) Meas Sci Technol 13(10):1643–1650
10. Fisher W, Burkhardt H (1990) Proc SPIE 1349:96–105
11. Gilabert G, Gang Lu, Yan Y (2007) 3D tomo recon of luminosity dist of combust flame. IEEE
Trans Instru Measure 56(4):1300–1306
12. Subbarao PMV, Munshi P, Muralidhar K (1996a) NDT & E International 30:359–370
13. Minerbo G (1978) Comput Grap Image Proc 10:48–68
14. Goyal A, Chaudhry S, Subbarao PMV (2013) Direct 3-D tomography of flames using MENT.
Combust Flame 161:173–183
15. Jiang W, Zhang X (2017) Relax factor opt for common iterative algorithms in opt computed
tomography. Math Prob Eng 2017(8)
Computational Study of Aerodynamic
Performance of a Spinning Projectile
with Different AFT-Body Configurations
E. Aishwarya, T. J. Akhila, S. R. Amrutha, R. P. Sindhu, Karthik Sundarraj,

P. S. Kulkarni, Manoj Veetil, and R. Ganesh Pawar
Nomenclature
SSVG Sharp serrated vortex generators.

TSVG Trapezium-shaped vortex generators.
SBT1 Serrated boattail 1
SBT2 Serrated boattail 2
RC1 Rare cavity 1
RC2 Rare cavity 2
RC3 Rare cavity 3
MSAB Multistep after body
SPAB Splitter plate after body
E. Aishwarya · T. J. Akhila · S. R. Amrutha · R. P. Sindhu · M. Veetil

IIAEM, Jain University, Bengaluru, India
K. Sundarraj (B)
Hexagon | MSC Software, Bengaluru, India
e-mail: karthik_sundarraj@yahoo.com
P. S. Kulkarni
Department of Aerospace Engineering, IISc, Bengaluru, India
e-mail: psk@iisc.ac.in
R. G. Pawar
Research Intern, CML, IISc, Bengaluru, India
306 E. Aishwarya et al.
1 Introduction
Objects in flight are influenced by two major forces, namely lift and drag. Hence,
there is a need to find an effective method to minimize the drag generated due to
its motion in flight. In the study of ballistics, the major problem encountered is the
increase in base drag due to the formation of low-pressure wake region at the rear end
of bodies such as missiles and projectiles. This wake region contributes to a major
amount in increase of the total drag acting on the projectile. The other parameters
that have a considerable effect on the drag acting on the projectile are Mach number
and the profile of the body. Research activity was carried out to study the coefficients
of lift and drag for projectiles and missiles.
W. Jiajan carried out experimental and computational study to reduce the aero-
dynamic drag and also to improve the static and dynamic stability of a spinning
projectile.
P.R. Viswanath suggested various techniques such as aft body attachments, base
bleed, ventilated cavities which highlights locking of vortices to minimize the aero-
dynamic drag. From the related researches, it motivated us to predict the aerodynamic
characteristics of M549 155 mm projectile for various configurations.
The idea of breaking down of vortices which is called vortex bursting is imple-
mented by making use of multistep after body. Vortex bursting prevents shedding of
vortexes thereby providing an acceptable amount of drag reduction. This is obtained
by interfering the vortex-shedding process, for instance by avoiding the occurrence
of the separation of the boundary layer along the body. The coefficient of drag for
the multistep after body is expected to be lesser than all other configurations. This
will be clear only after satisfactory evidence [1–15].
2 Methodology
In the field of computational fluid dynamics, the key elements that rule the flow
physics are the Navier–Stokes equations and the energy equations. So, after the fluid
problem is defined the CFD tool must fulfil two important requirements, first one
being the mathematical model for the given flow problem and the numerical methods.
For mathematical model, the CFD tool makes use of Navier–Stokes equations which
act as governing equations, which are used to solve the flow problem. The Navier–
Stokes equations include the three conservation laws including the conservation of
mass, momentum and energy equations. When a flow is turbulent, we make use of
turbulent models which has the ability to predict the effects of turbulence in a given
fluid problem. These turbulent models act as mathematical models. In our present
work, the turbulent model that we utilized is the K-ω SST model [5, 8] Mair WA
[15].
Computational Study of Aerodynamic Performance … 307
2.1 Validation
To ensure accuracy in the results obtained for optimized design, [16] consists of
validated results considering CFD simulations for three different Mach numbers and
different optimizations.
2.2 Geometry
The 2D model and all other optimized geometries were designed using CAD tool.
Initial design of the standard projectile is as shown in the figure below, and the base
diameter of the projectile was maintained to be 155 mm (Figs. 1 and 2).
Axisymmetric body was chosen for all simulations.
• to minimize the computational time.
• to reduce the mesh size.
The domain that encloses the projectile ensures to encapsulate the formation of
shocks at the forebody and shedding of the vortex at the rear. Initially, the domain
size used in [17] was considered but reversed flow was observed. Hence, a larger
domain forming quarter circle maintaining 3D and 6D before and after the body was
well considered (D = 155 mm base diameter of the projectile) (Figs. 3, 4 and 5).
1. Multistep After Body (MSAB)

At the rear end of the projectile, a multistep after body is attached. Multistep
after body falls under the category of “separation control by trapping vortices”
according to [18]. The height of the multistep is maintained to be 77.5 mm,
the horizontal distance of the annular step is maintained to be 45 mm, and the
vertical height is maintained to be 15 mm. The main motive behind the multistep
Fig. 1 Standard M549 model
Fig. 2 Axisymmetric M549 model

Fig. 3 Domain
Fig. 4 Multistep after body
Fig. 5 Measurements of
multistep after body
design is to disturb the airflow so as to reduce the wake region. At the region
where the multistep begins, the vortices are strong but as these vortices pass
through the annular steps, they start becoming weak.
From [18] it is recommended that the annular steps’ height is to be maintained
small compared to its width. The ratio maintained in the geometry is.
Fig. 6 SPAB
Fig. 7 Measurements of
SPAB
2. Splitter Plate After Body (SPAB)

The height and width of the after body are same as that of the boattail of the
standard projectile. The thickness of the splitter plates is maintained to be 3.2 mm
(Figs. 6 and 7).
It is observed that, in the splitter plates, the vortices that are generated are trapped
between the base of the body and the splitter plates, due to which vortex shedding is
suppressed. This reduces the pressure loss at the base which leads to the reduction
in base drag.
Both the configurations involve locking the vortex technique.
The optimized geometries were run for Mach 1.3, 2 and 4.

Coefficient of drag for the rear cavity simulated for Mach 2 is found to be the
best-optimized configuration as the value is 0.21290 (Fig. 8).
Table 1 represents results from [16] which are considered as reference for the
present work.
Figure 9 shows the graph of coefficient of drag versus Mach number for different
shape modifications.
From the graph, it is observed that the rear cavity of thickness 15.50 mm has the
minimum aerodynamic drag.
Fig. 8 Rear cavity 2
Table 1 Cd values for

Design Coefficient of Coefficient of Coefficient of
different configurations
drag for Mach drag for Mach drag for Mach
1.3 2 4
SSVG 0.39815 0.36136 0.19755
TSVG 0.36025 0.32587 0.19497
BTS 1 0.36123 0.28563 0.16359
BTS 2 0.3715 0.29123 0.19499
RC 1 0.34726 0.27152 0.15552
(7.75 mm t)
RC 2 0.26457 0.21290 0.16466
(15.50 mm t)
RC 3 0.26457 0.29469 0.15522
(23.50 mm t)
Splitter plates 0.35730 0.27589 0.15932
Multi-step aft 0.35231 0.32142 0.15708
body
From the velocity contours shown in Figs. 10 and 11, we see that the velocity
is low at the base regions where the vortices are being captured (Fig. 10) and the
vortices energy is being reduced (Fig. 11), thus leading to reduction in drag.
4 Conclusions
From [16], it is clear that with appropriate modifications to the basic model there
will be considerable decrease in drag. Hence, simulations of the above geometrical
modifications under spinning conditions at different angles of attack are being carried
out.
Fig. 9 Graph of coefficient of drag versus Mach number
Fig. 10 Velocity contours of splitter plates

Fig. 11 Velocity contours of multistep
Acknowledgements We would like to thank Mr. Karthik Sundarraj and Ganesh Pawar R for guiding
and helping us throughout this work.
References
1. Lun SW (2011) Aerodynamic validation of emerging projectile configurations. Naval Post-

graduate School Monterey, CA 93943–5000, December 2011
2. Jiajan W. Aerodynamic characteristics of high-performance rounds at Mach 1.8 to 4
3. Suryanarayana GK. Bluff-body drag reduction by passive ventilation
4. Byregowda G. Base drag considerations and projectile optimization on 0.5-Caliber projectile
5. Viswanath PR, Narasimha R (1974) Two-dimensional boat-tailed bases in supersonic flow.
Aero Quarterly XXV 210–224
6. Viswanath PR (2001) Drag reduction of afterbodies by controlled separated flows. AIAA J
39(1):73–78
7. Silton SI. Navier-Stokes computations for a spinning projectile from subsonic to supersonic
speeds
8. Sundararaj K. Computational study on base body drag reduction using Locked Vortex Flow
Management Technique by attaching Double splitter plates at the base
9. Wheeler GO. Low Drag Vortex Generators, 2715 185th Ave. E., Sumner, Wash. 98390
10. Calfee CT, Calfee VT. Nose cone and tail structures for an air vehicle. Both of 3294 S. Polk
St, Dallas, Tex, Filed Jan. 30, 1962, Ser. No. 173,856 8 Claims
11. Boattail plates with non-rectangular geometries for reducing aerodynamic base drag of a bluff
body in ground effect, Jason M. Ortega, Kambizsabari
12. Bullet, Gregory J. Giannoni, 1816 Stoney Crest Dr. Elberton, GA (US) 30635
13. Projectile with reduced base drag, Robert J. Paterson, Simsbury, Michael J. Werle, West
Hartford, both of Conn: Walter M.Prez, Jr., Wibraham
14. Surface modification apparatus and method for decreasing the drag or retarding forces created
by fluids flowing across a moving surface, Hilbert F. P. Drews, 5640 S. 76th St., Greendale,
Wis.53129
15. Mair WA (1969) Reduction of base drag by boat-tailed afterbodies in low speed flow. Aero
Quarterly 99: 307–320
16. Computational Study of Aerodynamic characteristics of a projectile by varying Boat tail
configurations
17. Jiajan W, Chue RSM, Nguyen T, Yu S (2013) Optimization of round bodies for aerodynamic
performance and stability at supersonic speeds. Aeronaut J 177(1193): 661–685
18. Viswanath PR (1996) Flow management techniques for base and afterbody drag reduction.
Prog Aerospace Sci 32: 79–129
Mixed Convection MHD Flow
of a Nanofluid in a Rotating System
with Heat Generation/Absorption
P. Vimala and K. Manimegalai
1 Introduction
Stretching sheet phenomenon has many real-life applications in industrial areas such
as extrusion, glass blowing, production of papers, manufacture of plastics and spin-
ning of fibers. These applications involve enormous heat and mass transfer. Therefore,
many investigations on such flows have been carried out to study the heat and mass
transfer characteristics [1–5]. [1] investigated the flow past a linear stretching plate
with heat conduction. [2] studied the suction and injection flow of a viscous fluid in a
porous stretching sheet with heat and mass transfer. [3] studied the heat transfer char-
acteristics of electrically conducting fluid flow between two horizontal plates with
stretching and porosity. [4] investigated the flow of a viscous incompressible fluid
over a continuous stretching sheet with heat transfer characteristics. [5] obtained a
numerical solution of free convection flow over a porous stretching sheet.
In the above investigations, rotation phenomenon has not been considered.
However, this phenomenon is encountered in many practical applications such as
rotating machinery and cooling processes. Owing to the significance of rotational
geometries in many industries, [6] studied electrically conducting fluid flow over a
stretching sheet in a porous rotating geometry.
All the above papers considered flow of viscous incompressible fluids. However,
in many applications such as cooling processes in machines, heat exchangers and
chemical processing, nanofluids are used owing to their better heat transfer charac-
teristics. Nanofluids are high thermal conducting particle suspensions in low thermal
conducting base fluids. Such fluids have higher thermal conductivity when compared
to their base fluids. This property of nanofluids makes them more significant in appli-
cations involving enormous heat transfer. Hence, [7] adopted a homogeneous flow
model with conventional transport equations for pure fluids along with the physical
properties of nanofluids. [8] analyzed the flow and heat transfer of copper–water
P. Vimala (B) · K. Manimegalai

Department of Mathematics, Anna University, Chennai, India
316 P. Vimala and K. Manimegalai
nanofluid between stretching and porous surfaces using HAM in a rotating system.
The above articles [7, 8] used a single-phase model for the study of nanofluid flow.
These studies showed the enhancement of heat transfer characteristics of nanofluids
significantly.
Introduced by [9], a two-phase model has also been used in the literature [10] for
describing the flow behavior of nanofluids. [9] presented seven slip mechanisms
between nanoparticles and base fluids. [10] studied the nanofluid flow and heat
transfer in a rotating system using numerical method.
Although many investigations have been carried out in flow of nanofluids between
parallel plates, factors such as mixed convection flow in such geometries, presence
of internal heat source/sink and presence of a magnetic field may impact the flow
behavior as well as the heat and mass transfer characteristics. Therefore, the present
problem discusses a mixed convection steady MHD flow of a nanofluid between a
lower stretching plate and an upper porous plate in a rotating system. Further, the
flow is assumed to have an internal heat source/sink. The governing equations are
derived using two-phase model, and the formulation is presented in Section II. The
problem is then solved using SAM as in Section III. Section IV presents the results
with appropriate graphical representations. Section V gives the conclusion.
2 Mathematical Formulation
A mixed convection steady laminar flow of an incompressible nanofluid between two

horizontal parallel circular plates is considered as shown in Fig. 1. The lower plate is
stretched along its plane by equal and opposite forces keeping the origin unchanged.
The upper plate is porous and is at a distance of h(t) from the lower plate. Fluid flows
through the pores in the upper plate with a constant wall suction/injection velocity
Fig. 1 The flow geometry

Mixed Convection MHD Flow of a Nanofluid in a Rotating System … 317
v0 . Both the plates rotate with a constant angular velocity about the y-axis, which
makes the nanofluid between them to rotate, and there is net cross flow along the z-
axis. A uniform magnetic field of intensity B0 is applied in the direction of the y-axis.
It is assumed that there is an internal heat source/sink with heat generation/absorption
parameter Q 0 . Under these conditions, the nondimensional form of the governing
equations of the nanofluid MHD mixed convection flow along with heat source/sink
is given by
2
−h /axρ f ν ∂ P ∗ /∂ x

= − f − R f 2 − f f − 2K rg − M f + Grt θ + Grc φ (1)

−1/ρ f h ∂ P ∗ /∂η = a 2 h f f + f /R (2)

g − R f g − f g + 2K r f − Mg = 0 (3)

θ + Pr R f θ + N bθ φ + N tθ 2 + Qθ = 0 (4)
φ + (N t/N b)θ + RSc f φ = 0 (5)
where the following nondimensional variables have been used
y T − Th C − Ch
η= , u = ax f (η), v = −ah f (η), w = axg(η), θ = ,φ =
h(t) T0 − Th C0 − Ch
(6)
where u, v and w denote the fluid velocity components in the x, y and z direc-
tions, respectively, p ∗ is the modified fluid pressure, T is the temperature of the
fluid,Th is the temperature at the upper plate andρ f , ν and α are the fluid density,
kinematic viscosity and thermal diffusivity, respectively. Also, C is the concentration
of the nanoparticles,DT is the thermophoretic
diffusion parameter,D B is the Brow-
nian motion coefficient, ρc p p / ρc p f is the ratio of effective heat capacity of the
nanoparticle to the heat capacity of the fluid, g is the magnitude of acceleration due
to gravity, βT is the thermal expansion coefficient,βC is the concentration expansion
coefficient, and Q 0 is the uniform volumetric heat generation (Q 0 > 0) or absorption
(Q 0 < 0) parameter. R = ah 2 /ν is the Reynolds number, K r = h 2 /ν is the rota-
tional number, M = σ B02 h 2 /νρ f is the magnetic parameter or Hartmann number,
Grt = g̃βT h 2 (T0 − Th )\νax = Gr \Rex2 is the ratio between the modified Grashof
number Gr and square of the local Reynolds number Rex , which represents the
temperature mixed convection parameter, Grc = g̃βC h 2 (C0 − C h )/νax represents
the concentration
mixed convection parameter, Pr = μc p /k is the Pranl number,
Q = Q 0 h 2 / ρcρ f ν is the heat source/sink parameter, Sc = ν/D B is the Schmidt
number,

N b = ρcρ P D B (C0 − C h )/ ρcρ f ν is the Brownian motion parameter, and

N t = ρcρ P DT (T0 − Th )/ ρcρ f ν is the thermophoresis parameter.
Eliminating the pressure gradient terms, Eqs. (1) and (2) yield

f − R f f − f f − 2K rg − M f + Grt θ + Grc φ = 0 (7)
Thus, it is required to solve the governing equations Eqs. (3)-(5) and (7) of the
present problem with the boundary conditions
f (0) = 1, f (0) = 0, g(0) = 0, θ (0) = 1, φ(0) = 1 (8)
f (1) = 0, f (1) = λ, g(1) = 0, θ (1) = 0, φ(1) = 0 (9)
where λ = v0 /ah is the porosity parameter.
3 Solution by SAM
Expanding the unknown variables f, g, θ and φ in powers of R which is of the order

O(1), (as in [11, 12])

f (η) = f 0 + R f 1 + R 2 f 2 + O R 3

g(η) = g0 + Rg1 + R 2 g2 + O R 3

θ (η) = θ0 + Rθ1 + R 2 θ2 + O R 3 (10)

φ(η) = φ0 + Rφ1 + R 2 φ2 + O R 3
and substituting Eq. (10) in the reduced governing Eqs. (3)-(7) and the boundary
conditions (8) and (9), the final equations are solved for f (η), g(η), θ (η) and φ(η)
upto the second order approximation.
The zeroth-order solutions are
f 0 (η) = A0 η3 + B0 η2 + η, g0 (η) = 0, θ0 (η) = 1 − η, φ0 (η) = 1 − η,.
where A0 = 1 − 2λ,B0 = 3λ − 2.
The first-order solutions are5
1η 1η 1η 1η
+ E16η + F12η ,.
7 6 4 3 2
f 1 (η) = A840 + B360 + C120 + D24
4
g1 (η) = − 2KR r A04η + B03η + η2 − G 1 η ,.
3 2

A0 η5 B0 η4 H1 η3 I1 η 2
θ1 (η) = Pr + + − − J1 η ,
20 12 6 2

A0 η5 B0 η4 η3
φ1 (η) = Le + + − k1 η ,
20 12 6
where A1 , B1 , C1 ,D1 , E 1 , F1 ,G 1 ,H1 , I1 ,J1 and K 1 .

A1 = 48λ2 −48λ+12, B1 = −72λ2 +84λ−24„ D1 = 2B0 1 + MR + GrR t + GrR c ,
E 1 = − 28
A1
− 15B1
− 3C20
1
− D21 , F1 = 105 A1
+ 60B1
+ C301 + D121 ,G 1 = A40 + B30 + 21 ,
H1 = 1 + R Pr , I1 =
Q N b+N t+Q
R
,
J1 = 20
A0
+ 12
B0
+ H61 − I21 and K 1 = 20 A0
+ 12B0
+ 16 .
The second-order expressions are very lengthy, and hence, they are not presented
here. However, solutions upto second order have been obtained and presented in the
graphical representations. Using the zeroth-order, first-order and second-order solu-
tions in Eq. (10), approximate expressions upto second order for the flow variables
f (η) and g(η), temperature distribution θ (η) and concentration distribution φ(η)
can be obtained.
The steady MHD rotating flow of an incompressible nanofluid between a lower

stretching plate and an upper porous plate is studied. Using SAM, the problem is
solved, and the effects of various parameters on velocity, temperature and concen-
tration distributions are discussed. Further, a comparison of the present results is
made with those of the existing literature. As the flow is considered to be laminar
and between parallel plates, the value of the Reynolds number R is taken as 1. The
values of the rotation number K r and magnetic Hartmann number M are chosen to
be 1 each. Since mixed convection is considered, the values of temperature Grashof
number Grt and concentration Grashof number Grc are chosen to be equal to 1 each.
As the nanoparticle size is very small, the values of Brownian motion parameter N b,
thermophoresis parameter N t and Schmidth number Sc are taken to be equal to 0.1
each. Prandtl number Pr of water is 6.2. The value of heat generation/absorption
value Q due to the internal source/sink is taken as 0.3. The above values of the
parameters are used in obtaining the results, unless otherwise specified.
Figures 2 and 3 compare the present results when M = 0, Grt = 0, Grc = 0, Q =
0 with those of [8]. Here, represents the volume fraction of [8]. Figure 2 shows
the effects of the porosity parameter λ and Reynolds number R on the temperature
distribution θ (η). It is seen that an increase in the porosity parameter results in a
decrease in temperature distribution and an increase in Reynolds number results in
a further decrease in temperature distribution. On neglecting the Brownian motion
and thermophoresis effects (ie.N b = N t = 0) and the Schmidt number (Sc = 0),
the present results agree well with [8]. Further, the present results including these
effects (N b = N t = 0.1 and Sc = 0.1) show an enhanced temperature distribution.
Figure 3 shows a comparison of temperature distribution θ (η) between single-phase
model [8] and two-phase model of the present work. It is observed from the results
of [8] that as the volume fraction increases, temperature distribution increases.
Fig. 2 Effects of porosity parameters and Reynolds number on temperature distribution—A

comparison
Fig. 3 Comparison of the single-phase (Sheikholeslami [8]) and two-phase models (present model)
R = 0.5, K r = 0.5, Grt = 1, Grc = 1, M = 1, Pr = 6.2, Q = 0.3
Also, a further increase in temperature distribution is observed in the present work

due to the effects of Brownian motion N b, thermophoresis N t and Schmidt number
Sc.
In Fig. 4a, x-direction velocity component f (η) has a dual behavior. In the first
half (0 ≤ η ≤ 0.5) of the region, y-direction velocity component f (η) increases
with temperature Grashof number Grt while in the second half (0.6 ≤ η ≤ 1), f (η)
increases in the negative direction with temperature Grashof number. From Fig. 4b,
it is seen that y-direction velocity component f (η) increases with an increase in
temperature Grashof number. Distribution of z-direction velocity component g(η)
(Fig. 4c) is similar to that of x-direction velocity component (Fig. 4a). As temperature
Grashof number increases, the temperature distribution decreases as seen in Fig. 4d.
Fig. 4 Effects of temperature Grashof number on the velocity components (a, b, c) and temperature
distribution (d)
Fig. 5 Effects of heat source/sink parameters on the temperature distribution (a) and concentration
distribution (b)
The effect of temperature Grashof number on concentration distribution φ(η) is

negligible.Grc
Figure 5 presents the effects of heat generation/absorption parameter Q. Q > 0
represents the heat source or heat generation causing temperature (θ (η)) enhance-
ment, whereas Q < 0 represents the heat sink or heat absorption which reduces
the temperature distribution θ (η). In Fig. 5a, temperature distribution increases for
greater values of heat generation/absorption parameter from lower part to middle
of the channel, but the trend is reversed in the upper half of the channel due to the
porosity of the upper plate. In Fig. 5b, concentration distribution φ(η) decreases with
an increase in heat generation/absorption parameter.
5 Conclusions
An investigation of the mixed convection steady laminar flow of an incompressible

nanofluid between a stretching sheet and a porous plate is carried out considering
a two-phase model. The problem is solved by SAM. It is seen that mixed convec-
tion flow of nanofluid enhances the heat transfer characteristics, heat source/ sink
improves the temperature distribution in the nanofluid.
References
1. Crane LJ (1970) Flow past a stretching plate. Z Angew Math Phys 21:645–647
2. Gupta PS, Gupta AS (1977) Heat and mass transfer on a stretching sheet with suction or
blowing. Can J Che Eng 55(6):744–746
3. Borkakoti AK, Bharali A (1983) Hydromagnetic flow and heat transfer between two horizontal
plates, the lower plate being stretching sheet. Q Appl Math 41:461–467
4. Chen CK, Char MI (1988) Heat transfer of a continuous, stretching surface with suction or
blowing. J Math Ana App 135:568–580
5. Vajravelu K (1994) Convection heat transfer at a stretching sheet with suction or blowing. J
Math Anal Appl 188:1002–1011
6. Vajravelu K, Kumar BVR (2004) Analytical and numerical solutions of a coupled non-linear
system arising in a three-dimensional rotating flow. Int J Non-Linear Mech 39:13–24
7. Choi SUS, Eastman JA (1995) Enhancing thermal conductivity of fluids with nanoparticles.
ASME Int Mech Eng Con Exp 231:99–105
8. Sheikholeslami M, Ashorynejad HR, Domairry G, Hashim I (2012) Flow and heat transfer of
Cu-water nanofluid between a stretching sheet and a porous surface in a rotating system. J Appl
Math 2012:1–18
9. Buongiorno J (2006) Convective transport in nanofluids. Trans ASME 128:240–250
10. Sheikholeslami M, Ganji DD (2014) Numerical investigation for two-phase modeling of
nanofluid in a rotating system with permeable sheet. J Mol Liquids 194:13–19
11. Usha R, Vasudevan S (1993) A similar flow between two rotating disks in the presence of a
magnetic field. J Appl Phys 60:707–714
12. Usha R, Vimala P (1998) Maghetohydrodynamic disk braking. Z Angew Math Mech 4:283–288
RANS Simulations of Flow Past
a NACA0012 Airfoil
Deepak Kumar Singh, Rahul Prajapati, and Arjun Sharma
1 Introduction
Flow past a NACA0012 airfoil at moderate Reynolds number is studied using steady
RANS simulations. The chord-based Reynolds number is 5 × 104 , and the angle of
attack is 5°. The Mach number based on free-stream conditions is 0.4. The simula-
tion results are compared to the reference studies [1, 2] based on direct numerical
simulations done at the same flow configuration. The objective of the comparison
is to understand the accuracy with which certain flow features are predicted in the
present RANS simulations.
At the flow configuration mentioned above, the laminar boundary layer starts to
separate near the leading edge on suction side due to strong adverse pressure gradient.
The separated shear layer undergoes transition and reattaches to the airfoil further
downstream. A recirculation region is formed between the points of separation and
reattachment referred to as laminar separation bubble. The laminar flow separation
near the leading edge depends strongly on the Reynolds number and the angle of
attack.
The direct numerical simulations presented in the study of Jones et al. [2] and
Balakumar [1] are at the same flow configuration described before. In the study
of Balakumar [1], an additional simulation at a higher Reynolds number of 106 ,
Mach number of 0.2 and angle of attack 15◦ is presented. The observation is that the
separation bubble becomes smaller and shifted upstream closer to the leading edge.
The RANS simulation results in the present study indicate that it is possible to
obtain flow solutions with features associated with the laminar separation bubble
in good agreement with the detailed reference studies. The RANS model used in
D. K. Singh (B) · R. Prajapati · A. Sharma

Department of Applied Mechanics, Indian Institute of Technology, Delhi, India
e-mail: deepak@am.iitd.ac.in
A. Sharma
e-mail: arjunsharma@am.iitd.ac.in
326 D. K. Singh et al.
the present study is the one equation turbulence model of Spalart and Allmaras
[3]. Rather than focussing on inter-comparison of more detailed turbulence models
with comparatively higher computational cost, the objective is to understand the
sensitivity of the flow solutions to the parameters in the model as well as the numer-
ical methods. This will facilitate future studies which involve more complicated
geometrical configurations.
2 Methodology
The flow around a standard NACA0012 airfoil is studied using steady-state Reynolds-
averaged-Navier–Stokes (RANS) simulations. The Reynolds number based on the
chord length of the airfoil and free-stream velocity is Re = U∞ c/υ = 5 × 104 . The
Mach number based on the free-stream conditions is Ma = 0.4. The angle of attack
is α = 5◦ . No slip and adiabatic flow conditions are specified at the wall.
The simulations are conducted using a finite-volume-based commercial solver,
FLUENT. The RANS simulations are done using the Spalart–Allmaras (SA) turbu-
lence model [3, 4]. In SA model, an equation for the modified turbulent viscosity, υ̃,
is solved in addition to the mean flow equations.
The right hand side in the equation for υ̃ has production, dissipation and nonlinear
diffusion terms. The notation used for various model parameters is the same as used
in the reference paper [3]. The parameters (σ, Cv1 , Cw3 ) are independently varied in
the present study. Once σ is specified the model parameters, (Cb1 , Cb2 , Cw1 , Cw2 )
are fixed from the calibration guidelines provided in the reference paper [3]. The
destruction term coefficient Cw1 is related to (Cb1 , Cb2 , σ ) by a relation which follows
by imposing the constraints from the classical log layer in a flat plate boundary layer
[3].
The working fluid used is ideal air from the FLUENT database [5]. A steady,
pressure-based type and two-dimensional solver setting is used. The computational
domain extends to 12 chord lengths in the upstream and 16 chord lengths down-
stream directions from the leading edge. A multi-block structured O-grid is used.
The boundary conditions used in the simulations are shown in Fig. 1. At the wall and
in the free stream, υ̃ is specified as zero.
The grids used for grid convergence study are listed in Table 1. The number of
grid points in the inner block around the airfoil and the wake region is progres-
sively increased in grids 1–5 to examine the effects of grid refinement. Based on the
maximum skin friction coefficient given in the reference study [1, 5], C f,max = 0.1,
the friction velocity and the thickness of the viscous sublayer are calculated as
uτ
U∞
= 0.22, δcv = 8.6 × 10−5 , respectively. The values of minimum grid size in
‘ + ’ wall units listed in the table are based on the value of δv calculated above.
The effect of grid refinement in the wake region is examined by comparing results
from grids 1–4. The approximate locations of the peak turbulent viscosity ratio in the
streamwise direction are traced, and a straight line fit to these locations is obtained.
This survey line joins the points (c, 0) to (16c, 1.31c). The distribution of turbulent
RANS Simulations of Flow Past a NACA0012 Airfoil 327
Fig. 1 Computational domain with boundary conditions and grid distribution near leading and
trailing edge of airfoil is shown
Table 1 Details of the grids used in the grid convergence study are given
Grid points in the (y)+
min Wake block Trailing edge Leading edge
inner block (x)+ min (x)+
min
Grid 1 400 × 100 1.9 50 × 50 0.7 12.1
Grid 2 550 × 150 1.9 50 × 75 0.5 8.0
Grid 3 800 × 200 0.7 100 × 100 0.4 5.9
Grid 4 800 × 200 0.7 200 × 200 0.2 2.9
Grid 5 950 × 150 0.7 100 × 100 0.3 4.7
viscosity ratio along the survey line is shown in Fig. 2. No changes in distribution
of turbulent viscosity ratio along the survey line are observed when grids 3 and
4 are compared. The surface pressure and skin friction coefficients obtained using
grids 1, 3 and 5 are compared in Figs. 3a, b. No changes are observed as the grid is
progressively refined from 1 to 5. The results from grid 3 are shown in the remainder
of this study.
Flow features observed in the baseline simulations are described. The boundary layer
on the lower surface is observed to remain laminar. On the upper surface, the boundary
layer is observed to separate at xc ≈ 0.12 (see Figs. 4 and 6). The separated shear
layer undergoes rapid transition and reattaches to the airfoil surface at xc ≈ 0.55 (see
Figs. 4 and 6). The separation and reattachment points are identified as the locations
where the wall shear stress is zero.
25
Grid_1
Grid_2
Turbulent Viscosty Ratio
20
Grid_3
15 Grid_4
10
0
0 2 4 6 8 10 12 14 16
X/c
Fig. 2 Turbulent viscosity ratio in wake region along survey line
Fig. 3 a Pressure coefficient and b Skin friction coefficient distribution over airfoil
A zoomed view of the flow on the upper airfoil surface around the separation
bubble is shown in Fig. 4. The streamlines colored with velocity magnitude and line
contours of turbulent viscosity ratio, vt /v, are shown. After laminar separation, the
turbulent viscosity ratio increases rapidly indicating transition in the separated shear
layer. The location of peak turbulent viscosity ratio is observed to approximately trace
the initial separated shear layer. The separated shear layer grows as turbulence sets
in. Subsequently, the streamlines curve toward the airfoil surface, and reattachment
occurs. The growth in turbulent viscosity is much slower downstream of the reat-
tachment point. In Fig. 5, the streamlines colored with vorticity (normalized using
chord length and free-stream velocity) and line contours of Reynolds shear stress
(normalized using free-stream velocity) are shown. There is a region of high posi-
tive vorticity formed along the surface between the vortex core and the reattachment
Fig. 4 Streamlines colored with velocity magnitude and line contours of turbulent viscosity ratio
are shown
Fig. 5 Streamlines colored with normalized vorticity and line contours of Reynolds shear stress
are shown
point. This region contributes to a local peak in the skin friction coefficient in the
interior of the separation bubble (see Fig. 6). A local maximum of Reynolds shear
stress is also observed in the separated shear layer region, slightly away from the
wall, where the streamlines curve toward the airfoil surface. The local maximum
value of Reynolds shear stress is consistent with those in the reference study [1].
The mean surface pressure (C p ) and skin friction (C f ) coefficients are compared
to the results from the reference study [1] in
Fig.
6. The rescaled value of the magni-
tude of skin friction coefficient, 1.5 + 50C f , is shown. The coefficients, C p and
C f , obtained on the lower surface, where the flow is laminar, in the present study
are observed to agree well with the reference study [1]. On the upper surface, the
boundary layer separation point is slightly downstream compared to the reference
results [1]. The reattachment point as well as the local peak in skin friction coeffi-
cient is shifted upstream compared to the reference study [1]. Thus, the size of the
separation bubble is under predicted. The local peak value of C f observed inside the
Fig. 6 Comparison of mean surface pressure and skin friction coefficients with the reference study
[1] is shown
separation bubble is slightly lower compared to the reference study [1]. The C f values
in the region downstream of the reattachment point ( xc > 0.6) are observed to be
much lower than those in the reference study [1]. In the reattachment region, stream-
lines curve toward the airfoil surface, and C f values decrease from the peak value to
zero at the reattachment point. A slower rate of decrease of C f in the reattachment
region, xc = 0.4 − 0.6, is observed compared to the reference study [1].
A comparison of mean surface pressure coefficient in Fig. 6 on the lower surface
indicates good agreement between the results from the present simulations and the
reference study [1]. On the upper surface, C p is observed to remain approximately
constant inside the separation bubble. The pressure coefficient starts to increase in
the reattachment region. This location, xc ≈ 0.4, corresponds approximately to the
location of the local peak in skin friction coefficient in the separation bubble. The
differences in the pressure coefficient between the present and reference studies are
observed primarily in the reattachment region, xc = 0.4 − 0.6.
The total lift and drag coefficients obtained from the present simulations and
the reference study [1] are (5.8 × 10−1 ,2.6 × 10−2 ) and (6.3 × 10−1 ,3.5 × 10−2 ),
respectively. The differences in drag and lift coefficients are approximately 25% and
8%, respectively.
The local peak in C f in the separation bubble is associated with transition in the
separated shear layer, curving of streamlines toward the airfoil surface and subsequent
reattachment. The upstream location of peak skin friction in the present study relative
to that observed in the reference study [1] is associated with early growth of turbulence
in the separated shear layer. However, turbulence production is low, so that the process
of reattachment is slower and is indicated in a relatively slow decay of skin friction
from its peak value to zero (point of reattachment) (see Fig. 6). The turbulence levels
(Reynolds shear stress and turbulent viscosity) are low which contribute to lower
values of C f compared to the reference study [1].
The effects of model parameters on the flow features are studied next. Compar-
isons are made to the baseline simulation results in Fig. 7. The notation used is the
same as used in the reference paper [3]. Observing that the turbulence levels in the
region downstream of reattachment are low, the destruction term in the equation
for modified turbulent viscosity is reduced by lowering the values of the wall func-
tion, f w (r ), in the region r > 1. The observed effect is a desired increase in the
skin friction downstream of reattachment. But alongside, turbulence growth in the
separated shear layer is accelerated which results in a faster reattachment process.
Additional effects of curvature correction terms (related to streamline curvature in
the recirculation region) in the turbulence model are seen. The curvature correction
terms limit the production of turbulent viscosity which results in slower reattachment
of the separated shear layer (slower decay of skin friction) and lower values of skin
friction downstream. The diffusion coefficient (1/σ ) which multiplies the nonlinear
diffusion term is increased by lowering σ to 0.6 from its value of 0.67 in the stan-
dard model [3]. Although the other model coefficients are also linked to σ (from the
calibration guidelines in the paper [3]), the net effect of lowering σ is a slight delay
in the growth of turbulence (peak skin friction is shifted slightly downstream) and a
lowering of skin friction in the region downstream of reattachment point.
The net effect of increasing the ratio of production to destruction terms in the
equation for modified turbulent viscosity results in an increase in the skin friction
downstream of the reattachment point, faster turbulence growth in the separated
shear layer and a faster reattachment process. Thus, the comparison of skin friction
coefficient with the reference study [1] does not necessarily improve since the location
0.01
Baseline Baseline with sigma changed
Baseline with wall function changed Baseline with curvature correction
0.008
0.006
|Cf|
0.004
0.002
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
x/c
Fig. 7 Comparison of skin friction coefficients obtained from baseline simulation and with changed
model parameters
of peak C f and the reattachment point move further upstream while C f values
downstream of reattachment increase (see Fig. 6).
4 Conclusions
The flow features related to separation bubble with turbulent reattachment are
predicted well in the RANS simulations for the flow configuration presented in
this study. The presence of a separation bubble and the resulting flow downstream
strongly affect the lift and drag coefficients. The size of separation bubble is under
predicted (slightly downstream position of laminar separation and upstream position
of turbulent reattachment point). The displacement of streamlines that occurs due to
presence of the separation bubble alters the surface pressure distribution imposed by
the external inviscid flow. Both the effects, a reduction in peak suction pressure and
the presence of an extended region of constant pressure inside the separation bubble,
are correctly obtained in the simulation results. The mismatch with reference results
[1] is primarily in the slow but early growth of turbulence in the separated shear
layer and low turbulence levels downstream of the reattachment point. The former
is related to transition in the separated shear layer, whereas the latter results in lower
skin friction coefficient. Since this kind of flow configuration is not directly used in
calibrating the turbulence model [3], the effects of changes in the wall function and
model coefficients in the equation for turbulent viscosity are studied and considered
for future work.
Acknowledgements The authors acknowledge support from a research grant (ECR/2017/001770)

by the Science and Engineering Research Board, India.
References
1. Balakumar P (2017) Direct numerical simulation of flows over an NACA-0012 airfoil at low
and moderate Reynolds numbers. In: AIAA fluid dynamics conference, paper number AIAA-
2017-3978
2. Jones LE, Sandberg RD, Sandham ND (2008) Direct numerical simulation of forced and
unforced separation bubbles on an airfoil at incidence. J Fluid Mech 602:175–207
3. Spalart PR, Allmaras SR (1992) A one-equation turbulence model for aerodynamic flows. In:
AIAA paper number AIAA-92–0439, American Institute of Aeronautics and Astronautics
4. Allmaras SR, Johnson FT, Spalart PR (2012) Modifications and clarifications for the implemen-
tation of Spalart-Allmaras turbulence model. In: International Conference on Computational
Fluid Dynamics, Paper number ICCFD 7–1902
5. ANSYS FLUENT 18.0 Theory Guide, ANSYS, Inc., 2018
6. Spalart PR, Shur ML (1997) On the sensitization of turbulence models to rotation and curvature.
Aerospace Sci Tech 1(5):297–302
The Influence of Rotating Element
on Flow and Heat Transfer
in the Two-Dimensional Confined
Channel
J. Ramarajan, D. Sathish Kumar, Ratnanjali Tiwari, and S. Jayavel
1 Introduction
Flow over cylinder is a classical problem in fluid mechanics. Flow through channel
with inbuilt cylinder is the combination of external and internal flow problem, and its
application is widely found in many engineering fields, where heating and cooling
of fluids are involved. The two-dimensional unsteady flow over a stationary cylinder
causes the vortex shedding. On the other hand, wall confinement suppresses the
shedding. Many authors have studied the fluid flow and heat transfer characteristics of
flow over a cylinder with or without channel confinement. Rehimi et al. [1] conducted
experiment in water tunnel, and the flow behavior was captured in using PIV. In order
to analyze the flow instability in the presence of channel confinement, Reynolds
number was chosen in the range of 30 ≤ Re ≤ 277. Sahin et al. [2] have studied
numerically the flow over a stationary cylinder placed in a channel for a range of
Reynolds number 0 < Re ≤ 280. The authors analyzed the vortex shedding frequency,
hydrodynamic forces and wake behind the cylinder, and also, their study included
the effect of channel confinement. Similar studies were carried out by Chakaraborty
et al. [3] and Zovatto et al. [4].
The present study is focussed on the effect of rotating elements on flow and
heat transfer characteristics. The rotating element is confined within a channel. The
resulting flow characteristics are complex and interesting to analyze. Over a decade,
many authors have analyzed the fluid flow and heat transfer characteristics of flow
over rotating cylinder. Lam et al. [5] experimentally investigated the vortex shedding
behind rotating cylinder. Flow visualization and PIV measurements were carried out
J. Ramarajan · D. S. Kumar · R. Tiwari (B) · S. Jayavel

Department of Mechanical Engineering, IIITDM Kancheepuram, Melakottaiyur, Chennai
600127, India
e-mail: mpd15i019@iiitdm.ac.in
334 J. Ramarajan et al.
for the range of Reynolds number 3600 ≤ Re ≤ 5000. They found that the vortex shed-
ding frequency increases with cylinder rotational speeds. Meanwhile with increasing
cylinder rotational speeds, the width of the wake was observed to decrease, and vortex
formation length also decreases from 2.5 times of diameter of cylinder (stationary
cylinder) to 0.65 times of cylinder diameter (rotating cylinder). Kimura et al. [6]
showed the effect of cylinder rotation on vortex street. They increased spin param-
eter from 0 to 2, which is defined as the ratio of peripheral speed of cylinder surface
to uniform inlet velocity. The authors used hydrogen bubble method for flow visual-
ization, and for numerical analysis, they used discrete vortex method. The Re range,
200 ≤ Re ≤ 10,000, considered for both experimental and numerical studies. The
conclusion made from their analysis was that the spin parameter increases the shed-
ding frequency of vortices. At higher Re, the Strouhal number value becomes large,
and critical spin parameter at which the meandering effect vanishes becomes small.
Prasad et al. [7] carried two-dimensional numerical analysis for a range of Re, 35
≤ Re ≤ 177 to compare the effect of stationary and rotating cylinder confined in a
channel. Blockage ratio (ratio between cylinder diameter and distance between the
walls) was varied from 0 to 50%. Negative lift force was produced due to coun-
terclockwise rotation, which increases with increasing blockage ratio. The author
analyzed the vortex strength and vortex trajectory for various rotational velocity
of the cylinder. The cylinder was maintained at constant wall temperature, and air
is considered as working fluid. The authors concluded that with the influence of
channel confinement, the wake region and width of the shear layer decrease with
increasing blockage ratio for both stationary and rotating cylinder. Kang et al. [8]
investigated numerically two-dimensional flow past rotating cylinder to understand
the flow mechanism and also to control the vortex shedding. The numerical study was
performed for Re = 40, 60, 100 and 160 various rotating speed of the cylinder. Their
results showed that the vortex shedding suppression was effective due to rotation of
cylinder and also showed that the critical rotational speed depends on Re. As rotation
speed increases, the amplitude of lift fluctuation was found to be nearly constant, but
the drag fluctuation increased linearly.
Many previous studies focussed on flow past rotating circular cylinder. In the
present numerical study, the effect of different rotating domain in a channel flow
on heat transfer and fluid flow characteristics are studied. Three different rotating
elements (solid cylinder (SC), two straight blades (SB), and the third type (CB) is
combination of the first two types comprising two blades projecting from a small
cylinder) are chosen to compare the flow behavior and heat transfer characteristics.
2 Problem Statement
The two-dimensional flow through rectangular channel with working fluid air is
considered in this numerical study. The aim is to analyze the flow and heat transfer
characteristics due to rotating element present in a channel. The effect of rotating
element and rotational speed on wake, vortex shedding and non-dimensional Strouhal
The Influence of Rotating Element on Flow and Heat Transfer … 335
Fig. 1 Two-dimensional computational domain
number is studied. The schematic representation of the channel flow and location of
rotating element is shown in Fig. 1.
Flow direction is along x-axis (channel length). There are three different cases,
(i) solid cylinder (SC), (ii) two straight blades (SB), and (iii) the third type (CB) is
combination of the first two types comprising two blades projecting from a small
cylinder, are considered, and the enlarged view of three different rotating elements
is shown in Fig. 2. The flow Reynolds numbers considered are 100, 300 and 600.
The flow is assumed to be laminar, unsteady and incompressible. The rotational axis
of the rotating element with diameter, D, is placed at the center of the channel along
y-axis. The upstream and downstream lengths of the domain with respect to rotating
element axis are 12D and 18D, respectively. The numerical study
is carried out for
three different non-dimensional rotational speeds, α = ω R u = 1, 2 and 3. It is
taken as positive for counterclockwise rotation. All these cases are compared with
stationary elements to know the effect of rotating element in the channel flow.
Fig. 2 Types of rotating elements, a solid cylinder (SC), b two straight blades projecting from a
small cylinder (CB) and c two straight blades (SB)
The continuity and Navier–Stokes equations are the governing equations and are
expressed in the following general convection–diffusion-source integral form (Eqs. 1
and 2).

∂
ρdV + ρ u · d S = 0 (1)
∂t
V S
where d S is the differential area vector of the surface directed along normal at
the given point on its surface. Equation (1) is the mass conservation equation, which
can be deduced from the general equation given in Eq. (2).

∂
ρφdV + ρ uφ · d S = φ ∇φ · d S + Sφ dV (2)
∂t
V S S V
where and S are, respectively, the general transpo φ, rt variable, diffusion

coefficient, and source term. The equations for mass, momentum and energy conser-
vation are obtained by assigning appropriate values to the variable φ, and S as
presented in Table 1.
The boundaries of the computational domain are top and bottom no-slip stationary
walls, fully developed flow inlet and pressure outlet. Fully developed flow is spec-
ified at the inlet with uniform inlet temperature of 25 °C, and the average velocity
values correspond to the respective Re. Outlet condition is given as outflow. No-slip
boundary condition has been imposed on the cylinder and impeller walls. The top
and bottom boundaries are assigned as constant temperature wall with 70 °C. Inter-
face boundary condition has been created between contact zones of stationary and
rotating domains. The dimensionless speed values (α) are imposed to the moving
mesh, which accommodates the rotating elements.
The computations are carried out using finite volume-based ANSYS Fluent soft-
ware. SIMPLE algorithm is used to solve pressure–velocity coupling. Second-order
upwind scheme is used for solving momentum and energy equations. Unsteady
simulations are carried out with the time step size of 0.005 s.
The present numerical results are compared with Prasad et al. [7]. Optimum
number of elements is chosen for calculating the results. The variation of averaged
Nusselt number with non-dimensional velocity is shown in Fig. 3. The computational
Table 1 Variables for

Equation φ φ Sφ
general transport equation
Continuity 1 0 0
∂p ∂p
Momentum u, v μ ∂x , ∂y
Energy T k 0
5
Average Nu
3 Present Results at Re. 100

Prasad et al (2011) Re.100
2 Re. 70
Re. 70
1 Re. 40
Re. 40
0
0 1 2
Non dimensional rotational velocity
Fig. 3 Effect of rotational velocity on average Nu for solid cylinder (SC)
results match well with the results available in the literature at Re = 40, 70 and 100.
At α = 2, for Re = 40, acceptable deviation is observed.
This section presents the flow and heat transfer characteristics in the presence of
rotating element in two-dimensional channel. The results are time-averaged values
for ten complete revolutions of rotating member. Figure 4 shows the effect of rotating
element on heat transfer coefficient (h) for the range of Re, 100 ≤ Re ≤ 600. The
top wall ‘h’ value increases with Re, and at low Re, the values are almost same
irrespective of the type of rotating element present in the channel. Meanwhile, as
Re increases, the straight blade impeller case (SB) shows good performance than
the other two cases. For the computed range of Re, the performance of the rotating
elements is better compared to stationary cylinder. Among the rotating elements,
straight blade impeller case (SB) performs better. Figure 5 shows the variation of
‘h’ in both top and bottom plates. Both Figs. 4 and 5 are plotted for particular non-
dimensional velocity α = 3. The counterclockwise motion of the rotating element
enhances heat transfer in the upper region.
Figure 5 shows the effect of combined cylinder and blade (CB) rotating element.
It is clearly showing more heat transfer coefficient over top plate than bottom for all
computed Re values. Similar trend is observed for solid cylinder (SC) and straight
blades (SB) cases also. The effect of non-dimensional velocity α on ‘h’ is shown
in Fig. 6. Re = 300 is chosen to know the variation among the rotating element.
5.5 SC CB SB
4.5
h (W/m2K)
3.5
2.5
2
100 200 300 400 500 600
Re
Fig. 4 Effect of Re on average heat transfer coefficient for various cases (SC, SB and CB)
6
5.5
5
h (W/m2K)
4.5
4
3.5
Top plate
3
Bottom Plate
2.5
2
100 200 300 400 500 600
Re
Fig. 5 Effect of Re on average heat transfer coefficient for CB case
When rotating elements become stationary, the h values are lesser as expected.
As α increases, the heat transfer enhancement becomes favorable for all range of
computed Re values. The straight blade (SB) element shows more favorable heat
transfer performance than other two cases.
3.5
h (W/m2K)
3
SC
CB
2.5
SB
2
0 1 2 3
Non dimensional rotational velocity
Fig. 6 Effect of ‘α’ on ‘h’
The vortex shedding behavior has also been analyzed for all rotating domain for
all Re values. The shedding frequency has been calculated for all dimensionless
rotation velocities. The effect of Re on non-dimensional vortex shedding frequency,
Strouhal number St = f D u at α = 2, is shown in Fig. 7. The linear variation is
observed in solid cylinder (SC) case, but this kind of trend is not observed in other
two cases.
4 Conclusions
The effect of rotation on flow over different rotating element is studied for non-
dimensional velocity (α = 0, 1, 2, 3). The flow behavior between stationary and
rotating element has been analyzed. Non-dimensional shedding frequency and heat
transfer characteristics with respect to top and bottom plates are presented as time-
averaged quantities. The rotating straight blade (SB) shows good heat transfer perfor-
mance for all Re values (100, 300 and 600) than other two cases (SC and CB). The
top plate shows higher heat transfer coefficient than bottom plate due to counter-
clockwise rotation of elements. The non-dimensional St values are calculated for all
Re solid cylinder (SC) case that shows linear variation.
0.3
0.25
0.2
0.15
St
0.1
0.05
SC CB SB
0
0 200 400 600 800
Re
Fig. 7 Effect of Re on St
5 Application and Scope
The applications include various domains where existence of fluid surroundings

occurs. From fluid mechanical point of view, underwater, aerodynamic experiments
and vehicle design for scientific exploration require enough theoretical parameters to
design in order to avoid failure. From thermal point of view, electronic elements can
implement the heat transfer caused due to the presence of bluff bodies to facilitate
cooling.
References
1. Rehimi F, Aloui F, Nasrallah SB, Doubliez L, Legran J (2008) Experimental investigation of a

confined flow downstream of a circular cylinder centered between two parallel walls. J Fluids
Struct 24:855–882
2. Sahin M, Owens RG (2004) A numerical investigation of wall effects up to high blockage ratios
on two-dimensional on flow past a confined circular cylinder. Phys Fluids 16:1305–1320
3. Chakraborty J, Verma N, Chhabra RP (2004) Wall effects in flow past a circular cylinder in a
plane channel a numerical study. Chem Eng Process 43:1529–1537
4. Zovatto L, Pedrizzeti G (2001) Flow about a circular cylinder between parallel walls. J Fluid
Mech 440:1–25
5. Lam K (2009) Vortex shedding flow behind a slowly rotating circular cylinder. J Fluid Struct
25:245–262
6. Kimura T, Tsutahara M, Wang ZY (1992) Wake of a rotating circular cylinder. AIAA J 30(2):555–
556
7. Prasad K, Paramane SB, Agrawal A, Sharma A (2011) Effect of channel-confinement and
rotation on the two-dimensional laminar flow and heat transfer across a cylinder. Numerical
Heat Transf Part A Appl 60(8):699–726
8. Kang S, Choi H (1999) Laminar flow past a rotating circular cylinder. Phys Fluids 11:3312–3321
Hydrodynamic Effect of Tsunami Wave
on Oscillating Water Column (OWC)
Type Wave Energy Converter (WEC)
Rujal D. Patel, Sagar G. Nayak, and Jyotirmay Banerjee
1 Introduction
In twentieth century, mankind has seen the Second Industrial Revolution. The require-
ment for energy grew exponentially, which leads to the exploitation of non-renewable
natural resources. Overuse of those natural resources turns out to be very harmful
to the environment. As the time passed, the effect of this misuse became visible on
the environment. The requirement of renewable and clean energy grew to lessen the
damage.
The ocean waves are an important source of renewable energy. Worldwide 93,000
TWh/year of energy is available in the form of ocean waves. As water has 800 times
more density than air, it can carry way more energy in the wave per unit volume
of fluid. So, research on extracting energy from ocean waves has made tremendous
progress. These devices are generally known as wave energy converters (WEC) [1].
Among these, the OWC type WEC is widely accepted because of its robust
construction and minimal impact on marine creatures. The operation of these type
of WEC is to keep an air-trapped pocket. The motion of the wave compresses the
trapped air like a piston. The generated pressure is utilized to power an air turbine.
When the wave moves downward, the suction generates. The reverse flow of air
takes place which again powers the air turbine. It is required for an air turbine to be
a bidirectional to get continuous motion of turbine.
The ocean provides very adverse condition. There were many studies carried out
to analyze the hydrodynamic effect of the regular waves on the OWC [2–4], but the
effect of adverse condition like tsunami waves on OWC is still an unexplored area
of study.
R. D. Patel · S. G. Nayak (B) · J. Banerjee

J. Banerjee
344 R. D. Patel et al.
Tsunami is a Japanese word which means the harbor wave. Tsunami is basically
a pressure wave. The waves which are generated by the displacement of large body
of water are known as tsunami wave.
The displacement of large water body may occur due to volcanic eruption, land-
slide, earthquake, meteorite impacts or any other type of disturbances. The most
common cause for the tsunami is the submarine earthquake [5]. Sudden motion of
tectonic plate displaces the water which generates the tsunami waves.
There are many approaches to analyze the hydrodynamic effect of the tsunami
waves on the OWC. For the interaction of tsunami wave with OWC, the nonlinear
approach is adopted [6].
Continuity and momentum equations in integral form are written as follows:

¨
−
→ − →
V · dA = 0cs (1)
˚ ¨ − ¨
∂ V → 1 →
−
d∀ + V V · dA = μ ∇ V · dA (2)
∂t ρ
C∀ CS CS
In the numerical simulation, the volume of fluid (VOF) method is employed for
tracing or capturing the free surface [7]. The governing equation is as follows:
∂C −
+→
v · ∇C = 0 (3)
∂t
C=1 ⇒cell is filled with primary fluid (full cell); 0 < C < 1 ⇒ interface is present
in the cell (mixed cell); C = 0 ⇒ cell is filled with secondary fluid (empty cell).
2.2 Tsunami Generation
The ocean floor motion displaces the water above the moving floor, and this displaced
water will generate the tsunami wave. The ocean floor can have many types of
motion for ocean floor uplifting such as linear, parabolic, exponential, sinusoidal
and fluctuating. For validation, two types of uplifting of ocean floor are used. Those
are as follows (Fig. 1):
(1) Half-sine floor motion
Hydrodynamic Effect of Tsunami Wave on Oscillating Water Column (OWC) … 345
Fig. 1 Floor motion a sinusoidal b exponential
(2) Exponential floor motion

The floor is lifted by the following equation which gives the position of floor at
given time.

πt
ξ(x, t) = ξ0 0.5 1 − cos × H (T − t) + H (T − t) × H (b − x) (4)
T

ξ (x, t) = ξ0 1 − e−t/tc × H (b − x) (5)
Here, the H(x) is the heavy-side function which is used to confine the motion
of the moving floor for required region and time interval. The implementation of
moving floor for tsunami generation can be given by Eqs. (4) and (5).
2.3 Domain Setup and Boundary Condition
The 15 m × 0.1 m domain is taken for the simulation with grid size of 6000 × 200.
The primary fluid is water with a density of 1000 kg/m3 and viscosity of 1.137×
10−3 kg/ms. The secondary fluid is air with a density of 1.226 kg/m3 and viscosity
of 1.78×10−5 kg/ms. The top boundary is taken as pressure outlet, and all other
boundaries are taken as no-slip wall. The moving boundary is considered for the
floor elevation. Ambient water depth of h = 0.05 m and the floor of size b = 0.61 m
from left side of the domain will move upward.
Fig. 2 Wave generated by sinusoidal floor motion
3.1 Validation of Tsunami Generation and Propagation
For validation, the calculation of wave elevation is carried out at 9 m away from
the generation of the tsunami waves. The numerical results are compared with the
experimental results [8]. The validation is carried for both sinusoidal and exponential
motion of floor elevation.
For sinusoidal floor elevation, the maximum floor rise is 0.01 m, and for expo-
nential floor motion 0.005m of maximum floor rise is considered. The time required
for sinusoidal floor elevation is 0.129 s, and for exponential floor elevation is 1.148
s.
From Figs. 2 and 3, it can be concluded that there is a good agreement between
the experiment and numerical results. It means the development of two-phase
model and implementation of moving boundary condition for tsunami generation
are successfully carried out.
3.2 Effect of Tsunami Waves on OWC
To analyze the effect of tsunami waves on OWC, the OWC is placed into the domain.
The tsunami is generated, and the hydrodynamic effect of the tsunami waves is studied
(Fig. 4).
The tsunami waves are generated, and it collides with the OWC at 3 s. When the
tsunami wave has impact on OWC, all the hydrodynamic pressure is converted into
Fig. 3 Wave generated by exponential floor motion
Fig. 4 Pressure contour near the OWC
the static pressure. The pressure contour and the streamlines at the OWC are shown
in Fig. 5.
When the tsunami wave has impact on the OWC, there would be a large pressure
difference in the rear wall of the OWC. This pressure difference generates the net
force on the rear wall which is shown in Fig. 6.
As shown in Fig. 5, the force applied on the walls of OWC is maximum at the rear
wall when the wave collides with the OWC. The net force applied on the walls of the
rear wall of the OWC is uniform. The OWC is mounted at the base of the rear wall,
so the net force applied on the walls of the OWC is the parameter which causes the
Fig. 5 Force applied on the walls of OWC
failure of the OWC. The moment is a driving parameter which leads to the failure of
the OWC. The moment generated due to this force is shown in Fig. 6.
The moment generated at the rear wall is maximum at the 3 s when the wave
collides with the OWC. The moment at the top portion of the OWC is higher due to
greater distance from the base of the OWC. The high moment generated at the top
portion of the rear wall is the cause of the failure of the OWC.
3.3 Improvement in Design of OWC
To reduce the moment, an improvement in the design of the OWC is proposed. The
pressure release valve is included to reduce the pressure difference in the rear wall.
The position of the pressure release valve should be as high as possible to minimize
the moment. The position of the pressure release valve chosen such that the front
wall of the OWC should not obstruct the flow.
To analyze the effectiveness of the pressure release valve, the simulation has been
carried on the same domain as the OWC without pressure release value.
Fig. 6 Moment generated on the walls of OWC
• The moment generated at the top portion of the rear wall is the driving parameter
which will lead to failure of the OWC.
• The inclusion of pressure release valve significantly reduces the net forces and
the generated moment at the walls of the OWC.
As shown in Figs. 7 and 8, the pressure release valve produces jet behind the OWC
so releases some of the pressure and reduces the forces as well as moment on the
rear wall of the OWC. The results are compared when the forces and the moment are
applied at the rear wall. The forces applied at the rear wall of the OWC are compared
at 3 s as shown in Fig. 9.
Significant reduction in the net force applied at the rear wall of the OWC can be
observed. Also, it can be observed that there is a reduction in the net force near the
pressure release valve. The effect of pressure release valve on moment is shown in
Fig. 9.
As shown in Fig. 10, there is a significant reduction in the moment generated at
rear wall of the OWC.
Fig. 7 Vorticity contour and streamline for OWC with pressure release valve
Fig. 8 Velocity contour and streamline for OWC with pressure release valve
Fig. 9 Comparison of the forces applied on rear wall of OWC with and without pressure release
valve
Fig. 10 Comparison of generated moment on rear wall of OWC with and without pressure release
valve
4 Conclusions
Two-phase model is developed to generate tsunami generation and propagation.

• The net force applied on OWC is maximum when the tsunami wave collides, and
the net force is uniform throughout the entire rear wall.
References
1. Falcão AFO, Henriques JCC (2016) Oscillating-water-column wave energy converters and air
turbines: a review. Renew Energy 85:1391–1424
2. Morris-Thomas MT, Irvin RJ, Thiagarajan KP (2007) An investigation into the hydrodynamic
efficiency of an oscillating water column. J Offshore Mech Arct Eng 129(4):273
3. Mahnamfar F, Altunkaynak A (2017) Comparison of numerical and experimental analyses for
optimizing the geometry of OWC systems. Ocean Eng 130:10–24
4. Kuo Y-S, Chung C-Y, Hsiao S-C, Wang Y-K (2017) Hydrodynamic characteristics of oscillating
water column caisson breakwaters. Renew Energy 103:439–447
5. Todorovska MI, Trifunac MD (2001) Generation of tsunamis by a slowly spreading uplift of the
sea¯oor. Soil Dyn Earthq Eng 17
6. Qi M, Kuai Y, Li J (2017) Numerical simulation of water waves generated by seabed movement.
Appl Ocean Res 65:302–314
7. Puckett EG, Almgren AS, Bell JB, Marcus DL, Rider WJ (1997) A high-order projection method
for tracking fluid interfaces in variable density incompressible flows. J Comput Phys 130(2):269–
282
8. Hammack JL (1973) A note on tsunamis: their generation and propagation in an ocean of uniform
depth. J Fluid Mech 60(4):769–799
Effect of Orifice Lip Geometry
on the Mean Flow Properties of a Plane
Wall Jet
S. V. H. Nagendra, D. V. S. Bhagavanulu, and Prasant Nanda
1 Introduction
When fluid is blown over a surface issued from an opening which can be an orifice
or a nozzle, the profile formed is known as wall jet. When a jet of air strikes a surface
at right angles and spreads out radially over it, it forms a wall jet [1]. Wall jet can be
classified as plane wall jet, radial wall jet, impinging jet, etc. The schematic diagram
of the wall jet is shown in Fig. 1. The major application of wall jet is like heating and
cooling of surfaces, exhaust of rocket nozzle, etc. In a wall jet, the most important
aspect is the variation of length scales in the longitudinal and spanwise directions
and decay of maximum velocity. It has been found that irrespective of inlet geometry
of the orifice, the spanwise growth of half width is 4 to 5 times higher than the
growth in the longitudinal direction [2]. The development of the flow in the potential
core region does depend on the geometry configuration, and it also has effect on the
formation of characteristic decay region. But in the radial decay region, the shape
of the orifice does not have much effect [3]. Further, many authors have investigated
on plane wall jet [1, 4–7]. They worked with different geometries of orifice like
slot-type orifice, nozzle type, sharp-edged orifice and diffuser, etc. Few authors have
compared turbulence model for resolving the flow dynamics of the wall jet [8–11].
It is understood from their studies that the Reynolds stress closure model is a better
choice for modeling the wall jet flow. Very little investigation has been done on
understanding the effect of different lip geometry on the mean flow properties [12].
They have investigated on the effect of orifice lip ratio on the generation of vortex
rings and their magnitude. The second-order model, Reynolds stress equations [13],
gave good results at wall when pressure strain relation was used. Results of RSM
S. V. H. Nagendra (B) · P. Nanda

VSSUT, Burla, Odisha 768018, India
D. V. S. Bhagavanulu
SVET, Tirupati 517102, India
354 S. V. H. Nagendra et al.
Fig. 1 Plane wall jet configuration
model give higher spread rate when compared to experimental results which is about
20% [14].
In the current investigation, different lip geometry of orifice is considered. The
orifice shape considered is square, and three different lip conditions, namely diffuser
type with 45 degree inclination, slot type and converging type (nozzle) are taken
and are shown in Fig. 2. In numerical modeling, a control volume is extracted for
all the three geometries of size 1600 mm × 500 mm × 1000 mm in x, y and z
directions, respectively. The geometry of orifice is shown in Fig. 2. The discritization
is done using tetrahedra structure mesh with 15 numbers of inflation layer based on
the exit Reynolds number which is 67,009. The Reynolds stress closure model is
implemented in the Fluent software with convergence criteria of 1e-4. The flow
properties are mean velocity profiles in the longitudinal and spanwise directions,
maximum velocity decay, turbulence kinetic energy and growth of half with in the
longitudinal and spanwise directions.
Effect of Orifice Lip Geometry on the Mean Flow … 355
Fig. 2 Different lip geometry of orifice
The numerical results are plotted as scattered graphs. For correlation, experimental
data is referred from the existing literature. The number of data in numerical results
is considered for 300 points at each station, i.e., 1 to 10 h, 12, 14, 16, 18, 20, 25, 30
and 40 h and are calculated.
2 Mean Velocity Profiles
The mean velocity profiles are drawn between U/U m versus y/b and Z/Z m/2 versus
U/U m where y and Z are directional axis, and b and Z m/2 are the point of U m /2 (half
of the maximum velocity) in the longitudinal direction and spanwise directions,
respectively. Figures 3, 4 and 5 show the mean velocity profiles in longitudinal
direction, and it is observed that the profile well correlated with the Glauert profile.
Figure 6 shows the comparison of mean velocity profile at 30 h for all the three types
of orifice considered in the present investigation. It is clearly seen that there is no
difference of shape of the profiles of wall jet. There is very little change in the outer
region after y/b = 1.4 which is due to over prediction growth in RSM model. The
2.0
diffuser 5 h
10 h
20 h
30 h
1.5 40 h
Glauert
y/b
1.0
0.5
0.0
0.0 0.2 0.4 0.6 0.8 1.0
U/Umax
Fig. 3 Mean velocity profiles in longitudinal direction for diffuser-type orifice
2.0
slot 5h
10 h
20 h
30 h
1.5
40 h
Glauert
y/b
1.0
0.5
0.0
0.0 0.2 0.4 0.6 0.8 1.0
U/Umax
Fig. 4 Mean velocity profiles in longitudinal direction for slot-type orifice
location of maximum velocity for all the three orifices is varying between y/b = 0.14
and y/b = 0.15 as the point of maximum velocity on the scale of y/b is typically
about 0.13 which is same for two-dimensional as well as three-dimensional wall jets
[15]. Figure 7 shows the mean velocity profiles at 30 h in the spanwise direction for
2.0
Nozzle 5 h
10 h
20 h
30 h
1.5 40 h
Glauert
y/b
1.0
0.5
0.0
0.0 0.2 0.4 0.6 0.8 1.0
U/Umax
Fig. 5 Mean velocity profiles in longitudinal direction for nozzle-type orifice
3.0
Diffuser 30 h
Slot 30 h
2.5 Nozzle 30 h
Glauert
2.0
y/b
1.5
1.0
0.5
0.0
0.0 0.2 0.4 0.6 0.8 1.0
U/Umax
Fig. 6 Mean velocity profile at 30 h for all the three orifice geometries
the three geometries chosen in the present investigation and compared well with the
free jet profile.
1.0 Diffuser CFD

Slot CFD
Nozzle CFD
U/U max 0.8 Free jet
0.6
0.4
0.2
0.0
-4 -2 0 2 4
Z/Z m/2
Fig. 7 Mean velocity profiles for all the three orifice geometries at 30 h in the spanwise direction
3 Growth of Half Width
Half width growth is plotted between x/h versus b/h. They numerical results are
compared with previous literature. Figures 8 and 9 show the half width growth in
longitudinal and spanwise direction. Using the linear fit, the value of constant for the
5
Diffuser CFD
Slot CFD
Nozzle CFD
Diffuser EXP Nagendra SVH 2018
4 Slot EXP Hall 2007
Nozzle EXP Poole 2014
3
b/h
0
0 10 20 30 40 50 60
x/h
Fig. 8 Growth of half width in longitudinal direction for all the three orifice geometries
4
Z/Zm/2
2 Diffuser CFD
Slot CFD
Nozzle CFD
1 Diffuser EXP Nagendra 2018
Slot EXP Hall 2007
0
0 10 20 30 40 50
x/h
Fig. 9 Growth of half width in spanwise direction for all the three orifice geometries
Table 1 Growth of half width

Type of Orifice lip b/h Z/Z m/2 b/h/Z/Z m/2
Diffuser 0.067 0.24 3.61
Slot 0.061 0.22 3.72
Nozzle 0.065 0.28 4.81
Diffuser (Nagendra 2018) 0.045 0.21 4.6
Slot( Hall 2007) 0.05 0.28 5.6
Nozzle (Poole 2014) 0.046 0.26 5.65
different lip geometries is shown in Table 1. The longitudinal half width is ranging
from 0.61 to 0.67 for all the three lip geometries. The spanwise spread is ranging from
0.22 to 0.28. The results show that the numerical results are higher in longitudinal.
It shows that the numerical results are higher in longitudinal direction when
compared to experimental result having difference of about 20% which is very high.
The spanwise results predict slower growth when compared to literature same as the
longitudinal results.
4 Maximum Velocity Decay
Decay of velocity is over the different x/h positions where h is width of orifice and
x is the direction vector. It is plotted between x/h and U m /U j . As the jet is three
Table 2 Decay of maximum

Type of Orifice Lip CD Region RD Region
velocity
Diffuser( CFD) −0.21 −1.18
Slot ( CFD) −0.16 −1.09
Nozzle ( CFD) −0.22 −0.89
Diffuser (Nagendra 2018) − −1.12
Slot (Hall 2007) − −1.03
Nozzle (Poole 2014) − −0.79
dimensional, the potential core, characteristic decay and radial decay region will
exist. The three-dimensional jet follows the power law, i.e.,
Um x −n
∝ (1)
Uj h
The values are listed in Table 2, it is observed that decay of maximum velocity
is slower in nozzle when compared with diffuser and slot, and the same behavior is
observed in literature where the constant is −0.79 (Fig. 10).
1
U/Umax
0.1
Diffuser CFD
Slot CFD
Nozzle CFD
Slot EXP Hall 2007
Diffuser EXP Nagendra SVH 2018
0.01
1 10
x/h
Fig. 10 Maximum velocity decay for all the three orifice geometries
50
Turbulence Kinetic Energy [ J kg^-1 ]

diffuser
slot
nozzle
40
30
20
10
0
0 10 20 30 40
x/h
Fig. 11 Distribution of turbulence kinetic energy over the center line for all the three orifice
geometries
5 Turbulence Kinetic Energy and Shear Stress Distribution
Turbulence kinetic energy over the center line is shown in Fig. 11. Graph is plotted
between J/kg and y/h, and the x-axis is normalized by width of the orifice. Overall the
turbulence kinetic energy is more in case of nozzle-type lip orifice when compared
to other the lip types. The, initially, turbulence is more in slot and diffuser but as
the flow passes 4 h the turbulence kinetic energy is more which is due to increase in
the frequency of eddies generated due to the sharp edge of nozzle-type lip geometry.
The turbulence kinetic energy at 40 h for nozzle-type lip is about the twice the rate of
slot and diffuser-type lip. Figure 12 shows the normalized Reynolds stress UV with
square of maximum velocity at 20 h along the y-direction which is normalized by
h, i.e., the width of orifice. The position of zero shear stress is at 0.18 for slot- and
diffuser-type lip geometry, but it is little bit higher in case of nozzle-type lip which
is about 0.22.
6 Conclusion
The investigation provides optimistic results to understand the effect of lip geom-
etry of orifice on the mean flow characteristics of three-dimensional wall jet flows.
It is clear that there is no significant variation in the mean velocity profile. Both
longitudinal and spanwise profile show self-similarity and agreed well with existing
literature. The growth of half width is poorly predicted by the RSM model but
0.020
Diffuser CFD
Slot CFD
Nozzle CFD
0.015
UV Stress /U 2 max
0.010
0.005
0.000
-0.005
0 2 4 6 8 10 12
y/h
Fig. 12 Shear stress distribution at 20 h for all the three orifice geometries
correlates well with literature. The decay of maximum velocity is slower in case of
nozzle-type lip geometry, and then other two geometries and the behavior resemble
the referred literature. The turbulence kinetic energy and shear stress distribution are
also predicted and found that the shear stress distribution is increased in the case of
diffuser type.
References
1. Glauert MB (1956) The wall jet. J Fluid Mech 1(06):625. https://doi.org/10.1017/S00221120

5600041X
2. Launder BE, Rodi W (1983) The turbulence wall jet measurements and modeling. Annu Rev
Fluid Mech 15(1):429–459. https://doi.org/10.1146/annurev.fl.15.010183.002241
3. Padmanabham G, Gowda BHL (2017b) Mean and turbulence characteristics of a class of
three-dimensional wall jets—Part 2 : turbulence characteristics 113: 629–634
4. Dunn M (2010) An experimental study of a plane turbulence wall jet using
5. Gowda HL, Durbha VSB (2017) Mean and turbulence characteristics of three- dimensional
wall jet on convex cylindrical surfaces 121: 596–604
6. Hall JW, Ewing D (2005) The rectangular wall jet. Part 1 : the effect of varying aspect ratio
(January), pp 10–13
7. Sforza PM, Steiger MH, Trentacoste N (1966) Stud Three-Dimensional Viscous Jets 4(5):800–
806
8. Dey S, Nath TK, Bose SK (2010) Fully rough submerged plane wall-jets. J Hydro-Environ
Res. https://doi.org/10.1016/j.jher.2010.01.002
9. Dutta R, Dewan A, Srinivasan B (2013) Comparison of various integration to wall (ITW)
RANS models for predicting turbulence slot jet impingement heat transfer. Int J Heat Mass
Transf 65:750–764. https://doi.org/10.1016/j.ijheatmasstransfer.2013.06.056
10. Kechiche J, Mhiri H, Le Palec G, Bournot P (2004) Application of low Reynolds number k-ε
turbulence models to the study of turbulence wall jets. Int J Therm Sci 43(2):201–211. https://
doi.org/10.1016/j.ijthermalsci.2003.06.005
11. Voegele AP, Trouve A, Cadou C, Marshall A (2010) RANS modeling of 2D adiabatic slot film
cooling. In:46th AIAA/ASME/SAE/ASEE Joint Propulsion Conference & Exhibit, July, 1–14
12. Jankee GK, Ganapathisubramani B (2019) Influence of internal orifice geometry on synthetic
jet performance. Exp Fluids 60(4):51. https://doi.org/10.1007/s00348-019-2697-2
13. Launder BE, Reece GJ, Rodi W (1975) Progress in the development of a Reynolds-stress
turbulence closure. J Fluid Mech 68(3):537–566. https://doi.org/10.1017/S0022112075001814
14. Hanjalić K, Launder BE (1972) A Reynolds stress model of turbulence and its application to
thin shear flows. J Fluid Mech 52(4):609–638. https://doi.org/10.1017/S002211207200268X
15. Padmanabham G, Gowda BHL (2017a) Mean and turbulence characteristics of a class of
three-dimensional wall jets—Part 1: mean flow characteristics 113, December 1991
Role of Inhibitor Flexibility
in the Mitigation of Pressure Oscillations
in Large Solid Rocket Motors
Kali Prasad V. A. S. N. Kallakunta, Sankaran Sathiyavageeswaran,

and Satyanarayanan R. Chakravarthy
1 Introduction
In large segmented solid rocket motors (SRM) having considerable length over diam-
eter ratio L/D, the coupling of flow instabilities with the acoustic modes of the
chamber during its operation leads to sustained low-amplitude pressure oscillations
[1–3]. Because of the large port diameter, these pressure oscillations transform into
very high percentage of thrust oscillations causing serious concern to launch vehicle
operability, especially in the latter half of burn time [4, 5].
The recent works have provided good understanding on different mechanisms
involving flow–acoustic coupling and its applicability to large segmented SRM like
Shuttle RSRM, Ariane P230 and Titan SRMU. The flow instabilities were categorized
into obstacle vortex shedding (OVS), corner vortex shedding (CVS) and surface
vortex shedding (SVS). Oscillations in pressure are the result of the interaction of
these large-scale flow vortices with the acoustic modes of the combustion chamber.
ISRO’s large boosters, S139 and S200 motors, have also shown oscillations in
both pressure and thrust during static tests/flights. The peak amplitudes are typically
< 0.86% in pressure oscillations, which correspond to 8% in thrust oscillations. While
the frequency of these oscillations consistently matches with the acoustic modes of
combustion chamber, their amplitudes are observed to vary for each static test/flight
[5]. In fact, most of the cold flow studies used rigid inhibitors to understand the effect
of different parameters on OVS. On the contrary, the actual inhibitors in full-scale
K. P. V. A. S. N. Kallakunta (B) · S. Sathiyavageeswaran

Satish Dhawan Space Centre SHAR, Indian Space Research Organisation, Sriharikota, India
e-mail: kvkprasad@shar.gov.in
S. Sathiyavageeswaran
e-mail: sankaran.s@shar.gov.in
S. R. Chakravarthy
Department of Aerospace Engineering, Indian Institute of Technology Madras, Chennai, India
e-mail: src@ae.iitm.ac.in
366 K. P. V. A. S. N. Kallakunta et al.
motors are flexible, which bend under flow loading and vibrate at the excitation
frequencies of unsteady flow and acoustic feedback. The time of occurrence of peak
oscillations in pressure is found to be dependent on the initial inhibitor thickness. The
ST-01 motor having highest inhibitor thickness of 32 mm has experienced oscillations
very late into the burn at 90 s. The ST-02 motor, with inhibitor thickness of 24 mm, did
not show any significant pressure oscillations after 50 s [5]. Therefore, it is prudent
to understand the effect of parameters of flexible inhibitor that effect the pressure
oscillation amplitudes.
The present work concentrates on understanding the role of inhibitor thickness-
to-height ratio t/h and its natural vibration frequency f ns on the pressure oscillations
in a sub-scale model due to OVS (refer Fig. 2) using cold flows.
2 Obstacle Vortex Shedding in SRM
As the segment inhibitors are inert and have lower burn rate, they do not recede
along with the propellant burn surface and protrude inside the motor’s port, leading
to periodic OVS in the mean flow at this location. In segmented SRM, the vortices
initiated at unburnt segment inhibitors travel downstream and produce an acoustic
wave upon impingement on nozzle dome that travels back upstream. A feedback
loop will further cause these vortical perturbations to grow making the self-excited
oscillator to tend toward its natural frequency [6] (Fig. 1).
Fig. 1 Feedback loop for pressure oscillations in segmented SRM
Fig. 2 Sub-scale cold flow model

Role of Inhibitor Flexibility in the Mitigation of Pressure Oscillations … 367
The physics of the obstacle vortex shedding mechanism inside SRM combustion
chamber, which is often referred as the acoustic feedback model [3], involves the
following steps:
(a) The shear layer is initiated at the segment inhibitor.
(b) Its vortex rolls up and grows in magnitude as it travels downstream.
(c) The impingement of the vortex on the nozzle dome.
(d) The vortex impingement generates an acoustic wave. This will try to establish
a standing wave inside the chamber.
(e) As acoustic wave travels back, at the initiation point, it perturbs the shear layer
at an appropriate time such that it aids in the generation of subsequent vortex.
(f) This leads to formation of a new vortex roll-up, establishing the feedback loop.
The above feedback loop is mathematically represented as
mT = l/kU + l/(a − U ) + t (1)
In Eq. (1), T is the vortex shedding time period, l is the length of travel between
the initiation and impingement point, U is the mean velocity of the flow, a is the
speed of sound in the medium, m represents the integer number of vortices that exist
between the initiation and impingement locations, k is the factor which represents
the ratio of the vortex convection velocity to the mean velocity of the flow, and t
or αT is the empirical factor representing a time increment.
For scaled model studies, Anthoine [4] derived an analytical model to predict the
suitable conditions for flow–acoustic coupling to occur. Considering the acoustic
feedback (Fn = na/2L), resonance occurs for the condition where the vortex shed-
ding frequency f matches with the duct acoustic frequency F n , and internal Mach
number at this condition is given by
2
Cvc n l d
M0 = (2)
2kv m − α L D
where C vc is the vena contracta coefficient for the jet inside the chamber, issued at
the inhibitor inner edge, and k v is the ratio of the vortex convection velocity to this
jet velocity.
When resonance occurs, the excitation of the duct acoustic mode n depends on
the relative location l/L of the inhibitor as compared to its mode shape. The above
coupling is expected to occur only for integer number of vortices m between the initi-
ation point and the impingement surface with peak resonance at given M 0 subjecting
to the condition that the geometrical parameters l/L and d/D satisfy the relationship
mentioned in Eq. (2).
2.1 Details of Test Setup and Scaling Parameters
The experimental facility uses air as surrogate fluid, which is compressed inside
a receiver of 4 m3 at 12 bar. The flow is regulated to a required pressure using
a regulator unit. The average working temperature of the medium inside the test
section is 298 K and is monitored real time during the experiments. A minimum
static pressure of 1.8 bar is ensured in the chamber to attain sonic conditions at
nozzle throat. To isolate the acoustic noise from the inlet hoses, a porous plate is
used at entry of the test section. The value of M, computed for the actual full-scale
solid rocket motor, is in the range of 0.05–0.11. Since the experiments are conducted
with flow at nozzle throat in choked condition, the internal Mach number is varied by
changing the throat area continuously by using a pintle mechanism to plug the nozzle.
Therefore, the static pressure inside the scaled model varies according to the mean
velocity of the flow, depending on the achieved pressure drop at the inlet because of
the porous media. To negate this effect, all the unsteady pressure measurements are
to be normalized by dividing it with the static pressure Ps . Thus, p /Ps becomes the
variable to investigate as the measure of the pressure oscillations (Fig. 2).
The test section is a 1:25 scale axisymmetric cold flow model of S200 motor. The
test section is made up of a number of segments to accommodate varying chamber
diameters of the downstream segment. This is very important for replicating the
vortex–nozzle interaction in the full-scale motor. The internal diameter of the cylin-
drical region of the test section upstream of inhibitor is chosen such that the similarity
in Mach number to the full-scale motor is at 70% of the burn time. The ratio l/h is
maintained between 7 and 10 as per criteria established by Stubos et al. [7] for cold
flows simulating large SRM. Three different diameter ratios (d/D = 0.65, 0.7 and
0.75) are chosen to match expected inhibitor heights during motor burn [8]. For each
diameter ratio, inhibitor thickness is varied between 1 and 5 mm (corresponds to t/h
= 0.057 to 0.4) to study the influence of inhibitor structural modes on flow–acoustic
coupling (Table 1).
The total duration of the experiment is 10 s. During this duration, the data is
recorded for all the channels at a sampling rate of 10 kHz using data acquisition
system. The evaluation of pressure oscillation is carried out using the unsteady pres-

sure data recorded near the head end ( pHead End ), and its location is closer to the
acoustic pressure antinode. The test duration is divided into individual segments of
0.41 s, and FFT is carried out on this segment data. Each segment contains 4096
data samples on which FFT is performed with a frequency resolution of 2.4 Hz. The
Table 1 Parameters for

Parameter Value(s)/Range
scaled cold flow model
l/L 0.42
d/D 0.65, 0.7, 0.75
t/h 0.057–0.4
M 0.03–0.072
average of amplitudes (0-peak) of 25 segments is reported as the pressure oscillation

p̂ along with dominant frequency for the corresponding internal Mach number M.
These amplitudes are normalized using the static pressure of test section PS .
The measurement accuracy of static pressure is ±0.4%. Piezoelectric-type
unsteady pressure sensor (make: PCB, model: 106B, nonlinearity < 1%) with a very
high resolution is used for the measurement of unsteady pressure. The mean flow
velocity is calculated using flow meter (make: Rosemount 3051S, accuracy: ±2%).
The spectral analysis of the experimental data (d/D = 0.7 for rigid inhibitor) shows
flow–acoustic coupling near first acoustic mode at 423 Hz and second acoustic mode
at 846 Hz (Fig. 3).
The lock-on of flow and acoustics happens in a narrow range of frequencies
near the chamber’s acoustic modes, generally referred as ‘frequency track.’ Peak
oscillations are observed at the middle of the track when the frequency of vortex
shedding and acoustic mode is at near match.
To understand the effect of structural coupling, the rigid inhibitors are replaced
with flexible ones made out of nitrile and neoprene rubber. Several experiments
are conducted by varying the thickness of inhibitor, thereby modulating its natural
frequency. The internal Mach number corresponding to maximum pressure oscilla-
tions is found to be in agreement with acoustic feedback model (Eq. 2). For a given
thickness of inhibitor, it is seen that M0 at resonance increases with the increase in
the diameter ratio d/D.
Fig. 3 Unsteady pressure measurements near first two acoustic modes for d/D = 0.7, rigid inhibitor
Fig. 4 Pressure oscillations 0.4

near first two acoustic modes t/h=0.057
0.35
for d/D = 0.65, flexible t/h=0.114
inhibitor 0.3 t/h=0.171
t/h=0.228
0.25
0.2
0.15
0.1
0.05
0
0.03 0.035 0.04 0.045 0.05 0.055 0.06 0.065
Mach number (M)
For all the diameter ratios, pressure oscillation amplitudes near first acoustic mode
show a decreasing trend as the thickness of inhibitor is increased. On the contrary,
near the second acoustic mode, amplitude of pressure oscillations increases with
the increase in the thickness of the inhibitor (Fig. 4). To understand the possible
reasons for this varying trend, corroboration of the above results is made with natural
frequency of the inhibitors.
The inhibitor natural vibration frequency f ns increases along with its d/D and thick-
ness (proportional to t 3/2 ) [9]. However, none of the selected inhibitors (with d/D =
0.7) has its first natural vibration frequency f 00 in a near match with the region of
flow–acoustic coupling at both acoustic modes and is found to be <350 Hz even for
the highest thickness ratio. A new inhibitor made out of neoprene rubber with higher
shore-A hardness was selected such that its natural vibration frequency was 404 Hz,
which is very close to first acoustic mode of the model. Results shown in Fig. 5 indi-
cate that amplitude of pressure oscillation increases by about 20% when compared to
rigid inhibitor and is more than double when compared to inhibitor of same thickness
ratio t/h with no structural coupling.
4 Conclusions
The present study has focused on understanding the role of flow–structure coupling
on pressure oscillations due to obstacle vortex shedding in segmented SRM. The
experiments have shown a clear dependence of inhibitor natural vibration frequency
on the amplitude of pressure oscillations.
When the inhibitor natural vibration frequency is far less when compared to the
acoustic frequency, which is generally the case for large boosters like S200 motor,
the pressure oscillations grow as the inhibitor thickness is gradually increased. The
results near the first acoustic mode also show the influence of structural coupling
when inhibitor natural frequency exactly falls in the region of resonance due to
Fig. 5 Effect of structural 1.6

coupling on pressure Rigid Inhibitor
oscillations near first 1.4 t/h=0.24, shore-A hardness =70 ( 00= 331.8 Hz)
acoustic mode t/h=0.24, shore-A hardness =75 ( 00= 403.9 Hz)
1.2
0.8
0.6
0.4
0.2
0
0.05 0.055 0.06 0.065 0.07
Mach number (M)
flow–acoustic coupling. Otherwise, the increase in inhibitor thickness seems to have

a damping effect causing a reduction in amplitude levels at resonance. This can be
possibly due to the change in relative position of the flapping inhibitor at the instant
of vortex shedding, based on its natural vibration frequency. The physics of this
phenomenon can be further explained only by carrying out vibration measurements
on inhibitor in the presence of flow to quantify the actual displacements during the
periodic motion of the inhibitor due to the flow–acoustic coupling.
The analysis on the expected inhibitor natural frequencies during the opera-
tion of the full-scale motor shows that the first vibration mode of the inhibitor is
always lower when compared to the first acoustic mode frequency of the combus-
tion chamber. Also, with the progression of burn time, the natural frequency of the
inhibitor decreases based on its thickness. Therefore, the natural frequency of the
inhibitor is far less when compared to the first acoustic mode late into the burn time.
In such a scenario, increase in the inhibitor thickness may also cause higher-pressure
oscillations similar to the observations noticed near the second acoustic mode in
the scaled model experiments. The comparison of pressure oscillation amplitudes
between the static tests [5] indicates such a trend.
If the pretest thickness of the inhibitor of full scale motor is less than 25 mm,
based on the inhibitor ablation rates, the first vibration frequency of the inhibitor
is expected to be <8 Hz during the latter half of the burn time. This eliminates the
chances of the structural coupling of the inhibitor with the acoustic mode even with
higher radial erosion of inhibitor, aiding in mitigation of pressure oscillations.
Acknowledgements The authors thank the management of SDSC SHAR and ISRO for their
support to this R&D study and constant encouragement.
References
1. Blomshield FS (2006) Lessons learned in solid rocket combustion instability. In: Proceedings of
American Institute of Aeronautics and Astronautics Missile Sciences Conference, ADM202095,
Monterey, CA, November 2006
2. Flandro GA, Jacobs HR (1973) Vortex generated sound in cavities, AIAA Paper 73-1014,
October 1973
3. Dotson KW, Koshigoe S, Pace KK (1997) Vortex shedding in a large solid rocket motor without
inhibitors at the segment interfaces. J Propulsion Power 13(2), March–April 1997
4. Anthoine J, Lema MR (2009) Passive control of pressure oscillations in solid rocket motors:
cold-flow experiments. J Propulsion Power 25(3), May–June 2009
5. Ramanaiah V, Kali Prasad VASN Kallakunta, Ananda Raj J, Seshagiri Rao V , Prasad MYS ,
Dathan MC (2013) Investigation of thrust oscillations in large segmented solid rocket motors
during ground tests, Paper No 21005, National Propulsion Conference (NPC-2013) at IIT
Madras, Feb 2013
6. Rockwell D (1983) Oscillations of impinging shear layers. AIAA J 21(5):645–664
7. Stubos AK, Benocci C, Palli E, Stoubos GK, Olivari D (1999) Aerodynamically generated
acoustic resonance in a pipe with annular flow restrictors. J Fluids Struct
8. McWhorter B, Ewing ME, Bolton DE, Albrechtsen KU, Earnest TE, Noble TC, Longaker M
(2003) Real-time inhibitor recession measurements in two space shuttle reusable solid rocket
motors, AIAA 2003–5107, 39th AIAA/ASME/SAE/ASEE Joint Propulsion Conference and
Exhibit. Huntsville, Alabama
9. Leissa AW (1969) Vibration of plates, NASA SP-160, Washington DC
Numerical Study of Fluid Flow and Heat
Transfer in Stepped Micro-Pin–Fin Heat
Sink
Prabhakar Bhandari and Yogesh K. Prajapati
1 Introduction
The usage of many electronics equipments in engineering and technology depends

on several factors among them size is an important parameter. The smaller size equip-
ment has advantage of low material and operational cost and ease in transportation.
The main hindrance which comes in the way of miniaturization is high heat removal
rate. If heat dissipation rate is less, then it will have adverse effect on the equipment.
To prevent over-heating problem, an effective heat management system is needed.
Various techniques like confined jet impingement, heat pipe, micro-heat sink, and
microchannel have been found effective for high heat dissipation rate in these
systems [1]. Tuckerman and Pease [2] were the first to pitch the idea of using
microchannel heat sink in heat management. The reason behind high heat transfer
rate of microchannel heat sink is large heat transfer surface area per unit fluid flow
volume.
The heat transfer rate can also be increased by passively triggering non-laminar
flow patterns instead of following laminar flow, for which the thickness of the devel-
oping thermal boundary layer is the reason for deterioration of heat transfer. The
technique for heat transfer augmentation is by using pin fin by which heat transfer
surface area increases and also obstructions in the flow field to originate a regular
thinning of the thermal boundary layer. Yang et al. [3] investigated five different
shapes of pin fin in a staggered fashion in a microchannel. The five geometries
of cross sections of fins are triangle, square, pentagon, hexagon, and circle. They
conducted both experimental and numerical study using deionised water as a coolant.
They observed that pin fins with triangular cross section has maximum, and circular
cross section has minimum blocking effects, and also hexagonal cross section fins
have lowest thermal resistance. Yadav et al. [4] performed numerical study on rectan-
gular microchannel with three different configurations of cylindrical microfins. These
P. Bhandari · Y. K. Prajapati (B)

National Institute of Technology Uttarakhand, Srinagar (Garhwal), Uttarakhand, India
374 P. Bhandari and Y. K. Prajapati
configurations are upstream finned microchannel, downstream finned microchannel,

and complete finned microchannel. They concluded that heat transfer performance
of upstream finned microchannel is better than other cases at low Reynolds number,
but at high Reynolds number complete finned microchannel performs best. Yu et al.
[5] has conducted experimental and numerical study on microchannel heat sink with
Piranha pin fins (PPFs) to enhance the heat transfer rate.
Al-Neama et al. [6] carried out experimental and numerical investigation on
serpentine minichannel heat sinks with chevron fin structure. The fluid transfer
between main and the secondary channel helps in reduction of thermal resistance
and pressure drop in comparison with the only serpentine minichannel.
In this work, the effects of parameters and location of pin fin on heat transfer
augmentation and flow are numerically simulated. The simulation is performed
using commercial CFD software ANSYS.
2 Simulation Model
The present micro-pin–fin heat sinks are of open microchannel type, as pin–fin height
is not equal to the channel height, and there is fluid flow over the top surface of the
pin–fin. Table 1 shows the dimensions of the channel used in the simulation, and other
dimensions are shown in Fig. 1. In present model, it is assumed that microchannel
Table 1 Details of the microchannel (all dimensions are in mm)

channel Total Width Total Height channel channel
length(Lch ) (channel + (channel + Width(Wch ) Depth(Hch )
substrate) substrate)
27 10 3 9 2
Fig. 1 a Channel cross

sectional view, b Channel
2
1 1 1 1 1
top view for uniform (a)

1
micro-pin–fin heat sink (All

are in mm)
9
0.5 0.5
(b)
2 2
27
Numerical Study of Fluid Flow and Heat Transfer … 375
Lch
Outlet
Inlet
Hch
Increasing
Outlet
Inlet
Decreasing
Outlet
Increasing
Inlet
and
Decreasing
Outlet
Inlet
Hch
Uniform
Fig. 2 Variation in pin-fins height for different cases
is made of copper and water as a working fluid. The heat flux is being applied from
the bottom wall of the channel. The dimensions of the channel are kept constant for
all the different configuration of fins in heat sink.
The micro-pin–fin used in the channel is of square cuboid shape. The base area
of the fin is of 1 mm by 1 mm for all cases. The fins are arranged in an inline fashion
with spacing of 1 mm in between two fins in longitudinal and transverse direction.
Total number of fins used are 48(4 × 12). Figure 1 shows cross sectional and top
view of the uniform micro-pin–fin heat sink.
There are four different cases of pin–fin height are under consideration in this
work. Firstly, pin–fin height is in increasing order from 0.45 mm in first row to
1.55 mm in last row with 0.1 mm increment. In second case, fin height is in decreasing
fashion from 1.55 mm in first row to 0.45 mm in last row with 0.1 mm decrement.
In third case, fin height is in increasing fashion up to middle row and then follows
decreasing fashion up to last row with height step of 0.1 mm. The sequences of
pin–fin height are 0.75, 0.85, 0.95, 1.05, 1.15, 1.25, 1.25, 1.15, 1.05, 0.95, 0.85, and
0.75 from first row to last row. In fourth case, pin–fin height is uniform for all rows
which is 1 mm. The heights of the different arrangement of the micro-pin fins are
under volume constraint; i.e., the total volume occupied by all pin fins remains same,
so that the fluid flow area does not change. Figure 2 shows the variation of pin fins
height for all cases.
2.1 Assumptions
Following assumptions are used to simplify the numerical simulation:

1. Flow is steady, laminar, and Newtonian.
2. At wall there is no slip boundary condition.
3. Heat flux is assumed to be uniform and constant throughout the bottom wall of
the channel.
4. The heat transfer by radiation is being neglected.
Based on the above assumptions, law of conservation of mass, momentum, and
energy form continuity equation (incompressible) and Navier–Stokes equations to
describe the fluid flow and heat transfer phenomena.
∇ · (ρl V ) = 0 (1)
∇ · (ρl V · ∇ V ) = −∇ p + ∇ · μl [(∇ V + ∇ V t ) − 2/3∇ · V ] + ρl g (2)
∇ · (ρ f c p,l (V Tl ) = ∇ · (kl ∇Tl ) (3)
ks ∇ 2 T = 0 (4)
where V is the velocity matrix and k f , μ f and ρ f is thermal conductivity,

dynamic viscosity, and density of working fluid. The thermodynamic properties like
density, specific heat, and viscosity of working fluid water are polynomial function
of temperature. The correlations are taken from Yadav et al. [4]
At the inlet of the microchannel, working fluid water is assumed to have constant
velocity (Uinlet ) and constatnt temperature (Tinlet ).
Tinlet = 300 K; and Uinlet is calculated for constant Reynolds number value of
200, 400, 600 and 800.
The heat flux of 100 kW/m2 is applied at the bottom wall uniformly. The flow is
fully developed at the channel outlet.
2.3 Solution Method
The current simulation has been carried using ANSYS Fluent software which is based
on finite volume method. To solve the governing differential equations, SIMPLE
algorithm has been applied. For convergence of solution, residual criteria of 1 ×
10–4 are used for continuity equation, 1 × 10–6 for the velocity in x, y, and z direction
and 1 × 10–7 for energy equation.
Table 2 Details of different mesh configurations used in grid independence test

Grid 1 Grid 2 Grid 3 Grid 4 Grid 5
Number of elements 207,501 226,856 669,310 1,098,516 1,713,024
Fig. 3 Variation of outlet 310.26

fluid temperature with no of 310.24
element for uniform case
Outlet Temperature (K)

310.22
310.20
310.18
310.16
310.14
310.12
310.10
310.08
0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
No of Elements( in Million)
2.4 Mesh Optimization
To eliminate the error, the grid independence test has been done. Table 2 shows
the details of five grid system used for the test. In the present case, the variation of
outlet fluid temperature with various element sizes for uniform type pin–fin has been
plotted in Fig. 3. The computational domain has been discretised using structured
grids of hexahedral volume elements.
The last two grids have minimum difference in the outlet temperature; hence,
to save the computational time, grid 4 has been used in the current simulations for
different cases.
3 Result and Discussion
3.1 Validation of the Present Model
Numerical model for the plain channel is compared with the correlations given by
Yu et al. [5] and shah and London [7] for Nu in plain channel. Figure 4 shows the
comparison of Nu at different Reynold number. The correlations given by Yu et al.
[5] and shah and London [7] are given in Eqs. 5 and 6.
NuYu et al. = 2.1219Re0.3204 (5)

Fig. 4 Comparison of 22
Nusselt number for plain
channel with the correlations 20
given by Yu et al. [5] and
Shah and London [7]
Nusselt Number
18
16
14
Present Simulation
12 Correlation by Yu et.al.
Correlation by shah and London's
10
100 200 300 400 500 600 700 800 900
Reynold Number
NuShah and London = 2.2348Re0.3333 (6)
Nu for present simulation shows the good agreement with the correlations value.
Hence, the similar numerical method can be applied to evaluate the performance of
the microchannel.
3.2 Heat Transfer
For estimating overall heat transfer characteristics, Nusselt number (Nu) is evaluated.
The overall Nusselt number is calculated as:
h Dh
Nu = (7)
Kf
where h is heat transfer coefficient and K f is thermal conductivity of fluid, i.e., water.
Figure 5 shows variation of Nu with Re for different pin–fin height case at heat flux
of 100 kW/m2 . As the Re increases, there is increase in Nu for all the cases, but rate of
increase in Nu decreases with Re for all cases. Prajapati [8] has also reported similar
type of observation in his numerical work. Increasing pin–fin pattern has shown
highest value of Nu for all the cases of Re. After that increasing and decreasing case
has shown the higher value of Nu. Also, it can be seen from the plot that up to Re
value of 300, the decreasing pattern pin–fin has higher value of Nu than after that
uniform pin–fin has higher value of Nu. It is due to the development of boundary
layer at the wall of fin, increase in heat transfer area and induced mixing of fluid due
to obstruction created by fin.
Figure 6 shows the variation of average bottom wall temperature (T b,wall ) with
Re for all different cases of micro-pin–fin heat sink for heat flux of 100 kW/m2 .
Fig. 5 Variation of Nu with 35

Reynold number for different Increasing
fin height at heat flux of 100 Decreasing
KW/m2 30 Increasing & Decreasing
Uniform
25
Nu
20
15
10
200 300 400 500 600 700 800
Re
Fig. 6 Variation of average 328

bottom wall temperature 326
(T b,wall ) with Re for all Increasing
the cases of micro-pin–fin 324
Decreasing
heat sink at heat flux of 322 Increasing & Decreasing
Uniform
(°K)
100 kW/m2 320

b,wall
318
316
T
314
312
310
308
200 300 400 500 600 700 800
Re
The lowest bottom wall temperature is obtained for increasing pattern pin–fin heat
sink. This is because maximum value of Nu is for increasing pattern pin–fin which
indicates maximum amount is carried away by water in this case. The value of bottom
wall temperature is dependent on value of Nu. One with higher value of Nu has lower
value of bottom wall temperature.
3.3 Fluid Flow
Figure 7 shows the variation of pressure drop with Reynold number for different
fin height at heat flux of 200 kW/m2 . The pressure drop has been calculated as
the difference of pressure value at inlet and outlet. With the increase in Re, the
pressure drop (p) increases because more fluid has to flow through channel. Fin
Fig. 7 Variation of pressure 150

drop with Reynold number
for different fin height cases 130 Increasing
at heat flux of 100 kW/m2 Decreasing
110 Increasing & Decreasing
Uniform
Δp (Pa)
90
70
50
30
10
200 300 400 500 600 700 800
Re
with increasing pin–fin height pattern has shown highest value of pressure drop for
all value of Re. The other three cases have very less difference in pressure drop
between them.
4 Conclusion
Heat transfer performance of micro-pin-finned channel depends upon the location

and configuration of the pin–fin in the channel. The following points can be concluded
from the study:
1. Pin–fin with increasing pattern has highest value of Nu. It can be concluded that
along the channel length, the fin effectiveness increases.
2. There is very less effect of micro-pin–fin on inlet side than on outlet side. So,
higher pin–fin may not be at inlet.
3. Variable pin–fin height along channel has higher value of Nu compared to the
uniform height pin–fin height.
4. Height of the fins plays a crucial role in heat transfer characteristics and pressure
drop across the channel.
References
1. Gururatana S (2012) Heat transfer augmentation for electronic cooling. Am J Appl Sci 9(3):436–
439
2. Tuckerman DB, Pease RFW (1981) High performance heat-sinking for VLSI. IEEE Electron
Device Lett 2:126–129
3. Yang D, Wang Y, Ding G, Jin Z, Zhao J, Wang G (2017) Numerical and experimental analysis
of cooling performance of single-phase array microchannel heat sinks with different pin-fin
configurations. Appl Therm Eng 112:1547–1556
4. Yadav V, Baghel K, Kumar R, Kadam ST (2016) Numerical investigation of heat transfer in
extended surface microchannels. Int J Heat Mass Transf 93:612–622
5. Yu X, Woodcock C, Plawsky J, Peles Y (2016) An investigation of convective heat transfer in
microchannel with Piranha Pin Fin. Int J Heat Mass Transf 103:1125–1132
6. Al-Neama AF, Khatir Z, Kapur N, Summers J, Thompson HM (2018) An experimental and
numerical investigation of chevron fin structures in serpentine minichannel heat sinks. Int J Heat
Mass Transf 120:1213–1228
7. Shah RK, London AL (1978) Laminar flow forced convection in ducts: a source book for compact
heat exchanger analytical data, Acad. Press, p 477.
8. Yogesh K Prajapati (2019) Influence of fin height on heat transfer and fluid flow characteristics
of rectangular microchannel heat sink. Int J Heat Mass Transf 137:1041–1052
Numerical Investigation of Cavitating
Flow over S-Blade in Acoustic Field
Pankaj Kumar and R. Bala Anand
1 Introduction
The scientific community has focused its attention and efforts on harvesting clean,
non-polluting energy, on a global level. Tidal energy is seen as a viable option as it
is renewable and available free of cost. It would be economical if a single machine
can perform the respective work of both pump and turbine during both forward and
reverse operations. This creates a necessity to develop a suitable fully reversible
turbomachine to work as a pump as well as a turbine in both directions of flow [1].
Due to this, the normal flow becomes very complicated even at a very low angle of
attacks causing streamwise varied pressure gradients and so cavitation occurs.
An extensive literature study has been undertaken on the cavitating flow on hydro-
foils blades. Based on an experiment with pressure and velocity measurement on a
hydrofoil, Kjeldsen et al. [2] reported dynamic characteristics of cavitation varied
considerably with a change in cavitation number. They explained the variation of
Strouhal number with cavitation number as almost linear. Gopalan and Katz [3]
explained the flow structure with the help of PIV and high-speed photography. They
reported that when the cavity is thin, there is no reverse flow downstream and reen-
trant jet starts developing at a region where adverse pressure gradient exists. Leroux
et al. [4] reported partial cavitation over hydrofoil and observed that reentrant jet
and the cavity interface influenced cavity shedding. The cavity collapse leads to an
increase in pressure which affects the growth of the residual cavity. Kumar et al. [5]
did the experimental and numerical simulation on the S-shape blade and explained
the detail flow physics by lift, drag, and pressure fluctuations. They observed that
pump mode at a negative angle of attack more prone to cavitate. Recent works are
more focused to resolve local information near the cavitating field by using numerical
P. Kumar (B) · R. B. Anand

Department of Mechanical Engineering, SRM Institute of Science and Technology, SRM Nagar,
Kattankulathur, Chennai 603203, Tamilnadu, India
e-mail: pankajkr@srmist.edu.in
384 P. Kumar and R. B. Anand
methods. Several different modeling approaches have been pursued such as Singhal
et al. [6], Sauer, and Schnerr [7]. Walters and Davor [8] calculate velocity fluctua-
tion in the region of transition. This model used the additional transport equation of
laminar kinetic energy. Genc [9] used k-k L -omega transition model by which varying
characteristics are better predicted than other models.
It is well known that ultrasound increases the bubble activity, i.e., acoustic cavi-
tation. Some observers suggested that the bubbles at higher frequency act as nuclei
and thus increases the cavitation activity Lie and Hsieh [10]. Ultrasound was used
to manipulate the nuclei size in the flow, and cavitation control was achieved in
venturi by Chatterjee and Arakeri [11]. Kujur et al. [12] reported the ultrasonic
effect on the cylinder at driving frequency 50KHz and observed the cavitation activity
enhancement due to the presence of ultrasound.
The unsteady flow physics in terms of frequency domain and Strouhal number is
not addressed in the past which is the scope of the work. The wide range of imposed
frequency of wake cavitation is not done in the past, which is another motivation of
this work. The effect of ultrasound on flow using monitor points on the upper and
lower is systemically processed and explained with flow physics.
2 Methodology
Governing equations for unsteady, non-cavitating, and cavitating flow over single
S-shaped blade with and without ultrasonic field is discussed in this section.
2.1 Numerical Scheme
Blade surfaces and computational domains are created and meshed using ANSYS
ICEM-CFD and simulated in fluent. The computational domain is as shown in Fig. 1.
The hybrid mesh is used to generate relatively finer near the blade surface while far
away from the blade, the mesh could be made coarser. The chord length of the blade
is 60 mm. The mesh structure is shown in Fig. 2. The inlet boundary conditions
as velocity inlet (8 m/s), at the outlet boundary condition, as pressure outlet, at the
Fig. 1 Computational domain of S-blade with ultrasound position

Numerical Investigation of Cavitating Flow Over … 385
Fig. 2 Structured mesh for the computational domain
top and bottom walls as symmetry, top and the bottom surface of the blade as no-
slip boundary conditions are applied. A grid sensitivity study is carried out, and
the optimum number of cells required is found to be 22,000. The size of the time
step is taken such that frequencies could be captured, and total time duration was
chosen in such a way that it ensures that all variables become stationary with no
significant cycle-to-cycle variation. Thus, in this present work, a time step size of
10–4 (s) is chosen, and the number of time steps adopted is 7000. The resolution
of frequency domain data is taken sufficient by the use of a large number of data
points for sampling. There are eight monitor points (four on the upper side and four
on the lower side) on the blade profile to measure the pressure, vapor fraction. The
numerical simulation was done at the Re = 4.8 × 105 based on chord length. The
simulation is carried out for various angles of attack (AOA) including both positive
and negative values. It is further extended to different cavitation stages by lowering
test-section pressure and keeping Reynolds number constant.
Governing equations solved are the mass and momentum conservation equations.
Unsteady Reynolds averaged equations (RANS) are carried out with multiphase
mixture model. Schnerr and Sauer model [7] is used here which connects the vapor
volume fraction to number of bubbles per volume of liquid.

∂ρm ∂ ρm u j
+ =0 (1)
∂t ∂x j

∂(ρm u i ) ∂ ρm u i u j ∂ ∂u i ∂u j 2 ∂u k ∂P
+ = (μm + μt ) + − δi j −
∂t ∂x j ∂x j ∂x j ∂ xi 3 ∂ xk ∂ xi
(2)
The acoustic cavitation was created with help of pressure variation by dynamic
wall motion. For the ultrasound modeling, approach to Sutkar and Gogate [13] is
adopted. Rayleigh Plesset Noltingk Neppiras equation has been used.

d2 R 3 dR 2 1 4μ dR 2σ
R + = Pi − − −P (3)
dt 2 2 dt ρL R dt R
where Pi pressure inside cavity, R is instantaneous cavity radius, μ is viscosity, and

σ is surface tension and P. The pressure modulation is done by upper wall motion as
P = P0 − PA sin(2π f t) (4)
√
where PA = (2ρ I C), f is frequency of ultrasound, I is intensity, and C is speed
of sound.
Validation of results
For the validation of numerical approach, pressure coefficient (C P ) versus theta
is obtained for upper and lower surface at zero AOA and shown in Fig. 3. It can be
seen from the plot that the numerical simulation matches well with the experimental
results presented by Kumar et al. [5]. It is also able to capture the positive and negative
lift suggested by Kumar et al. [5] as the first area enclosed is smaller than the second
area enclosed in the plot. This will produce a net result in the negative lift. Since the
cavitation inception largely depends on body shape, surface quality, water quality,
nuclei content, and dissolved gas, it is advisable to present the results in normalized
form (σ /σ i ) where sigma (σ ) is defined similar way [5].
Fig. 3 Variation of pressure

coefficient (C P ) with respect
to non-dimensional length
(x/c) and shown with the
results of Kumar et al. [5]
The drag coefficient (C D ) value in time and the corresponding frequency by fast-
Fourier transformation is obtained. The peak amplitude is seen at the frequency of
399 Hz which corresponds to Strouhal number of 0.3, which corresponds to the
similar frequency obtained experimentally and numerically reported in the earlier
research study [5].
Test-section pressure is varied to achieve different cavitation numbers for each
AOA. The variation of lift to drag coefficient (L/D) for different cavitation number
(σ /σ i ) for different AOA is as shown in Fig. 4. The L/D value increases with the
cavitation number for 0 AOA. For the positive 8° AOA, this trend is not similar but
decreases with cavitation number. This phenomenon is attributed to the increase of
vapor due to a reduction in cavitation number which modifies the recirculation zone
on the blade surface.
There are some monitor points created on the top and bottom wall along a stream-
wise direction to capture flow parameters like pressure and vapor fraction. Figure 5
shows the Strouhal number (f*C/U) where f is the dominating frequency, C is chord
length, and U is freestream velocity.
The frequency variation is different for the upper and lower surface, and it varies
with the cavitation number. The reason for the difference is the formation of the
circulation zone and reentrant flow pattern for different angle of attacks. Strouhal
number increases with cavitation number, and one possible reason is the formation
of cavity delay the shedding process. A similar trend was also observed in the case of
the cylinder [14]. Figure 6 expresses cavity dynamics by the effect of ultrasound at
the different driving frequencies (50 kHz, 500 kHz). It was found that drag increases
for lower frequency which shows the higher cavitation activity. The lift coefficient
reduces under the influence of ultrasound.
Fig. 4 Variation of lift to

drag ratio at different
cavitation number
Fig. 5 Variation of
normalize frequency
(Strouhal number) obtained
from pressure monitor points
(top surface and bottom
surface) for different
cavitation number
4 Conclusions
The numerical work is carried out to study the effect of ultrasound on cavity dynamics
over the S-blade. The grid-independent and validation are done and found a good
match with the results available in the existing literature. The pressure coefficient,
lift to drag ratio, and Strouhal number are also validated for cavitating flow. Time
variation of pressure and vapor fraction is obtained at different points along the chord
length for the upper and lower surface. Strouhal number variation shows unique
behavior for different AOA but in most of the cases, it increases with cavitation
number. The effect of ultrasound (50 kHz, 500 kHz) in terms of Strouhal number,
drag, and cavity area is compared for different AOA and found that the effect of
ultrasound is significant, and at least, at the lower frequency, it enhances the cavity
dynamics.
Fig. 6 Variation of drag

coefficient and lift coefficient
with cavitation number and
ultrasonic frequency
Acknowledgements The author acknowledge the support from Department of Mechanical

engineering, SRMIST Kattankulathur..
References
1. Ravindran M (1978) Design and flow investigation on a fully-reversible pump-turbine. PhD

thesis, IIT Madras
2. Kjeldsen M, Arndt REA, Effertz M (2000) Spectral characteristics of sheet/cloud cavitation.
Trans ASME J Fluid Eng 122:481–487
3. Gopalan S, Katz J (2000) Flow structure and modeling issues in the closure region of attached
cavitation. Phys Fluids 12(4):895–911
4. Leroux J, Astolfi J, Billard J (2004) An experimental study of unsteady partial cavitation. Trans
ASME J Fluids Engg 126:94–101
5. Kumar TMP, Kumar P, Chatterjee D (2014) Cavitation characteristics of s-blade used in fully
reversible pump-turbine. Trans ASME J Fluid Eng 136(5):1–15
6. Singhal AK, Athavale MM, Li H, Jiang Y (2002) Mathematical basis and validation of the full
cavitation model. Trans ASME J Fluid Engg 124:617–623
7. Sauer J, Schnerr GH (2001) Development of a new cavitation model based on bubble dynamics.
Z Angew Math Mech 81:561–562
8. Walters DK, Davor C (2008) A three—equation eddy-viscosity model for reynolds-averaged

navier-stokes simulations of transitional flow. Trans ASME J Fluid Eng 130:1–14
9. Genc M (2010) Numerical simulation of flow over a thin aerofoil at a high reynolds number
using a transition model. Proc IMechE Part C J Mech Eng Sci 224:2155–2164
10. Liu H-L, Hsieh C-M (2009) Single-transducer dual-frequency ultrasound generation to enhance
acoustic cavitation. Ultrason Sonochem 16:431–438
11. Chatterjee D, Arakeri VH (1997) Towards the concept of hydrodynamic cavitation control. J
12. Kujur A, Kumar P, Chatterjee D (2012) Effect of ultrasound on wake cavitation. In: 8th
international symposium on cavitation, 2012, 13–16 August, Singapore
13. Sutkar VS, Gogate PR (2010) Mapping of cavitational activity in high frequency sonochemical
reactor. Chem Eng J 158:296–304
14. Kumar P, Chatterjee D, Bakshi S (2017) Experimental investigation of cavitating structure in
the near wake of a cylinder. Inter J Multiphase Flow 89: 207–217, 03/2017
Study on the Effective Platooning
of Indian Trucks Using Computational
Fluid Dynamics
R. Praveen Kumar, S. A. Prabhaharan, and G. Vinayagamurthy
1 Introduction
Platooning is a mechanism where two or more vehicles travel close together in

tandem to reduce the drag force faced by them. By varying the separation distance
for different velocities of the vehicle, drag force experienced by trucks can be altered
[1]. By obtaining suitable data points of velocity and distance values, effectiveness in
platooning can be attained. Recent research studies have demonstrated the wind flow
features over the bluff bodies (buildings and trucks) using CFD [2–4]. Platooning
provides benefits such as improved fuel economy and reduction in harmful exhausts.
Grouping vehicles into platoons is a method of increasing the vehicular capacity
on roads. An automated highway system can be proposed if platooning of vehicles
is implemented. This capability will allow trucks to accelerate or brake simultane-
ously at the same time reducing fuel consumption. Platoon capability might require
buying new vehicles, or it may be something that can be fitted in existing vehicles.
Platooning provides benefits such as greater fuel economy due to reduction in wind
resistance, reduced congestion, shorter commutes during peak periods and fewer
traffic collisions.
R. P. Kumar · S. A. Prabhaharan
School of Mechanical Engineering, VIT, Chennai, India
G. Vinayagamurthy (B)
School of Mechanical Engineering & Center for Innovation and Product Development, VIT,
Chennai, India
e-mail: vinayagamurthy.g@vit.ac.in
392 R. P. Kumar et al.
Table 1 Factors influencing

Factors Reduction in drag coefficient
drag coefficient
Gap between cab and trailer 0.01–0.03
Trailer base flaps 0.08
Side skirts 0.05
Under body wedge 0.01
2 Geometrical Model
Geometrical set-up for this project is a standard dimensions of an Indian truck. The
truck dimensions were obtained from the Internet, and modelling was done as per the
dimensions. Considering the aerodynamic aspects of the truck, a parabolic deflector
is added on top of the truck to improve flow around that region [5]. The factors
affecting the drag coefficient in trucks are as follows (Table 1)
2.1 Drag Equation and Drag Coefficient
Ground vehicles have mainly two major resistances to overcome: friction (internal
and external) and air resistance. At lower velocities, frictional forces dominate;
however, with increasing velocity, the aerodynamic effects become more impor-
tant. The drag force, which refers to the force acting on a moving body in the fluid
flow direction (i.e. the opposite direction of motion of the body), is defined as
D = 0.5 ∗ ρ ∗ ν 2 ∗ A ∗ Cd
where D is the drag force, A is the reference area of the object, and Cd is the drag
coefficient of the object. In vehicle aerodynamics, the reference area A is usually
defined as the area of the orthographic projection of the object in the plane perpen-
dicular to the direction of motion. The drag coefficient Cd is the ratio of drag per
unit area to dynamic pressure. Above equation is called the drag equation and is only
accurate for objects with a blunt form factor and inflow cases where the Reynolds
number Re is large enough to form turbulence behind the body (Fig. 1).
3 Methodology
The computations were made using ANSYS FLUENT 18 and meshing using ICEM
CFD 18. The governing equations used are Reynolds Averaged Navier Stroke
(RANS). The domain generated is in accordance to the dimensions of the wind
tunnel of the cross section 1 m * 1 m. The turbulence model used is two equations
Study on the Effective Platooning of Indian Trucks … 393
Fig. 1 Geometrical
modelling of truck
k – ω SST. The blockage ratio of the computational analysis is 5%. The pressure
velocity coupling equations are solved by SIMPLE algorithm.
3.1 Computational Simulation
The equations are solved using FLUENT 18. The case is considered as the steady
state, and pressure-based solver coupled with absolute velocity formulation. Then, the
material condition is chosen as air with the appropriate boundary conditions at inlet,
outlet and wall are given as velocity inlet, the pressure outlet and no-slip conditions
are assigned, respectively. In the reference value, the length of the truck 8.95 m and
the area projected 11.72 m2 , are considered for evalution of drag coefficients. In
the solution method, the standard solution initialization and the pressure velocity
coupling equations are solved by the SIMPLE scheme, and the spatial discretization
for pressure is second order, and other parameters like momentum, turbulent kinetic
energy, turbulent dissipation energy and energy follow second order upwind scheme.
SIMPLE scheme usually starts with the initial guess of pressure, velocities. Then,
the pressure correction equation is carried out, and then, velocity correction is done.
Then, the discretized momentum equation is solved and checked for the convergence.
If the solution is not converged, then the obtained solution of this step is taken as
initial guess and iterated.
Solver details
There are different turbulent models. Of which the two equations RANS models are
k –ε model and k – ω model which are generally used to solve for solutions in viscous
layer. The k – ω turbulence model can be integrated to the wall without wall function,
i.e. the model equations do not contain any terms which are undefined at the wall.
The most widely used k – ω model is k – ω shear stress transport to gradual transition
from k – ω model near the wall to k –ε model in the outer portion of the boundary
layer by using a blending function. k – ω shear stress transport is less suitable for
free shear flows because of the dependency on the wall distance. Mostly applied in
the aerospace applications and turbo machinery applications.
domain of a single lead truck
computational test cases
Meshing details
The meshing is done in ICEM CFD commercial package, and the computational
domain is shown in Fig. 2. For reducing the complexity in solving, unstructured
meshing is applied throughout the geometry of the model. Tetrahedral type of
meshing is applied with scale factor of 0.08. With necessary and sufficient cell
smoothening, the cell count arrives to a number of 0.4 million.
Computational test cases
The main objective of this study is to analyse the advantages of platooned trucks of a
lead vehicle followed by a trail truck [3, 5]. Since platooning is considered to reduce
one of the major factors affecting the efficiency of vehicles that is fuel consump-
tion, which ultimately accounts to the computation of drag coefficients which are
inevitable. Hence, velocity contours, and stream-lined patterns are captured in order
to compute the drag coefficient and the aerodynamic advantage of the platooned
trucks, respectively. For the test cases of platooning of tucks, three studies are
performed at various velocities of the trucks, considering the safe distance between
the trucks for the respective speeds. The safe distance varies for each case as follows
(Table 2):
Table 2 Test case details

Case number Velocity (km/hr) Distance between trucks (m)
1 50 1.605
2 60 2.133
3 70 2.775
Case 1
In this case wind velocity of 50 kmph and 1.605 m distance (which is considered to
be safe) between the lead and the trailing truck, are assumed respectively. Along the
vicinity of flow, it is seen that the part of wind sheds to the rear end of the lead truck
downstream. After the shedding of the airflow, it can be observed from the figure
that a kind of considerably stabilized wind impinges on the windshield of the trailing
truck.
Due to reduced turbulence, the drag coefficient reduced drastically compared to
the leading truck. The net reduction in drag was computed around 0.19 in comparison
to the leading vehicle (Fig. 3).
Case 2
In this computational case which has a wind velocity of 60 kmph and 2.133 m
seperation distance between trucks, stabilized flow phenomena can be observed.
The results show that towards the downstream of the wind flow, there is sufficient
space for most of the wind separations and shedding. Hence, the air becomes vitally
Fig. 3 a Velocity contour for velocity of 50 kmph and distance between trucks is 1.6 m, b streamlines
for velocity of 50 kmph and distance between trucks is 1.6 m
Fig. 4 a Velocity contour for velocity of 60 kmph and distance between trucks is 2.13 m,
b streamlines for velocity of 60 kmph and distance between trucks is 2.13 m
stabilized compared to the previous case. In accordance to drag reduction, the lead
truck experiences adequately less drag due to the increase in velocity. This provides
a low air resistance for the trail vehicle, and hence, drag force of lead truck is reduced
to a very small margin (Fig. 4).
Case 3
In this case wind velocity of 70 kmph and 2.775 distance between trucks is
assumed and it is seen that the wind sheds off completely. This provides no air
resistance for the trail vehicle, and there would be no change in the drag force. Thus,
the concept of platooning fails in this case (Fig. 5).
With the acquired results from the graph, it is concluded that the principle of
platooning works distinctly for different vehicle ranges that vary from commercial
vehicles to heavy vehicles. By examining the trend of the graphical plots, it can
be understood that the platooned vehicles do not experience the expected outcome,
i.e. reduced drag, eventually leading to lesser fuel consumption, in a same manner.
Furthermore, in this case for platooning of trucks, in spite of having aerodynamic
advantages, the principle does not work beyond speed range of 60 kmph. Analysing
the cases deeply a factor called drag bucketing comes into picture. Drag bucketing is
the phenomena where the drag coefficients approach towards the minimum value for
Fig. 5 a Velocity contour for velocity of 70 kmph and distance between trucks is 2.77 m,
b streamlines for velocity of 70 kmph and distance between trucks is 2.77 m
a certain speed range and increases beyond. Here, in our study, the drag bucketing
occurs beyond 60 kmph. So in order to get better results and good fuel economy, this
kind of semi-aerodynamic trucks performs better, and platooning is effective from
the speed range of 40–50 kmph.
5 Conclusions
Platooning is effective at the velocity of 40 kmph compared to the velocities of 50

kmph and 60 kmph which is mostly followed speed for the container trucks in India.
Thus, it is proven that the individual truck which moves at 40 kmph individually can
be platooned for better fuel efficiency (Fig. 6).
Fig. 6 a Plot between drag

coefficient and velocity of
the truck, b plot between
combined drag coefficient of
both the trucks and distance
between the trucks
References
1. Bhoopalam AK, Agatz N, Zuidwijk R (2018) Planning of truck platoons: a literature review and
directions for future research. Trans Res Part B: Methodological, Elsevier 107(C):212–228
2. Rajasekarababu KB, Vinayagamurthy G, Selvi Rajan S (2019) Experimental and computational
investigation of outdoor wind flow around a setback building. Build Simul 12:891–904
3. Rajasekarababu KB, Vinayagamurthy G (2020) CFD validation of wind pressure distribution
on a tall building under the influence of upstream terrain. Prog Comput Fluid Dynam Int J
20(5):284–298
4. Siemon M, Smith P, Nichols D, Bevly D, Heim S (2018) An integrated CFD and truck simulation
for 4 vehicle platoons (No. 2018-01-0797). SAE Technical Paper
5. Chowdhury H, Moria H, Ali A, Khan I, Alam F, Watkins S (2013) A study on aerodynamic drag
of a semi-trailer truck. Procedia Eng 56:201–205
Review of Pool Boiling Heat Transfer
During Power Transients
John Pinto and Janani Srree Murallidharan
1 Introduction
Boiling can be categorized into steady boiling and transient boiling. Transient boiling
can further be subcategorized into boing under power transience, flow transience and
pressure transience [1]. There is a shallow understanding of boiling under challenging
conditions of power transience i.e. under conditions of rapid heating such as those
encountered in the nuclear reactors. Transient boiling is also encountered in appli-
cations like cryogenics, metallurgy processes, during the starting and stopping of
heat exchangers. With advances in the knowledge in this area, we will be able to
design safe and more efficient nuclear power plants. In all these applications power
transients are most commonly encountered. With the above objective in mind this
review is limited to boiling under power transience.
2 Review of Literature
The literature can be best interpreted by looking at the effect of transient heating on
the different aspects of boiling. A systematic study of the same has been done and
reported in the following subsections.
J. Pinto · J. S. Murallidharan (B)

Department of Mechanical Engineering, IIT Bombay, Powai, Maharashtra 400076, India
e-mail: js.murallidharan@iitb.ac.in
400 J. Pinto and J. S. Murallidharan
2.1 Mechanisms of Heat Transfer
Su
et al. [2] conducted
transient pool boiling experiments under exponential heating
q = q0 × et/τ . Analytical treatment of the transient conduction in the fluid was
done. It was found that, when the bulk of the fluid is at saturation condition, the
thickness of the superheated layer (SHL) is large. This SHL thickness decreases
with increasing degree of subcooling. Depending upon the thickness of the SHL,
two different phenomena were observed. When the bulk of the fluid is at saturation
condition, the SHL thickness is large. The growth rate of the bubbles is proportional
to the thickness of the SHL. This resulted in large bubbles being generated which
grow by taking heat from the SHL. It was also found that in such condition, the
wall superheat and heat flux at the onset of nucleation is small. Thus the number
of bubbles formed on the surface is few and the heat transfer is governed by the
formation of a few large bubbles.
On the other hand; when the bulk of the fluid is at subcooled condition, it was
found that the superheated layer (SHL) thickness is small, the wall superheat and heat
flux at the onset of nucleation is large. The bubble generated cannot grow beyond
the SHL and condense. This resulted in the formation of small size bubbles. Thus
large number of bubbles are formed on the surface and the heat transfer is governed
by the formation of a large number of small sized bubbles. It was observed that these
bubbles undergo a cycle of growth and condensation at large frequencies. This gives
an enhanced rate of heat transfer which is ten times more than other cases.
Two different boiling curves were also observed under transient heating: with
temperature overshoot and without temperature overshoot. This difference is
observed due to the different conditions of subcooling of the fluid and the different
periods of power excursion (τ).
Figure 1 shows one representative case. As the heating commences, the wall
superheat increases with increase in heat flux. In this phase the wall superheat follows
Fig. 1 Typical boiling

curves with temperature
overshoot for the test at 5 ms
and 75 K subcooling [2]
Review of Pool Boiling Heat Transfer During Power Transients 401
closely the trend predicted by the analytical transient conduction solution in the fluid.
At an instant when critical wall superheat is reached, bubbles begin to nucleate and
Onset of Nucleate Boiling (ONB) occurs. The boiling curve starts to deviate from the
transient conduction curve. With the increase in heat flux and wall superheat more
bubbles start to nucleate. The heat transfer associated with the bubble formation
becomes large and a sudden rise in the heat flux is observed. This is denoted as Onset
of Boiling Driven heat transfer (OBD). Part of the heat supplied by the heater goes
to the water while the remaining part goes to the substrate. After OBD, the heat flux
to water keeps increasing while the heat flux to the substrate keeps decreasing. With
further enhancement in boiling, a point is reached when the heat flux to the substrate
becomes zero. This is the point where a local maxima of temperature occurs. This
is known as temperature overshoot (OV). At this stage the boiling heat transfer is so
rapid that the heat transferred to water exceeds the heat supplied by the heater. This
difference in heat is taken from the substrate. This results in a decrease in the heater
temperature. The temperature of the heater keeps decreasing till the superheated layer
around the bubbles is exhausted. This slows down the bubble formation. By this time
the exponential increase of power catches up to the requirement of heat flux. A new
balance between the heat transferred to water and substrate is attained and boiling
progresses towards Fully Developed Nucleate Boiling (FDNB). In the case of low
degree of subcooling (25 K), the boiling phenomenon is same till OBD. In this case,
OV is not observed. The reason for this is that boiling is not vigorous enough and
the heat flux to water never increases beyond the heater input. This boiling curve is
observed at low periods and low subcooling.
It can be safely concluded that the phenomenon observed on the macroscopic
scale can all be linked to the formation, growth and departure of a single bubble.
Thus single bubble studies will shed more light on the underlying physics involved
in transient heating conditions.
2.2 Effect of Subcooling and Time Period
Su et al. [2] solved transient conduction equation in the fluid to obtain the time
dependent temperature profile in the fluid. Their mechanistic ONB model is given
below:

εw rc
qw,ONB = √ [Tsat ( patm + 2σ/rc ) − Tbulk ]exp √ (1)
τ aw τ

rc
Tsat,ONB = [Tsat ( patm + 2σ/rc ) − Tsat ( patm )]exp √
aw τ

rc
+ Tsub exp √ −1 (2)
aw τ
The model clearly shows that the wall superheat and heat flux at ONB increases
with increasing degree of subcooling and decreasing time period. The trend of the
experimental results is captured well by the model.
Walunj and Sathyabhama [3] also conducted transient pool boiling under expo-
nential heating but on a thick heater. In their experiments it was also observed that the
heat flux and wall superheat at ONB increased with increasing transience. Similar
results were also observed by Sakurai et al. [4] and Park et al. [5].
Su [6] proposed that with increasing transience, the time lag between the increase
in power and the development of thermal boundary layer becomes significant. Thus
for the same cavity, a shorter period of heating results in higher heat flux and superheat
at ONB. Though this sounds like a good justification of the observed phenomenon,
a thorough experimental investigation of the time scale of heating and of thermal
boundary layer formation is needed.
On the macroscopic level, degree of subcooling and time period is seen to control
only the boiling onset conditions. However, a link can be made with the studies of
the previous section. Degree of subcooling and the time period control the thickness
and the rate of formation of the SHL. And SHL as seen before controls the bubble
formation and hence the heat transfer rate.
2.3 Effect of Different Heating Techniques
Visentini et al. [7] conducted transient boiling experiments in a setup that was specif-
ically designed to simulate the rapid transient boiling behaviour on nuclear fuel rod.
The transient heating was achieved by supplying square or triangular current signal.
The boiling curve changed depending on the signal shape and power levels.
For the rapid rise in power in case of the square pulse, the experimentally observed
wall temperature is lower than the predicted value. This shows that the heat transfer
mechanism observed is superior to transient conduction. It was proposed that a
combined effect of transient conduction and convection may be the mechanism of
heat transfer. Thus we can conclude that the way in which the wall is heated governs
the temperature distribution at the wall and in the fluid which in turn governs the
nucleation phenomenon.
It was observed that the ONB time decreased with increase in dT /dt. To have a
better understanding of this, the amount of energy transferred to the liquid till ONB
was evaluated by integrating the heat flux till the ONB time.
tONB
E ONB = ∅(t)dt (3)
0
This data suggests that the energy ranges between 104 J/m2 to 4 × 104 J/m2 . This
range is quiet narrow for the range of the experimental conditions. This condition is
even satisfied by rapid transient tests (heating time 0.2–0.5 s) at high power (1500 W).
This shows that a minimum threshold of energy is required to initiate boiling which is
not dependent on the heating rate. This is a strong statement and needs to be analysed
in detail.
Specific tests with very short heating time (0.1 s) at very high power (1500 W)
were also conducted. For such tests the wall temperature profile shows no change
in the slope even at ONB. The ONB time is found by high speed video. The rate
of wall temperature rise for these tests is very high (close to 1800 K/s). The wall
temperature at ONB was 80 °C (Fluid used for experiments was HFE3000, T sat =
34 °C at atmospheric pressure). Within a few milliseconds, the temperature rose to
100 °C exceeding the spinodal limit of 98.5 °C. This caused an explosive vapour
formation. E ONB in this condition was found to be 16.5 × 104 J/m2 , which is an
order of magnitude larger than the other tests. This shows that for such tests the
underlying phenomenon is different and needs to be explored.
2.4 Effect of Heater Size
Duluc et al. [8] conducted transient boiling experiments using stepwise heat
generation. Two heater samples were analyzed: a thick flat sample and a thin wire.
For the thick sample it was found that with the increase in stepwise heat generation
rate, the ONB superheat increases and the time required for ONB decreases. It was
suggested that transient heating does not alter the ONB conditions significantly in
case of a thick flat sample.
Walunj and Sathyabhama [3] also conducted transient pool boiling experiments
on a thick sample. In their study it was observed that the ONB conditions changed
significantly in transient heating. ONB superheat varied by 6 °C during transient
heating. Thus a detailed study is needed on transient boiling on a thick sample.
Experimental data expressed in the form of boiling curve shows that with change
in transience, there is no change in the convective boiling regime. Due to the large
thermal inertia of the thick flat sample, the heat flux supplied to water is controlled
even during sudden step heat input. A higher heat flux shifts the boiling curve to the
right. From the study it can be concluded that transient step heating has little effect
on the boiling phenomenon on a thick flat sample. For the case of thin wire, the
convective regime lasts longer and heat flux at ONB is higher in comparison to the
thick flat sample. This is due to the thin geometry of the wire which enhances natural
convection. It was also noted that the wire has a low thermal inertia and all the heat
generated is released almost instantly to the water. Thus the time scales involved is
much smaller in comparison with the thick flat sample.
Few things to note from this study are that the effect of transience is more
pronounced in the case of thin wire than in the case of thick flat sample. The results
obtained on the thick flat sample by the two studies are contradictory to each other.
This may be due to the outdated measurement techniques used in this study that allow
for only average temperature measurement of the surface. The reason for the increase
in ONB superheat with heating step is not clearly analyzed and only a hypothesis is
proposed. Microscale visualization and single bubble studies will help to understand
this phenomenon.
2.5 Effect of Surface Roughness
Walunj and Sathyabhama [3] conducted transient pool boiling experiments on a

thick sample. Exponential heating (Q ∝ eγ ) was employed and boiling was studied
at different pressures. Effect of surface roughness was also studied by creating a
range of surfaces with unidirectional scratches of different roughness values. It was
observed that the boiling curve shifts to the left with increasing roughness value.
This shows that there is enhancement in heat transfer for the same wall superheat.
This enhancement was also observed by Duluc et al. [8]. Walunj and Sathyabhama
[3] found that the absolute depth of the scratches (Rz ) plays an important role in
boiling incipience. Rz value was found to increase with increasing roughness. Thus
the number of cavities on the surface increased which resulted in a decrease in the
ONB superheat. The unidirectional scratches also contributed significantly to the
enhancement of heat transfer by acting as narrow passages supplying liquid to the
active cavities. The scratches acted as capillary tubes which would replenish the
cavities with liquid by capillary action. This mechanism enhanced heat transfer and
also delayed the CHF phenomenon.
Here, the upper limit of this enhancement with increasing roughness is not
discussed. Increasing roughness may increase the contact area thereby increasing the
surface tension force holding the bubble. This may reduce the departure frequency
and lead to a decrease in the CHF. This can be ascertained by studying the effect of
roughness on the liquid–vapour interface movement.
3 Summary and Conclusions
• A lot of work has been done on transient boiling but with a lot of incongruence.
Different authors work with different heating techniques, heating time, heater size,
subcooling and so on. A coordinated research on the effect of different heating
techniques on boiling under a range of conditions is needed.
• Literature review also highlighted that heater size contributes significantly to the
observed heat transfer phenomenon. Su et al. [2] observed that the heat transfer
coefficient increased with decreasing time period in case of a thin film heater.
Whereas Walunj and Sathyabhama [3] observed that heat transfer coefficient
decreased with decreasing time period in case of a thick heater. This ambiguity
can be addressed by studying the local bubble structure on the two heaters.
• Initial heat generation prevalent before the onset of transience is a very important
parameter which has not been given due attention in the literature. It will have
a huge contribution on the outcome of the boiling phenomenon after transient

heating and needs to be looked at in detail.
These conclusions point to a need for a systematic study of transient boiling. This
study should look at the single bubble growth using microscale visualization and
infrared thermometry. It should account for the effect of different heating techniques,
subcooling and roughness. Such a study will bridge the gap between the vast research
and shed light on the physics which is still not uncovered. It will also help to develop
models for prediction of heat transfer coefficient and CHF.
References
1. Walunj A, Sathyabhama A (2016) Dynamic pool boiling heat transfer due to exponentially
increasing heat input—a review. Proc Technol 25:1137–1145
2. Su G-Y, Bucci M, McKrell T, Buongiorno J (2016) Transient boiling of water under exponentially
escalating heat inputs. Part I: pool boiling. Int J Heat Mass Transf 96: 667–684
3. Walunj A, Sathyabhama A (2018) Transient CHF enhancement in high pressure pool boiling on
rough surface. Chem Eng Process 127:145–158
4. Sakurai A (2000) Mechanisms of transitions to film boiling at CHFs in subcooled and pressurized
liquids due to steady and increasing heat inputs. Nucl Eng Des 197:301–356
5. Park J, Fukuda K, Liu Q. Transient CHF phenomenon due to exponentially increasing heat
inputs. Nuclear Eng Technol 41(9): 1205–1214
6. Su G-Y. Experimental study of transient pool boiling heat transfer under exponential power
excursion on plate-type heater. Unpublished Master’s Thesis. Massachusetts Institute of
Technology
7. Visentini R, Colin C, Ruyer P (2014) Experimental investigation of heat transfer in transient
boiling. Exp Thermal Fluid Sci 55:95–105
8. Duluc M-C, Stutz B, Lallemand M (2004) Transient nucleate boiling under stepwise heat
generation for highly wetting fluids. Int J Heat Mass Transf 47:5541–5553
Studies on Ignition Over Pressure
Characteristics Related to Typical
Launch Vehicle Lift-off
R. Venkatraman, Sathiyavageeswaran Sankaran, V. Venkata Ramakrishna,

Debopam Das, and T. Sundararajan
1 Introduction
During the ignition transient of solid rocket motors, unsteady pressure waves are
generated in the combustion chamber. These pressure waves emanate through the
conventional convergent–divergent (C-D) nozzle and propagate outward and interact
with the launch vehicle and its surroundings [1–5]. This will result in spatial-
and time-dependent pressure disturbances. The pressure disturbance (dP) could be
defined as,
dP(x, y, z, t) = P(x, y, z, t) − Pα
where P(x, y, z, t) is the transient pressure at any location (x, y, z) and time (t), and
Pα is the atmospheric pressure. This pressure disturbance is termed as the ignition
over pressure (IOP). It is typical shock front, which propagates to a large distance
R. Venkatraman (B)
Vehicle Assembly and Launch Facilities, SDSC SHAR ISRO, Sriharikota, India
e-mail: rvraman@shar.gov.in
S. Sankaran · V. V. Ramakrishna
Test Facilities, SMP&ETF, SDSC SHAR ISRO, Sriharikota, India
V. V. Ramakrishna
e-mail: vvrk@shar.gov.in
D. Das
Department of Aerospace Engineering, IIT Kanpur, Kanpur, India
T. Sundararajan
Department of Mechanical Engineering, IIT Madras, Chennai, India
e-mail: tsundar@iitm.ac.in
408 R. Venkatraman et al.
Fig. 1 Typical IOP signature of solid rocket motor
initially at supersonic speeds but gradually reduces to subsonic speeds with time and
also the magnitude decays accordingly. The typical IOP signature is shown in Fig. 1.
As shown in Fig. 1, the peak pressure magnitude of an IOP pulse is a strong
function of motor pressure rise rate (dp/dt) and the frequency is found to be scaling
for different class of motors as defined by Strouhal number (Str = f * De /U e ), where
‘f ’ is the IOP relaxation wave frequency, De is the nozzle exit diameter, and U e is the
nozzle exit velocity. It is imperative to understand the IOP characteristics for various
class of motors for which an effort is being made to measure the signatures. Also,
time series analysis has been performed to understand its propagation characteristics.
Further, in order to understand the effect of ignition skew between a pair of boosters,
flow visualization has been captured for a typical case.
2 Experimental Methodology
For identification and characterization of IOP signatures of different class of solid

rocket motors, unsteady pressure measurements were performed during each class of
rocket motor static firing. Measurements were taken at several locations spatially at
specific intervals scaled to nozzle exit diameter. The peak pressure and frequency of
IOP pulses of three classes of solid rocket motors with nozzle exit diameters of 0.14
m, 0.95 m, and 3 m have been studied in order to study the scaling effect, and a time
series analysis has been performed to understand the propagation characteristics of
IOP pulse. Further, a cold flow simulation of twin solid rocket boosters is performed
to understand the effect of skew in IOP generation at ignition transient.
Studies on Ignition Over Pressure Characteristics Related … 409
2.1 IOP Measurements During Small, Medium, and Large

Solid Rocket Motors
Experimental setup for the measurement of IOP of solid rocket booster ignition
transient looks as shown in Fig. 2. The rocket motor is fired in the horizontal condition,
and the unsteady pressure measurements are taken in the midplane of the nozzle at
different spatial locations.
Kulite make unsteady pressure measurement sensor (model: XCE-062 series)
with 5 psig measurement range was used for the measurements. For measuring the
motor combustion chamber pressure, Druck make pressure transducer is utilized.
Figure 3 shows the locations of the Kulite sensor (field unsteady pressure) during
the static firing of rocket motor with 0.14 m nozzle exit diameter at a radial location
of 20De from nozzle exit and at various angles from nozzle axis. The IOP pulse
Fig. 2 Experimental setup for rocket motor static firing
Fig. 3 IOP measurement

locations at various spatial
locations
recorded is found to be happening at the peak chamber pressure rise rate and it has a
positive peak pressure followed by negative relaxation waveform as shown in Fig. 4.
However, the negative relaxation waveform is affected by the ground reflections,
resulting in superimposed peaks on it.
As indicted in Fig. 4, the initial peak (first event) is due to the igniter firing and
the second event results in a major IOP pulse which is matching in time scale to the
peak chamber pressure rise. The frequency of the relaxation waveform of major IOP
pulse is found to be ~625 Hz.
For intermediate class of rocket motor, i.e., with nozzle exit diameter of 0.95 m and
for large solid rocket motor, i.e., with nozzle exit diameter of 3 m, IOP measurements
were performed at the similar location as that of small rocket motor shown above.
First event
Second event
Fig. 4 IOP measurement at 20De and 30 degree location for a small rocket motor, b medium-scale
rocket motor, c large rocket motor
The signature of the IOP waveform recorded for all three motor cases is shown in
Fig. 4. Peak pressure of IOP pulse recorded is 48, 80, and 100 mbar, respectively,
for the small, medium, and large rocket motors, respectively, for the corresponding
chamber pressure rise rates, which are of the same order. Though the pressure rise rate
is comparable even for the large and small motors, the IOP peak pressure recorded
is more for the large motor.
The variation in the IOP peak amplitude at a similar location may be attributed
to the change in nature of IOP generation as the scale of the motor changes. The
apparent IOP source will be elongated axially as the scale increases as observed by
Varnier [5], and this might have contributed to the increased IOP magnitude at a
given radial location.
The frequency of the relaxation waveform observed in the case of medium class
rocket motor and large rocket motor firing is in the order of 100 and 30 Hz, respec-
tively. As per the Strouhal number criteria, from the small rocket motor case results,
the frequency of IOP waveform for intermediate class motor shall be ~95 Hz, which
is matching very close to that of the experimentally recorded value.
2.2 Time Frequency Analysis of IOP Signal
The IOP measurements during solid rocket motor horizontal firing in general will
have different temporal characteristics compared to the rocket motor firing on the
launchpad. To understand this in a better way, IOP measurements were performed
during the 1:12.8 scale model hot test of a typical launch vehicle. The scale model
that consists of twin solid rocket boosters with nozzle exit diameter of 240 mm is
fired simultaneously. A piezo-electric field microphone was placed near the bottom
of launch pedestal facing the nozzle exhaust jet. The waveform recorded during the
ignition transient of the firing is analyzed using wavelet transform.
The initial IOP pulse seen can be attributed to the igniter firing event, as there is
no nozzle shutter in this motor. The frequency of this IOP pulse is found to be of the
order of ~250 Hz. Subsequent to the igniter firing, due to abrupt expansion of the
combustion products, an apparent source formed at the downstream of the nozzle
acts as a source of high-intensity pressure waves or shock waves. The shock waves
propagating into the flame trench are known as source over pressure (SOP), and the
shock waves coming out of the flame trench are known as duct over pressure (DOP).
Time series analysis has been performed to decompose these events using Morley
wavelet. Figure 5 illustrates the different shock waves originating at different time
scales with varying frequency components.
From Fig. 5, it is found that the IOP pulse propagates with 250 Hz trailing wave
right at the instant of igniter firing, whereas the SOP shock waves travel with 30 Hz
trailing component and the two events occurring at different time intervals can be
attributed to the ignition delay between the two motors.
Fig. 5 Time frequency analysis of transient signal
2.3 Simulation of Twin Rocket Booster Ignition Transient
As evident from the numerous experimental measurements, unsteady pressure fluc-

tuations arising at ignition transient of a solid rocket booster have complex signature
and involve temporal shifted events. When two solid rockets boosters are firing
simultaneously, analysis of the ignition transient pressure fluctuations in the field
will be a further more complex job. A cold flow experiment using shock tubes has
been performed at IIT Kanpur under ISRO-SDSC SHAR RESPOND program with
different time lags of two diaphragm ruptures to simulate the ignition delay between
the motors. Figure 6 shows the test setup used for simulation of twin rocket motor
ignition transient.
Flow visualization of the twin nozzle start-up has been performed to capture the
interaction of the jets/shock fronts. For the selected case of launch vehicle model, a
time gap of less than 5 ms is required for the two jets to interact before impinging
on the launchpad jet deflector. Figure 7 shows the interaction between the two jets
before impinging on the launchpad. The shock front and acoustics measured by the
field microphone for this case are presented in Fig. 8. It is clear that the right shock
tube shock front is reaching the microphone first and has the highest amplitude.
The shock front from the left shock tube emanates with a time delay and it reaches
the corresponding microphone on its side after a time delay with two diffractions
and hence leads to a smaller amplitude.
Fig. 6 Scaled model setup for twin rocket motors ignition transient on launchpad
Fig. 7 PIV flow visualization of twin nozzle case with 1.3 Mach t = 1600 µs and time between
bursting is 200 µs
IOP characteristics of three classes of solid rocket boosters have been identified and
found that the peak amplitude varies across the scales but the frequency content
is following the Strouhal number shifting. IOP and SOP shock waves generated
during the ignition transient of a solid rocket booster have been analyzed using time
frequency analysis and its rarefaction frequencies are identified for a typical class of
Fig. 8 Microphone responses of the twin nozzle case with time delay
motor. A shock tube experiment with flow visualization has been performed on the
twin nozzle case for simulating the possibility of shock front interaction for various
Mach numbers and it is found that the interaction is minimal.
4 Conclusion
IOP is a shock front associated with the ignition transient of a solid rocket booster.
In order to predict the IOP of a full-scale motor through scaled model experiments,
the chamber pressure rise rates and the effect of apparent source location are to be
considered. The frequency of IOP among the scales is found to be closely following
the Strouhal criteria. Time frequency analysis of the transient signal clearly gives
the various components associated with the IOP generation and will be useful in
assessing the impact on the launch vehicle dynamic behavior.
Acknowledgements Authors thank AD, SDSC SHAR & Director SDSC SHAR for their technical
support.
References
1. Broadwell JE, Tsu CN (1967) Transient pressures caused by rocket start and shutdown in ducted
launchers. J Spacecraft 4(10):1323–1328
2. Dougherty NS, Nesman TE, Guest SH (1982) Shuttle SRB ignition overpressure: model suppres-
sion test program and flight results. JANNAF 13th Plume Technology Meeting Houston, Texas,
USA, CPIA Publication 357:217–243
3. Lai S, Laspesa FS (1982) Ignition overpressure measured on STS lift-off and correlation with
subscale model tests. JANNAF 13th Plume Technology Meeting Houston, Texas, USA, CPIA
Publication 357:207–216
4. Canabal F, Frendi A (2002) A computational study of the ignition over-pressure for launch vehi-
cles. In: AIAA 2002-2541, 8th AIAA/CEAS aeroacoustic conference and exhibit, Brekenridge,
Colorado, USA
5. Varnier J (2002) Blast wave at ignition of the rocket engines. ISMA 2002, Louvain (Belgique),
TP 2002-175
Revisiting T+ Versus Pry+ Scaling
in Turbulent Flows With 0.01 ≤ Pr ≤ 760
Using RANS
Kunal Kumar, Kuldeep Aswal, Manish Kumar Rajwade, Ankit Kumar,

S. S. Mondal, and L. Chandra
Nomenclature
α Thermal diffusivity of fluid

αt Turbulent thermal diffusivity
i, j Indices for Newtonian summation
k Von Karman constant (=0.41) in wall laws
v Kinematic viscosity
vt Eddy viscosity
ϕ A flow variable
Pr Prandtl number
Pr t Turbulent Prandtl number
Re Reynolds number
u1 Time-average velocity component along x i
U∞ Free stream velocity
U∞ Free stream temperature
x 1 or x Axial direction
x 2 or y Cross-stream direction
T Time-averaged temperature
T+ Non-dimensional statistical temperature
U∞ Free stream temperature
y+ Non-dimensional height
() Fluctuating quantities
K. Kumar · K. Aswal · M. K. Rajwade · A. Kumar (B) · S. S. Mondal · L. Chandra

Department of Mechanical Engineering, Indian Institute of Technology (BHU) Varanasi,
Varanasi, Uttar Pradesh 221005, India
e-mail: ankit.kumar.mec15@iitbhu.ac.in
418 K. Kumar et al.
1 Introduction
The turbulent flow over a smooth flat surface is encountered in numerous industrial
and environmental applications. This includes cooling/heating of metals during the
heat treatment and removal of heat from the surface of a fuel pin in the nuclear
reactor; see, e.g., [1, 2]. The analysis of hydraulically and thermally developing
boundary layer is necessary to investigate such practical problems. Analysis of a
hydraulic boundary allows estimating the shear stresses on the surface of interest.
The turbulent heat transfer, for instance, plays the major role in limiting the surface
temperature of fuel rod in nuclear reactor. However, the prediction of heat transfer
in thermally and hydraulically developing boundary layer is still a challenge owing
to the involved complexity [3]. The law of the wall has been extensively used to
predict wall shear stresses and turbulent heat flux; some of these are being discussed
in [1–6]. Suga et al. [7] used turbulence models for validating the results obtained
with such wall functions, like Craft et al. [8] with flow over a flat plate. Vigodorovich
et al. [9] deduced based on analysis two new formulations of u+ and T + with y+ in the
respective boundary layers. Most of the investigations have been performed using the
commonly encountered fluid such as air (with Pr = 0.7) and water (with Pr = 7). For
a Prandtl number ranging from 0.025 to 100, investigations of turbulent heat transfer
using the direct simulation concluded the importance of Pr >> 1 or Pr << 1 [10, 11].
They demonstrated the T + varies with Pry+ in the heat conduction region for 0.025
≤ Pr ≤ 100 and suggested the need for corrections to wall laws depending on Pr.
Dong et al. [12] used large eddy simulation (LES) technique coupled with dynamic
subgrid scale model for predicting the turbulent heat transfer at 0.1 ≤ Pr ≤ 200 in the
fully developed channel flows. Grötzbach [4] has outlined the challenges associated
with the modeling of turbulent liquid metal flow having Pr << 1 and proposed the
use of variable turbulent Prandtl number instead of a fixed value. Therefore, it may
be argued that the adaptation of wall laws with Pr << 1 and Pr >> 1 needs special
attention. This includes various oils with a very high Pr ~ 500–700 and is widely
employed in applications such as concentrated solar thermal system. In practice, the
RANS-based approaches are widely employed and therefore a revisit to the scaling
of T + with Pry+ is necessary. With this motivation, the paper aims at investigating
the thermally and hydraulically developing turbulent boundary layer over a flat plate
with 0.01 ≤ Pr ≤ 760. The findings allow the intended reassessment in an extended
Pr range beyond the status quo. The paper is organized as follows: Section II deals
with the methodology including numerical framework, Section III deals with results
and discussion including validation and finally the conclusions with future outlook
are discussed.
Revisiting T+ Versus Pry+ Scaling … 419
2 Methodology
The two-dimensional hydraulically and thermally developing turbulent, steady,

incompressible fluid flow over an impermeable smooth flat plate is analyzed. This
assumes its infinite span allowing the use of a single plane with symmetry. The
considered rectangular domain is having a length of 1 m and a height of 0.2 m. The
schematic and the associated coordinates are shown in Fig. 1. The height is consid-
erably larger than the estimated thermal boundary layer thickness (~0.09 m) along
the plate length for the highest Reynolds number, Re = 25 × 105 and the lowest
Prandtl number, Pr = 0.01. The selected length is sufficient to ensure turbulent flow
over a wide length along the stream-wise (x). A uniform velocity is adopted as the
inlet boundary condition at x = 0. The free stream velocity and temperatures are
represented by U ∞ and T ∞ , respectively.
To compute the two-dimensional velocity and temperature fields, the Reynolds-

averaged Navier–Stokes (RANS) equations along with the statistical energy equation
are solved over a flat plate with the zero pressure gradient. The statistical equations
are given by:
∂u i
= 0, i = 1, 2 (1)
∂ xi

∂u i ∂ ∂u i
uj = v − u j ui , (2)
∂x j ∂x j ∂x j

∂T ∂ ∂T
uj = α − u j T .

(3)
Fig. 1 Schematic of the computational domain with relevant dimensions (not to scale)
420 K. Kumar et al.
where u 1 and u 2 are the mean velocities along the stream-wise (x orx 1 ) and cross-
stream (y-direction orx 2 )-direction, respectively;T is the mean temperature;v is the

kinematic viscosity;α is the thermal diffusivity;() denotes the fluctuating compo-
nents. In Eq. (2) and (3), the Reynolds stress and turbulent heat flux are modeled
using the gradient approximation as,
∂u i
−u j u i = vt , (4)
∂x j
∂T
−u j T = αt . (5)
∂x j
where vt and αt are the turbulent eddy viscosity and thermal diffusivity,
respectively.
2.2 Computational Framework
The mesh generation considers that the variation of a flow variables (ϕ) along the
stream-wise direction is much smaller than that of the cross-stream direction, i.e.,
∂ϕ ∂ϕ
. (6)
∂x ∂y
Therefore, a wall-resolved mesh with y+ ≤ 1 is desired along the y direction

to avoid the use of wall function. A stretching factor of 1.2 is maintained along
this direction. Furthermore, the flow is made to attain the turbulent condition at an
early axial position along the x-direction. Thus, without loss of generality, a uniform
meshing is adopted along this direction. Based on these considerations, structured
mesh was generated. An example of the generated mesh is shown in Fig. 2.
The well-known Spalart–Allmaras eddy viscosity transport model as in ANSYS-
FLUENT is solved for the present investigation. This is given by,
Fig. 2 Generated mesh over the entire domain (a) and near the wall (b)
∂ ∂
(ρ + νt ) + (ρνt u i ) = G ν
∂t ∂ xi

1 ∂y ∂νt ∂νt 2
+ (ρν + ρνt ) + Cb2ρ − Yν (7)
σνt ∂ x j ∂x j ∂x j
where G ν is the production of eddy viscosity, and Yν is the destruction of

eddy viscosity that dominates in the near-wall region due to wall blocking and
viscous damping. The redistribution of eddy viscosity is modeled assuming the
gradient approximation. The coefficients σνt and Cb2ρ are set to the default values
for simplicity. Indeed, a more detailed Reynolds stress transport equation-based
modeling approach may be preferred. However, considering our objective to assess
the applicability of the scaling T + versus Pry+ with 0.01 ≤ Pr ≤ 760 using the wall-
resolved mesh, the model should provide a clue to the same. Once this is found,
the effect of turbulence models on the scaling may be assessed at a later stage. The
expressions for some of the well-known wall functions as in Table 1 are considered
in the analysis. For the velocity boundary layer, the standard wall law and for the
thermal boundary layer three different wall laws, including the standard, are selected.
For the numerical analysis, momentum and energy equations are solved using
the second-order upwind scheme. The turbulent eddy viscosity transport equation is
solved with the first-order upwind scheme. For the pressure and velocity coupling,
SIMPLE approach is preferred for simplicity. Uniform velocity and free stream
temperature are applied as the inlet boundary conditions. No slip and isothermal
boundary conditions are applied on the flat plate. Zero gauge pressure is employed
Table 1 Expressions for wall function

Velocity u+ = 1
k ln y + + B, where B = 5.1 – 5.5
Temperature Kader ( Pr ≤ 0.7)
T+ ≈ Pr t
k ln y + + A(Pr),
2
A(Pr ) = 13 Pr 3 −7
Standard
T + = Pr y + , 0 ≤ y + ≤ 5

T + = Pr k1 ln E y + + P , y + ≥ 13
t
3
Pr
4 −0.007 Pr
P = 9.24 Pr
Prt − 1 1 + 0.28e t
Bejan (0.5 ≤ Pr ≤ 5.5)

⎧ ⎫
⎪
⎪
⎨ Pr y + y + < yC+S L ⎪
⎪
⎬
+ + Pr t y+ + +
T = Pr yC S L + k ln y + y > yC S L
⎪
⎪ C SL ⎪
⎪
⎩ ⎭
where Prt ∼
= 0.9, k ∼
= 0.41, yC+S L ∼
= 13.2
422 K. Kumar et al.
Table 2 Conditions for

Fluid U∞; T ∞ Plate temperature (K)
numerical simulation
Sodium 1.63 m/s; 400 K 450
Air 40 m/s; 300 K 350
Water 2.13 m/s; 300 K 350
Ethylene glycol 26.2 m/s; 300 K 350
Therminol 14 m/s; 300 K 350
Table 3 Generated mesh for

Mesh x-direction y-direction
simulation with air
Coarse 250 25 (y+ ~ 2)
Medium 500 50 (y+ ~ 1)
Fine 1000 100 (y+ ~ 0.5)
at the outlet and symmetry boundary condition is applied at the free surface. The
inlet velocity is chosen to ensure that the transition to turbulence takes place at x
= 0.2 m. The temperature-dependent thermo-physical fluids properties are used in
these simulations. For convergence, a relative change of 10–4 is achieved for the
continuity equation and beyond 10–6 for the momentum and energy equations. The
employed remaining conditions for simulations are summarized in Table 2.
Mesh Independence
Mesh independence studies are performed separately for each of the considered
fluids. For instance, the generated meshes for air (Pr = 0.7) are shown in Table 3.
The y+ values for the first grid point near the wall are generally maintained within 2
to capture the viscous sublayer.
Rectangular mesh is generated to mitigate the issues pertaining to gradient calcu-
lation. The aspect ratios are in the range of 100–300 for the different meshes and
for all the fluids, which is slightly at the higher side. Figure 3 provides a compar-
ison between the computed u+ and T + with the different near-wall region resolving
meshes at an axial position x = 0.6 m and at a Rex = 15 × 105 . These are well
within turbulent regime and allow a fair comparison. Computed values with medium
and fine meshes are identical and therefore the medium mesh is selected for further
analysis.
For validation and subsequent analysis, the medium mesh resolution is selected with
air as fluid. A comparison between the computed and standard wall law analyzed
values of u+ and T + is provided in Fig. 4 at an axial position x = 0.6 m and at a Rex =
15 × 105 . This clearly illustrates that the generated mesh is capable of capturing the
Fig. 3 Variation of u+ and 25

T + with y+ for air with Pr =
0.7 at x = 0.6 m and Rex = Medium mesh
20
15 × 105 Fine mesh
15 Coarse mesh
u+
10
0
1 10 y+ 100 1000
20
Fine mesh
Coarse mesh
15 Medium mesh
T+
10
0
1 10 100 1000
y+
viscous sublayer extending up to y+ ~ 5. Also, the computed u+ values agrees quite

well with that of the standard log-law, which is consistent with literature. Similarly
the computed values of T + are assessed. For this purpose, three different log-laws,
viz. standard, Bejan, and Kader are selected. The analysis reveals that the RANS
approach is indeed capable of predicting the statistical non-dimensional temperature
T + with a reasonable accuracy. Therefore, the selected computational scheme is
applied for analyzing u+ and T + for liquid metal to oil as fluid having 0.01 ≤ Pr ≤
760.
The RANS computed and wall law-based T + values are plotted against Pry+ in
Fig. 5. It is usually a customary to use y+ instead of Pry+ , however a close inspection
of the values reveals that using this scaling approach not only allows understanding
the heat conduction in the near-wall sublayer but collapsing the start of log-law region
within a narrow band (15 ≤ Pry+ ≤ 40) as for instance evident from Fig. 5a with 0.01
≤ Pr ≤ 5. However, the same is not fully applicable for a Pr >> 1 in Fig. 5b. Also, a
careful analysis allows us to conclude that the slope of T + is fairly well represented
by the selected wall laws.
The analysis reveals, especially, for Pr << 1 and Pr >> 1 there is a need of correction
P in the standard wall function, depending on the Pr. For instance, P is greater than
424 K. Kumar et al.
Fig. 4 Comparison between 30

RANS analyzed and
wall-based u+ and T +
profiles with air, at x = 0.6 m 20
and Rex = 15 × 105
u+
10 Simulation
u+ = y+, viscous sublayer
Standard log-law
0
1 10
y+ 100 1000 10000
20
15
T+
Simulation
10
Standard log-law
5 Wall law, Bejan
Wall law, Kader

0
1 10 y+ 100 1000 10000
Fig. 5 Computed and wall a 50

Simulation, air Simulation, water
function-based T+ versus Simulation, sodium Standard log-law, air
Pry+ for a sodium, air, and 40 Standard log-law, water Kader log-law, sodium
water and b ethylene glycol
and therminol with 0.01 ≤ Pr 30 ∆P
≤ 760
T+
20
10
0
1 10 Pr y+ 100 1000 10000
b 2400 Simulation, therminol

2000 Standard log law, therminol
Simulation, ethylene glycol
1600
Standard log law, ethylene glycol
1200
∆P
T+
800
400
0
100 1000 10000 100000
Pr y+
0 for water and otherwise for sodium. However, this trend seems to disappear moving
away for Pr >> 1 as depicted in Fig. 5b. Therefore, due care should be taken while
selecting wall laws in simulating a very high and low Pr. Apparently, corrections to
turbulent Prandtl number as in Grötzbach [4] for liquid metals are necessary for a
high Pr fluid.
4 Conclusions
The paper deals revisits the applicability and limitations of well-known wall func-
tions for statistical temperature for fluids having a Prandtl number (Pr) in the range
0.01 to 760. For this purpose, the widely employed Reynolds-averaged Navier–
Stokes equations are solved with a wall-resolved mesh. Comparison with standard
wall functions for velocity and temperature demonstrates the suitability of adopted
numerical framework. Further investigation with 0.01 ≤ Pr ≤ 5 reveals that the repre-
sentation of T + against Pry+ indeed limits the initiation of log-law region within 15
≤ Pry+ ≤ 30. The need of correcting wall law is also presented with a trend for 0.01
≤ Pr ≤ 5. However, the same was not obvious for Pr >> 1. In general, the slope of T +
is satisfactorily represented by the standard or specific wall functions. This analysis
allows concluding the need of adapting the standard wall laws with corrections based
on Pr and turbulent Prandtl number for extreme Pr. Thus, it is recommended that the
application of wall function for Pr << 1 and Pr >> 1 must be done with caution.
References
1. Pardeshi R, Tiwari S, Sharma P, Pardeshi R, Chandra L, Shekhar R (2015) Solar convective

furnace for metals processing. JOM 67(11):2697–2704
2. Chandra L, Roelofs F, Houkema M, Jonker B (2009) A stepwise development and validation
of a RANS based CFD modeling approach for the hydraulic and thermal-hydraulic analyses
of liquid metal flow in a fuel assembly. Nucl Eng Des 239:1988–2003
3. Wang X, Castillo L, Araya G (2008) Temperature scalings and profiles in forced convection
turbulent boundary layers. J Heat Transf 130 (021701): 1–17
4. Grötzbach G (2013) Challenges in low-Prandtl number heat transfer simulation and modelling.
Nucl Eng Des 264:41–55
5. Kader BA (1981) Temperature and concentration profiles in fully turbulent boundary layers.
Int J Heat Mass Transf 24(9):1541–1544
6. Bejan A (2013) Convection heat trasfer. Wiley, Hoboken, New Jersey
7. Suga K, Craft TJ, Iacovides H (2006) An analytical wall-function for turbulent flows and heat
transfer over rough walls. Int J Heat Fluid Flow 27:852–866
8. Craft TJ, Gerasimov AV, Iacovides H, Launder BE (2002) Progress in the generalization of
wall-function treatments. Int J Heat Fluid Flow 23: 148–160
9. Vigdorovich II (2017) Turbulent thermal boundary layer on a plate. reynolds analogy and heat
transfer law over the entire range of Prandtl numbers. Fluid Dyn 52(5): 631–645
10. Nagano Y, Shimada M (1996) Deveopment of a two-equation heat transfer model bsaed on
direct simulations of turbulent flows with different Prandltl number. Phys Fluids 8:3379–3402
426 K. Kumar et al.
11. Kasagi N, Kuroda A, Hirata M (1989) Numerical investigation of near wall turbulent heat
transfer taking into account the unsteady heat conduction in the solid wall. J Heat Transf
111:385–392
12. Dong YH, Lu XY, Zhuang LX. An investigation of the Prandtl number effect on turbulent heat
transfer in channel flows by large eddy simulation. Acta Mechanica 159: 39–51
Double Diffusive Mixed Convection
in a Lid-Driven Enclosure
with the Influence of Shape
of Nanoparticle and Magnetic Field
Ratnadeep Nath and K. Murugesan
1 Introduction
Heat transfer using nanofluid is a well-established technique introduced by Choi [1]

implemented in various engineering applications. Sufficient research is available on
enclosure type geometry using nanofluid as a working medium [2, 3]. Nowadays, the
modern cooling devices are incorporated with some electric and magnetic field that
may effects the fluid flow phenomenon and thus convective heat and mass transfer.
Therefore, a research is needed to explore the physics behind it. Aghaei et al. [4]
and Mahalakshmi et al. [5] worked on enclosure type problem with nanofluid and
magnetic field. Both of them agreed that with increasing in Hartmann number (Ha),
convective heat transfer decreases. The authors did not consider any mass transfer
which actually present in actual case in terms of moisture transport. Recently, Reddy
and Murugesan [6] worked on square cavity influenced by thermosolutal buoyancy
and magnetic field. They found suppression of Nuavg and Shavg with Ha varies
depending upon the inclination of magnetic field. The study did not consider the
effect of nanoparticle.
Most of the researchers used spherical shape of nanoparticle which may not be the
case always. The properties of nanofluid are controlled by the shape of the nanopar-
ticle [7] also. Hence, the present study is an attempt to understand the influence of
shape of nanoparticle on convective heat and mass transfer phenomenon in presence
of magnetic field.
R. Nath (B) · K. Murugesan

Mechanical and Industrial Engineering Department, IIT Roorkee, Roorkee 247 667, India
K. Murugesan
e-mail: krimufme@iitr.ac.in
428 R. Nath and K. Murugesan
2 Problem Formulation
The present study considered a square cavity (length L) shape as a geometry with
its upper moving plate, as shown in Fig. 1. The square is filled with single phase
Al2 O3 —water/ethylene glycol nanofluid. Thermophysical properties of nanofluid
used are listed below [4, 5]:
Density ρnf = (1 − χ )ρf + χρp

Heat capacitance (ρcp )nf = (1 − χ )(ρcp )f + χ (ρcp )p
Volume expansion coefficient (thermal) (ρβT )nf = (1 − χ )(ρβT )f + χ (ρβT )p
Mass diffusivity Dnf = (1 − χ )Df
Thermal conductivity knf = kf (1 + Ck χ )
Dynamic viscosity μnf = 1 + Aχ + Bχ 2 μf
Elc. conductivity σnf = (1 − χ )σf + χ σp
(1)
The flow is assumed to be Newtonian, steady, 2D, and there is a thermal equilib-
rium between base fluid (water-EG) and Al2 O3 nanoparticles. The thermophysical
properties of nanofluid are constant and are listed in Table 1. There are two shapes

Y
of moving lid square cavity Θ=Ф=V=0 ; U=1
Θ=Ф=1 ; U=V=0 X
Table 1 Thermophysical
Properties Water Ethylene glycol Al2 O3
properties of nanofluid
cp (J/kg K) 4182.2 2415 765
ρ (kg/m3 ) 998.203 1114.4 3970
k (w/m K) 0.613 0.252 40
βT × 10−4 (1/K) 2.06 6.50 0.08
σ ( m)−1 0.05 1e−10
Double Diffusive Mixed Convection in a Lid-Driven Enclosure … 429
Table 2 Coefficient of
Platelets Blades
thermal conductivity and
viscosity for two shapes of A 37.1 14.6
nanoparticle B 612.6 123.3
Ck 2.61 2.74
of alumina nanoparticle such as platelet and bricks are used in this study where the
values of constant used in viscosity and thermal conductivity is listed in Table 2,
Timofeeva et al. [8] model. Moreover, change in density caused by the difference in
temperature and concentration is taken care of by assuming Boussinesq approxima-
tion, and it can be expressed as ρ = ρ 0 [1 − βT,nf (T − T c ) − βC,nf (C − C c )]. The
term buoyancy ratio (N) used in the present analysis represents the relative strength
of solutal buoyancy force over thermal buoyancy force where positive sign means
aiding and negative sign means opposing nature of thermosolutal buoyancy force.
Some parameters are kept constant throughout the analysis such as Reynolds number
(Re) = 100, Prandtl (Pr) and Schmidt (Sc) number = 6.2, magnetic field inclination
angle = 0° Richardson number (Ri) = 0.1, and nanoparticle volume fraction (χ ) =
6%.
3 Governing Equation and Methodology
The Navier–Stokes equation is transformed into velocity–vorticity form of equation

in order to avoid the pressure term. The governing equation consists of vorticity
transform equation, energy equation and concentration equation and using the prop-
erties of nanofluid [as listed in Eq. (1)] it can be expressed in dimensionless form as
below:
Vorticity transport

∂ζ ∂ζ ∂ζ 1 (1 + Aχ + Bχ 2 ) ∂ 2 ζ ∂ 2ζ
+U +V = +
∂τ ∂X ∂Y Re (1 − χ ) + χρp ∂ X2 ∂Y 2
ρf
⎡ χρp βT
⎤
(1 − χ )ρf + ρf βT p ∂
σnf
H a2
σf
+Ri ⎣ f ⎦
(θ + N ) + χρp
(2)
(1 − χ )ρf + χρp ∂ X (1 − χ ) + ρf Re

∂ ∂
(U sin ϕ cos ϕ − V cos2 ϕ) − (V sin ϕ cos ϕ − U sin2 ϕ)
∂X ∂Y
Velocity Poisson
∂ 2U ∂ 2U ∂ζ ∂ 2 V ∂2V ∂ζ
+ = − ; + = (3)
∂X 2 ∂Y 2 ∂Y ∂ X 2 ∂Y 2 ∂X
Energy
2
∂θ ∂θ ∂θ 1 (1 + Ck χ ) ∂ θ ∂ 2θ
+U +V = + (4)
∂τ ∂X ∂Y Re. Pr (1 − χ ) + χ (ρcp )p ∂ X2 ∂Y 2
(ρcp )f
Concentration

∂ ∂ ∂ (1 − χ ) ∂ 2 ∂ 2
+U +V = + (5)
∂τ ∂X ∂Y Re.Sc ∂ X 2 ∂Y 2
Initial condition [@τ = 0]
U =V =ζ =θ ==0 (6)
Boundary conditions [@τ > 0]
Top wall: U = 1, V = θ = = 0;
Bottom wall: U = V = 0; θ = = 1;
∂θ ∂
Side walls: U = V = 0; = =0 (7)
∂X ∂X
The average Nusselt (Nu) and Sherwood (Sh) number along the bottom wall can
be expressed by the following equations:
L L
1 knf ∂θ 1 Dnf ∂
Nuavg = − dX ; Shavg = − dX (8)
L kf ∂Y Y =0 L Df ∂Y Y =0
0 0
The governing equations for fluid flow, thermal and solutal field are solved using
Galerkin’s weighted residual finite element method with given initial and boundary
conditions. An isoparametric formulation is used to discretize computational domain
using bilinear quadrilateral elements. The time discretization is carried out by Crank–
Nicholson scheme, a second order accurate method. The convergence criteria set for
any variable is set as 10−5 .
4.1 Validation of Results
In order to run the simulation for the present analysis, it is necessary to validate
the current formulation developed on in-house built FORTRAN code. For that, a
Table 3 Comparison on
(Nuavg ) Relative error (%)
Nuavg for validation
Present study Amiri et al. [9]
N = −50 1.63 1.6 1.87
N = 50 2.85 2.8 1.78
validation has been done with Amiri et al. [9] paper where the relative error is below
2%, as shown in Table 3.
The validation result with good accuracy gives sufficient confidence to proceed
for other test cases for this analysis.
4.2 Streamline Pattern
The streamline pattern of moving plate enclosure for Ha = 10 at buoyancy ratio (N)
= −10, 10 for the shape of platelets and blade is shown in Fig. 2a, b. From Fig. 2a,
it is observed that there is a secondary recirculation zone developed at the bottom
left corner for both shape of nanoparticle where blade shape nanoparticle has larger
secondary recirculation region and platelet has smaller recirculation zone. There
are two things simultaneously responsible for such pattern, one is opposing nature
of thermosolutal buoyancy force and another is variation in viscosity for different
1 1 1
0.75 0.75 0.75
0.5 0.5 0.5
0.25 0.25 0.25
0 0 0
0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 1
Blade (N= −10) Platelet (N= −10) Blade (N= +10)

1
0.75
0.5
0.25
0
0 0.25 0.5 0.75 1
Platelet (N= +10)
Fig. 2 Effect of shape of nanoparticle on streamline pattern at N = −10, 10

shapes of nanoparticle. It is very well understood that the first recirculation zone is
developed due to inertial force caused by the moving lid. On the other hand, because
of negative buoyancy ratio the solutal buoyancy force tries to hold the fluid near the
bottom plate though the thermal bouncy force tries to lift up the fluid flow towards
upside. Because of stronger solutal buoyancy force the inertial force is unable to reach
at the bottom. Secondly, the shape of nanoparticle is responsible for the variation
in dynamic viscosity as seen from Table 2. Platelet shape offers more resistance
in fluid flow motion as the value of dynamic viscosity is high whereas the same
thing is lower for the case of blade shape of nanoparticle. These two things such
as opposing nature of thermosolutal buoyancy force and higher viscosity made a
smaller secondary recirculation zone developed adjacent to the bottom left corner.
But for the case of blade shape, the value of dynamic viscosity is less and hence,
a larger secondary recirculation zone is observed. In Fig. 2b, streamline pattern at
buoyancy ratio (N) = 10 is shown for platelet and blade shape of nanoparticle. Here,
one can easily identified that the size of secondary recirculation zone is very small
compared to that of N = −10 for both the shape of nanoparticle. The reason behind
such pattern is due to the aiding nature of thermosolutal buoyancy force, the fluid is
forced to move in upward direction whereas the inertial force reaches upto a depth
of the enclosure. In this case, the variation in dynamic viscosity is also present as
discussed for N = −10 case. Therefore, a smaller secondary recirculation is formed
for platelet shape and a large shape is developed for blade shape of nanoparticle. It
is very interesting to observe that the strength of thermosolutal buoyancy force is
good enough to overcome the viscous force and hence a larger primary recirculation
is observed for both the shape of nanoparticle.
4.3 Average Nusselt and Sherwood Number
The variation of average Nusselt number with Hartmann number for platelet and
blade shape of nanoparticle at N = −10 and 10 is shown in Fig. 3. It is observed that
for any shape and buoyancy ratio, magnetic field supress the convection process that
results reduction in average Nusselt number. As the value of buoyancy ratio increases
from −10 to 10 the fluid motion also increases that enhances the heat transfer from
bottom plate due to aiding nature of thermosolutal buoyancy force which is already
observed in streamline pattern. In addition, the variation in thermal conductivity for
two shapes of nanoparticle is also present. Overall, 15 and 14% decrement in average
Nusselt number is noticed for blade shape of nanoparticle at N = 10 and −10 whereas
8 and 11% is noticed for platelet shape.
In Fig. 4, the variation in average Sherwood number with Hartmann number at N
= −10, 10 for blade and platelet shape of nanoparticle. The trend of change in average
Sherwood number with Hartmann number is similar to that of average Nusselt
number because of suppression of convection due to augmentation in magnetic field.
The physics behind variation in average Sherwood number for N = −10 and 10 is
similar to that of average Nusselt number. Moreover, the mass diffusivity of nanofluid
6 6
5 5
4 4
Nuavg
Nuavg
3 3
2 2
N=-10 N=-10
N=10 N=10
1 1
0 5 10 0 5 10
Ha Ha
Blade Platelet
Fig. 3 Nuavg versus Ha for different shape of nanoparticle at N = −10, 10
6 6
5 5
4 4
Shavg
Shavg
3 3
2 2
N=-10 N=-10
N=10 N=10
1 1
0 5 10 0 5 10
Ha Ha
Blade Platelet
Fig. 4 Shavg versus Ha for different shape of nanoparticle at N = −10, 10
is decreased with doping of nanoparticle. Hence, convective mass transfer is reduced

than that of convective heat transfer. There is a decrement of 17% and 14 in convective
mass transfer with Hartmann number at N = 10 and −10 in blade shape nanoparticle
whereas 8 and 11% is noticed in platelet shape.
5 Conclusion
After analysing the results, following conclusions can be made:

1. The aiding or opposing nature of thermosolutal buoyancy force causes formation
of secondary recirculation zone whereas the size varies depending upon the fluid
viscosity offered by different shape of nanoparticle.
2. The Magnetic Field at 0° Inclination Angle Rescind the Convective Heat and
Mass Transfer Phenomenon.
3. The blade shape of nanoparticle gives maximum value of Nuavg and Shavg at Ha
= 0 and N = 10 compared to that of platelet shape. The minimum 8% reduction
in convective heat and mass transfer is observed in platelet shape of nanoparticle
at N = 10.
References
1. Choi SUS, Eastman JA (1995) Enhancing thermal conductivity of fluids with nanaoparticles.
In: International mechanical engineering Congress and exhibition
2. Izadi S, Armaghani T, Ghaseminasl R, Chamkha AJ, Molana M (2019) A comprehensive review
on mixed convection of nanofluids in various shapes of enclosures. Powder Technol 343:880–907
3. Oztop H, Estelle P, Yan W, Salem K, Orfi J, Mahian O (2015) A brief review of natural convection
in enclosures under localized heating with and without nanofluids. Int Commun Heat Mass
Transfer 60:37–44
4. Aghaei A, Khorasanizadeh H, Seikhzadeh G, Abbaszadeh M (2016) Numerical study of
magnetic field on mixed convection and entropy generation of nanofluid in a trapezoidal
enclosure. J Magn Magn Mater 403:133–145
5. Mahalakshmi T, Nithyadevi N, Oztop H, Hamdeh N (2018) MHD mixed convective heat transfer
in a lid-driven enclosure filled with Ag-water nanofluid with center heater. Int J Mech Sci
142–143:407–419
6. Reddy N, Murugesan K (2017) Magnetic field influence on double diffusive natural convection
in a square cavity–a numerical study. Numer Heat Transfer Part A 71:448–475
7. Khodashenas B, Ghorbani HR (2005) Synthesis of silver nanoparticles with different shapes.
Arab J Chem, 1–16
8. Timofeeva EV, Routbort JL, Singh D (2009) Particle shape effects on thermophysical properties
of alumina nanofluids. J Appl Phys 106:014304-1–14310
9. Al-Amiri AM, Khanafer KM, Pop I (2007) Numerical simulation of combined thermal and mass
transport in a square lid-driven cavity. Int J Therm Sci 46:662–671
Effect of Gust Intensity in Flow Past
Cube with Built-In Wings at Low
Reynolds Number
Ramachandra Vijjapu and Shaligram Tiwari
1 Introduction
The study of flow past bluff bodies is always active and on-demand because of
its applications in architecture and aviation. The pillars of bridges and skyscrapers
are constructed in such a way that the drag forces and vortex shedding effects are
minimized or at least countered. The spacings left in between the huge apartment
structure to facilitate the airflow, thereby reducing the form drag. Raul et al. [1]
had done one of the earliest works on laminar flow past a cube for Re ranging in
between 10 and 100. They used vorticity-vector potential method for solving the
Navier–Stokes and continuity equations. Further, they used finite difference method
for discretization. Chirag et al. [2] investigated on gusty flow past a circular cylinder.
In this study, they focused on flow field features by varying frequency and Re of wind
gust. In the present study, we are presenting the effect of gust intensity (amplitude)
on flow characteristics. Klotz et al. [3] studied on flow past cube for moderate Re
of value in between 100 and 400. They presented the three well-defined regimes
based on the wake characteristics. The initial basic flow regime was found to be
steady with orthogonal symmetry, characterized by the existence of four pairs of
streamwise counter-rotating vortices originating at the lateral edges of the cube.
This type of flow regime having the orthogonal symmetry is noticed in present study.
Saha [4] has investigated the transition of flow past a cube. He compared the wake
characteristics of flow past a cube with other bluff bodies such as sphere and square
cylinder (2D) and also presented the variation of drag and side force coefficients with
Re. Patil et al. [5] investigated effect of blockage ratio on wake transition for flow
past square cylinder. In this study, they reported the dependence of critical Reynolds
number (for onset of planar vortex shedding) on blockage ratio.
R. Vijjapu (B) · S. Tiwari

S. Tiwari
436 R. Vijjapu and S. Tiwari
The previous studies mainly focused on uniform inlet flow and various cylinder
shapes, and the literature pertaining to gusty flow is limited. This forms the motivation
for the present study. Considering these gaps in the literature, authors have carried
out investigations to study the behaviour of gusty flow and cube with wings. Present
study on flow past cube with wings under the influence of gusty flow is a simplified
version of a bird or plane undergoing a gliding motion in the air.
2 Problem Definition and Numerical Methodology
2.1 Geometry and Mesh Details
The computational domain is shown in Fig. 1 and has dimensions as L 1 = 18 d, L 2

= 10 d and L 3 = 10 d in X, Y and Z-directions, respectively. It considers flow past a
cube with rigid wings placed at a distance (L u ) of 4 d from inlet face of the domain.
The side length of the cube is d which is also considered as the characteristic length.
The length and thickness of the wings are d and 0.1 d, respectively.
The fluid flows along X-direction. Mesh is generated using ICEM CFD 18.1, and
further refinement is done as shown in Fig. 1, i.e. in the regions where flow properties
vary steeply. This could include the wake region and the regions above and below
the wing surface.
Fig. 1 Schematics of
a computational domain
b mesh
Effect of Gust Intensity in Flow Past Cube with Built-In Wings … 437
Navier–Strokes and the continuity equations are solved together. The fluid considered
is air, and it is assumed as incompressible.
∂u i
=0 (1)
∂ xi
2
∂u i ∂ ui u j ∂P 1 ∂ ui
+ =− + (2)
∂t ∂x j ∂ xi Re ∂ x j ∂ x j
The first equation is the continuity equation for incompressible fluid flow, and
the second equation represents the indicial form of Navier–Stokes equations. The
equations are in non-dimensionalized form with u i being the velocity component in
xi -direction, Re the Reynolds number and P the static pressure. Here the velocities
are non-dimensionalized with the mean velocity corresponding to Re chosen, all the
lengths are non-dimensionalized with the characteristic length (d), time with d/u mean
and pressure by ρ u 2mean .
The inlet and outlet boundary conditions are velocity inlet and pressure outlet,
respectively. Top, bottom and the two side surfaces are treated as free-slip boundaries.
The surface of the geometry is assumed to be no-slip and impermeable. SIMPLE
(Semi-Implicit Method for Pressure-Linked Equation) algorithm is employed to
decouple the momentum and continuity equations. Second-order upwind scheme
is used for discretization of the convective term. First-order implicit scheme is used
for discretisation of the transient term. Least squared method is used for computing
the gradients. Gust effect is introduced into the computations by using user-defined
functions (UDFs) in ANSYS Fluent 18.1. The velocity profile at the inlet is defined
as
u(t)inlet = u mean + IF × u mean sin(2π f t) (3)
where u mean is the mean velocity corresponding to the value of Re chosen, IF is wind
gust intensity factor, I F × u mean represents the amplitude of wind gust, f is the
frequency of flow at the inlet (gust frequency) which is considered to be a fixed. The
Reynolds number for the present study may be defined as
ρu meanl
Re = (4)
μ
where ρ is the density of fluid (air), u mean is the mean velocity, l is the characteristic
length and μ is the dynamic viscosity.
Table 1 Comparison of drag

Re Drag coefficient (Cd)
coefficient with results in the
literature Saha [4] Raul et al. [1] Present study
20 3.03 2.96 3.24
50 1.78 1.82 1.90
2.3 Validation of Results
However, the geometry in the present study is unique, and the literature pertaining
to flow past cube with wings is not available to the authors’ best knowledge. The
present simulation results are validated against the flow past cube (Table 1).
All the co-ordinate axes presented in this discussion are non-dimensionalized. The
X, Y and Z axes are non-dimensionalized using the characteristic length. The time
axis is non-dimensionalized using the mean inlet velocity corresponding to the Re
used and characteristic length (Table 2).
From Fig. 2, it is evident that drag is periodic at Re values of 100 and 200. At both
Reynolds numbers, the amplitude of the drag signal is increasing with an increase
in gust intensity (IF). It is also clear from the above table that the mean value of
drag is decreasing as the Re is increasing at any given intensity of wind gust which
is in agreement with flow past cube as mentioned in Saha [4]. At a given Re (say
Re = 100), the mean value of drag is decreasing with increase in value of IF. The
frequency of the drag signal is almost the same for all intensities of wind gust, and
it is equal to flow frequency (gust frequency) at inlet.
As for the lift coefficients (Cly and Clz ), their fluctuation is in the order 10−6 and
their mean value is almost equal to zero for all the values of IF and Re considered.
However, the nature of fluctuations is periodic and the amplitude of lift signals (Cly
and Clz ) increase with increase in gust intensity (IF) at Re equal to 100.
Figure 3 shows the evolution of Y-vorticity and Z-vorticity at Re = 100 with gust
intensity (IF). From all the vorticity contours, it is clear that the flow in the wake
Table 2 Variation of mean drag coefficient and amplitude of drag signal with intensity factor (IF)
and Re
Intensity factor Re = 100 Re = 200
Cd Amplitude of Cd Cd Amplitude of Cd
0.1 2.1435 0.6128 1.5605 0.3847
0.2 2.0985 1.1997 1.5593 0.7591
0.3 2.0588 1.7643 1.5567 1.1165
5 IF=0.1
IF=0.2
IF=0.3
4
3
Cd
180 190 200

(a) FlowTime
3.5 IF=0.1
IF=0.2
3 IF=0.3
2.5
Cd
1.5
0.5
400 420 440

(b) FlowTime
Fig. 2 Variation of drag coefficient with respect to time for different intensities of wind gust at
a Re = 100 and b Re = 200
zone is symmetric in both Y-plane and Z-plane and due to this the lift generated in Z
and Y-directions is almost zero. At Re = 100 for steady uniform flow at the inlet, the
flow is almost steady, and oscillation of flow in the wake region is hardly noticeable.
However, as the value of IF increases, the flow becomes unsteady. At a lower value
of IF (=0.1), the flow symmetrically oscillates to small extent. At IF = 0.2, the flow
oscillation increases to a noticeable value, but at IF = 0.3 the flow oscillates at a
higher amplitude and a symmetric vortex shedding is found to take place.
4 4
2 2
Y
Z
0 0
-2 -2
-4 -4
-4 0 4 8 12 -4 0 4 8 12
X (a) X
4 4
2 2
Z
Y
0 0
-2 -2
-4 -4
-4 0 4 8 12 -4 0 4 8 12
X (b) X
4 4
2 2
Y
Z
0 0
-2 -2
-4 -4
-4 0 4 8 12 -4 0 4 8 12
X (c) X
Fig. 3 Y-vorticity contours in XZ plane (left side), Z-vorticity contours in XY plane (right side) at
Re = 100 and frequency = 2 Hz for various intensities of wind gust a IF = 0.1, b IF = 0.2 and c IF
= 0.3
Figure 4 depicts the streamline contours at Re value of 100 and inlet flow frequency
of 2 Hz for various intensities of wind gust. It is clear that at all the intensities of
gust mentioned, the flow separates at leading edge without any reattachment on the
top and bottom surfaces of the geometry. The separated shear layers (flow) seem to
rejoin in the downstream, forming two recirculation bubbles. Saha [4] also reported
similar findings in flow past cube for Re = 50–150. The bubble size is quantified by
recirculation length. The recirculation length or vortex formation length is defined as
the streamwise distance from the centre of the geometry to saddle point. The saddle
point is the location where the streamwise velocity changes its sign or where the
shear layers rejoin. At a given Re and frequency as the gust intensity is increasing,
the bubble size or recirculation length is decreasing. The centre of the bubble seems
to move in a streamwise direction and close to centreline (longitudinal axis) as the
gust intensity is increasing.
Fig. 4 Streamline contours in XZ (left) and XY plane (right) at Re = 100 and frequency = 2 Hz
for various intensities of wind gust a IF = 0.1, b IF = 0.2 and c IF = 0.3
Figure 5 shows the Fourier spectra for drag signal at Re values of 100 and 200
for different values of gust intensity. The dominant frequency is found to be 2 Hz
which is also the frequency of flow at the inlet (wind gust). Despite being a vanishing
value, the dominant frequency of lift-Y and lift-Z is also 2 Hz, but the power spectral
density is very low. The above two diagrams having a single dominant peak justify
the fact that the variation of drag for all gust intensities at Re values of 100 and 200
is periodic and almost sinusoidal fluctuation.
2400
IF=0.1
2000
Power Spectral Density
IF=0.2
IF=0.3
1600
1200
800
400
0
2 2.1 2.2 2.3
(a) Frequency
1500
IF=0.1
IF=0.2
Power Spectral Density
IF=0.3
1000
500
0
1.9 2 2.1 2.2 2.3
(b) Frequency
Fig. 5 Fourier spectra for drag signal at frequency = 2 Hz and different gust intensities at a Re =
100 and b Re = 200
4 Conclusion
Three-dimensional numerical investigations of flow past cube with wings at gust

frequency of 2 Hz for three different values of gust intensity (IF) at Re values of 100
and 200 have been studied. The variation of Cd is periodic, and almost sinusoidal at
both Re values and at each value of wind gust intensity (IF). Moreover, the frequency
of gust at inlet and the frequency of drag signal are equal for each value of IF. As
the gust intensity increases, the mean drag coefficient decreases and the amplitude
of the drag signal is increasing. The flow is symmetric about the centreline, and
even the vortex shedding is symmetric because of which vanishing values of lift
coefficients are produced. Wake oscillations are observed to increase, and finally,
vortex shedding is taking place when the intensity factor (IF) is raised beyond a
certain value. The critical IF at Re = 100 is in between the value of 0.2 and 0.3.
The streamline contours clearly show the flow separation at the leading edge of the
cube surface (top and bottom) without any further reattachment on the cube surface.
The separated shear layers form two recirculation bubbles. With increase in value of
IF, the bubble size is found to decrease with bubble centre moving downstream and
approaching closer to centreline.
Acknowledgements Authors would like to thank IIT Madras for providing the required computa-
tional resources and support.
References
1. Raul R, Bernard PS, Buckley FT Jr (1990) An application of the vorticity-vector potential method
to laminar cube flow. Int J Numer Methods Fluids 10:875–888
2. Parekh CJ, Roy A, Harichandan AB (2018) Numerical simulation of incompressible gusty flow
past a circular cylinder. Alexandria Eng J 57:3321–3332
3. Klotz L, Goujon-Durand S, Rokicki J, Wesfreid JE (2014) Experimental investigation of flow
behind a cube for moderate Reynolds numbers. J Fluid Mech 750:73–98
4. Saha AK (2004) Three-dimensional numerical simulations of the transition of flow past a cube.
Phys Fluids 16:1630–1646
5. Patil PP, Tiwari S (2008) Effect of blockage ratio on wake transition for flow past square cylinder.
Fluid Dyn Res 40:753–778
Design and Fabrication of a Portable
Stagnant Water Drainer
R. Naveen, T. Prem Kumar, K. Sridharan, and D. Sandheep
1 Introduction
Water is one of the basic needs of human beings apart from clothing, food and
shelter. Many researches are being undergone to find new methods to produce water.
However, there is a great need to come out with ways to use the available water effi-
ciently. There have been many interesting studies in this field including Authorship
Pattern of Rain Water Harvesting Research Management [1]. Relative Economics
of Irrigation through Percolation Tank and Irrigation Tank presents an interesting
comparison between the practicalities of the methods [2]. The run-off analysis for
Sustainable Storm water Drainage has been done for the City of Madinah, Saudi
Arabia [3]. An insight on the surface water drainage technique [4] has been studied
and designed effectively by many researchers [5]. Percolation tanks have been effi-
ciently used in paddy fields [6]. The response of soil water and deep percolation
has been studied in northern China [7]. Infiltration of rain water through the soil
[8] has been useful in obtaining the strategy and issues involved in the designing
aspect. Storm water management by microfiltration and ultra-filtration [9] provides
a deeper understanding on the treatments that are existent. Cumulative impacts of
residential rainwater harvesting on storm-water discharge [10] have been studied
through a semi-urban drainage network. Surface water is being innovatively treated
R. Naveen (B) · T. Prem Kumar · K. Sridharan · D. Sandheep

Department of Mechanical Engineering, PSG Institute of Technology and Applied Research,
Coimbatore, India
e-mail: 16m134@psgitech.ac.in
T. Prem Kumar
e-mail: premkumar@psgitech.ac.in
K. Sridharan
D. Sandheep
446 R. Naveen et al.
Table 1 Average
Soil type Permeability (cm/h)
permeability for different soil
textures in cm/hour Sand 5
Loam 1.3
Clay loam 0.8
Silty clay 0.25
Clay 0.05
with lotus plant [11]. Variation of micro-organisms in drip irrigation systems [12]
has been observed in-depth using high-sand surface water. After studying the above
processes, an innovative idea is to be discussed that can play an instrumental role in
draining the stagnant water in areas where the top-most layer of the soil is clay (up to
15 cm) and loamy or sandy soil underneath. The permeability values are discussed
in Table 1.
Thus, the system will be suitable for a surface layer of clay and a lower layer of
sand/sandy loam. It can greatly reduce the time required for seeping by accelerating
the flow using a portable stagnant water drainer.
The primary objective of the equipment is to drain out the stagnant surface water.
This not only reduces the amount of stagnant water which causes excessive breeding
of mosquitoes, but this also replenishes the ground water.
2 Methodology
2.1 Design
The equipment is fabricated by welding of the hand pump with the threaded pipes
as per the proposed model using arc welding process. The 3-D model is shown in
Fig. 1.
The components required for the above model are listed in Table 2.
The soil sample’s permeability is tested using constant head permeability method
[13]. The material chosen is mild steel as it is easily available in the market and can
be welded using arc welding process. The NRV valves are made of brass.
But before finalising the final model to be fabricated, the conical section has been
optimised for better performance of the equipment. The optimised throttling section
is shown in Fig. 2.
This modified fitment provides easier penetration, avoids blocking of nozzle by
sand and therefore improves the handling of the equipment by making the process
of penetration easy for the operator.
Design and Fabrication of a Portable Stagnant Water Drainer 447
Fig. 1 3-D model of the

drainer
Table 2 Components and its

Components Specification
specification
Hand pump 50 mm(dia)
Non-return valve (NRV) 0.5”
Cylindrical pipe for outer casing 0.5”
Pipe with threads on either side- 2 0.5”
Throttling section Custom
Reducer 0.5”
GI pipe 0.5”
Fig. 2 3-D model of

modified throttling device
The permeability of the soil is calculated based on Eq. (1):

Q L 1
K = . . (1)
t H A
where
K permeability of soil (cm/s)
Q quantity of water collected (in cm3 )
T time taken (in seconds)
L height of sample (in cm)
H constant head maintained (in cm)
A cross-sectional area (in cm2 )
Two soil samples (clay soil and permeable soil) have been taken from the site and
checked for the permeability using constant head permeability method. The measure
and calculated values of bottom permeable soil are given as below.
Readings of Constant Head Permeability method for bottom Permeable soil

Q = 1000 cm3
t = 170 s
L = 11 cm
H = 44 cm
D = 10 cm

A = 3.14 ∗ d 2 /4 (2)
Using Eq. (2)
A = 18.5 cm2
From Eq. (1);

1000 11 1
K = ∗ ∗
170 44 18.5
K = 0.18669 cm/s
The value obtained above is for permeable soil, and similar readings have been
taken for clay soil too and the permeability, k is found to be 0.051 cm/s.
For the purpose of testing the set-up, a rectangular tank of dimensions 43*70*110 cm
respectively has been constructed. The tank is constructed using bricks and mortar.
The bricks are lined up in single row as a wall.
This tank consists of high permeable soil at the bottom-most layer, and the height
of the permeable soil from the bottom is 50 cm. The above layer is made up of low
permeable soil (clay), and the depth of the soil is 29 cm. The depth that is allowed
for the water to stand is 20 cm. Three sides of the tank are made of bricks and mortar,
and the other side is covered with an acrylic sheet for better optical vision to view
the functions happening inside the tank. Initially, the set-up is checked thoroughly
for the leakage.
Water is poured into the set-up above the clay soil, and the water is allowed to
stand on the clay soil. It is observed that water does not percolate into the clay soil,
so that the water remains stagnant on the soil surface. The device is now inserted
into the set-up through the clay soil. It is inserted till the modified throttling device
reaches the second layer of soil. Now, it is observed that the inlet hose is completely
submerged in the water such that the water can be sucked and pressurised.
The water enters into the cylinder through the NRV when the lever (or piston) is
moved from BDC (bottom dead centre) to TDC (top dead centre), i.e. suction stroke.
Water is sucked into the throttling device through the NRV. The water does not return
back to the inlet pipe due to the usage of NRV. Then, the piston is moved from TDC
to BDC (delivery stroke), and the water inside the pump will be pressurised due
to closure of both inlet and outlet valves. The pressure inside the cylinder is raised
slightly higher than the back pressure due to the soil. Once the pressure exceeds the
back pressure due to soil at outlet, the outlet valve opens and the water is forced into
the permeable soil through the outlet NRV. The actual and schematic of the testing
set-up is shown in Figs. 3 and 4 respectively.
It can be observed from Table 3 that the volumetric efficiency is 55.6%. The possible
reasons for less efficiency could be the inertia effect of the non-return valve and the
back pressure resistance offered by the soil at outlet. At an average rate of 6 seconds
per cycle, the system produces a water draining rate of 4.9 litres per minute. Thus,
Fig. 3 Actual experimental

set-up
Fig. 4 Schematic of the

set-up
Table 3 Volumetric efficiency calculation

Stroke length Bore dia Area Vtheoretical (m3 /stroke) Vactual Volumetric efficiency
(m) (m) (m2 ) (m3 /stroke) (%)
0.45 0.05 0.0019 8 × 10–4 4.9 × 10–4 55.6
the water seeping rate has subsequently increased due to presence of the stagnant
water drainer compared to natural penetration which will be very less.
4 Conclusions
The equipment is a prototype model for a portable stagnant water drainer (diameter
= 50 mm, stroke = 500 mm, reciprocating type) that pumps into the lower layer
of the soil about 1–1.5 feet which can drain the surface water at the rate of 4.9 L
per minute (55.6% volumetric efficiency). With the advancements of automation and
material sciences, we can reduce the weight of the component by using materials
that provide corrosion resistance and higher mechanical strength at lower weight.
Motorised mechanism can also be used for commercial usage.
References
1. Manju PU, Raja S (2019) Authorship pattern of rain water harvesting research management
publications 39(1):31–37
2. Nayak A, Ananth GS, Chandrakanth MG (2017) Relative economics of irrigation through
percolation tank and irrigation tan
3. AlSaleem SS (2017) Rainfall–runoff analysis for sustainable stormwater drainage for the city
of Madinah, Saudi Arabia
4. Dobbins DV (1961) Surface water drainage. Norte Dame Law Rev 36(4)
5. Strzepek KM, Marks DH, Wilson JL, Grossman DS (1980) Planning and design of agricultural
drainage systems under uncertainty. IFAC Proc Vol 13(3):243–252
6. Baoli Xu, Shao D, Fang L, Yang X, Chen S, Wenquan Gu (2019) Modelling percolation and
lateral seepage in a paddy field-bund landscape with a shallow groundwater table. Agric Water
Manage 214(1):87–96
7. Liu X, He Y, Zhao X, Zhang T, Li Y, Yun J, Wei S, Yue X (2016) The response of soil water
and deep percolation under Caragana microphylla to rainfall in the Horqin Sand Land, northern
China. CATENA 139:82–91
8. Borsi I, Farina A, Fasano A (2004) On the infiltration of rain water through the soil with runoff
of the excess water. Nonlinear Anal Real World Appl 5(5):763–800
9. Sandoval ADO, Brião VB, Fernandes VMC, Hemkemeier A, Friedrich MT (2019) Storm-water
management by microfiltration and ultrafiltration treatment. J Water Process Eng 30:100453
10. Deitch MJ, Feirer ST (2019) Cumulative impacts of residential rainwater harvesting on
stormwater discharge through a Peri-urban drainage network. J Environ Manage 243(1):127–
136
11. Abd Rasid NS, Naim MN, Che Man H, Abu Bakar NF, Mokhtar MN (2019) Evaluation of
surface water treated with lotus plant; Nelumbo nucifera. J Environ Chem Eng 7(3):103048
12. Zhou Bo, Li Y, Xue S, Feng Ji (2019) Variation of microorganisms in drip irrigation systems
using high-sand surface water. Agric Water Manag 218(1):37–47
13. Sandoval GFB, Galobardes I, Teixeira RS, Toralles BM (2017) Comparison between the
falling head and the constant head permeability tests to assess the permeability coefficient
of sustainable pervious concretes. Case Stud Constr Mater 7:317–328
Specific Heat
of Nanofluids—An Experimental
Investigation
Tushar Anand and Soumya Suddha Mallick
1 Introduction
Nanofluids are a new class of engineered fluids obtained by suspending nano-size

(10−9 m) particles with an average size below 100 nm in heat transfer fluids [1].
Oxides, metals, nitrides, and non-metals, such as carbon nanotubes are used as
nanoparticles, while water, ethylene glycol, oils, and polymer solutions, and conven-
tional coolants are used as base fluids. The smaller size of nanofluids offers several
advantages over conventional heat transfer fluids, such as long-term stability, low
abrasion, low pumping power, homogeneity, and minimum clogging in flow passages
[2, 3]. These benefits make nanofluids potentially attractive to various industries
having heat transfer applications like microelectronics, transportation, biomedical,
micro-fluids, nuclear, automobile, power generation, X-ray, refrigerators, etc. [4,
5]. Miniaturized systems will reduce heat transfer fluid inventory and successful
employment of nanofluids will result in significant energy and cost savings because
heat exchange systems can be made smaller and lighter [6].
Despite such merits and widespread potential applications of nanofluids, nanofluid
technology is still limited for commercial use because there is yet no proven standard-
ized design process for accurately predicting important heat transfer properties, such
as the nanofluids specific heat, thermal conductivity, and viscosity because of the
influence of various particle and fluid properties, such as the shape and size distribu-
tion of nanoparticles [7, 8], the volume concentration of nanoparticles in base fluids
[9, 10], ultrasonication and storage time to prepare nanofluids [9], use of surfactants
[7, 10], pH value [3] and temperature [7, 10]. Developing an accurate fundamental
model for the specific heat of nanofluids is a challenging task. As a thermodynamic
T. Anand (B)
Satyug Darshan Institute of Engineering and Technology, Faridabad 121002, India
S. S. Mallick
Mechanical Engineering Department, Thapar University, Patiala 147001, India
e-mail: ssmallick@thapar.edu
454 T. Anand and S. S. Mallick
property, the specific heat capacity of a nanofluid dictates the nanoparticle and fluid
temperature changes, which affect the temperature field of the nanofluid, and hence,
the heat transfer and flow status in any application, e.g., systems for utilizing low-
temperature solar thermal energy, include means for heat collection, usually heat
storage, either short-term or interseasonal and distribution within a structure or a
district heating network [11].
Hence, the purpose of this paper is to first construct a setup and validate it.
Secondly, to carry out experimental work to find out the effect of different particle
sizes, volumetric concentration, sonication time, frequency of the specific heat of
different nanofluid. Thirdly, to develop an improved model for specific heat using
the data obtained.
2 Experimental Setup and Procedure
The experimental setup used is shown in Fig. 1. The apparatus consists of a 20 cm

long and 7 cm inside borosilicate container. The apparatus is designed to hold about
900 ml of liquid. The nanofluids are heated from about 298–323 K by using an
electrical immersion heater. Ten K type copper–constantan thermocouples place in
the apparatus. Four thermocouples are placed within the liquid volume to get the
average temperature of the liquid. One on the heating coil, two thermocouples are
placed on the outer surface of the container, two on the outer surface of second
insulation, and one at the midpoint of the last insulation coil. These thermocouples
are connected to a data logger that records the temperature data at every 15 s interval.
Fig. 1 Layout of experimental setup for specific heat

Specific Heat of Nanofluids—An Experimental Investigation 455
The container is well insulated to minimize heat loss. The specific heat of a
nanofluid was calculated from the following equation:
Qt − m C CPC TC − m co CPCO TCO − m in CPIN − qL t

CPnf = (1)
m nf Tnf
where Q is the heat applied to the electrical heater in watts determined from the power
meter. The time interval t is measured by the data logger s, T is the temperature rise
K, m is the mass in kg, Cp is the specific heat kJ/kgK, and qL is the heat transfer to
the environment (W). The subscripts C represents the container, CO the heating coil,
and IN the insulation. The masses of the container, coil, and insulation are measured
individually by an electronic mass balance. The temperature range T is recorded
every 15 s intervals. The data values used are taken from standard sources. The setup
has been validated satisfactorily with experiments on water and ethyl glycol (25, 30,
and 50%).
3 Modeling
There are two specific heat models widely used in the nanofluid literature. The model
I is similar to mixing theory for ideal gas mixtures [12]. This is macroscopic, that is,
the specific heat capacity of a nanofluid is equal to the average of the specific heat
capacities of base fluid and nanoparticles.
CPnf = φCPs + (1 − φ)CPbf (2)
Model II [7] is based on the assumption of thermal equilibrium between the

particles and the surrounding fluid. The is microscopic, which assumes the base fluid
and the nanoparticles are in their thermal equilibrium. The nanofluid specific heat
capacity per unit mass of nanofluid, that is, the nanofluid specific heat, is
Cps φρs + (1 − φ)ρbf CPbf

CPnf = (3)
ρnf
where ρ s is the density of the solid nanoparticle, ρ bf is the density of the base fluid,
and ρ nf is the density of the nanofluid. The product of density and specific heat is
the volumetric heat capacity of each constituent and that of the nanofluid.
3.1 Evaluation of Existing Models
Models given in Eqs. (2) and (3) are being compared with the experimental data for
the same range of volume fraction, temperature, and particle size. Lines of best fit
have been drawn through the experimental data to indicate their trends. The following
parameters are considered for calculation in this paper (Figs. 2 and 3):
Specific heat and density of water areas 4.186 kJ/kgK and 1000 kg/m3 , respec-
tively; (b) density of Al2 O3 particles were as mentioned in their respective references,
such as ρ p : 3965 kg/m3 ; (c) specific heat of Al2 O3 is 880 J/kg K;
With an increase in the nanoparticle volume fraction, the predicted values increase
for all the models; also, the number of inaccuracy increases. It is seen that the above
models do not fit the experimental data; therefore, a model using the dimensional
analysis was used to develop a correlation fitting the experimental data. The variables
like diameter, temperature, the viscosity of the fluid, the density of the particle, base
Fig. 2 Al2 O3 nanoparticles,

d: 50 nm, T: 35 °C
Specific Heat (Cp)
Fig. 3 Al2 O3 nanoparticles,

d: 50 nm, T: 55 °C
fluid, volume fraction, the specific heat of nanoparticle and that of nanofluid are
taken. Specific heat of nanofluid is believed to be depending on various particle and
fluid properties, as given by:
CPnf = f (d, T, μ, ρf , ρP , CPP , CPf , φ) (4)
Or
f 1 (d, T, μ, ρf , ρP , CPP , CPf , φ)
Using Buckingham Pi theorem, the following dimensionless groups have been

obtained:
d 2 TρP2 Cpf ρf d 2 Tρp2 C Pp

π1 = , π2 = , π3 = ,
μ2 ρp μ2
d 2 TρP2 C pf
π4 = , π5 = φ
μ2
Experimental data for Al2 O3 -water nanofluids ZnO-water nanofluids and SWCNT
for a wide range of volume fraction, particle size, and temperature, the following
models have been derived using regression analysis. Finally, the specific heat of
nanofluid from the dimensionless analysis is as shown in (Fig. 4).
2.28 d 2 Tρ 2 −1.62 d 2 Tρ 2 2.81

μ2 ρf pC Pp pC pf
Cnf = 0.152 2 2 φ (5)
d TρP ρp μ2 μ2
The model given in Eq. (5) is for Al2 O3 -water nanofluids, ZnO, and SWCNT.
The temperature range for which the relation is valid up to 338 K.
4.8
4.7
4.6
Cp (Predicted)
4.5
4.4 +2.5%
4.3
-2.5%
4.2
4.1
4
4 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8
Cp (Experimental)
Fig. 4 Experimental versus predicted values of specific heat for Al2 O3 -water (alpha) nanofluids
using a new model given (Eq. 5)
The above models have been used to predict the nanofluid specific heat between
predicted and experimental values for Al2 O3 -water (alpha) nanofluids. The compar-
ison plots show that the new model generally predicts within ±2.5% accuracy range
for Al2 O3 -water (alpha) nanofluids
Figure 5: Comparison between experimental and predicted plots shows that the
new model generally predicts within ±5% accuracy range for Al2 O3 -water (gamma)
nanofluids
new model generally predicts within ±5% accuracy range for ZnO (14 nm) nanofluids
new model generally predicts within ±3% accuracy range for ZnO (14 nm) nanofluids
Fig. 5
Al2 O3 -water(gamma)
nanofluids using a new
model given Eq. (5)
Fig. 6 ZnO (14 nm)

model given Eq. (5)
Fig. 7 ZnO (24 nm)

model given (Eq. 5)
Fig. 8 SWCNT (1 nm)

model given (Eq. 5)
new model generally predicts within ±2.5% accuracy range for SWCNT (1 nm)
nanofluids
4 Uncertainty Analysis
The uncertainty in the specific heat of the experimental data measurement in Eq. (1)
can be determined from the standard approach presented by [13]. The parameters
measured are the rate of heat input to the nanofluid, the temperature, and the mass
of several objects and dimensions of insulations.
2
δCpnf δQ δm c 2 δTC 2 δm CO 2 δTCO 2
= + + + +
Cpnf Q mc mC m CO TCO
1/2
δm IN 2 δTIN 2 δqL 2 δm nf 2 δTnf 2
+ + + + + (6)
m IN TIN Q m nf Tnf
Specific heats of the container, heating element, and insulations were directly read
from tabulated values in books and were not measured quantities in this experiment.
The data acquisition system was capable of sampling temperatures at intervals of
microseconds. Therefore, the uncertainty in t of 15 s was considered negligible.
For the power meter, δ Q/Q is about 1%. For the electronic precision, mass balance
is δm/m about 0.5%. The uncertainty in measurements of temperature for copper–
constantan thermocouple used in this apparatus is 0.5 °C between −100 and 400 °C.
Therefore, at the mean temperature of 50 °C within the range of measurements, δt/t
= 1%. The uncertainty in calculating the heat loss through the insulation can be
expressed as Eq. (7)
2 2 2 2 2 1/2
δqL δA δTC δTO δ X1 δ X2
= + + + + (7)
qL A TC TO X1 X2
where X 1 and X 2 are thicknesses of insulations, and T C and T O are the surface
temperatures of insulations. The thermal conductivities of insulations are taken from
Incropera, [14]. The uncertainty in length measurement δL/L by the modern meteo-
rological gauge is about 0.5%. The area A is proportional to the square of the length
1/2
dimension L, so the uncertainty in area measurement is (2(δL/L))2 . Using the
above numbers, δqL /qL = 1.87%. Finally, combining all the uncertainties together in
Eq. (6), the uncertainty in measurement of the specific heat of nanofluid is δCpnf /Cpnf
= 3.1%.
5 Conclusions
1. The specific heat of nanofluids increases with an increase in temperature and

volume fraction. For example, Al2 O3 (0.0014%) at 302 K is 4.4 kJ/kgK, and at
303 K, it is 5.1 kJ/kgK, ZnO; at 0.001%, volume fraction is 3.7 kJ/kgK, and at
0.002% is 4.03 kJ/kgK.
2. A new specific heat model has been developed using dimensionless analysis.
The new model has been found to generally predict specific heat of nanofluids
within
±2.5% accuracy range for Al2 O3 -water nanofluids,

± 5% accuracy range for ZnO-water nanofluids and
± 5% accuracy range for SWCNT.
3. The newly developed model gave good results when compared with models of
different nanofluid of SWCNT, Al2O3, ZnO, ethyl glycol up to the temperature
range of 298–308 K.
Acknowledgements The authors are very thankful to Dr. N J Dembi for his important suggestion
regarding this paper and Thapar University for its seed money.
References
1. Choi SUS (1999) Nanofluid Technology: current status and future research. Energy technology
division, Argonne national laboratory, IL, p 60439
2. Gallego MJP, Lugo L, Legido JL, Pineiro MM (2008) Thermal conductivity and viscosity
measurements of ethylene glycol-based Al2 O3 nanofluids. Nanoscale Res Lett 6(211):1–11
3. Xie HQ, Wang JC, Xi TG, Liu Y, Ai F, Wu QR (2002) Thermal conductivity enhancement of
suspensions containing nanosized alumina particles. Appl Phys 91(7):4568–4572
4. Chandrasekar M, Suresh S, Bose AC (2010) Experimental investigations and theoretical
determination of thermal conductivity and viscosity of Al2 O3 water
5. Khanafer K, Vafai K (2011) A critical synthesis of thermophysical characteristics of nanofluids.
Int J Heat Mass Trans 54:4410–4428
6. Parashivamurthy KI (2010) International conference on advanced materials, manufacturing,
management and thermal science, Department of Mechanical Engineering, Siddaganga
Institute of Technology, Tumkur-572104, Karnataka, India, vol 1
7. Murshed SMS, Leong KC, Yang C (2005) Enhanced thermal conductivity of TiO2 -water-based
nanofluids. Int J Therm Sci 44:367–373
8. Wang ZL, Tang DW, Liu S, Zheng XH, Araki N (2007) Thermal-conductivity and thermal-
diffusivity measurements of nanofluids by 3ω method and mechanism analysis of heat transport.
Thermophys 28:1255–1268
9. Timofeeva EV, Gavrilov AN, McCloskey JM, Tolmachev YV, Sprunt S, Lopatina LM, Selinger
JV (2007) Thermal conductivity and particle agglomeration in alumina nanofluids: experiment
and theory. Phys Rev E 76(061703):1–16
10. Madhu P, Rajasekhar PG (2017) Measurement of density and specific heat capacity of different
nanofluids. IJARIIT. ISSN: 2454-132X. 165–170
11. Vajjha RS, Das DK (2009) Specific heat measurement of three nanofluids and development of
new correlations. J Heat Transfer 131(7):071601
12. Das SK, Choi SUS, Yu W, Pradeep K (2007) Nanofluids science and technology, Wiley
interscience. Appl Phys 9:131–139
13. Coleman HW, Steele WG (1999) Experimentation and uncertainty analysis for engineers, 2nd
edn. Wiley, New York
14. Incropera FP, DeWitt DP (1996) Introduction to heat transfer, 3rd edn. Wiley, New York
15. Chon CH, Kihm KD, Lee SP, Choi SUS (2005) Empirical correlation finding the role of
temperature and particle size for nanofluid (Al2 o3 ) thermal conductivity enhancement. Phys
Lett 87(153107):1–3
16. Mintsa HA, Roy G, Nguyen CT (2007) New temperature-dependent thermal conductivity data
of water-based nanofluids. In: 5th IASME/WSEAS international conference on heat transfer,
thermal engineering and environment, Athens, Greece, pp 290–294
17. Das SK, Putra N, Thiesen P, Roetzel W (2003) Temperature dependence of thermal conductivity
enhancement for nanofluids, heat transfer. J Heat Transf 125:567–574
Real-Time Detection of Droplet Velocity
Using Open-Source Computer Vision
on EWOD Device
Vandana Jain and Rajendra M Patrikar
1 Introduction
Nowadays, new method for microscale liquid control is being developed due to
advancement of scaling in technology in which manipulating fluid movement by
using surface tension is getting more attention [1]. DMF is become promising droplet
handling technology that control droplets on an array of electrodes pads. DMF on
a single platform can perform multiple operations parallel like dispensing, trans-
porting, splitting, and merging [2]. This chip can perform as a glucose test [3], human
physiological fluid [4], polymerase chain reaction (PCR) [5], multiplexed proteomic
[6]. EWOD has been shown as a most powerful technique in biomedical applica-
tion due to its reconfiguring ability, flexibility, dynamic nature, high automation and
processability using optical and electrical techniques.
1.1 Electrowetting-on-Dielectric (EWOD) Theory
In EWOD, wetting behavior of a conducting droplet is changed by electric field

[7, 8]. After rigorous theory and experiments, the relationship between the wetting
behaviors of droplet with the droplet contact angle [8] is given by the Lippmann–
Young. The Lippmann–Young equation (EWOD equation) is given in Eq. (1)
V. Jain (B)
Department of Mechanical Engineering, Indian Institute of Technology Kanpur, Kanpur, Uttar
Pradesh 208016, India
e-mail: vandanaj@iitk.ac.in
R. M. Patrikar
Centre for VLSI and Nanotechnology, Visvesvaraya National Institute of Technology, Nagpur
440010, India
e-mail: rajendra@computer.org
464 V. Jain and R. M. Patrikar
Fig. 1 EWOD system with catenae a without change of contact angle b with change of contact
angle
εr ε0 2
cos θ (V ) = cos θ0 + V (1)
2dγLG
Here, εr : relative dielectric constant, ε0 free space permittivity, γ LG : solid–liquid

surface tension d: dielectric thickness, V is the applied voltage and θ (V ) andθ 0
contact angle with and without applied voltage. The EWOD system with catenae is
shown in Fig. 1.
1.2 Dynamic Aspects of EWOD
In EWOD devices, droplet movement is occurred due to capillary force which takes
place in between actuated and non-actuated electrode pad as shown in Fig. 2. The
capillary line is given by [9].
Fx = LγLG (cos θ (V ) − cos θ0 ) (2)
The electrowetting effect is translated in to a capillary effect by using Lippmann–

Young equation. Therefore, electrowetting force is rewritten by following equation
[10]. L is the effective contact line length.
1 εr ε0 L 2
Fx = V (3)
2 d
Fig. 2 Droplet motion in open EWOD system a without actuation voltage (yellow electrode) b with
actuation voltage (red electrode)
Real-Time Detection of Droplet Velocity Using Open-Source … 465
The analytical droplet velocity expression for 105° static contact angle is derived
by using [11]
εr ε0 L 2
V
U= d
εr ε0 L
(4)
π K 1 K 2 K c Cv μR(1.48 + dγLG
· V 2)
Here K 1 : time factor, K 2 : top ground wire dragging effect. K C : damp factor, C V :
empirical parameter, μ: droplet viscosity, and R: droplet radius.
In this paper, we have presented PDMS-based EWOD system with real-time
detection of droplet velocity using OpenCV. This work provides a comprehensive
approach for detecting droplet velocity using of digital image processing, which is
eliminated system complexity.
2 Device Fabrication
The open EWOD device is shown in Fig. 3. We have used UV lithography technique
for the fabrication of square electrode array.
In this study, three Cr/Gold (50 nm/150 nm thickness) metal electrodes array
having size 2 × 2 mm, 1.5 × 1.5 mm and 1 × 1 mm with 80 µm separation are
realized on a 50 mm × 50 mm thick glass substrate. The 1 mm × 1 mm electrode
pad through 100 µm full metal line is fabricated for providing connection of each
bottom electrode pad.
Fig. 3 Open EWOD device;

square electrode size (1)
2 mm × 2 mm (2) 1.5 mm ×
1.5 mm (3) 1 mm × 1 mm.
For each case, the gap
between two electrodes is
80 µm
We have used PDMS (Sylgard 184) polymer solution both as a dielectric and a
hydrophobic material. PDMS has a contact angle of 105°; hence, this can be useful
for droplet operation [12]. The PDMS solution is prepared by taking 10 to 1 ratio of
base and hardener and mixed it for 10 min. Then mixed solution is kept for 10 min
in a vacuum chamber. The thin layer of PDMS coating on the glass is achieved by
spin coater (5000 rpm, 60 s). After that, coated device is cured at 100 °C, 1 h. A
Zeta 3D microscope is used for PDMS layer thickness (15 µm) measurement. The
PDMS coating is removed from contact pads to get the contacts for droplet actuation.
Then double-sided tape is used to stick the EWOD glass device on the printed circuit
board, and female connector is used to provide the contact. A 100-µm-diameter-
thick aluminum is used as a top ground wire, and it arranges parallel with bottom
electrode using small PCB patch.
3 System Specification
3.1 High-Voltage Control Unit
In the EWOD, the variable high DC voltage source (0–400 V DC ) is required for
performing the basic droplet operation. Hence, a boost converter is designed for
getting actuation voltage up to 400 V DC as shown in Fig. 4. Eight high-voltage source
lines are connected to the electrode pads through eight electromagnetic relays. These
relays are connected to GPIO pins of Raspberry Pi . For controlling the GPIO pins, a
GUI is developed using Qt software in C++. The complete setup (see Fig. 4) consists
of USB digital microscope, sample holder XYZ stage. Top ground electrode and
Display
High Voltage Unit
Z Stage
Microscope
Sample
X-Y Stage
Fig. 4 Test setup

Fig. 5 Droplet transporting with 2 mm × 2 mm square electrode at 350 V DC
bottom electrode pad are connected with power supply by using male connector
[13].
3.2 Image Processing
The image analysis is done by using OpenCV whose libraries are linked with Qt soft-
ware. OpenCV image processing functionalities are improving our analysis results.
Each frame of the live stream is correlated through image processing algorithm.
4.1 Droplet Transportation
The droplet motion by using the electric field is examined in the device. In this
experiment, 5 µL water droplet is used. To make water droplet conducting, 0.1 M
KCL is used. By using the high-voltage control unit, electrodes are sequentially
energized from P1 to P5 to get forward direction droplet motion. In initial experiment,
we have not found any droplet movement from one pad to another pad due to PDMS
pining effect. Therefore in order to get droplet transportation, a thin layer of silicone
oil is dispersed over the PDMS layer. Silicone film minimizes the tension required
to move the droplet [14]. Therefore, all the experiments are performed with silicone
film. The droplet transporting on square electrodes is shown in Fig. 5.
4.2 Droplet Velocity Measurement
The droplet velocity is playing an essential role for various digital microfluidics
applications. The real-time measurement of droplet velocity is challenging task in
EWOD system. In this paper, we have demonstrated the real-time droplet velocity
measurement using OpenCV libraries. The droplet centroid is detected by applying
the color threshold method on the droplet as shown in Fig. 6.
Fig. 6 a Camera captured droplet image b threshold droplet image
After getting the droplet centroid point, we have started timer count 1 when the
droplet centroid overlap with pad1 centroid and timer count 2 is started when the
droplet centroid overlap with pad2 centroid. Then, difference between the two timer’s
times and two droplet centroid point is calculated. The droplet centroid difference is
multiplied by calibration factor to get the distance in mm. The ratio of distance and
time is calculated. Using this technique, we can track efficiently droplet velocity in
real time at any pad.
We have calculated the average droplet velocity for the 150 V DC to 350 V DC
applied voltage range at 80 µm electrode gap with different square electrode sizes
as shown in Fig. 7.
We have found that no droplet motion occurred below the threshold voltage. Here,
the minimum threshold voltage required for the droplet motion is 160 V DC. After
160 V DC the significant droplet motion is happened. For droplet motion, enough
electric field has to be taken to reduce the interfacial tension [15, 16]. The droplet
Fig. 7 Droplet velocities for

different electrode size at
80 µm with respect to
voltage
Fig. 8 Comparison between

experiment and theoretical
results
velocity is 9.75 ± 0.5 mm/s, 6.76 ± 0.5 mm/s, 3.3 ± 0.5 mm/s for 2 × 2 mm 1.5 ×
1.5 mm and 1 × 1 mm respectively at 350 V DC with 80 µm electrode gap.
To verify the experiment results with analytical results, the average droplet
velocity and applied voltage for 2 mm × 2 mm square electrode size at 80 µm
with analytical value are plotted which is shown in Fig. 8.
The analytical droplet velocity is calculated using Eq. (4). We have found that
the velocity is showing square relationship with the applied actuation voltage (R2 =
0.99). The obtained results are well matched with the analytically calculated droplet
velocity. The relationship is also showing the good matches with the other analytical
models [7, 15–17].
5 Conclusions
In this paper, we have illustrated the open EWOD device for square electrode patterns
with different sizes at 80 µm electrode gaps. The average droplet velocities are
calculated in real time by using OpenCV libraries and Qt software. The average
droplet velocity is having the square relationship with the applied actuation voltage
which is investigated both experimentally and analytically. This work gives a useful
platform of characterizing EWOD devices in real time, which can be used for various
digital microfluidics applications.
Acknowledgements We would like to thank to all CEN staff at IIT Bombay for their constant
support for improvement in EWOD device fabrication under INUP program.
References
1. Fair RB (2007) Digital Microfluidics: is a true lab-on-a-chip possible? Microfluid Nanofluid

3:245–281
2. Cho SK, Moon H, Kim CJ (2003) Creating, transporting, cutting, and merging liquid droplets
by electrowetting-based actuation for digital microfluidic circuits. J Microelectromech Syst
12(1):70–80
3. Srinivasan V, Pamula VK, Fair RB (2004) Droplet-based microfluidic lab-on-a-chip for glucose
detection. Anal Chim Acta 507(1):145–150
4. Srinivasan V, Pamula VK, Fair RB (2004) An integrated digital microfluidic lab-on-a-chip for
clinical diagnostics on human physiological fluids. Lab Chip 4(4):310–315
5. Chang Y-H, Lee G-B, Huang F-C, Chen Y-Y, Lin J-L (2006) Integrated polymerase chain
reaction chips utilizing digital microfluidics. Biomed Microdevices 8(3):215–225
6. Moon H, Wheeler AR, Garrell RL, Loo JA, Kim C-JC (2006) An integrated digital microfluidic
chip for multiplexed proteomic sample preparation and analysis by MALDI-MS. Lab Chip
6(9):1213–1219
7. Berthier J (2013) Micro-drops and digital dicrofluidics, 2nd ed. Elsevier
8. Mugele F, Baret J-C (2005) Electrowetting: from basics to applications. J Phys Condens Matter
17(28):R705–R774
9. Berthier J, Dubois P, Clementz P, Claustre P, Peponnet C, Fouillet Y (2007) Actuation potentials
and capillary forces in electro wetting based microsystems. Sens Actuators A Phys 134(2):471–
479
10. Ren H, Fair RB, Pollack MG, Shaughnessy EJ (2002) Dynamics of electro-wetting droplet
transport. Sens Actuators B Chem 87(1):201–206
11. Cui W, Zhang M, Duan X, Pang W, Zhang D, Zhang H (2015) Dynamics of electrowetting
droplet motion in digital microfluidics systems: from dynamic saturation to device physics.
Micromachines 6(6):778–789
12. Mata A, Fleischman AJ, Roy S (2005) Characterization of polydimethylsiloxane (PDMS)
properties for biomedical micro/nanosystems. Biomed Microdevices 7(4):281–293
13. Jain V, Raj TP, Deshmukh R, Patrikar R (2015) Design, fabrication and characterization of low
cost printed circuit board based EWOD device for digital microfluidics applications. Microsyst
Technol 21:1–9
14. Bavière R, Boutet J, Fouillet Y (2008) Dynamics of droplet transport induced by electrowetting
actuation. Microfluid Nanofluid 4(4):287–294
15. Jain V, Devarasetty V, Patrikar R (2017) Effect of electrode geometry on droplet velocity in
open EWOD based device for digital microfluidics applications, J Electrostat 87
16. Song JH, Evans R, Lin YY, Hsu BN, Fair RB (2009) A scaling model for electrowetting-on-
dielectric microfluidic actuators. Microfluid Nanofluid 7(1):75–89
17. Baird ES, Mohseni K (2007) A unified velocity model for digital microfluidics. Nanoscale
Microscale Thermophys Eng 11:109–120
Stability Analysis of Two-Phase Slug
Flow Using OpenFOAM
Rohit Singh Gulia, Siddharth Sharma, and Jyotirmay Banerjee
1 Introduction
Due to interaction between gas and liquid at interface, the interface is deformed
in different shapes leading to different flow patterns for gas–liquid two-phase flow.
Three distinct flow patterns are: intermittent, separated and dispersed flow pattern.
Nature of Navier–Stokes equation is nonlinear. Hence, any rise in velocity grows
instability with time. A small interfacial instability grows or decays with time. Flow
instability behaviour is observed to be similar to that of amplifier. External energy
is supplied, and the primary instability is transformed to secondary instability. For
the initiation of slug flow regime, having high flow rate of gases analysis is required
for slug stability rather than stratified flow stability. Slug initiation at higher gas
velocities is by merging of two K–H waves rather than due to the growth in smaller
disturbances. It is the growth of primary instability due to the presence of shear
or relative velocity on the interface between two moving fluids which initiates the
perturbation on interface.
Gas–liquid two-phase flows are the complex due to high relative density. Avail-
ability of geometric scheme, i.e. isoAdvector for volume fraction calculation, is an
advantage for their analysis using OpenFOAM. Another advantage for IsoAdvector
algorithm is to retain the accuracy of geometric schemes while keeping the geometric
operations to a minimum.
R. S. Gulia · S. Sharma (B) · J. Banerjee

J. Banerjee
472 R. S. Gulia et al.
Fig. 1 Flow regime map by Thaker and Banerjee [3]
2 Literature Review
IsoAdvector algorithm is elaborated by Roenby [1]. It is geometric-based algorithm

and behaves accurately compared to MULES (multi-dimensional universal limiter
for explicit solution) [2].
Flow pattern map given by Thaker and Banerjee suggests the transition boundaries
between different flow patterns are as shown in Fig. 1. Of these slug flow pattern is
chosen for current work. Ruder [4] stated that for the stable slug, the slug front is
defined as the hydraulic jump which in turn gives a minimum gas velocity at which
slug is expected to occur. Woods and Hanratty [5] stated the relation of shedding
rate at slug front with slug stability. Numerical simulations using OpenFOAM are
performed by Thaker and Banerjee [6], and results are compared with the experi-
mental facility available in Advanced Fluid Dynamics Laboratory, SVNIT, Surat.
Shuard [7] simulated flows through a symmetric pipe geometry having 0.052 m
diameter and 4 m length using OpenFOAM. Comparison of simulated results with
that of the mechanistic model results for pressure and liquid hold-up reported by
Petalas and Aziz [8] is done.
3 Methodology
To validate IsoAdvector algorithm, the experiments [3] are performed in Advanced

Fluid Dynamics Laboratory, SVNIT. Two-phase flow is modelled for 0.025-m-
diameter and 8-m-long pipe. Slugging frequency and average slug velocity correla-
tions given by Thaker and Banerjee [9] are used to test the accuracy of CFD results.
Default Euler scheme for temporal terms followed by Gauss linear upwind scheme
Stability Analysis of Two-Phase Slug Flow Using OpenFOAM 473
Table 1 Grid size

Mesh R1 R2 R3 R4 R5
Cells(*103 ) 1 1.8 2.8 3.55 3.95
((X*Y*2*L) + (X*Z*2*L))/104 15 27 42 53.25 59.25
Fig. 2 Average liquid hold-up at different conditions
are used for advection. Courant number is set to 0.3 for better flow of information
through cells. IsoAdvector algorithm (to track the transport of volume of fluid across
mesh face) is selected as interface method by setting nCorrectors to 2 in PIMPLE
loop. K-Omega SST two-equation model is used to resolve turbulence. At inlet, 50%
area to both gas and liquid is given for entry of fluid into the pipe through groovyBC
using swak4Foam tool. Pipe is assumed to be completely filled with liquid in the
beginning.
First of all, grid independence analysis is carried out considering average liquid
hold-up as grid sensitivity parameter. Five different test cases are performed by
changing cells along length only, as the cells along cross section are not found out
to be so influencing in the simulation results. Mesh R 4 is selected as the final grid
size for analysis (Table 1; Fig. 2).
Mass flow rates for both the phases at inlet are controlled by regulating control valves
with help of SCADA system. To get a proper two-phase flow mixture at inlet, a two-
phase flow mixer near inlet is installed. For validation, a 25-mm-diameter pipe (from
available 12-, 25- and 50-mm-diameter pipes) is considered, and data is recorded via
experiments performed at room temperature; test section for analysis is between 6
and 8 m length. Velocities with star as superscript in Table 2 are considered as inlet
velocity conditions (Table 3).
Table 2 Inlet conditions

UG (m/s) UL (m/s) USG (m/s) USL (m/s)
2.20869* 0.441739* 1.10434783 0.2208696
*The velocities are taken as inlet velocities
Table 3 Fluid properties

Fluid ρ(kg/m3) υ(m2/s) σ (N/m)
Liquid 995 1.5541e−05 0.0724
Gas 1 0.8040128e−06
Validation of IsoAdvector algorithm shown in Fig. 3f or average slug velocities along

length is elaborated. Simulation results are in accordance with that of the experiments
performed (Table 4).
Fig. 3 Average slug 2.5

Average Slug Velocity
velocities along length 1.89183 1.722786 1.93352

2
1.5 1.722786 1.5812 1.722786
1
0.5
0
1 2 3
Slugs Along Length
Experiments CFD
Table 4 Boundary conditions

Parameter Inlet Outlet Wall
U groovyBC PressureInlet OutletVelocity noSlip
P_rgh fFPa totalPressure fFP
K TIKEIb inletOutlet KWFc
Omega inletOutlet inletOutlet OmegaWallFunction
a Fixed flux pressure. b Turbulent intensity kinetic energy inlet. c kqRWallFunction
Fig. 4 Hydraulic jump
Fluid
Average velocity and mixing velocity calculations are below. It is found out that
mixing velocity is always less than average by the multiplication factor of 1.3 experi-
mentally. However, results are slightly different than that of Hubbard and Gregory but
are in accordance with numerical. Values of average slug velocity via experimental
results and CFD results are plotted in Fig. 3. Errors are below 5% with respect to
experimental results.
Vmix = VSL + VSG = 1.32522 m/ s
Vavg = 1.3 × Vmix = 1.722786 m/ s
1.802183333 − 1.722786
Error = × 100 = 4.608%
1.722786
There are various energy levels at different points along interface, and all of these
are attached to the interface via force balance. The primary instability grows to
the secondary instability, and this gives rise to the eddy formation, turbulence and
makes fluid more chaotic. This results into shooting nature of flow as shown in Fig. 4,
reduction in velocity of fluid and increase in the depth of fluid. This rise in depth of
fluid is termed as hydraulic jump or standing wave. Froude’s number calculations
are shown below.
Section 1(Rapid Flow) Taken at a distance of 6.76357 m from inlet and in the
middle of cross section of pipe.
v1 0.541395
Fr1 = √ =√ = 1.08438079
g × h1 9.81 × 0.0254
Section 2 (Tranquil Flow) Taken at a distance of 6.79096 m from inlet and in the
middle of cross section of pipe.
v2 0.484554
Fr2 = √ =√ = 0.970712945
g × h2 9.81 × 0.0254
There is a difference in depth of fluid between Sects. 1 and 2. According to the

continuity equation, the discharge should be same on both the section. But both
sections have different depth of fluid. Hence, the discharge per unit width at Sect. 2
is less than that of the Sect. 1. The specific energy of the flow is directly proportional
to the discharge per unit width (height of fluid from pipe base). Hence, there is a loss
of specific energy of flow when hydraulic jump occurs within the flow.
Fig. 5 Pressure contours
4.1 Condition for Occurrence of Hydraulic Jump
In a small region along pipe length due to presence of relative velocity in the bottom
fluid, accumulation of heavier fluid is observed. Due to this, the fluid depth starts
rising. This leads to the conversion of dynamic pressure (high velocity stream) to
static pressure (high depth wave), and as a result, the depth of fluid increases. Forma-
tion of standing wave or hydraulic jump here is driven by the variation in pressure
along pipe diameter as shown in Fig. 5.
4.2 Slug Front as Hydraulic Jump
During flow through horizontal pipe, there are various pressure transition regions
along length. Some of these are sudden transition zones, and some are gradual. Out
of these, the hydraulic jump comes under the sudden pressure transition. In complete
slug body, slug front is the region which is on verge of sudden transition as ahead
of it there is a constant pressure region while before it there is a gradual shift in
pressure throughout the slug body. Here fluid accumulation is due to the velocity
difference along length within same fluid. This whole process of energy loss and
energy extraction continuously happens at slug front which helps in making this
unstable slug travel throughout pipe length.
As shown in Fig. 6, after pressure transition within slug body there is a sudden
pressure rise in bottom fluid at slug front. This is the cause of hydraulic jump.
4.3 Stability of Slug Flow via Hydraulic Jump
Figure 7 explains the slug length varying mechanism, which is periodic in nature.
At time 38.575 s, the slug length is 0.83 m, which is the maximum length among
all time steps. Slug travels with same length in next time step but at time 38.625 s,
there is decrement of 0.02 m in slug length. The possible cause behind this is that
Slug Front
Slug
Body
Fig. 6 Contours and field of pressure within slug
Fig. 7 Mechanism of variation of slug length
the velocity gained by slug front is transmitted to the slug tail through slug body
which leads to decrease in slug front velocity. Due to this, the slug tail propagates
at faster rate than before and this faster movement of slug tail causes a compression
effect in slug length. Again, as the slug front approaches the interface ahead, it loses
its velocity and falls back on interface ahead. This leads to decrease in velocity of
that region and the surge coming from behind generates the standing wave, which
again influences the velocity. Velocity rise of slug while coming in contact with that
of standing wave causes an increment in length. Length at time 38.65 s reaches to
0.83 m. This cycle repeats in a periodic manner. Also due to standing wave, there
is an obstruction in the region of slug which lies below wave crest level. Hence, it
leads to the accumulation of water and leads to nozzle effect, i.e. water is pressurized
towards wave crest causing overriding of flow. It increases the pressure and in effect
height of wave increases.
Figure 8 shows the shedding and picking rate of fluid from interface. Due to
interfacial tension, there is a kind of obstruction of flow as high moving fluid (slug)
hits the slow moving fluid ahead at interface. This causes the accumulation of fluid
over this small region, and static pressure at bottom rises. This rise in static pressure
increases the depth of fluid. This wave will travel with less velocity than that of
slug. Initially, it acts as an obstruction for slug. Slug accumulates this liquid inside,
and some part of it overrides at the same time due to shear between fluid layers.
Fig. 8 Shedding and picking rate at slug.
This creates venturi effect above wave crest. When slug front passes this region, its
velocity increases. After some time, inertia is transferred throughout the slug body.
Slug front travels at faster than rest of slug body which leads to its breakage from rest
of the slug body. This phenomenon is known as the shedding and picking mechanism
of slug.
5 Conclusion
• The cause of hydraulic jump in flow is increase in static pressure at the bottom
of pipe due to accumulation of fluid because of relative velocity within heavier
fluid.
• Stability of slug flow depends on the rate of formation of hydraulic jump at slug
nose.
• It is very difficult for stable slug to travel through long pipe; regular energy
extraction or conversion process helps it in doing so.
• Slug flow is a periodic phenomenon, and regular extraction or release of energy
causes an increment or decrement in its length.
References
1. Roenby J, Bredmose H, Jasak H (2016) A computational method for sharp interface advection.
R Soc Open Sci 3(11):160405
2. Deshpande SS, Anumolu L, Trujillo MF (2012) Evaluating the performance of the two-phase
flow solver interFoam. Comput Sci Discov 5(1):014016
3. Thaker J, Banerjee J (2015) Characterization of two-phase slug flow sub-regimes using flow
visualization. J Pet Sci Eng 135:561–576
4. Ruder Z, Hanratty PJ, Hanratty TJ (1989) Necessary conditions for the existence of stable slugs.
Int J Multiph Flow 15(2):209–226
5. Woods BD, Hanratty TJ Relation of slug stability to shedding rate, p 20
6. Thaker JP, Banerjee J CFD simulation of two-phase flow phenomena in horizontal pipelines
using openfoam, p 15
7. Shuard AM, Mahmud HB, King AJ (2016) Comparison of two-phase pipe flow in OpenFOAM
with a mechanistic model. IOP Conf Ser Mater Sci Eng 121:012018
8. Petalas N, Aziz K A mechanistic model for multiphase flow in pipes, p 16
9. Thaker J, Banerjee J (2016) Influence of intermittent flow sub-patterns on erosion-corrosion in
horizontal pipe. J Pet Sci Eng 145:298–320
Characterisation of Steady Flow Regime
and Drag Force on the Forward
and Backward Facing Trapezoidal
Cylinders: A Numerical Study
Akash Bhunia, Prabir Sikdar, Sunil Manohar Dash, and Kim Boon Lua
1 Introduction
The study of the fluid dynamics around the bluff bodies is of great interest over the
last few decades due to its several engineering applications like the design of the
buildings, towers, bridges, road vehicles, etc. In general, beyond a critical Reynolds
number (Re) the flow past, a bluff body creates a periodic vortex shedding at its
downstream that imparts a fluctuating load and may cause structural damage on the
bluff body. Many factors like Reynolds number [1, 2], shape [3, 4] and orientation [5]
of the bluff body can affect the flow characteristics and vortex shedding phenomena
of the bluff body. Sen et al. [3] numerically studied flow past a square cylinder and
showed that the separation bubble length varies with Re. They have also reported
that the drag on the circular cylinder is less than the square one at the same Re. Flow
past square cylinder at different inclination angles was analysed by Sohankar et al.
[5]. They have investigated the unsteady flow features for Re ranging from 45 to
200. Dhiman et al. [1] and Sharma et al. [6] numerically studied the heat transfer
characteristics of the unconfined square cylinder. Many other researchers [1, 4, 7]
have changed the shape of the bluff body to see the geometric effect on the vortex
shedding pattern, shedding frequency as well as flow-induced drag, lift force. Lee
et al. [4] have investigated the wake regime behind a tapered trapezoidal cylinder for
A. Bhunia (B) · P. Sikdar · S. M. Dash

e-mail: akashbhuniavms@iitkgp.ac.in
P. Sikdar
e-mail: iamprabirsikdar@gmail.com
S. M. Dash
K. B. Lua
Department of Mechanical Engineering, NCTU Taiwan, Hsinchu, Taiwan
e-mail: engp4324@nctu.edu.tw
482 A. Bhunia et al.
Re range from 25 to 1000. They noted that Re and tapered angle changes the growth
rate, length and width of the separation bubble. Chung and Kang [7] investigated the
laminar vortex shedding from the trapezoidal cylinders having different height ratio
at different Re. Despite the variety of applications, a limited study on the unconfined
flow regime over the trapezoidal cylinders is available in the literature where the
flow field depends on Reynolds number, and the ratio of the size of the front and
rear faces and trapezoid length in the axial direction. In the present study, by varying
the above-mentioned parameters, a systematic numerical investigation on the steady
flow past trapezoidal cylinders is conducted using a finite volume solver. Also, the
effects of these parameters on the drag coefficient (C d ) and the separation bubble
length are analysed.
domain for flow past (top)
TCFF and (bottom) TCBF
cylinder
Characterisation of Steady Flow Regime and Drag Force … 483
2 Methodology
2.1 Problem Definition
As shown in Fig. 1, a computational domain of size 90D × 60D is followed in

this study. Here ‘D’ is the reference length scale, i.e. one of the vertical faces of
the trapezoidal cylinder. The axial length ‘H’ and size of the other vertical face
‘Y’ of the trapezoidal cylinder is gradually varied between 0.5D–2.0D and 0.125D–
1.0D. The trapezoidal cylinders are defined as a forward facing (TCFF) or backward
facing (TCBF) based on the direction of tapering (refer to Fig. 1). Here, the boundary
conditions are shown in Fig. 1, and as the initial condition, the computational domain
is initialised with inlet freestream velocity U∞ . The Reynolds number Re = U∞ D/ν
in this study is selected as 20 and 40.
2.2 Numerical Solver
The governing equations of the present two-dimensional, laminar and incompressible

flow past trapezoidal cylinder are shown in Eqs. (1)–(2). These equations are solved
using a commercial finite volume, Ansys Fluent solver with second-order upwind
discretisation scheme and SIMPLE pressure-velocity coupling algorithm.
∂u i
=0 (1)
∂ xi
∂u i ∂p 1 ∂ 2ui
uj =− + (2)
∂x j ∂ xi Re ∂ x j ∂ x j
For analysis of the aerodynamic forces on the trapezoidal cylinder, the following
non-dimensional parameter is defined.
Fdrag
Cd = 2 D
(3)
0.5ρU∞
Here, Cd is the steady-state drag coefficient. F drag is the steady-state drag force
experienced by the trapezoidal cylinder.
2.3 Validation of the Solver
To validate the Ansys Fluent solver, at first, a square cylinder, i.e. H = Y = D is

simulated at Re = 20 and 40 in the same computational domain as defined in Fig. 1.
Table 1 Mesh independency test

Square cylinder % Error
No. of mesh elements Re = 20 Re = 40 Re = 20 Re = 40
53,917 (M1) 2.1976 1.6512 – –
82,477 (M2) 2.2045 1.6481 0.3139 0.1877
118,237 (M3) 2.2049 1.6474 0.0181 0.0424
Table 2 Validation of C d and

Investigator Blockage L/D(Re = 40) C d (Re = 20)
L/D for square cylinder
Dhiman et al. [1] 0.500 2.822 –
Sen et al. [3] 0.067 2.735 2.2140
Present 0.016 2.776 2.2045
Three different mesh sizes as shown in Table 1 are followed to perform the mesh
independence study. The mesh M2 is followed for the rest of the simulation since
the percentage of error, in this case, is < 0.5%. Here, the convergence criteria are
specified to be below O (10–6 ) for both mass and momentum residuals. In Table
2, drag coefficients obtained from the present simulation are compared with the
literature which is found to be in excellent agreement.
3.1 Flow Regime Description
In the present study, the incompressible laminar flow over the trapezoidal cylinder
encounters a large curl angle at the sharp corners where the flow separates due to the
adverse pressure gradient forming free shear layers. These shear layers curl back and
attach on the rear side of the cylinder creating two steady vortices, also known as the
separation bubbles, as shown in Fig. 2. Note that the point of the flow separation and
curling angle of the shear layers depend upon the shape, size and orientation of the
trapezoidal cylinders. Here, we have classified three different flow regimes based on
the point of the flow separation. The flow regime is called Type 1 and Type 2, when
the flow separates from the leading edge and trailing edge corners of the trapezoidal
cylinder, respectively.
If no flow separation takes place on the surface of the trapezoidal cylinder, then
the flow regime is called Type 3. In Fig. 2, streamlines for different flow types are
shown. In Fig. 3, the flow types observed for different H/D, Y/D and Re are plotted
for both forward and backward facing trapezoidal cylinders.
In the following section, we have determined the variation of the length of the
separation bubble and drag coefficient for different trapezoidal cylinders. Here, the
Fig. 2 Streamlines past TCBF cylinders
Fig. 3 Categorisation of
flow types in TCBF for a Re
= 20 b and Re = 40
Fig. 4 X-velocity plotted

against the centreline to find
the separation bubble length
from the X-intercept
separation bubble length (L/D) is measured as the distance from the centre of the rear
edge of the cylinder to the point along the horizontal centreline where the X-velocity
changes its sign from negative to positive as shown in Fig. 4.
The variation of L/D as functions of H/D, Y/D and Re are shown in Figs. 4 and 5,
for both forward and backward facing trapezoidal cylinders. It is noticed that with the
increase in Re and presence of higher inertia effect, L/D increases for both forward
and backward facing trapezoidal cylinders. Furthermore, when L/D of the forward
facing trapezoidal cylinder is plotted as functions of Y/D for different H/D, it shows
a parabolic type variation for both Re = 20 and 40 and attains maximum when Y/D
is 0.125 or 1.0 (see Fig. 5). A minimum L/D is noticed when Y/D is in the range 0.6
to 0.8, for both Re = 20 and 40, at all H/D. In case of backward facing trapezoid,
the variation of L/D as functions of Y/D and H/D is not unique unlike the forward
facing trapezoidal cylinder cases (see Fig. 6). Note that at Re = 20, for Y/D < 0.375
and H/D ≥ 1.5, there is no separation bubble formation (i.e. L/D = 0).
Fig. 5 Variation of (L/D) of

TCFF with (Y/D) for
different (H/D) at Re = 20
and Re = 40
Fig. 6 Variation of (L/D) of

TCBF with (Y/D) for
and Re = 40
In Figs. 7 and 8, the variation of the drag coefficient (Cd ) as functions of H/D, Y/D
and Re is shown for both forward and backward facing trapezoidal cylinders. Since
the viscous effect is reduced when the Re is changed from 20 to 40, a decrease in
Cd is noticed for both forward and backward facing trapezoidal cylinders. Moreover,
Cd variation for both forward and backward facing trapezoidal cylinders shows an
increasing trend when Y/D is increased (i.e. the trapezoids having higher bluff shapes)
for most of the H/D cases at both Re = 20 and 40.
Fig. 7 Variation of Cd of
TCFF with Y/D for with
different H/D at Re = 20 and
Re = 40
Fig. 8 Variation of Cd of
TCBF with (Y/D) for
and Re = 40
4 Conclusions
In this investigation, the drag coefficient, separation bubble length and wake flow
regime of the trapezoidal cylinder are analysed as functions of the axial length,
taper angle and flow Reynolds number. It is noticed that, in presence of higher
inertia effect, the drag coefficient and separation bubble length of both forward and
backward facing trapezoidal cylinder decreases and increases, respectively, with the
increase in Reynolds number. This finding is in line with the findings of Dhiman et al.
[1] and Sen et al. [3]. Moreover, both drag coefficient and separation bubble length
are strongly affected by the change of the shape and size of the trapezoidal cylinder.
Additionally, we have classified the wake flow regime into three types based on the
location of the flow separation point on the trapezoidal cylinder.
Acknowledgements The first author is thankful to IIT Kharagpur, India and National Chiao Tung
University, Taiwan for providing the necessary support to perform this research work.
References
1. Dhiman A, Amit MH (2013) Flow and heat transfer over a trapezoidal cylinder: steady and
unsteady regimes. Asia-Pac J Chem Eng 3:433–434
2. De AK, Dalal A (2006) Numerical simulation of unconfined flow past a triangular cylinder. Int
J Numer Meth Fluids 52:801–821
3. Sen S, Mittal S, Biswas G (2011) Flow past a square cylinder at low Reynolds numbers. Int J
Numer Meth Fluids 67:1160–1174
4. Lee TS (1998) Early stages of an impulsively started unsteady laminar flow past tapered
trapezoidal cylinders. Int J Numer Methods Fluids 26:1181–1203
5. Sohankar A, Norberg C, Davidson L (1998) Low-Reynolds-number flow around a square

cylinder at incidence: study of blockage, onset of vortex shedding and outlet boundary condition.
Int J Numer Methods Fluids 26:39–56
6. Sharma A, Eswaran V (2004) Heat and fluid flow across a square cylinder in the two-dimensional
laminar flow regime. Numer Heat Transfer, Part A: Appl 45:247–269
7. Chung YJ, Kang SH (2000) Laminar vortex shedding from a trapezoidal cylinder with different
height ratios. Phys Fluids 12:1251–1254
Forced Convection Analysis
in a Horizontal Pipe in the Presence
of Aluminium Metal
Foam—A Numerical Study
Prakash H. Jadhav, Banjara Kotresha, N. Gnanasekaran,

and D. Arumuga Perumal
1 Introduction
Porous metal foam materials are considered by many investigators due to its unique
properties such as light in weight, high strength structure and ability to increase
the instability of the flowing fluid which helps to enhance the heat transfer. Metal
foams are used in several engineering applications such as heat exchanger, electronic
cooling and geothermal. Shokouhmand et al. [1] executed numerical exploration on
the effect of porous insert position which is partly filled in the channel. The porous
medium is attached at different positions of the channel including the core and near
the wall. It was observed that the Darcy number and thermal conductivity were
found to be higher at insert position near the channel wall compared to core position.
Sener et al. [2] performed experiments on rectangular channel with partially filled
aluminium metal foam for a wide range of Reynolds numbers. The metal foam of pore
density 10 and 20 was placed inside the rectangular channel for enhancing thermal
performance. The result shows that as filling rate of metal foam increases the thermal
performance increases and decreases the free flow area in the channel. For particular
pumping power, the metal foam of 10 PPI gives high heat transfer coefficient related
to 20 PPI metal foam. Baragh et al. [3] conducted experimental exploration on forced
convective heat transport in circular pipe with different arrangements of wire mesh
for different flow regimes. The porous media is arranged in six different positions. It
was noticed that the heat transfer rate and heat transfer enhancement ratio are higher
for fully filled wire mesh case compared to other cases for both laminar regime and
turbulent regime of flow. It is also observed that increase in diameter of the porous
medium increases the thermal performance for three different regimes. Li et al. [4]
conclude that the adding of metallic foam increases the thermal performance at the
P. H. Jadhav · N. Gnanasekaran (B) · D. Arumuga Perumal

Department of Mechanical Engineering, NIT Karnataka, Surathkal 575025, India
B. Kotresha
Department of Mechanical Engineering, Government Engineering College, Haveri 581110, India
492 P. H. Jadhav et al.
wall compared to non-porous region. The results also reveal that the 75% filling rate
of metal foam shows better results for 40 PPI metal foam matched to 50% filling
rate of 10 and 20 PPI metal foam. Kotresha and Gnanasekaran [5] numerically
investigated the effect of thickness of copper and aluminium metal foams filled in a
heated vertical channel. The metal foams of 10-mm thickness, 20-mm thickness and
30-mm thickness were considered for the numerical study. The result reveals that
the performance factor was found to be better for 30-mm thickness metal foam than
10- and 20-mm thickness metal foams. It was also noticed that copper metal foam
gives 4% higher rate of heat transport compared to aluminium metal foam. Rabbani
et al. [6] performed experiments considering tubes partly filled with open-cell metal
foam using magnesium oxide (Mgo) nanofluid. The Mgo nanofluid was blended to
the base fluid water and ethylene glycol in volume concentrations of 0.0625, 0.25
and 0.5% to enhance the thermal performance. It was found that the use of nanofluid
increases convection heat transfer by 9 and 15% for both tube profiles compared to
base fluid. Also, it is observed that the porous media has more significant effect on
the heat transfer and pressure fluctuation compared to nanoparticle and found that
the tube II gives 22% higher Nusselt number compared to tube I.
Based on the above literature, many authors have studied partially filled metal
foam in a tube as well as in vertical channels. Due to the advent of high-performance
computing facility, it is possible to perform more numerical studies in order to unveil
the nuances of fluid flow in a metal foam and to perform parametric studies to
enhance heat transfer. Therefore, the aim of the current numerical examination is to
investigate the effect of metal foam fully filled in horizontal pipe for different flow
regimes. The aluminium metal foam of 10 PPI and 0.85 porosity is considered for
the investigation.
2 Computational Domain and Boundary Conditions
The problem domain considered consists of a horizontal pipe of 0.1-m diameter and
7-mm thickness made of aluminium material. The length of the pipe is 1 m, and the
metal foam is fully filled at the centre of the pipe having a length of 0.6 m. Since the
pipe is symmetrical, a two-dimensional computation field is selected which entails
upstream and downstream of the pipe, thickness of the pipe and metal foam-filled
region for further computations. The wall of the pipe is defined with continuous
heat flux of 275 W/m2 , and the flow velocity is varied so as to perform numerical
simulations in laminar, transition and turbulent regions. The inlet to the pipe is defined
with constant velocity inlet, while the outlet of the pipe is defined with zero pressure.
The axis of the pipe is defined with symmetry boundary condition, and part of the
wall which is not exposed to heater is defined with adiabatic wall condition. The
schematic of the computational domain along with boundary conditions is shown in
Fig. 1. Table 1 shows the boundary conditions used for the numerical investigation.
Forced Convection Analysis in a Horizontal Pipe in the Presence … 493
Fig. 1 Computational domain for numerical simulation
Table 1 Boundary
S. no. Boundary conditions
conditions
1 Inlet velocity u = u∞ , v = 0, T = T ∞
∂u ∂v
2 Pressure outlet P = 0, ∂x = 0, ∂x = 0, T = T ∞
3 Adiabatic wall u = 0, v = 0, ∂∂Ty =0
∂T ∂u
4 Symmetry ∂y = 0, ∂y = 0, v = 0
2.1 Numerical Simulation
The numerical investigations are examined using the commercially existing software
ANSYS FLUENT 15.0 [7]. Air is treated as the operative fluid which streams through
metallic foam region as well as foam-free region of the pipe. Properties of air are
computed at inlet temperature of 300 K. The inflow velocities of the operative fluid
range are from 0.17 to 0.95 m/s, and the Reynolds number is evaluated based on the
hydraulic diameter of pipe that varies from 1125 to 6437. The fluid flow assumed
laminar flow regime when Re < 2300, transition flow regime when 2300 < Re < 4500
and turbulent flow regime when Re > 4500.
The governing equations are adopted to solve the k-kl-ω transition, and k-E turbu-
lent models are given in FLUENT [7–9]. The metallic foam region is assumed to be
isotropic homogeneous porous medium and modelled using DEF flow model. In the
momentum equation, the DEF model is considered as source term which comprises
two parts such as inertial and viscous terms which are due to the permeability and
form drag coefficient of the porous medium, respectively. The following governing
equations used for computing the porous region are as follows (Lin et al. [10]).
Continuity equation

∂ ρ f ε ui
= 0. (1)
∂ xi
Momentum equation

∂ ρ f ui u j ∂p ∂ ∂u i ∂u j
= + μeff +
∂x j ∂ xi ∂x j ∂x j ∂ xi
μ
eff
− u i + ρ f C|u|u i (2)
K
Energy equation: LTNE

∂ ρ f Cpf u j T ∂ ∂Tf
i. For fluid ε = αfe + h s f as f Ts − T f (3)

∂ ∂ Ts
ii. For solid αse = h sf asf (Ts − Tf ) (4)
∂x j ∂x j
Grid independence study: The grid independence exercise was performed to fix
the ideal size of the grids for the numerical calculations. To attain this, the numerical
simulations are carried out for the 10 PPI metal foam with uniform heat flux of
275 W/m2 . The grid sensitivity analysis is performed for four dissimilar grids of
47,751, 67,281, 86,811 and 106,341 to optimize the optimum grid size of the mesh.
The variation of temperature on the wall of the pipe for the grid sizes is shown in
Fig. 2. From the results, it was observed that the deviation of 2.72%, 0.83 and 0.23%
in the average wall temperature for the grids of 47,751, 67,281 and 86,281 is obtained
compared to the 106,341 grid size. The grid size of 86,281 is preferred for the further
numerical examination as it shows less deviation compared to other two grid sizes.
3.1 Validation
For the current numerical exploration, the porosity of 0.85 and 10 PPI metallic foam is
considered. The variation of wall temperature found in the present research is matched
with the experimental results of Baragh et al. [3], for a Reynolds number of 3500
is shown in Fig. 3. The numerical results fairly agree with the experimental results,
and the deviation obtained is because of the different porous mediums. Baragh et al.
[3] considered a wire mesh porous structure which is inserted at different locations
of the pipe, but in the present study metal foam is fully filled up to a length of 0.6 m.
Wall temperature: The variation of wall temperature of the pipe for laminar, transi-
tion and turbulent flows along the dimensionless length of the pipe is shown in Fig. 4.
Fig. 2 Wall temperature

variation for different grids
Fig. 3 Validation of
numerical results
From the results, it is noticed that the wall temperature declines with rise in the flow
Reynolds number. The effect of metallic foam on the wall temperature is clearly seen
from the plot. The temperature of the wall decreases in the metal foam-attached pipe
region. For transition flow, the decrease of 19.65% in the average temperature on
the wall of the pipe is obtained compared to laminar and for turbulent flow regimes
Fig. 4 Wall temperature

variation for different fluid
flow regimes
the decrease of 31.5% in the mean temperature on the wall of the pipe is obtained
compared to the laminar flow.
Nusselt number: The variation of local Nusselt number along the wall of the pipe for
different flow regimes is shown in Fig. 5. From the figure, it is noticed that the local
Nusselt number increases with increase in flow Reynolds number at the metal foam
Fig. 5 Local Nusselt

number for different flow
regimes
Fig. 6 Pressure drop for

different inlet velocities
region and then decreases at the non-porous region. This is owing to the presence of
metal foam that acts as an extended surface which increases thermal performance.
Thermal performance increases to 23.54% and 75.19% in case of transition and
turbulent flow, respectively, compared to the laminar flow.
Pressure drop: Metal foam porous material is used to improve the heat transfer rate
in a heat exchanger. Figure 6 shows the variation of pressure drop along the length of
pipe for clear channel and foam-filled channel for different inflow velocities. Due to
increase in inflow velocity, the pressure drop increases for the foam-filled pipe when
compared to the clear pipe due to obstruction in the flow.
4 Conclusions
A two-dimensional numerical examination was done for a pipe filled with metal foam
using ANSYS FLUENT. The metallic foam was assumed to be isotropic homoge-
neous medium, and investigation is made using aluminium metal foam of pore density
of 10 with porosity of 0.85. The flow and thermal features are captured for laminar,
transition and turbulent flows. Initially, the computed numerical results were vali-
dated by comparing with experimental results. The pressure drop increases with
increase in the inflow velocity of the flowing fluid. The wall temperature decreases
in the aluminium metal foam region because of the more conduction and convection
effects. It is observed that heat transfer rate increases to 23.54% and 75.19% in case
of transition and turbulent flow, respectively, compared to the laminar flow at the
expense of pressure loss.
References
1. Shokouhmand H, Jam F, Salimpour MR (2011) The effect of porous insert position on the
enhanced heat transfer in partially filled channels. Int Commun Heat Mass Transf 38(8):1162–
1167
2. Sener M, Yataganbaba A (2016) Forchheimer forced convection in a rectangular channel
partially filled with aluminum foam. Exp Therm Fluid Sci 75:162–172
3. Baragh S, Shokouhmand H, Ajarostaghi SSM, Nikian M (2018) An experimental investigation
on forced convection heat transfer of single-phase flow in a channel with different arrangements
of porous media. Int J Therm Sci 134(August):370–379
4. Li Y, Wang S, Zhao Y (2018) Experimental study on heat transfer enhancement of gas tube
partially filled with metal foam. Exp Therm Fluid Sci 97(April):408–416
5. Kotresha B, Gnanasekaran N (2019) Effect of thickness and thermal conductivity of metal
foams filled in a vertical channel–a numerical study. Int J Numer Methods Heat Fluid Flow
29(1):184–203
6. Rabbani P, Hamzehpour A, Ashjaee M, Naja M, Houshfar E (2019) Experimental investigation
on heat transfer of MgO nano fluid in tubes partially filled with metal foam. Powder Technol
354:734–742
7. Fluent ANSYS (2017) ANSYS V R [ANSYS Fluent], 15.0 help system, user’s guide/theory
guide. ANSYS, Inc., Canonsburg, PA. https://www.ansys.com/Products/Fluids/ANSYS-Fluent
8. Walters DK, Cokljat D (2008) A three-equation eddy-viscosity model for reynolds-averaged
Navier–Stokes simulations of transitional flow. J Fluids Eng 130
9. Ahsan M (2014) Science direct numerical analysis of friction factor for a fully developed
turbulent flow using k e ε turbulence model with enhanced wall treatment. Beni-Suef Univ J
Basic Appl Sci 3(4):269–277
10. Lin W, Xie G, Yuan J, Sundén B (2016) Comparison and analysis of heat transfer in aluminum
foam using local thermal equilibrium or nonequilibrium model. Heat Trans Eng 7632
Mathematical Modeling of the Effects
of CD147/EMMPRIN on Tumor
Angiogenesis
P. Vimala and S. Bhooma
1 Introduction
Angiogenesis, the process of growth of new blood vessels from existing vessels, is
essential for providing the cells with nutrients and oxygen and removal of waste
products from the cells. A tumor can grow only up to 1–2 mm in diameter with the
help of diffusion from the nearby blood vessels after which the tumor relies on tumor
angiogenesis for its growth. Tumor angiogenesis is followed by invasion and metas-
tasis [1–3]. The avascular state is transformed into a dangerous malignant vascular
state (cancer) through tumor angiogenesis, which leads to the further growth of the
tumor. Tumor angiogenesis starts with the secretion of tumor angiogenesis factor
(TAF) into the surrounding tissues. Among TAFs, VEGF plays a major role in tumor
angiogenesis. The response of ECs to these TAFs is termed chemotaxis. The migra-
tion of ECs takes place due to chemotaxis, giving rise to cluster of cells, which then
forms sprouts. These sprouts then move to form loops and pierce the tumor, thus
forming a tumor vasculature [4–6]. The tumor microenvironment consists of the
ECM which contains macromolecules such as collagen, fibronectin, and vitronectin.
During tumor angiogenesis, the ECs interact with the ECM, and this effect is exam-
ined via fibronectin macromolecule. It helps in the adhesion of cells to the ECM.
The migration of ECs is due to the gradient created by the bound fibronectin. This
response of ECs to non-diffusible substances is termed as haptotaxis [7].
Another important transmembrane glycoprotein is cluster of differentiation 147
(CD147), which belongs to the immunoglobulin (Ig) family of receptors. It is present
on all cell surfaces, but its levels are quite high on tumor surfaces, and this high
expression relates to poor prognosis in cancer patients. Its main functions are cell
P. Vimala (B) · S. Bhooma

Department of Mathematics, Anna University, Chennai, India
500 P. Vimala and S. Bhooma
adhesion, response to inflammation, and tumor progression. It helps in tumor angio-

genesis by various methods—by inducing VEGF and increasing VEGF expres-
sion (which is done by activating the vascular endothelial growth factor receptor-
2 (VEGFR-2)), by the secretion of MMPs and by EC proliferation and migration.
Tumor cells also shed fragments of cleaved CD147 through vesicles, and these frag-
ments are found to interact with fibroblasts to induce matrix metalloproteinases-2
(MMP2). The cleaved fragments are also released independently of membrane vesi-
cles [8–10]. Many mathematical models related to tumor angiogenesis are available
in the literature. In [11], the authors have considered modeling tumor angiogenesis in
the absence of EC proliferation. They have constructed a one-dimensional conserva-
tion equation for EC migration with diffusion, chemotaxis, and haptotaxis terms and
have applied no-flux boundary conditions. The model proposed in [11] was extended
to a two-dimensional model in [12], where the authors have considered both contin-
uous and discrete models. In [12], the authors have shown using continuum model
how the ECs migrate in response to the gradients of TAF and fibronectin. The discrete
model is used to explain the process of branching (formation of sprout) and anas-
tomoses (loop formation). They have shown that for the initiation of a sprout, the
response to TAF gradient has to be strong.
In relevance to these models, this paper presents a one-dimensional model incor-
porating the concentration of CD147 in the continuum model proposed in [12]. The
experimental results on the role of CD147 in tumor angiogenesis are verified using
the model presented in the paper. Numerical simulations are carried out, and various
parameter values are borrowed from previous models. The simulations are compared
with the results from in vivo experiments.
2 Mathematical Model
The model presented in this paper is similar to the continuum model given in [12]. The
EC density per unit area is denoted by n ∗ , the TAF concentration by c∗ , the fibronectin
concentration by f ∗ , and the CD147 concentration by b∗ . The corresponding non-
dimensional quantities are denoted by n, c, f , and b. The ECs respond to CD147 in
chemotactic manner and haptotactic manner. This is because of the many functional
aspects of CD147 [9]. Hence, the conservation equation for EC density is same as
in [12], taken in one-dimensional form. The CD147 increases and induces VEGF
(via VEGFR-2), which is incorporated in the model. The interaction of CD147 with
fibronectin is modeled as an uptake term. CD147 is a protein present on the surface
of the cells, and they interact with ECs. It is also shed by the tumor cells in the form
of fragments of cleaved CD147 [10]. These are also included in the model. Hence,
the non-dimensional system of equations is given by
Mathematical Modeling of the Effects of CD147/EMMPRIN … 501
⎫
∂n
= D ∂∂ xn2 − ∂ χ
n ∂c ∂
ρn ∂∂ xf ⎪
2
− ⎪
∂t ∂x 1+αc ∂ x ∂x ⎪
⎬
∂c
∂t
= −ηnc + φb, (1)
∂f
= βn − γ n f − ξ b f ⎪
⎪
∂t ⎪
⎭
∂b
= Db ∂∂ xb2 − δb − nb
2
∂t
These equations are assumed to be valid in an interval of length 1, and no-flux

boundary conditions are imposed on the interval, as in [12]. The boundary conditions
(BCs) are, therefore, given by

−D ∂∂nx
x=0,1 = 0, n ∂∂cx
x=0,1 = 0, n ∂∂ xf
= 0, −Db ∂∂bx
x=0,1 = 0 (2)
x=0,1
where the non-dimensional quantities are given by
Dn τ χ0 c0 τ ρ0 f 0 τ
D= 2
,χ = 2
,ρ = ,
L L L2
c0 b0
α = ∗ , η = η∗ τ n 0 , φ = φ ∗ τ ,
k c0
n0 ∗
β = ωτ , γ = μτ n 0 , ξ = ξ τ b0 ,
f0
∗ τ
Db = Db 2 , δ = δ ∗ τ, = ∗ τ n 0 .
L
The non-dimensionalization was carried out using appropriate reference variables
for x ∗ , t ∗ , n ∗ , c∗ , f ∗ , and b∗ as L, τ, n0 , c0 , f 0 , and b0 , respectively, taking
x∗ t∗ n∗ c∗ f∗ b∗
x= ,t = ,n = ,c = , f = ,b = .
L τ n0 c0 f0 b0
The initial conditions (ICs) for the above system are taken to be similar to that in
2 2
− ∫x − (1−x)
n(x, 0) = e 2 , c(x, 0) = e ∫ 1 ,
[12], given by 2
.
− (1−x)
2
− x∫
f (x, 0) = k1 e , b(x, 0) = k2 e ∫ 3 , x ∈ [0, 1]
where k1 = 0.75, ε = 0.45, ε1 = 0.45, ε2 = 0.001 [12], k2 = 0.8 [13], and ε3 = 0.45
[present work].
3 Simulation and Results
Discretization of Eq. (1) with their corresponding initial and boundary conditions
(BCs (2)) was carried out by using finite difference scheme. Many of the parameter
values have been borrowed from the literature and previous experimental results.
The values used in the simulations (unless otherwise specified) are given below.
L = 2 mm, τ = 1.5 days, D = 0.00035, χ = 0.38, ρ = 0.34, α = 0.6, η = 0.1, β

= 0.05, γ = 0.1 [12], Db = 0.0084138 [12], δ = 0.005, φ = 0.025, ξ = 0.1, =
0.1 [present work].
A graph of the initial concentrations was plotted and was found to coincide with the
graph of Anderson and Chaplain [12] for t = 0. Using the numerical finite difference
scheme, four different time plots (Fig. 1) for the concentrations are obtained. As the
time progresses, the plots show that there is a considerable reduction in the concen-
trations of CD147 and fibronectin. However, as CD147 is considered to increase the
level of VEGF as suggested in [9] and as this is incorporated in the model, the TAF
concentration is higher, and as c increases, the migration of ECs increases; that is, the
ECs move faster toward the tumor. Comparison of plots with that of [12] is carried
out in each figure. It is observed that the ECs have reached 0.8 along the x-direction
at 15 days in [12]. However, in the present work, it is seen that the ECs have reached
0.9 along x-direction at 15 days (t = 10). However, as t is increased further (Fig. 2),
it is seen that the ECs have moved much closer to the tumor to approximately x =
0.98.
Graphs showing the effects for different sets of parameter values were also
obtained. It was seen that for varying values of , the change in f was negligible.
However, there was a slight variation in c. As increased, b reduced slightly, and
this reduced c after x = 0.9 to a negligible extent which in turn slowed down the
migration of ECs very slightly.
For different values of ξ , the effects are shown in Fig. 3. It is seen that as ξ
increases, f decreases drastically, and this has an effect on the ECs. The ECs are seen
to be moving faster toward the tumor with the reduction in f . It is also seen that the
changes in c and b are negligible.
The graph for the effects of different values of φ is shown in Fig. 4. As CD147
produces and upregulates TAF [9] and as this is incorporated, the graph for increasing
values of φ obviously shows an increase in c to a significant level which in turn
increases the motion of ECs toward the tumor.
The effects of different values for δ are plotted in Fig. 5. As δ increases, b decreases
which reduces c, and this makes the ECs to move slower. As δ is the decay rate of
CD147, this is to be expected. There is also a negligible change in f up to x = 0.1
and after x = 0.9.
In case of both and δ, it is seen that as b reduces c also reduces. This is
in accordance with the experimental result [14] which states that “knockdown of
CD147 resulted in the down regulation of VEGF expression.”
As the model considered in this paper is based on the continuum model of [12],
it is seen that as in [12], the ECs do not reach the tumor, and in [12], the ECs have
“attained quasi-steady state at approximately x = 0.9”. In this model, however, the
quasi-steady state is reached at approximately x = 0.98 (Fig. 2).
Fig. 1 Concentrations of CD147 b‚ ECs n, TAF c and fibronectin f at times t = 2, t = 4, t = 7

and t = 10
Fig. 2 Concentrations of CD147 b, ECs n, TAF c and fibronectin f at times t = 14 and t = 18
Fig. 3 Effects of ξ on the concentrations of CD147 b, ECs n, TAF c and fibronectin f at time t =
10
Fig. 4 Effects of φ on the concentrations of CD147 b‚ ECs n, TAF c and fibronectin f at time t =
10
Fig. 5 Effects of δ on the concentrations of CD147 b‚ ECs n, TAF c and fibronectin f at time t =
10
4 Discussion and Conclusion
The simulation results show that the concentration of TAFs is significantly increased
by CD147, a result incorporated in the model following [9]. Also, it reduces the
concentration of fibronectin as compared to [12]. The increase in TAF concentration
and reduction in the fibronectin concentration has led to faster migration of ECs
toward the tumor under the action of CD147, as compared to [12], initiating faster
angiogenesis. The variation in different parameter values was carried out, and it
was seen that the reduction in CD147 expression resulted in the decrease of TAF
concentration in par with the result proposed in [14]. The quasi-steady state for the
ECs is at x = 0.98, showing that the ECs have moved closer to the tumor at a faster
rate compared to [12] due to the effect of CD147.
Extending this model for two-dimensions can make it to comply with realistic
experimental results. Further, anti-angiogenesis strategies can also be investigated.
References
1. Eichhorn ME, Kleespies A, Angele MK, Jauch K-W, Bruns CJ (2007) Angiogenesis in cancer:
molecular mechanisms, clinical impact. Langenbeck’s Arch Surg 392:371–379
2. Folkman J (1971) Tumor angiogenesis: therapeutic implications. N Engl J Med 285(21):1182–
1186
3. Folkman J, Kalluri R (2003) Tumor angiogenesis (Chap. 11). In: Kufe DW, Pollock RE,
Weichselbaum RR et al. (eds) Holland-Frei, Cancer medicine, 6th edn. Hamilton (ON), BC
Decker
4. Kerbel RS (2000) Tumor angiogenesis: past, present and the near future. Carcinogenesis
21(3):505–515. https://doi.org/10.1093/carcin/21.3.505
5. Roussos ET, Condeelis JS, Patsialou A (2011) Chemotaxis in cancer. Nat Rev Cancer
11(8):573–587
6. Paweletz N, Knierim M (1989) Tumor-related angiogenesis. Crit Rev Oncol Hematol 9(3):197–
242
7. Lamalice L, Le Boeuf F, Huot J (2007) Endothelial cell migration during angiogenesis. Circ
Res 100(6):782–794
8. Xiong L, Edwards CK III, Zhou L (2014) The biological function and clinical utilization of
CD147 in human diseases: a review of the current scientific literature. Int J Mol Sci 15:17411–
17441
9. Wang Z, Zhao Z, Jiang T, Chen Y, Huang C (2016) The multistep functions of
EMMPRIN/CD147 in the tumor angiogenesis, review article. Cell Mol Med Open Access
2(1):12
10. Hatanaka M, Higashi Y, Fukushige T, Baba N, Kawai K, Hashiguchi T, Juan Su, Zeng W, Chen
X, Kanekura T (2014) Cleaved CD147 shed from the surface of malignant melanoma cells
activates MMP2 produced by fibroblasts. Anticancer Res 34:7091–7096
11. Anderson ARA, Chaplain MAJ (1997) A mathematical model for capillary network formation
in the absence of endothelial cell proliferation. Appl Math Lett 11(3):109–114
12. Anderson ARA. Chaplain MAJ (1998) Continuous and discrete mathematical models of tumor-
induced angiogenesis. Bull Math Biol 60:857–900
13. Weidle UH, Scheuer W, Eggle D, Klostermann S, Stockinger H (2010) Cancer-related issues
of CD147, review. Cancer Genom Proteomics 7:157–170
14. Chen Y, Gou X, Ke X, Cui H, Chen Z (2012) Human tumor cells induce angiogenesis through
positive feedback between CD147 and insulin-like growth factor-I. PLoS ONE 7(7):e40965
Experimental Investigation on the Effects
of Canalization for Ignition Over
Pressure Measurement
V. Venkata Ramakrishna, S. Sankaran, and R. Senthil Kumar
1 Introduction
Ignition over pressure is an unsteady phenomena associated with sudden abrupt

expansion of gases from the combustion chamber through the nozzle exit during
the ignition transient of a rocket motor. It has a compression front with subsequent
rarefaction and propagates through the launch vehicle support structure holes and jet
deflector on the launch pad. These high-amplitude unsteady fluid-dynamic perturba-
tions can adversely affect the vehicle and surrounding launchpad structures. There
were several over pressure related issues for the launch of the first space shuttle
flight, Space Transportation System 1 (STS-1). Accelerations measured at the wing,
body flap, vertical tail, and crew cabin exceeded the 3σ lift-off design environment
[1]. The basic mechanism of IOP generation and the factors governing the IOP peak
pressure were discussed by counter [2]. The chamber pressure rise rate, chamber
pressure, chamber pressure peak, and chamber pressure impulse were investigated
to determine the influence of each on IOP production. Study results indicate that
IOP for the 1% solid rocket motor of space shuttle is basically controlled by the
chamber pressure peak and resultant steady-state pressure with lesser dependence
on the chamber pressure rise rate. As it is very difficult to simulate the full scale
rocket motor ballistics during the scaled model IOP experiments, Broadwell and Tsu
[3] proposed a transient pressure scaling model for IOP amplitude and frequency
scaling between different scales of rocket motors.
V. Venkata Ramakrishna (B) · S. Sankaran · R. Senthil Kumar

SMP&ETF, SDSC SHAR, Sriharikota, India
e-mail: vvrk@shar.gov.in
S. Sankaran
R. Senthil Kumar
e-mail: senthil@shar.gov.in
508 V. Venkata Ramakrishna et al.

Px Py
Px+ + = Py+
Dx Ṗx D y Ṗy
where x and y refers to two different scales

Ṗ Chamber pressure rise rate
P Steady-state chamber pressure
P+ Over pressure peak amplitude
D Effective duct diameter at over pressure source.
As the region between nozzle exit plane and jet deflector duct contains the source
of IOP, very high-amplitude pressure disturbances emanate from here [4]. This source
over pressure travels upwards as IOP and can induce severe pressure loading on the
base of rocket due to the confinement of rocket nozzles on support structure on the
pad. It can adversely affect the thermal protection system of rocket base shrouds and
may damage the stage components. This paper tries to study the correct methodology
of measuring the peak pressure in these confinement zones through the canalization
of sensors. In open literature, no such effort is made to measure the shock wave
amplitude using canalization. Hence, an experimental work is initiated on this aspect
using cold and scaled hot flow experiments. The shock wave amplitude correction
factors from the scaled experiments are applied to the full scale rocket measurements.
2 Methodology
To evaluate the effects of using canalization in IOP measurements, it is assessed

through cold flow and sub-scale hot flow experiments.
2.1 Cold Flow Experiments
In the case of cold flow experiments, a shock tube is used to create an over pressure
pulse. Shock tube consists of two sections namely driving and driven and are sepa-
rated by a diaphragm. The driving section is pressurized using gaseous nitrogen till
the rupture of diaphragm. The shock tube set-up used for the experiments is shown
in Fig. 1.
Driving section (chamber 1) of the shock tube is fitted with a 30 bar. Driving section
(chamber 1) of the shock tube is fitted with a 30 barA range pressure transducer to
record the chamber pressure history. Prior to canalization experiments, direct field
measurements were performed to ascertain that the over pressure pulse is produced
at the rupture of diaphragm. Also, to understand its propagation characteristics, six
PCB make microphones, model 377A14, starting from 5De and at an interval of 5De
Experimental Investigation on the Effects of Canalization … 509
CD Mylar Driving
nozzle diaphragm section
Fig. 1 Shock tube set-up for cold flow over pressure studies
were placed at 30° from the nozzle axis as shown in Fig. 2. The CD nozzle used for
these experiments has an exit diameter (De) of 30 mm.
After ensuring the generation of over pressure pulse and its spatial decay charac-
teristics, canalization measurements were performed with the same set-up. For these
experiments, less expensive Druck make pressure transducers, model no PDCR 4000,
with response time of 0.2 ms were used. These sensors are connected to a tube of
300 mm length (corresponds to 10De) and were positioned in the horizontal plane
of nozzle with tube inlet placed at 5De from nozzle exit. Hence, effectively the
sensor is placed at a distance of 15De from nozzle exit. The experimental set-up with
canalization is shown in Fig. 3.
To get the comparison of the measured over pressure magnitude using canalized
sensor, a direct measurement is taken using Kulite make unsteady pressure sensor,
model no XCE-062, at 15De from the nozzle exit at 30° from the nozzle axis. To
Fig. 2 Schematic of microphone locations

Druck sensor
Kulite
sensor
Tube inlet
15 De
5De
Nozzle
exit plane
Fig. 3 Canalization measurements using shock tube
determine the effect of varying tube inner diameter on the measured over pressure
magnitude, three different sizes of tubes, viz. 4, 6, and 8 mm were used. In the
experimental set-up as shown in Fig. 3, two sizes of tubes (4 and 6 mm) were used
for each test run. Tube with 6 mm diameter is used as reference for obtaining the
measurements of over pressure with 8 mm tube. Similar type of diaphragm is used
for all the experiments to ensure the same pressure drop rate while rupturing.
2.2 Hot Flow Experiments with Solid Rocket Motor
Canalization experiments for IOP measurements were performed during static firing
of rocket motor to validate the findings from the cold flow experiments. A 6 mm tube
with Druck make pressure transducer is used for the canalization measurements and
Kulite make unsteady pressure transducer is used for the measurements similar to
the cold flow experiments. Hot flow experiment set-up is shown in Fig. 4.
Direct field measurements during cold flow experiments (ref Fig. 2) are indicating
the formation of over pressure at the diaphragm rupture and the magnitude of over
pressure is decaying with distance as per the inverse law. Figure 5 shows the over
pressure time histories starting from a distance of 5De till 30De from the nozzle exit.
Fig. 4 Rocket motor static firing set-up and canalized sensors fixing on stand
Fig. 5 Over pressure time history with distance from nozzle exit
From the above figure, it is observed that the magnitude of over pressure is
having linear inverse dependency with distance from the nozzle exit in the measured
orientation of 30°.
Subsequently, experiments were performed to understand the canalization effect
on the measurement of over pressure. Three tubes of ID 4, 6, and 8 mm were used
with the Druck make pressure transducer and Kulite make unsteady pressure sensor.
Fig. 6 Over pressure time history with canalized Druck make pressure transducers at 15De
Figure 6 shows the measured time history with canalized Druck sensor. The canalized
over pressure measurements were compared with that of direct measurement.
From the above figure, it is seen that the over pressure has attenuated in the
4 mm diameter tube, amplified in 8 mm tube, whereas, 6 mm tube has given almost
same value of over pressure as compared to the direct measurement. The rise time
of over pressure peak is of the order of 40 µs in the case of direct measurement
compared to the 100 µs as seen in the canalized measurements. It indicates that the
shock front thickness is increased due to the canalization. This may be due to the
continuous reflection of shock front at tube boundaries and subsequent coalescence
of the reflected shock pulse with the main shock front.
To assess the behaviour of the over pressure pulse on the sensor being canalization,
similar experiments were repeated by canalizing 1/16 Kulite sensor. Figure 7 shows
the measured time history with canalized Kulite sensor. It is observed that the over
pressure measurements with all tubes are showing amplification when measured
with Kulite sensor. It indicates a strong dependency of the type of sensor being
used in the canalization on the measured value of over pressure. Table 1 gives the
percentage variation of over pressure magnitude with canalization compared to the
direct measurement.
The results obtained from the cold flow shock tube experiments are further vali-
dated in the scaled down rocket motor static firing. A 6 mm ID tube of 800 mm
length is considered for canalization measurements and the results are compared
with direct measurements. The tube inlet is positioned at 20De from the nozzle
exit (De is 144 mm) and 300 from the nozzle axis. Kulite pressure sensor is placed
Fig. 7 Over pressure time history with canalized Kulite make pressure transducers at 15De
Table1 Comparison of canalized and direct over pressure measurements with Druck and Kulite
sensors
Sensor % variation compared to direct measurement
8 mm tube 6 mm tube 4 mm tube
Druck sensor (model: PDCR 4000) + 172.8 −1.3 −35.9
Kulite (model: XCE-062) + 182.8 + 157.1 –
spatially at a distance equal to that of Druck sensor from the nozzle exit. Figure 8
shows the IOP measurements during static firing of scaled rocket motor.
From the above figure, it is observed that during static firing of solid rocket motor,
the IOP magnitude measured using 6 mm tube canalization with Druck sensor is
matching closely with that of the direct measurement. Rise time of the IOP pulse is
seen as 1.6 ms for the direct measurement and 4.4 ms for the canalized measurement,
supporting the observation made in the cold flow experiment that the canalization is
increasing the shock front thickness. Also, even with the variation of over pressure
rise time from 40 µs in cold flow to 1.6 ms in the hot flow experiment, the canalization
measurement of IOP magnitude is same. It reveals that the canalization measurements
are independent of the rise time of over pressure pulse, in turn on the size of the rocket
motor.
Fig. 8 Comparison of canalized and direct IOP measurements
4 Conclusions
Experimental study on measurement of IOP with canalization has resulted in the

following conclusions.
1. Over pressure measurement with canalization has strong dependency on the type
of sensor being used. From the cold flow shock tube experiments, it is observed
that canalization of Druck pressure transducers with 6 mm ID tube is giving the
accurate measurement of over pressure peak. Same tube when canalized with
Kulite pressure transducer, over pressure peak amplified by ~150%.
2. Size of the tube used for canalization is significantly influencing the measure-
ment of over pressure peak. Three tubes of 4, 6, and 8 mm ID are used in the
present experiments. It is observed that 4 mm ID tube is attenuating and 8 mm ID
tube is amplifying the over pressure peak, whereas, 6 mm ID tube is measuring
the peak pressure accurately.
3. Cold flow experimental finding on effectiveness of Druck sensor canalization
with 6 mm tube for over pressure measurement is validated using the hot flow
experiments.
4. Variation in rise time of the over pressure pulse is not affecting the characteristics
of canalization measurements; hence, similar measurement philosophy can be
applied for the IOP measurements for any size of rocket motor.
Acknowledgements Authors acknowledge the support provided by the E&IS, SMP&ETF team
for instrumentation and data acquisition. Also, thank SDSC SHAR management for providing the
opportunity to carry out the experiments.
References
1. Casiano MJ, Alvord DA, McDaniels DM (2011) Ares I scale model acoustic test over pressure
results. NASA/TM-217452, MSFC, NASA
2. Counter DD, Ignition over pressure study from solid rocket motor firings. NASA TM-86587,
Marshall Space Flight Center
3. Broadwell J, Tsu C (1967) Transient pressures caused by rocket start and shutdown in ducted
launchers. J Spacecraft Rockets 4(10):1323–1328
4. Nance DK, Liever PA (2015) Space launch system scale model acoustic test ignition over
pressure testing. In: 29th aerospace testing seminar, MSFC, NASA
Numerical Investigation of Rotating
Convection in a New Configuration
with Bidirectional Thermal Gradients
Shivam Swarnakar, Ayan Kumar Banerjee, Amitabh Bhattacharya,
and Sridhar Balasubramanian
1 Introduction
Atmospheric general circulation being an example of rotating convection, is instru-

mental in understanding large-scale synoptic weather pattern [1–3]. However, it is
difficult to study it in the real atmospheric situations due to the fact that in real
atmospheric situation, different atmospheric parameters (e.g. temperature, pressure,
density, etc.) could vary randomly and hence cannot be studied in a controlled manner.
But these large-scale atmospheric flow phenomena can be studied numerically and
in laboratory conditions controlling the governing physical parameters. The differ-
entially heated rotating cylindrical annulus of fluid, generally known as baroclinic
annulus experiment, is accepted as an exquisite experimental set-up for studying
baroclinic instability mechanism, which is responsible for heat and momentum trans-
port in mid-latitude. However, mutual feedback and interaction between the fully
developed baroclinic instabilities and background stratification cannot be studied
in baroclinic annulus configuration due to intense boundary layer dominated over-
turning circulation. Hence, a new system is proposed consisting of an inhomoge-
neously heated fluid annulus where the inner wall is maintained at a constant colder
temperature and the periphery is heated on the bottom using a “thin” aluminium strip.
Therefore, this new configuration is subjected to thermal gradients both in radial and
vertical directions. This system is expected to provide a better understanding of the
heat transport from the strong convection near the tropics to the stably stratified
mid-latitude [1–3].
S. Swarnakar (B) · A. K. Banerjee · A. Bhattacharya · S. Balasubramanian

Department of Mechanical Engineering, IIT Bombay, Mumbai, India
518 S. Swarnakar et al.
2 Methodology
In this work, OpenFOAM is used to simulate the problem. The problem is initially
considered to be laminar (Re = 40−400), so no turbulence model is applied. A modi-
fied buoyancy-based heat transfer solver is developed and used. A 2D axisymmetric
vertical plane (Fig. 1) is considered for simulation.
Navier–Stokes equations in non-inertial frame of reference are solved in this work.

The additional terms which come in solver code because of non-inertial frame of
reference are: Coriolis force (2 × r ) and centrifugal force ( × × r ). The solver
uses PISO algorithm which stands for pressure implicit splitting of operators. In
this algorithm, one predictor loop and two corrector loops are present to predict the
velocity and correct the pressure (two times), respectively.
For meshing, a structured mesh is used with 338 × 338 nodes in r and z direction,
respectively. The mesh has uniform spacing of 3.55 × 10−4 in both r-direction and z-
direction, which is sufficient to this problem and grid independence study is carried
on. Time step is kept as adjustable to run time and to ensure the stability of the
simulation Courant–Friedrichs–Lewy (CFL) condition is used, which ensures that
in a single time step fluid will pass through a single control volume. So, maximum
Courant number was set to be 0.5.
Fig. 1 Schematic of the configuration and the vertical plane for simulation
Numerical Investigation of Rotating Convection … 519
(a) (b)
(c)
Fig. 2 Contour of axial velocity at varying rotations
3.1 Axial Velocity Contour
Figure 2 shows axial velocity contour. In most of the cases, buoyancy-driven upward
flow (convective plume) is found to have existed near-outer periphery and sinking
downward motion exists near-cold inner wall. Bulk flow field in between inner and
outer wall has near-zero velocity.
3.2 Radial Velocity Contour
Figure 3 shows radial velocity contour. In general, radial velocities exist only near
boundary (towards inner wall on bottom and towards outer wall on top) for rotating
(a) (b)
(c)
Fig. 3 Contour of radial velocity at varying rotations
cases. Bulk flow field in between inner and outer wall has near-zero relative radial
velocity. Therefore, the iso-contour plots of relative velocities (radial and axial) depict
that in the absence of baroclinic instability, convection occurs mostly within boundary
layers. The convective patterns are characterized by upward moving (downward
moving) relative axial velocity in vertical direction over the heating strip (on the
cold side) being confined to a narrow boundary layer. Similarly the relative radial
velocities are also confined within Ekman boundary layers (top and bottom). The
fluid interior remains nearly motionless.
3.3 Temperature Contour
Figure 4 shows temperature contour. It can be seen that impact of heating decreases
with elevation, which can be better understood from T versus z plot as shown in Fig. 5.
(a) (b)
(c)
Fig. 4 Temperature contour at varying rotations
At zero rotation rate, isotherms are horizontal. Rotation leads to the spreading of the
temperature field due to a combination of quasi-hydrostatic and quasi-geostrophic
balance in the interior of the flow domain.
Figure 5 shows the variation of temperature along the z-axis at r = 0.118 m from
the inner wall. The temperature profiles indicate an unstable fluid layer up to z = 1–
2 cm from the bottom, followed by a uniform temperature profile. The manifestation
of this temperature profile is a result of the localized heating provided at the bottom,
which allows for the evolution of a background stratification and associated thermal
instabilities.
Fig. 5 Temperature
variations along z
4 Conclusions
Based on the contours presented along with T versus z plot and also based on exper-
imental results [1–3], we present a schematic of the flow (Fig. 6), wherein a rising
convective plume is generated over the heating strip (Z − I), and a sinking thermal
boundary layer is present on the cold side (Z − II). Both the radial and vertical temper-
ature gradients reduce with height, and the domain within the bulk fluid represents
Fig. 6 Schematic of the 3D

flow dynamics on a r − z
plane. [2] Reproduced from
“Banerjee et al.; licensed
under a Creative Commons
Attribution (CC BY) license”
the thermal wind induced by the combination of radial temperature gradient and
Coriolis force.
References
1. Banerjee AK, Tirodkar S, Bhattacharya A, Balasubramanian S (2016) Convection in rotating

flows with simultaneous imposition of radial and vertical temperature gradients. arXiv preprint
arXiv:1611.00807
2. Banerjee AK, Bhattacharya A, Balasubramanian S (2018) "Experimental study of rotating
convection in the presence of bi-directional thermal gradients with localized forcing". AIP adv
8(11)
3. Banerjee AK, Balasubramanian S, Bhattacharya A (2020) "Investigation of heat transfer char-
acteristics in a rotating convection system with bi-directional thermal gradients ". J Heat Transf.
doi: https://doi.org/10.1115/1.4048825
Thermal Hydraulic Investigations
on Fast Reactor Fuel Subassembly Under
Blocked Condition—Using Porous Body
Approach
S. Srihari, M . Naveen Raj, and K . Velusamy
1 Introduction
There is a tremendous enthusiasm for fast breeder reactor (FBR) innovation because
of its enormous breeding capacity, high burn up and potential to transmute heavy
isotopes. In India, FBR remains as a workhorse in the second phase of the atomic
power program. A multitude of experimental and computational investigations have
been carried out in the reactor core, particularly, on fuel pin bundle. Most of them
deal with the CFD analysis of helical wire wrapped 217 pin bundle. Naveen Raj and
Velusamy [1] explore the transverse and axial flow distributions in different types
of subchannels. The studies show that the axial mass flow rate in any subchannel
oscillates periodically in the fully developed region of the bundle. Naveen Raj et al. [2]
also explore the effect of porous blockage in the flow and temperature characteristics
in a fuel pin bundle. They found that a large re-circulating wake zone exists in the
downstream side of the porous blockage and it is proportional to the radius of the
blockage. In addition, they also observed that the peak clad temperature is increased
in cases of low porosity of blockage, thereby making temperature a strong function
of porosity. Di Piazza et al. [3] performed various CFD studies on the pin bundle
with planar blockage and observed that the clad temperature exceeds 1000 °C if
the blockage occupies more than 30% of flow area. Though CFD studies give very
detailed prediction, it is computationally expensive and time consuming. To perform
large amount of parametric studies, simplified models based on the results obtained
from CFD must be developed. Gajapathy et al. [6] studied the characteristics of flow
S. Srihari (B)
Department of Chemical Engineering, NIT Trichy, Trichy, India
M. . Naveen Raj
FRTG/IGCAR, Kalpakkam, India
K. . Velusamy
NSAG/RDG/IGCAR, Kalpakkam, India
e-mail: kvelu@igcar.gov.in
526 S. Srihari et al.
mixing among the peripheral and central zones in 7, 19 and 37 pin bundles. Then, the
characteristics were extended to a 217 pin bundle. Govinda Rasu et al. [4] investigated
the development of cross-stream velocity as a function of helical pitch length and
number of pins. They found the existence of periodic spatial oscillations in friction
factor and Nusselt number. Recently, Gajapathy et al. [7] conducted a CFD-based
computation on a 217 pin bundle and analyzed the effect of helical pitch on the flow
and temperature distributions and the variation in friction factor and Nusselt number
as a function of helical wire parameters.
In the current study, a robust porous body model is developed to understand
the thermal hydraulics within the normal and partially blocked sodium cooled fuel
subassembly. The results of this three-dimensional model are compared against that
of CFD results reported in open literature. This model employs sink terms (Blasius
correlation) in the momentum equations for flow over (axial and crossflow) fuel pin
bundles. Using this model, low by-pass toward the peripheral channels, local sodium
temperature rise in regions occupied by porous blockage and radial temperature
non-uniformity have been predicted.
2 Methodology
For better elucidation, validations for each case are discussed under the results and
discussion section.
In the present study, numerical analysis is performed to understand the flow and
temperature fields within the pin bundle under unblocked and blocked conditions.
Thermal hydraulics of sodium in the main vessel cooling system is governed by the
conservation equations of mass, momentum and energy. The Blasius equation with
the modified multiplication factor is also used [5].
∇v = 0 (1)
ρv.∇(v) = μ∇ 2 v − ∇ p + ρg + S (2)
ρv.∇(T ) = (k + kt )∇ 2 (T ) (3)
f = 0.342 × Re−0.25 (4)

Thermal Hydraulic Investigations on Fast Reactor Fuel … 527
where S includes Sx, Sy and Sz, which are the source terms for x, y and z directions,
respectively. Though the Blasius correlation captures the effect of pressured drop in
a channel, the effect of helical wire cannot be captured properly. To accommodate
this, the coefficient (0.316) in Blasius correlation is changed to 0.342. This modified
correlation is used in the axial source term.
In our case, inlet boundary condition is velocity inlet. The sectors inlet face is
taken as the velocity inlet. Outlet boundary condition is pressure outlet, which is
assigned to the exit face of the subassembly. The inlet temperature is given as 670 K.
We also have a solid material in order to account for the density of the fuel pins. The
density of the fuel rods is given as 2719 kg/m3 .
3.1 Unblocked Case
For the unblocked case, the validation of results is done with already published
data [1]. The pressure drop and friction factor values are considered as a validation
parameter. The pressure drop as a function of mass flow rate predicted in the porous
body model is compared with that obtained from CFD analysis of wire wrapped 217
pin bundle [1]. These results are depicted in Fig. 1. It is seen that the agreement
between the porous body and CFD data is good in the low flow rate regime. Further,
even for the large flow rate of 35 kg/s, the deviation is very small and therefore is
considered satisfactory
The fully developed friction factor as a function of Reynolds number predicted in
the porous body model is compared with that obtained from CFD analysis [1]. These
results are depicted in Fig. 2. It is seen that for Reynolds number above 50,000 the
comparison is very close.
The plots of average axial velocity distribution and bulk average temperature
distribution from center to periphery of the subassembly (see inset in Figs. 3 and
4) obtained from porous body and CFD analysis are compared to understand the
relation between velocity field and temperature in Figs. 3 and 4. These distributions
are taken at an elevation of 700 mm from inlet.
Fig. 1 Comparison of
Pressure Drop (ΔP)
predicted pressure drop in 25000

0
porous body model against
CFD results [1] 20000
0 Present
Data
15000
0 Publishe
d Data
0
0 2 4 6
Mass Flow Rate (Kg/s)
Fig. 2 Comparison of 0.04

computational results with
published friction factor 0.03
correlations for 5 Present
Friction Factor
wire-wrapped pin bundles 0.03
[1] 0.02
5
0.02 Previous
0.01
5
0.01
0 50000 100000 150000
Fig. 3 Average axial velocity in various distances from the periphery to the center for a height of
700 mm
Fig. 4 Average sodium temperature in various positions in the sector for a height of 700 mm
It is seen that the average temperature is higher in the central zone where the
velocity is lower, and the temperature value maintains a nearly constant value for
the entire interior part of the central region and then sharply falls in the peripheral
region. This low value of temperature in the peripheral regions is due to high value of
average velocity in this region. Therefore, the phenomenon of flow getting diverted
to the periphery of the subassembly is called flow bypassing.
Further, it is seen that the average temperature in the central region increases with
increase in elevation and in the periphery. This suggests that thermal development is
incomplete. It is seen that the comparison between the porous body and CFD results
are very good for unblocked bundle.
3.2 Blockage cases
In the framework developed for unblocked bundle, additional source term (Ergun’s
correlation) for blockage is added. Blockages of different lengths are set up in the
center of the subassembly. A set of 14 cases as seen in Table 1 are studied in the
present analysis. The pressure drop and maximum temperature of the coolant are
found and reported in Table 2. These parameters are then compared with previous
CFD data [2] in order to understand its degree of accuracy. The pressure drop and
peak clad temperature results are calculated and compared with results obtained from
the original CFD model (Table 3).
From these results, it can be inferred that the porous body model does not give
exact results for all the cases. But the comparison with the CFD results is quite
satisfactory in most of the cases.
This can be attributed to the inability of the porous body model to correctly
simulate and predict the peak temperature around the wake region. This is clearly
seen in cases where the length of the blockage is small and in cases where the porosity
is low.
Table 1 Boundary
Boundary conditions Type Values
conditions
Liquid sodium entry Velocity inlet 7.56
(central)
Liquid sodium entry Velocity inlet 7.56
(periphery)
Zero shear adiabatic Wall Specified shear (X, Y,
wall Z directions) = 0
Inner sides of sector Symmetry
The system of algebraic equations is solved using Gauss–Seidel
iterative method.
Table 2 Various cases of porous blockage

Case Length (mm) Radius (mm) Porosity Diameter (mm)
1 80 8.28 40 0.5
2 35 16.56 40 0.5
3 14.5 24.84 40 0.5
4 14.5 33.12 40 0.5
5 6.25 8.28 5 0.5
6 10.5 8.28 10 0.5
7 19 8.28 20 0.5
8 400 8.28 60 0.5
9 25 8.28 40 0.25
10 200 8.28 40 0.75
11 5 8.28 40 0.5
12 5 16.56 40 0.5
13 5 24.84 40 0.5
14 5 33.12 40 0.5
Table 3 Comparison of different cases of blockage with previous CFD models

Case Pressure drop (Pa) Pressure drop (CFD) Peak temperature (K) Peak temperature
(Pa) (CFD) (K)
1 184,017 184,700 1023 1160
2 184,531 187,000 943 1155
3 185,710 192,900 880 1187
4 188,396 196,600 885 1218
5 183,828 184,300 948 1213
6 183,828 184,300 900 1229
7 183,852 184,400 895 1203
8 184,881 186,200 1047 1070
9 183,852 184,300 868 1172
10 184,346 185,200 1075 1173
11 183,826 185,400 868 1070
12 184,326 187,800 873 1172
13 185,710 193,500 880 939
14 188,396 206,200 885 997
4 Conclusions
This study is conducted to determine the flow and temperature characteristics of

fast reactor subassembly by using the porous body approach. The model based on
this approach is developed to reduce the computational time and to perform many
parametric study. Unblocked case and multiple cases of blockages are investigated
in the present study. The results of the various blocked and unblocked cases are
compared with published CFD data. In the case of unblocked bundle, the results
obtained from the present model are in very good comparison with the CFD results.
This shows that when fuel bundle is unblocked, the porous body model can predict
the scenario accurately.
For blockage cases, thermal hydraulic features of blocked fuel subassemblies with
various porous blockage configurations have been investigated through porous body
approach. A wide range of blockage radius, porosity, mean particle diameter and
location of blockage has been considered in this parametric analysis.14 cases are run
and simulated, and the results are collected. The results are compared with published
CFD data.
When simulating subassembly with blocked condition, there are some limitations.
Satisfactory comparison with CFD results are not obtained in some of these cases. In
the cases, blockage with large length and high porosity, the comparison is satisfactory.
But, when the blockage has low porosity and small lengths, this model cannot be
used.
Acknowledgements The authors would like to thank the Reactor Design Group for providing
computing facilities in the SAMDO Laboratory.
References
1. Naveen Raj M, Velusamy K (2016) Characterization of velocity and temperature fields in a 217
pin wire wrapped fuel bundle of sodium cooled fast reactor. Ann Nucl Energy 87(2):331–349
2. Naveen Raj M, Velusamy K, Maity R (2016) Thermal hydraulic investigations on porous
blockage in a prototype sodium cooled fast reactor fuel pin bundle. Nucl Eng Des 303:88–108
3. Di Piazzaa I, Magugliani F, Tarantinoa M, Alemberti A (2014) A CFD analysis of flow blockage
phenomena in ALFRED LFR demo fuel assembly. Nucl Eng Des 276:202–215
4. Govindha Rasu N, Velusamy K, Sundararajan T, Chellapandi P (2014) Simultaneous develop-
ment of flow and temperature fields in wire-wrapped fuel pin bundles of sodium cooled fast
reactor. Nucl Eng Des 267:44–60
5. Govindha Rasu N, Velusamy K, Sundararajan T, Chellapandi P (2014) Thermal hydraulic effect
of porous blockage in fuel subassembly of sodium cooled fast reactor. Ann Nucl Energy 70:64–81
6. Gajapathy R, Velusamy K, Selvaraj P, Chellapani P, Chetal SC (2009) A comparative CFD
investigation of helical wire-wrapped 7, 19 and 37 fuel pin bundles and its extendibility to 217
pin bundle. Nucl Eng Des 239:2279–2292
7. Gajapathy R, Velusamy K, Selvaraj P, Chellapandi P (2015) CFD investigation of effect of
helical wire-wrap parameters on the thermal hydraulic performance of 217 pin bundle. Ann
Nucl Energy 77:498–513
Multiscale Simulation of Droplet Wall
Interaction
M. Ananthan and Gaurav Tomar
1 Introduction
The need for multiscale approach in direct numerical simulation (DNS) of multiphase
flow is evident due to the existence of highly disparate time and length scales in the
flow which can vary over two orders of magnitude. E and Engquist [1] have classified
multiscale problems in multiphase flows into two categories:
Type A problems: Dealing with isolated localised issues involving multiple length
scales.
Type B problems: Constitutive modelling based on microscopic models.
Figure 1 shows both categories of problems for a bubbly flow in an inclined
channel. Problem B is the classical averaging method in multiphase flow modelling,
which results in a model which captures the aspects of the flow relevant to the present
study without completely resolving the flow. Problem type A is how we deal with
the unresolved thin films, threads and droplets that appear as the flow evolves.
In this study, we are interested in type A problems especially those involving thin
films, which are formed when two fluid masses collide or when a fluid drop hits a
surface. As we have already discussed, the length scales involved in the film drainage
process vary over a wide range of length scales from external flow at the order of
droplet diameter to nanometre scales where van der Waals forces become active.
It is computationally impractical to capture this using a fixed grid method. So, in
any grid-based method, the final stages of film drainage will always be at sub-grid
scale, and in conventional interface capturing methods, two approaching interfaces
will automatically coalesce when the two interfaces are close to each other and that
M. Ananthan (B) · G. Tomar

Department of Mechanical Engineering, Indian Institute of Science, Bangalore, India
e-mail: ananthanm@iisc.ac.in
G. Tomar
e-mail: gtom@iisc.ac.in
534 M. Ananthan and G. Tomar
Fig. 1 Schematic of a
bubbly flow in an inclined
channel where one can
observe two different kinds
of multiscale problems
involved in multiphase flows
Fig. 2 Figure showing the

schematic of the
computational domain and
its relation to a larger
multiphase problem
distance is in the order of grid dimension [2]. This is called as numerical coalescence,
and it is unphysical in nature.
A solution to this problem is to use an interface capturing technique like coupled
level set volume of fluid (CLSVOF) to capture the interface up to the sub-grid level,
then coupling it with an analytically derived thin film model to capture the evolution
of the film and drainage in the final stages (Fig. 2). Thomas et al. [3] used this
method for a front tracking/finite volume code where a thin film equation ignoring
the effects of surface tension and van der Waals forces was used. The results capture
the evolution of films thinner than the grid spacing reasonably well. In the present
study, we simulate a droplet sliding down an inclined plane, and we resolve the
thin fluid film in the sub-grid regime using a simple film model to give boundary
conditions to the flow solver.
2 Multiphase Flow Solver
Both phases are assumed to be incompressible with a sharp interface separating

them. The dynamics of the two phases are governed by the one fluid formulation of
Navier–Stokes equation given by

∂u
ρ(F) + u.∇u = −∇ p + ∇ · μ(F) ∇u + (∇u)T + ρ(F)g + fvγ (1)
∂t
Multiscale Simulation of Droplet Wall Interaction 535
where u and p are the velocity vector and pressure, respectively, ρ(F) and μ(F)
are the fluid density and viscosity which are a function of void fraction field, F.
We use Chorins projection to solve Eq. (1) where we discretise the advection term
using a second-order ENO scheme and diffusion term using central differencing.
γ
Surface tension forces, fv , are surface forces acting only at the interface and have
been modelled as volumetric body force using the continuum surface force model
of Brackbill et al. [4]. The interface is captured using CLSVOF algorithm given by
Sussman and Puckett [5]. This algorithm is mass conserving and calculates curva-
ture and surface normal with high accuracy which is used for surface tension force
calculation. The interface is advected by solving the advection equations for level-set
function, φ, and void fraction, F. Void fraction, F, is the volume fraction of liquid in
a cell and is equal to 0 for pure gas cell, 1 for pure liquid cell and assumes a value
between 0 and 1 for a two-phase cell.
3 Thin Film Equation
Using long wave approximation [6] on Navier–Stokes equations for a thin film on a
plane solid surface with another ambient fluid on top, we get
∂P ∂ 2u
=μ 2 (2)
∂x ∂ y
∂P
=0 (3)
∂y
where P = p + φ, and negative of φ(= −) is the disjoining pressure of the film
due to intermolecular forces.
Using kinematic boundary condition and continuity equation, we get

∂h 1 d dP h 3
= (4)
∂t μ dx dx 3
Using the expression for pressure we get,

3
∂h 1 ∂ ∂ h ∂ ∂h 3
=− γ 3+ h (5)
∂t 3μ ∂ x ∂x ∂h ∂ x
Equation (5) is the sum of three forces, which are viscous forces, surface tension
forces due to local curvature of the free surface and sum of excess intermolecular
forces. While the viscous forces retard the growth of instability, surface tension
forces have a stabilising influence on the interface until the film ruptures, but once
the film rupture happens, surface tension aides in the growth of the hole. And negative
Fig. 3 Evolution of random 2.5

perturbations in an
unstable 3D thin film: L = 2
8.48 at non-dimensional time
25.0
1.5
0.5

diffusivity ∂∂h
> 0 engenders the interfacial instability in the form of growth of
thicker regions at the expense of thinner regions of the film.
The disjoining pressure is obtained by a pairwise summation of interactions among
the molecules of the thin film and among molecules of the film and substrate and is
given by

A 8B
=− 3
+ (6)
6π h h9
where A is an effective Hamaker constant. The non-physical divergence of the

hydrodynamic model at the three-phase contact line is removed by the inclusion
of extremely short-range Born repulsion [6], which provides an equilibrium cut-off
distance and thus prevents the non-physical penetration of liquid into the substrate
at the point of rupture (Fig. 3).
The successive central differencing of Eq. (5) in space was performed on a uniform
grid. The resulting set of differential equations are integrated in time using backward
differentiation formulas using CVODE solver [7] due to the stiffness introduced
by widely separated length and time scales close to film rupture. Numerical results
agree well with the linear stability analysis results of Eq. (5), as shown in Fig. 4, thus
validating the solver.
4 Problem Formulation and Results
In the current study, we simulate a drop sliding down an inclined plane. The drop
diameter is 0.5, and the domain height is twice the drop diameter and the length
four times. We have repeated the numerical experiment from Thomas et al. [3].
The governing non-dimensional numbers are Bond number, ρgd
2
σ
= 7.955, the
Ohnesorge number, √ρμddσ = 0.1414, the ratios of the densities, ρρda = 10 and the
d
viscosities, μμda = 10. The surface tension coefficient is taken as σ = 0.1. The drop
starts from a small height from the surface and slides down the plane. We observe
that as we refine the domain the drop slides down further (refer Fig. 5), and thus,
Multiscale Simulation of Droplet Wall Interaction 537

growth coefficient values for 0
different wave numbers
obtained from simulations to -0.5
the analytical result obtained
Growth Coeffcient
from linear stability analysis -1
-1.5
-2
-2.5
-3
Numerical
-3.5 Analytical
-4
0 0.5 1 1.5
Wave Number
1.0
At time T = 0
At time T = 2, ∆x =1/64 At
time T = 2, ∆x =1/128
0.8
At time T = 2, ∆x =1/64 with Film coupled
0.6
Y
0.4
0.2
0.0
0.0 0.5 1.0 1.5 2.0
X
Fig. 5 Drop shape at time T = 2 for x = 1

64 (violet), x = 1
128 (red) and the green contour
represents coarse grid with coupled thin film
it is numerically expensive to reach a grid independent solution. Here, we have

solved a simple thin film model, Eq. (4), neglecting the effect of surface tension and
intermolecular forces. Those points in the drop contour which goes into the sub-grid
region have an imposed velocity boundary condition coming from evolving the thin
film equation.
τfilm y
u( j = 0) = u( j = 1) −
μdrop
where we obtain the shear stress from the film τfilm from τfilm = μ0 u film
h
. As observed,
the coupled model in the coarse grid gives results which are much closer to the
resolved results.
5 Conclusions
In this study, we have used a simple thin film model to capture a multiscale
phenomenon when a droplet slides down an inclined plane. As observed from the
results, incorporating this model reduces the computational expense considerably
since we need not refine the domain further to capture the flow physics. We have
successfully simulated a complex film model which captures the effect of surface
tension and intermolecular forces, which we plan to couple with the macroscopic
multiphase flow solver in the immediate future.
References
1. Weinan E, Engquist B, Huang Z (2003) Heterogeneous multiscale method: a general method-

ology for multiscale modeling. Phys Rev B 67(9):092101
2. Tryggvason G, Scardovelli R, Zaleski S (2011) Direct numerical simulations of gas–liquid
multiphase flows. Cambridge University Press
3. Thomas S, Esmaeeli A, Tryggvason G (2010) Multiscale computations of thin films in multiphase
flows. Int J Multiph Flow 36(1):71–77
4. Brackbill JU, Kothe DB, Zemach C (1992) A continuum method for modeling surface tension.
J Comput Phys 100(2):335–354
5. Sussman M, Puckett EG (2000) A coupled level set and volume-of-fluid method for computing
3D and axisymmetric incompressible two-phase flows. J Comput Phys 162(2): 301–337.
6. Khanna R, Sharma A (1997) Pattern formation in spontaneous dewetting of thin apolar films. J
Colloid Interface Sci 195(1):42–50
7. Cohen SD, Hindmarsh AC, Dubois PF (1996) CVODE, a stiff/nonstiff ODE solver in C. Comput
Phys 10(2):138–143
Numerical Study of Effect of Asymmetry
on Performance of Bio-mimetic Caudal
Fin Shapes
K. L. Vidhu Manohar and Ranjith Maniyeri
1 Introduction
Adapting the propulsion systems existing in the biological world to man-made util-
ities is one of the topics which gains interest at a faster pace. Although the physics
behind the biological propulsion systems is very complex, attempts are made to
analyse simple models which will pave for future research.
The authors would like to concentrate on propulsion systems to be employed in
autonomous underwater vehicles (AUVs) and autonomous surface vehicles (ASVs).
The body and/or caudal fin (BCF) propulsion systems are found most suitable for
efficiency and speed considerations. The effect of fin shapes on the performance of
caudal fins is not widely studied. Krishnadas et al. [1] have performed such kind
of a study which aimed to arrive at an optimal shape for caudal fin shape based on
numerical simulations for optimum propulsion efficiency. The study was conducted
on eight fin shapes developed, from which one is identified as optimal.
Feilich and Lauder [2] had conducted experimental studies on robotic fins for the
performance. Since they studied the shapes of different surface areas, the conclu-
sions were a function of problem-specific. Geder et al. [3] observed that the thrust
developed increased nine times as area was increased threefold.
From conclusions of Blake [4] who studied pectoral fins of the same area and from
studies by Krishnadas et al. [1], fin shapes for the present work have been chosen with
the same surface area. Further in the present work, the variation of thrust generated
is mainly studied if the asymmetry is present in the fin. Geometries from [1] are
used as the reference shape (shown in Fig. 1), and geometric, motion and numerical
models are assumed to be the same as that of [1]. Also, the rigidity condition is
K. L. V. Manohar · R. Maniyeri (B)

Biophysics Laboratory, Department of Mechanical Engineering, National Institute of Technology
Karnataka, Surathkal, Mangalore, Karnataka 575025, India
540 K. L. Vidhu Manohar and R. Maniyeri
Fig. 1 Fin shapes with

symmetry [1]
unaltered, since the flexibility of the fins has a great effect of performance of the fins
as described by Esposito et al. [5] and Feilich and Lauder [2].
2 Methodology
The methodology consists of computational analysis of caudal fin shapes for the effect
of asymmetry of fin’s performance. The movement of caudal fins for propulsion in
thunniform and carangiform swimming modes is described by a combination of linear
and angular movements. Different species of fishes will have heaving and pitching
or swaying and yawing depending upon the plane of movement. The present work
analyses fins in heaving and pitching motion.
To choose a domain of analysis for the study, the non-dimensional parameter
Strouhal numbers selected are 0.2, 0.25 and 0.3 from the studies conducted by
Triantafyllou et al. [6]. Selected Strouhal number is within the range of optimal
performance for the fins.
2.1 Geometrical and Motion Models
The geometry selected for the present study is from [1]. All the fin shapes used in
[1] are of the same surface area and symmetric. Keeping the surface area same, two
Numerical Study of Effect of Asymmetry on Performance … 541
Fig. 2 Geometries designed with asymmetry. Case-1 represents geometries with same forking
angle, and Case-2 represents geometries with variable forking angle
sets of three models each are modelled in which symmetry is altered with more area
concentrated on the upper limb. The geometries are made with 62 to 82% of the area
concentrated in one of the limbs in three equal steps.
Figure 1 shows the reference fin shapes, and Fig. 2 shows the fin shapes designed
with asymmetry. The first set has been designed with the same forking angle and
second set with variable forking angle. Although the length of the fins has increased,
it is expected that flow will be still laminar.
Motion model
The transverse motion of the model is referred to as heaving which is given by
h(t) = Asin(2π f t + ψ) (1)
where A is the flapping amplitude, f is the flapping frequency and ψ is the phase
difference between heaving and pitching. Pitching about a mid-span axis is given by
θ (t) = θmax sin(2π f t) (2)
where θmax is the pitching amplitude which is assumed to be 25°.
2.2 Details of CFD Simulation Implementation
ANSYS Fluent 16.2 (ANSYS Inc. 2015) [7] is used for simulating physics. The
implementation used pressure-based solver for the incompressible flow problem.
The laminar flow field is modelled in similar fashion as in [1] considering operating
regions involved in [8] and [9] with Re = 20,000 for parameters of a flapping foil.
Although the flow is laminar, the simulation results in [1] are in good agreement with
the experimental results of Anderson et al. [10] even at Re = 40,000. The governing
equations are discretized using first-order implicit formulation in the time domain
and second-order upwind scheme in space, with SIMPLE algorithm employed for
pressure velocity coupling.
All boundaries except inlet are set to pressure outlets. The velocity corresponding
to each Strouhal number is assigned for the inlet.
To simulate the movement of fins, a user-defined function (UDF) is used in
ANSYS. To conduct an analysis of moving boundaries, dynamic meshing of ANSYS
is used with the combination of smoothing, layering and re-meshing to adapt to the
fin movement. Negative cell volumes are avoided with a spring constant of 1 in
smoothing function. Re-meshing values also set to the default values of the mesh
generated to avoid the negative cell volume. From [1], it was clear that the CFD
scheme gave close results between fine and coarse meshes in heave and thrust forces.
Near the fin, the meshing method used is body of influence which is set from 0.003 m
with 8% growth rate. To have moderate accuracy, a medium mesh is used elsewhere.
The drag force measured provides the thrust developed by the fin at each instant
of time. The thrust developed over a cycle is the time average value of the drag force
measured
T
1
Thrust force = Fx dt (3)
T
0
The simulations are conducted for St = 0.2, 0.25 and 0.3 with flapping frequency,
f = 1 Hz, and amplitude of flapping, A = 0.1 m, in the Strouhal number expression.
The swimming speed U is calculated for each St.
U
St = f (4)
A
Validation of results
The developed computational model is first validated for the case of the symmetric
fin by comparing the variation of thrust force for the Strouhal numbers as shown in
Fig. 3, with that of Krishnadas et al. [1]. A good agreement is obtained between our
results and that of [1].
The variation of the thrust force over a cycle is shown in Fig. 4. The time average of
thrust is calculated for the geometries for Case-1 and Case-2 and is shown in Fig. 5.
0.2
Thrust force(N) 0.15
Krishnadas et al. [1]

0.1
present study
0.05
0
0.15 0.25 0.35
Strouhal number
Fig. 3 Average thrust force in comparison with [1]
[1]
0.3
Thrust(N)
Case-1A
0 0.5 1 1.5
-0.2 Case-1B
Time (s)
Case-1C
-0.7
Fig. 4 Variation of thrust in a cycle for Case-1 (for St = 0.2)
The average thrust generated in Case-1 is less than that of original geometry as
compared to Case-2. The variation of thrust does not vary much with redistribution
of the area in Case-1. But in Case-2C (Fig. 5b) where forking angle increases, it is
observed to have a decrease in thrust force as compared to lesser forking.
4 Conclusions
With the aim to identify the effect of asymmetry of caudal fin shapes on thrust gener-
ated through numerical simulations, we modelled six geometries keeping the surface
area constant. The asymmetry causes a decrease in the average thrust generated.
0.22
0.18
Thrust force (N) [1]
0.14
Case-1A
0.1
Case-1B
0.06
Case-1C
0.02
0.15 0.25 0.35
Strouhal number
(a)
0.22
0.18 [1]
Thrust force(N)
0.14 Case-2A
0.1
Case-2B
0.06
Case-2C
0.02
0.15 0.25 0.35
Strouhal number
(b)
Fig. 5 Variation of the thrust reported in reference geometry [1] with that of geometries in Case-1
in a and in geometries of Case-2 in b
Distribution of area to any limb does have a marginal effect in thrust generation if
forking angle is not kept the same. The thrust produced by the symmetric fin still
is the highest. Future studies need to be conducted to understand the variation in
input power and efficiency due to asymmetry. The study will be extended for higher
flapping frequency and amplitude of oscillation in future.
Acknowledgements This research was supported by the Science and Engineering Research Board,
a statutory body of the Department of Science and Technology (DST), Government of India, through
the funded project ECR/2016/001501.
References
1. Krishnadas A, Ravichandran S, Rajagopal P (2018) Analysis of biomimetic caudal fin shapes

for optimal propulsive efficiency. Ocean Eng 153:132–142
2. Felich KL, Lauder GV, Passive mechanical models of fish caudal fins: effects of shape and
stiffness on self-propulsion Bioinspiration Biomimetics 10
3. Geder JD, Ramamurti R, Edwards D, Young T, Pruessner M (2014) Development of a robotic
fin for hydrodynamic propulsion and aerodynamic control. In: Oceans-St. John’s, pp. 1–7
4. Blake RW (2014) Influence of pectoral fin shape on thrust and drag in labriform locomotion.
J Zool 194:53–66
5. Esposito CJ, Tangorra JL, Flammang BE, Lauder GV (2012) A robotic fish caudal fin: effects
of stiffness and motor program on locomotion performance. J Exp Biol 215:56–67
6. Triantafyllon GS, Triantafyllon MS, Grosenbaugh MA (1993) Optimal thrust development in
oscillating foils with application to fish propulsion. J Fluids Struct 7:205–224
7. ANSYS Inc 2011, ANSYS fluent user’s guide
8. Mittl R, Dong H, Bozkuttas M, Lauder G, Maden P (2006) Locomtion with flexible propulsor:
II. Computational modelling of pectoral fin swimming in sunfish. Bioinspiration Biomimetics
1:514–535
9. Heathcote S, Gurusul I (2007) Flexible flapping airfoil propulsion at low Reynolds number.
AIAA J 45:1066–1079
10. Anderson JM, Streitien K, Barett DS, Triantafyilou MS (1998) Oscillating foils of high
propulsive efficiency. J Fluid Mech 360:41–72
Inertial Migration of Neutrally Buoyant
Particle in Confined Channel
Atul M. Bhagat and Partha S. Goswami
1 Introduction
The flow of suspended particles is ubiquitous in engineering processes. The flow may
be either near-uniform or shear flow. The motion of particle can be predicted through
flow visualisation experiments or using CFD considering different forces exerted on
the particle, like a drag, lift or particle–particle hydrodynamic interaction. The finite
inertia of fluid in channel flow is the main source of a lift force which is responsible
for the particle lateral migration. Rotation of particle due to torque exerted by the fluid
plays an important role in controlling migration of particle across the width of the
channel. The experimental observations of this behaviour have been demonstrated by
several authors in case of macroscopic flows [1] as well as microfluidic devices [2].
A theoretical analysis has been performed by Asmolov [3] using matched asymptotic
expansion, and he came up with scaling for the lift which is based on the dimension of
a channel. CFD study has also been performed considering the reference frame on the
particle to predict the equilibrium position, and results are validated with microfluidic
experiments [2]. The effect of Reynolds number on equilibrium position in case of
flow through the channel of few millimetres in size has been studied numerically
[5]. The results have been validated with experiments performed by Muira [6]. All
of these studies are mainly focused on the prediction of equilibrium position for
non-interacting particles. Therefore, these investigations are valid only for a single
particle system with very low volume fraction. Till date, very few studies have been
carried out using a multi-particle system, where inter-particle interaction plays a
significant role. An experimental investigation of the inter-particle interaction in the
macroscopic channel reveals that particles form trains along the equilibrium position
A. M. Bhagat · P. S. Goswami (B)

Department of Chemical Engineering, IIT Bombay, Powai-76, Mumbai, India
e-mail: psg@che.iitb.ac.in
A. M. Bhagat
e-mail: atul.bhagat@iitb.ac.in
548 A. M. Bhagat and P. S. Goswami
[7]. It was observed that the trains formed by particles are affected by confinement
of channel and particle concentration. Similar kind of study has also been carried
out in microchannel [8]. The physics of inter-particle interaction is yet to be fully
understood. Therefore, it demands further investigations using the robust numerical
methodology which takes care of moving interacting particles in a flow field. In the
present study, we have set up a framework to investigate the multi-particle system
using a CFD method in which particle is modelled using an immersed boundary
method (IBM). To arrest the effect of inter-particle interaction (collisions, if at all),
we have included discrete element methodology. The results reported here address
the flow of two particles in a channel for different Reynolds numbers.
2 Resolved Particle Approach
2.1 Immersed Boundary Method (IBM)
This method is mainly used when the particle is considered to be of finite size that
means it encompasses several fluid mesh cells in it. These covered cells are assumed
to have the same velocity as that of particle covering it. The motion of incompressible
fluid in computational domain is governed by Navier–Stokes equations, along with
initial and boundary condition. Here, the same equations are applied to whole compu-
tational domain which also includes the domain inside the particle. More details on
IBM can be found elsewhere [11]. Equations of motion are summarised as follows
[9, 10]
∂u f
ρf + ρ f u f · ∇ u f = −∇ p + μ∇ 2 u f in f (1)
∂t
∇ · u f = 0 in f (2)
u f = u in (3)
u f = u p and σ · n = t p in p (4)
u f (x, t = 0) = u 0 (x) in f (5)
Time discretised form of momentum Eq. (1) is
u f − u (n−1)

+ ρ f u (n−1) · ∇ u (n−1) = −∇ p̃ + μ∇ 2 u (n−1)

f
ρf (6)
t f f f
Inertial Migration of Neutrally Buoyant Particle in Confined … 549
In Eq. (6), u (n−1)

f is velocity from the previous time step and u f is intermediate

solution. p̃ is the guessed value for pressure in the current time step that needs to be
corrected in the later stage using Poison equation. The data obtained from particle
equation of motion is then used to correct the velocity n domain occupied by the
particle. This process results in new velocity in the fluid domain ũ f , and the process
is similar to addition of equivalent force f to NS equation as

ũ f − u f
ρf = f (7)
t
With this process, the velocity in a computational domain now is no more
divergence-free, and to make it divergence-free the correction has to be carried out
as described below.
u (n)
f = ũ f − ∇φ (8)
Here, u (n)
f represents final velocity field in the current time step. This newly
obtained velocity field now guaranties divergence-free condition, i.e. ∇ · u (n)
f = 0,
and also results in Poison equation for scalar φ as
∇ 2 φ = ∇ · ũ f (9)
Particle velocity and position can be obtained through the solution of equation of
motion of the particles. This requires the estimation of forces exerted on particle by
fluid. Forces on the particles by fluid can be calculated as follows.

η(x, y, z, t)dV = η(x, y, z, t)ζ (x, y, z, t)dV (10)
p
Here,

1, if (x, y, z) ∈ p
ζ = (11)
0, else
Here, η can be used for both viscous and pressure components of forces as
η = ρυ∇ 2 u (n−1)
f (12)
And,
η = −∇ p (n−1) (13)
2.2 Particle Equation of Motion
In the present method, motion of the particle is described by Lagrangian descrip-

tion, in which position (xi ) and angular velocity (ωi ) of particles are calculated in
computational domain using Newton’s second law of motion. A simple force balance
equation for a single discrete particle is given by
m i ẍi = F i,n + F i,t + F i, f + F i,b (14)
dωi
Ii = ri,c × F i,t (15)
dt
Here, normal and tangential particle–particle contact forces are F i,n and F i,t , respec-
tively. The force exerted by surrounding fluid is given by F i, f , and F i,b is the
body force such as gravity. Mathematically, these discrete particles are modelled
as spheres, and its velocities and positions are calculated from respective momentum
equation. If at all any collision happens between two particles, during collision an
infinitesimal overlap of colliding particle is allowed in this soft sphere mode (DEM
approach). This small overlap δ p and relative velocity of these particles in normal
direction u n during collision are used to calculate the normal repulsive forces.
Using the linear spring–dashpot model, a normal force is given as
F n = −kn δn + cn u n (16)
The tangential forces are given by

t
F t = min kt
u t dt + ct u t , μF n (17)
tc,0
where the relative tangential velocity in between contacting particles is u t , and

spring and damping coefficient in normal and tangential direction is kn , kt , cn and ct .
2.3 Simulation Set-Up
In present simulations, the motion of a neutrally buoyant particle inside a channel is

considered as shown in Fig. 1. The domain is long duct with a square cross section.
Length of domain is 20d p along the flow direction (x-direction), and width (w) and
height (h) of channel are taken to be 4.5d p . No-slip boundary condition is imposed on
the side walls of channel. Periodic boundary condition is applied along x-direction.
A particle with diameter d p is initially placed at a random position in the channel
Fig. 1 Domain for

CFD-DEM study (a) and
position of particle in
channel (b)
along z-axis. The domain is discretised using uniform grids in all the directions.
Approximately 5000 grids were used to resolve the background laminar flow. Parti-
cles were resolved using very fine meshes. The maximum size of the grid near surface
of the particle was one-fourth of the background grid. Such a fine mesh is capable of
capturing the disturbance velocity field induced by the motion of the particle. Prop-
erties of fluid phase are as follows. Fluid kinematic viscosity isυ = 0.01m2 /s and
density ρ = 1kg/m3 . Similarly, properties of a particle are d p = 0.22m, and density
is adjusted to make particle neutrally buoyant. The restitution coefficient (e) and
coefficient of friction (μ) are considered as 0.9 and 0.3, respectively. A periodic flow
is driven through mapped boundary condition at inlet and outlet with given average
velocity in a domain. This resulted in a parabolic velocity profile in a channel. The
channel Reynolds number is defined as
u avg Dh
Rec = (18)
ν
Here, Dh = h+w 2hw

, that is, hydraulic diameter of channel. In present simulations,
the minimum inter-particle distance never becomes less than the particle diameter;
therefore, resolving lubrication layer is not required. The simulation methodology
has been validated through the prediction of single particle equilibrium position as
reported in the literature [2].
In this section, we have presented the particle migration and predicted the equilibrium
position of the particle which originates from different wall-normal locations. Next,
we have described the effect of channel Reynolds number (Rec ) on lateral migration
of the particle. Towards the end, we have also discussed the effect of hydrodynamics
interaction of the particle migration through the multi-particle simulation.
3.1 Particle Equilibrium Position
Figure 2a shows the particle lateral migration and attainment of equilibrium position
in the channel. It is observed from the figure that if a particle is released from an
initial position of 0.3rc , it migrates in the wall-normal direction and equilibrates at
≈ 0.6rc , in an annular region near a wall. At the equilibrium position, a particle is
found to rotate at a torque-free rotation rate of −1.0932 as shown in Fig. 2b. The
rotation rate is scaled with an ambient rotation rate of undisturbed flow. Stability of
present equilibrium position in channel is analysed through a series of simulations as
shown in Fig. 3. All tracks in the figure show merging in single line located at around
≈ 0.6rc that is the equilibrium position of a particle. Particles with initial condition
near ≈ 0.6rc tend to equilibrate at a rapid rate than that of others. The particle with
initial position very near to centre line (≈ 0.5rc ) is also found to migrate towards the
wall. Even though the channel centre plane is a plane of symmetry, it is not a stable
equilibrium position for the particle. A very small perturbation helps the particle to
migrate and attain the fixed equilibrium position at ≈ 0.6rc . All these observations
assure that the wall facing equilibrium position is most stable in a channel.
3.2 Effect of Channel Reynolds Number (Rec )
Several simulations are carried out with different channel Reynolds numbers to study
the effect of Reynolds number on particle migration. The Rec values are varied from
20 to 100, and results obtained are plotted in Fig. 4. The particle equilibrium position
is found to move towards the centre with increasing Reynolds number. Such a trend
contradicts with the existing literature [2, 4] except the numerical study of Nakagawa
0.8
0.7
0.6
z
0.5
0.4
0.3 Re c = 40, d/D = 0.22

z = 0.6430
z = 0.5780 COMSOL
0.2
0 20 40 60 80 100 120 140 160 180 200
(a)
2
Re c = 40, d/D = 0.22
1.5 y= 1.0
0.5
y
-0.5
-1
-1.5
0 20 40 60 80 100 120 140 160 180 200
(b)
Fig. 2 Particle migration at Rec = 40, a equilibrium position and, b particle equilibrium rotation
rate
0.8
0.7
0.6
0.5
0.4 Re c = 40, d/D = 0.22

z
z = 0.1
0.3 z = 0.2
z = 0.3
0.2 z = 0.4
z = 0.5
z = 0.6
0.1 z = 0.7
z = 0.74
0
0 20 40 60 80 100 120 140 160 180 200
Fig. 3 Particle migration from different positions along channel width
[5]. The authors have suggested that such behaviour may occur beyond some critical
Rec value. Their study is carried out at very high Rec , and also d/D ratio is also
less compared to the present study. For the present study, d/D ratio is set to 0.22.
Such an observation gives rise to the question that whether the critical Rec value
after which this type of behaviour is observed depends on d/D ratio. Another result
obtained from the present study is for rotation rate. A trend is shown in Fig. 4b. The
torque-free rotation rate at equilibrium position is found to vary linearly with Rec . It
is observed that the magnitude of rotation rate increases with Rec . This happens due
to an increase in shear across the particle with increasing channel Reynolds number.
3.3 Effect of Hydrodynamics Interaction
The effect of inter-particle hydrodynamic interaction is studied using two different

sets of simulation. In the first case, the particles are aligned horizontally in z-direction,
whereas in the second case, particles are aligned vertically. In both cases, the inter-
particle distance of 2.27d p is kept between the centres of two particles. The steady
particle position obtained from these two cases is shown in Fig. 5. Both cases are
simulated at channel Reynolds number of 40 with particles having diameter to half
channel height ratio as 0.22. Figure 6 shows the results obtained for horizontally
aligned particles.
0.7
d/D = 0.22
0.69
0.68
0.67
0.66
0.65
z
0.64
0.63
0.62
0.61
0.6
0 10 20 30 40 50 60 70 80 90 100 110
Re c
(a)
0
d/D = 0.22
-0.5
-1
y
-1.5
-2
-2.5
-3
0 10 20 30 40 50 60 70 80 90 100 110
Re
c
(b)
Fig. 4 Effect of channel Reynolds number on a equilibrium position and b particle equilibrium
rotation rate
Fig. 5 Steady equilibrium

position in multi-particle
system for both horizontally
and vertically aligned
particles Horizontal
Vertical
Both particles in this test are found to migrate in the same direction and equilibrate
at the same wall-normal location at 0.6477rc . During the lateral migration, it is
observed both the particles follow an equivalent trajectory in a wall-normal direction
which is very similar to the single particle case study. The difference between the
rotation rates of both particles is found in an initial transient state. In this region, the
rotation rate for particle-1 is lower than that of particle-2. This behaviour explains
the lower lateral velocity of particle-1 while reaching to the equilibrium position as
seen in Fig. 6a. Figure 7 shows the inter-particle distance as function of time. Both
particles are initially placed at an inter-particle distance of 2.27d p , which moves
further away as both particles migrate towards a stable equilibrium position. During
this migration phase, the distance between these two particles increases very rapidly
as time progress. Finally, the inter-particle distance is found to have a steady value
of ≈ 8d p . In the second case, the particles are aligned vertically in wall-normal
direction.
The steady equilibrium position acquired by these two particles is shown in Fig. 5.
Same inferences can also be drawn on particle tracks as shown in Fig. 8. Particles,
in this case, are found to attain the equilibrium position in opposite direction in the
channel. This equilibration process is also observed to be very rapid. The particles
are found to reach their respective equilibrium position within a short duration. One
possible reason behind such behaviour, in this case, may be due to strong hydrody-
namics interaction due to low surface clearance. The particle rotation obtained in this
case is shown in Fig. 8b, which suggests that the particles are rotating in opposite
direction to each other. This also confirms the assumed reason behind the equilibra-
tion in opposite direction. A very rapid change in rotation rate is observed in this
case, and it is also found that there is a significant difference in trends of rotation
rate with time for both particles. This change is mainly due to the respective position
of the particle in a channel and their finite size. Finally, both particles are found to
rotate with an equal and opposite magnitude of rotation rate.
0.8
0.7
0.6
0.5
z
0.4
Re c = 40, d/D = 0.22

0.3 Particle-1
Particle-2
single Particle
0.2
0 20 40 60 80 100 120 140 160 180 200
(a)
0.2
Re c = 40, d/D = 0.22
Particle-1
0
Particle-2
0.9303 (ambiant rotation rate)
-0.2
-0.4
y
-0.6
-0.8
-1
-1.2
0 50 100 150 200
(b)
Fig. 6 Comparison for horizontally aligned particles, a equilibrium position and b particle
equilibrium rotation rate
5
h
2
Re c = 40, d/D = 0.22
1
0 50 100 150 200
Fig. 7 Inter-particle distance in horizontally aligned particles in multi-particle system
4 Conclusions
The simulations are carried out using fully resolved CFD-IBM approach. The fluid
part is solved using CFD, and the particle motion is computed by solving particle
equation of motion. Particle is modelled using an immersed boundary method. From
the simulations, it is observed that the equilibrium position obtained is found to be
stable. The stability of the equilibrium position has been verified by releasing the
particle from different initial positions. It is observed that all the particles migrate to
the same final position. With increasing Reynolds number, we observed shifting of
equilibrium position towards the centre, though further studies are required to provide
a conclusive view. It is also observed the inter-particle hydrodynamics interaction
plays an important role to dictate final equilibrium position.
0.9
0.8 Re c = 40, d/D = 0.22

0.7 Particle-1
Particle-2
0.6 1
0.5
z
0.5
z
0.4
0.3 0
0 0.5 1
0.2
0.1
0
0 20 40 60 80 100 120 140 160 180 200
(a)
3
Re c = 40, d/D = 0.22

2
Particle-1
Particle-2
1 2
0 0
y
y
-2
-1
-4
-2 0 1 2
-3
-4
0 20 40 60 80 100 120 140 160 180 200
(b)
Fig. 8 Comparison for vertically aligned particles, a equilibrium position and b particle equilibrium
rotation rate
Acknowledgements This work is supported by IRCC IIT Bombay (No: RD/0512-IRCCSH0-017).
References
1. Silberberg A, Segre G (1961) Radial particle displacements in Poiseuille flow of suspensions.

Nature 189:209–210
2. Di Carlo D, Edd JF, Humphry KJ, Stone HA, Toner M (2009) Particle segregation and dynamics
in confined flows. Phyis Rev Lett 102:094503
3. Asmolov SE (1999) The inertial lift on a spherical particle in a plane Poiseuille flow at large
channel Reynolds number. J Fluid Mech 381:63–87
4. Di Carlo D, Edd JF, Irimia D, Tompkins RG, Toner M (2008) Equilibrium separation and
filtration of particle using differential inertial focusing. Anal Chem 80:2204–2211
5. Nakawaga N, Yobu T, Otomo R, Kase A, Makino M, Itano T, Sugihara-Seki M (2015) Inertial
migration of a spherical particle in laminar square channel flows from low to high Reynolds
numbers. J Fluid Mech 779:776–793
6. Muira K, Itano T, Sugihara-Seki M (2014) Inertial migration of neutrally buoyant sphere in a
pressure-driven flow through square channels. J Fluid Mech 749:320–330
7. Matas J, Glezer V, Guazzelli E, Morris JF (2004) Trains ofparticles in finite Reynolds number
pipe flow. Phys Fluids 16:4192–4195
8. Kahkeshani S, Haddadi H, Di Carlo D (2016) Preferred interparticlespacings in trains of particle
in inertial microchannel flows. J Fluid Mech 786:R3
9. Patankar NA, Singh P, Joseph DD, Glowinski R, Pan TW (2000) A new formulation of the
distributed lagrange multiplier/fictitious domain method for particulate flows. Int J Multiph
Flow 26:1509–1524
10. Shirgaonkar AA, MacIver MA, Patankar NA (2009) A new mathematical formulation and fast
algorithm for fully resolved simulation of self-propulsion. J Comput Phys 228:2366–2390
11. Kloss C, Goniva C, Hager A, Amberger S, Pirker S (2012) Models, algorithms and validation
for open source DEM and CFD-DEM. Prog comput fluid dyn 12:140–152
A Massively Parallel High-Order
ADER-WENO Finite-Volume Scheme
on Unstructured Meshes
Dipak Vaghani, Dasika Sunder, and Ratnesh K. Shukla
1 Introduction
Unstructured grids are widely used in science and engineering for their ability to
accurately represent complicated geometries. In this article, we present an adaptive
order weighted essentially non-oscillatory (WENO) finite-volume scheme of fourth-
order accuracy both in space and time for solving nonlinear hyperbolic systems on
unstructured meshes in two dimensions. The choice WENO schemes is justified
by the inherent high-order discretization associated with the WENO reconstruction
procedure which enables resolution of the intricate flow features such as propa-
gating vortices and unsteady slip lines on relatively coarse grids. The non-oscillatory
feature of the WENO schemes ensures that discontinuous features such as shocks
are captured with high fidelity. Prior work on unstructured mesh WENO schemes
has been principally focused on least squares reconstruction over a relatively wide
stencil [1–4]. Since our goal is to develop an efficient massively parallel code, the
stencil chosen for the WENO reconstruction is the most compact possible for a
given order of accuracy. This leads to decrease in the number of communications
required between the processors and the numerical dissipation of the scheme. We
evolve the solution in time using Arbitrary DERivative (ADER) method proposed by
Dumbser et al. [2]. Unlike the Runge–Kutta methods which are multi-step methods
and the number of steps required increases with increasing order of accuracy, the
ADER method involves only single step for any desired order of accuracy. Since the
ADER schemes are one-step methods, there is a significant decrease in the number
D. Vaghani · D. Sunder · R. K. Shukla (B)

Department of Mechanical Engineering, Indian Institute of Science, Bengaluru 560012, India
e-mail: ratnesh@iisc.ac.in
D. Vaghani
e-mail: vaghanih@iisc.ac.in
D. Sunder
e-mail: dasikasunder@iisc.ac.in
562 D. Vaghani et al.
of communications required between the processors for each time step, which is a
very desirable feature in a massively parallel code. In the subsequent sections, we
present the reconstruction procedure and its validation for two-dimensional inviscid
Euler equations of gas dynamics.
2 Methodology
We consider the inviscid compressible flow of an ideal gas. The governing two-
dimensional Euler’s equations for such flow scenarios are given by

∂
U dV + F dS = 0 (1)
∂t
V S
where V is the control volume and S is the surface bounding the volume.
= (ρ,
U ρu, ρv, E) denotes the vector
of the conserved variables
and F =
ρu, ρu 2 + p, ρuv, u(E + p) î + ρv, ρuv, ρv2 + p, v(E + p) ĵ is the flux
vector. Here u and v are the velocity components along x- and y-directions, ρ is the
density, p is the pressure and E is the total energy of gas. On using an appropriate
N-point Gaussian quadrature for the surface integral, the following semi-discrete
form of the governing equations is obtained
d U̇i −1 Nf

N gp
= L U̇˙ i = Sj wk Fk nk (2)
dt Vi j=1 k=1
where N gp is the total number of quadrature points. In our present work, an upwind
flux is calculated using the rotated HLLC Riemann solver [5]. The above semi-
discrete equation is updated in time using the fourth-order accurate ADER scheme.
A detailed implementation of ADER scheme is presented in [2].
2.1 Reconstruction Method
In order to find the values of the conserved variable at the quadrature points with
fourth-order accuracy, we assume that it is represented using a fourth-order (third
degree) two-dimensional polynomial. The reconstruction polynomial in a cell j0 can
be obtained by following procedure:
(1) Select a central spatial stencil containing enough distinct cells to obtain
a fourth-order polynomial. In two dimensions, the fourth-order polynomial
consists of ten degrees of freedom. Therefore apart from the target cell, there
A Massively Parallel High-Order ADER-WENO … 563
Fig. 1 Stencil for

reconstruction of a
fourth-order polynomial
should be at least nine cells in the stencil. A reference stencil is shown in Fig. 1.
In order to obtain the fourth-order reconstruction polynomial u r 4 (x, y) on the
stencil, we maintain the cell average of the conserved quantity exactly on the
target cell j0 (for making the scheme conservative) on rest of the cells, the cell
average is maintained in a least-squares sense.
(2) From the bigger stencil shown in Fig. 1, we choose S 3 number of smaller sten-
cils for obtaining the third order reconstruction polynomials u r 3:k (x, y), k =
1, 2, . . . , S3 . Note that in two dimensions, the third-order polynomial has six
degrees of freedom. Therefore, apart from target cell, there must be at least
five cells in each stencil. The stencils are chosen in various sectorial directions
to provide the necessary stable shock capturing capabilities.
(3) The final step is to hybridize the fourth-order and the third-order reconstruc-
tion polynomials. To do this, we first assign linear weights to each stencil.
Each stencil is given a linear weight: γ r 4 = α, γ r 3:k = (1 − α)/S3 . The
parameter α is taken as 0.85 throughout the paper. Smoothness indicators and
nonlinear weights for each stencil can be calculated for each stencil as specified
in [3]. If w represents the nonlinear weight associated with a stencil, the final
reconstruction polynomial can be written as:

wr 4 S3
u(x, y) = u (x, y) −
r4
γ u (x, y)
r 3:k r 3:k
γ r4 k=1

S3
+ wr 3 u r 3:k (x, y) (3)
k=1
2.2 Parallelization Strategy
In this section, some remarks on the parallelization of ADER-WENO scheme on

modern massively parallel systems are made. Since the ADER-WENO method leads
to an explicit one-step scheme in time, the total communication overhead is consid-
erably low compared to methods using high-order Runge–Kutta time integration
because data has to be exchanged only at the beginning of each time step. Then,
each subdomain can evolve the solution independently of its neighbour domains.
Unstructured mesh partitioning is done with the Z-order algorithm provided in the
deal.II framework.
The major task is to identify various neighbour cells along with their orientation
relative to target cell. The first step is to establish vertex-to-cell connectivity which
provides the information about the Voronoi neighbours (cells sharing a common
vertex with the target cell). Using this information, first layer of neighbours cells is
identified. Among these neighbour cells, some of them are vertex neighbours (shares
vertex only, coloured green in Fig. 1), and rest of them are face neighbours (coloured
yellow in Fig. 1). In order to keep the stencil small and hence decrease the cost of
computation, we do not choose all the Voronoi neighbours of the first layer of cells in
the stencil. Instead, only cells that share a common face with the face neighbouring
cells of the target cell are chosen. These cells are coloured in blue in Fig. 1. However,
in case there are not enough number of cells in the stencil for reconstruction, we also
add the Voronoi neighbours of the first layer of cells in the stencil.
In this section, we present three test cases. In the first test case, we verify if the
proposed ADER-WENO scheme converges at fourth order for smooth problems. In
the second test case, we study the parallel efficiency of the code. Finally in the third
test case, we test the ability of scheme to handle strong shocks and complex domains.
3.1 Order Property
In this test case presented by Balsara and Shu [6], a smooth isentropic vortex initially
located at origin is convected at an angle of 45° through a sinusoidally perturbed
domain (x, y) ∈ [−8, 8]×[−8, 8]. Periodic boundary conditions are used on all sides,
and the vortex is advected for five cycles. Since exact solution for this test case
is available, we compute solution for different grid sizes varying from h = 32 1
to
h = 192 and calculate L ∞ and L 2 norm for error in density field. It confirms that the
1
rate of convergence is as expected value for fourth-order accurate WENO scheme.

As can be seen from Fig. 2, the scheme reaches the desired convergence rate.
Fig. 2 Convergence rate for

isentropic vortex test
3.2 Weak and Strong Scalability Tests
We perform strong and weak scalability tests using the thirteenth configuration of the
two-dimensional Riemann problems proposed by Lax and Liu [7] as a reference case.
The problem initially consists of a vertical contact discontinuity and a horizontal
shock. In strong scalability test, we study how the solution time varies with the
number of processors for a fixed total problem size, whereas in weak scalability test
we study how the solution time varies with the number of processors for a fixed
problem size per processor.
(1) Strong scalability: In strong scalability test, the problem is simulated for 1 h
of CPU time. The number of cells in the domain is kept fixed at 20482 , and
number of processors used for simulation is varied from 240 to 3840. Since
the total number of cells is fixed, the number of time steps through which the
solution advances in one hour of CPU time must increase linearly with the
number of processors used. Figure 3 (top) confirms that the algorithm scales
well up to around 4000 processors.
(2) Weak scalability: In this test, the problem is again simulated for 1 h of CPU
time. However, the number of cells in the complete domain is set in such a
way that 2730 cells are present on each processor. The number of cells and
processors ranges from 2562 to 20482 and 24 to 1536, respectively. As the load
on each processor is fixed, we expect the solution to advance through the same
number of time steps for all the different configurations. As can be seen from
Fig. 3 (bottom), the solution advanced through approximately thousand time
steps for all the different configurations.
Fig. 3 Strong (top) and weak (bottom) scalability

3.3 Shock Diffraction from a Circular Cylinder
In this test, we consider interaction of strong Mach 10 shock with stationary cylinder
of unit diameter. We computed solution to this problem on an unstructured mesh with
2.8 million cells and 1056 processors. Since the cylinder is curved, we use higher-
order mappings to approximate the curved boundary more accurately. Initially, a
bow shock is reflected from the cylinder when the incident shock hits the cylinder.
As the incident and the reflected shocks meet, two Mach stems along with contact
discontinuities emanating from the triple points form. These shocks and contact
discontinuities undergo multiple interactions and reflections leading to formation
of complex vortical structures in the cylinder wake as shown in Fig. 4. Clearly, the
proposed WENO scheme captures these unsteady features well, and at the same time
is able to handle the strong Mach 10 shock robustly.
4 Conclusions and Future Work
A fourth-order accurate ADER-WENO scheme has been developed for general

two-dimensional unstructured grids with quadrilateral cells. The proposed scheme
combines fourth-order and third-order accurate stencils to maintain high order of
accuracy in smooth regions of the flow while maintaining essentially non-oscillatory
behaviour near strong shocks. The one-step ADER time update maintains high
temporal accuracy and also decreases the number of communications between
processors. The code has been shown to scale well up to thousands of processors.
Numerical tests showing the robust shock capturing capabilities and the ability to
capture intricate flow features have been presented. Future work concerns extension
of the scheme to three dimensions and to include physical effects like multiple phases
and viscosity.
Fig. 4 Evolution of flow

through numerical Schlieren
for Mach 10 shock
diffraction from a circular
cylinder. The vortices
formed in the wake region
can be clearly seen
Acknowledgements The entire WENO implementation on unstructured grids was performed in

the open-source framework deal.II [8].
References
1. Zhao F, Pan L, Wang S Weighted essentially nonoscillatory scheme on unstructured quadrilateral

and triangular meshes for hyperbolic conservation laws. arXiv:1712.08317 [math.NA]
2. Dumbser M, Boscheri W, Semplice M, Russo G (2017) Central WENO schemes for hyperbolic
conservation laws on fixed and moving unstructured meshes. SIAM J Sci Comput 39(6):A2564–
A2591
3. Hu C, Shu C-W (1999) Weighted essentially non-oscillatory schemes on triangular meshes. J
Comput Phys 150:97–127
4. Barth T, Frederickson P High order solution of the Euler equations on unstructured grids using
quadratic reconstruction. AIAA Paper No. 90-0013
5. Toro EF (2009) Riemann Solvers and numerical methods for fluid dynamics, 2nd ed. Springer
6. Balsara DS (2017) Higher-order accurate space-time schemes for computational astrophysics,
Part I: finite volume methods. L Rev Comput Astro 3:2
7. Lax PD, Liu XD (1998) Solution of two-dimensional Riemann problems of gas dynamics by
positive schemes. SIAM J Sci Comput 19:319–340
8. Bangerth W, Hartmann R, Kanschat G (2007) deal.II—a general purpose object oriented finite
element library. ACM Trans Math Softw 33(4):24/1–24/27
Experimental Investigation of Heat
Transfer Enhancement on Impinging Jet
Cooling Using Surface Laser Treatment
Maheandera Prabu Paulraj, Santosh Kumar Sahu,

Avadhesh Kumar Sharma, Anuj Kumar, Karthikeyan Sengadurai,
and I. A. Palani
1 Introduction
The impinging jet cooling process is widely used because of capability of higher heat
transfer rate. Air, water and nanofluids are used for senior temperature reductions.
In general, the water liquid is commonly used in many engineering industries for
sudden cooling process such as meal cooling process, glass industries, electronic
cooling systems and fuel road cooling in nuclear power plant to avoid loss of coolant
accident (LOCA) [1–3]. The LOCA occurs due to emergency shutdown on primary
unit because of this the fuel rod gets overheated. In order to avoid overheating of
fuel rod, the impinging jets are used to control the temperature of fuel rods. In
this emergency condition, the liquid water jets directly impinged to the fuel rods
to reduce the temperature of the rod. The contact between liquid water and the hot
surface creates vapour film is induced on the hot surface of the rods. It is necessary
to predict the heat transfer rate on hot surface due to impingement cooling in order
to control the cooling of nuclear fuel rods and avoiding loss of cooling accidents
(LOCA). The researchers Sahu et al. [4], Agerwarl and Sahu [5] investigated and
reported the rewetting phenomena for circular impinging jet on hot vertical roads and
plates, respectively their studies. Also the surface heat flux for various parameters
has been studied. Infrared thermal camera was utilized to capture the temperature
M. P. Paulraj (B) · S. K. Sahu · A. K. Sharma · A. Kumar

Fluid Mechanics and Machinery Lab, Discipline of Mechanical Engineering, Indian Institute of
Technology Indore, Simrol, Indore 453552, MP, India
e-mail: prabusita@iiti.ac.in
S. K. Sahu
e-mail: sksahu@iiti.ac.in
K. Sengadurai · I. A. Palani
Mechatronics and Instrumentation Lab, Discipline of Mechanical Engineering, Indian Institute of
Technology Indore, Simrol, Indore 453552, MP, India
e-mail: palaniia@iiti.ac.in
572 M. P. Paulraj et al.
variation during jet impingement cooling process by earlier researchers [6, 7]. The
similar procedure is followed in the present experimental investigations as well. The
effect of Reynolds number and nozzle to plate distance on heat transfer on downward
facing jet was described by various researchers [1, 8, 9]. The researchers reported
that by increasing nozzle to plate distance, the heat transfer decreased gradually [1].
Issa [8] conducted experiment for quenching process using oil for jet impingement
on bottom side of hot SS sheet and reported that the plate-to-jet distance affects
the heat transfer rate on hot sheet due to impalement cooling. The author reports
that the cooling rate increased when jet to plate distance decreased due to effect of
acceleration against in gravity force with momentum of water liquid [9]. The higher
heat transfer at stagnation area was noticed and reported for the case of plate to jet
space about six times the diameter of nozzle [10].
The ultimate aim of the present study is to find the unsteady heat transfer
phenomena of vertical impinging water jet on hot stainless steel (SS 304) plate.
Also the effect of artificially roughness by laser treatment on unsteady heat transfer
rate has been analysed. The temperature contour and profile for both smooth and
roughened surface have been predicted and compared for various parameters such
as Reynolds number (Re) and laser power (P).
2 Methodology
The actual experimental resource developed for the present analysis is illustrated in
Fig. 1.
The experimental set-up is having various components such as personal computer,
auto transformer, collector tank, gear pump, coolant tank, gear box, water nozzles,
stepper motor, run-out table, test specimen, infrared camera, computer cables, trans-
former cables, stand and water tubes. Stainless steel SS304 (matt finish) (used with
the length of 250 mm, width 60 mm and thickness 0.25 mm) is taken for experimental
analysis. The roughness average value of present thickness plate is Ra = 0.081 µm.
The experimental sheet is hold tightly along the length direction and two copper
bars used to clamp the SS sheet in both sides with power cables. The whole set-up
is fixed between the L shapes of frames. In order to avoid short circuit on clamping
section, four Teflon bushes are used two in each sides for pass up the power between
bolt and L-shaped frame. The entire assembly section is placed on the run-out table.
The infrared camera, FLIR camera, A655sc model having capable to measure the
temperature ranges from 0 to 2000 °C are used, and it can capture with scanning
speed of 200 frames/s. Due to fine thickness of the plate, the temperature drop along
normal direction is negligible and the temperature is found to be less than 0.5 °C.
Similarly, in past studies, researchers neglected the temperature gradient along the
lateral direction was neglected for 0.15 mm thickness SS 304 sheet of their exper-
iments. Also the emissivity of the plate has been calculated separately [3, 6], and
researchers taken 0.95 emissivity value for their entire calculations. In the present
Experimental Investigation of Heat Transfer Enhancement … 573
1 Personal Computer (P.C) 8 Stepper Motor
2 Auto Transformer 9 Run-Out Table
3 Collector Tank 10 Test Specimen

4 Gear Pump 11 Infrared Camera
5 Coolant Sump 12 Computer Cables
6 Gearbox 13 Transformer Cable
7 Water Nozzles 14 Stand
Fig. 1 Model of experimental set-up
study is also based on the emissivity value 0.95. The uncertainty of the temperature
is found to be +2 °C.
The test specimen is fixed horizontally on L-shaped frame and the thermal camera
placed on above the test specimen at 0.5 m distance. The IR camera captured the
data which is extracted by using Infrared V3.3 evaluating computerized software.
The water is impinged by using micro-gear pump, GJ-N27-DEEELE and the gear
pump are controlled via 12 V Arduino-UNO system. The aluminium tube is used
as nozzle with 1 mm inner diameter for water flow. The auto transformer is utilized
to heat the specimen up to 300 °C, and the power is supplied with steady in range.
The coolant (water) is supplied through the 1 mm aluminium nozzle via control unit.
The thermal camera is activated once the temperature reached 300 °C; the camera
captures the thermal images and it is saved as video with the dimension 640 by 480
pixels for all the cases. The temperature range +10 °C is considered as the human
error. The test has been conducted for varied rages from 5000 to 10,000. The plate
to nozzle distance (z/d) 10 is taken as constant for all the cases.
The transient heat flux (qt ) for different spatial locations is obtained by Eq. (1)

V dT
qt = −ρC (1)
A dt
where ρ, C, V, A represents the density, specific heat, projected area and volume of
the sheet, respectively.
Here the density and specific heat are dependent on temperature, and they can be
calculated by using Eqs. (2) and (3)
ρ = 9 − 9.5 × 10−4 Ts + 4.665 × 10−7 Ts (2)
C = 365.45 + 0.4065(Ts + 273) − 1.73 × 10−4 (Ts + 273) (3)
The surface heat transfer can be evaluated by using Eq. (4)

qt
h= (4)
Ts − T f
Reynolds number (Re) is described as
ρU d
Re = (5)
μ
where Ts and T f represent surface temperature and coolant temperature respectively.

Also, U, d, μ represent jet velocity, nozzle diameter and viscosity, respectively.
2.2 Investigation Parameters
The parameters used for the present investigation are given in Table 1.
Table 1 Operating parameters

Parameter Operating range
Standoff distance (z/d) 10
Jet diameter (d), mm 1
Coolant inlet temperature, °C 28
Initial plate temperature, °C 300 ± 10
Coolant thermal conductivity, W/mK 0.62
The present experimental study performed for various Reynolds number (Re = 5000–
10,000) and various surface roughness (smooth and laser treated roughness profiles)
of SS plates and nozzle to plate distance (z/d) 10 is taken.
The effect of Reynolds number and effect of surface roughness on bottom jet
impingement cooling process have been reported via temperature contours and profile
(T ), heat transfer coefficient (h). The results are measured by using IR camera with
the speed of 100 f/s, and the time step of 0.01 s and using computer with Infrared V3.3
evaluating software the IR image dates are extracted. Figure 2a depicts the transient
cooling behaviour of bottom impinging jet on hot surface. The results show that the jet
moves along the radial direction with respect to time. It is clearly visible that the flow
and heat transfer along radial direction of higher Re are always higher than that of
lower Re, and it occurs due to higher momentum effect on flow and heat transfer. The
effect of roughness on temperature contour for case Re = 5000, t = 1.0 s, laser power
30 W with speed 30 bits/ms, is shown in Fig. 2b. From figure, higher temperature
distribution is noticed on roughened surface compared with smooth surface. It may
owe fluid flow resistance of rough surface, and this leads to increase contact between
coolant and hot surface. Figure 3 shows the transient cooling profile at various radial
locations for smooth hot plate Re = 5000, z/d = 10. From the figure, it is observed
that the sudden decrement in temperature is noticed at stagnation point (r/d = 0).
Also the cooling time is gradually increased along the radial locations, and it may
be due to wetting delay along the radial direction. This wetting delay is due to loss
of momentum in impinging jet along the radial axis.
Surface heat flux for smooth surface and roughened surface is shown in Fig. 4a,
b. Figure denotes the behaviour of impinging jet on surface heat flux for various
surfaces. At the stagnation point, the surface heat flux is higher for both of the cases.
Along the radial direction, the heat flux value is gradually decreased. Due to delay in
fluid flow, the cooling time is increased along the radial direction. The heat flux value
of roughened surface (Fig. 4a) is higher compared with smooth surface (Fig. 4b) at
all the radial locations (r/d). Due to roughness surface, the contact between coolant
water and hot surface is increased, and it leads to increase of the heat transfer rate.
4 Conclusions
Experimental investigation is carried out in order to predict the transient heat transfer
characteristics of bottom jet impingement cooling on smooth and roughened surface
for Re = 5000–10,000, z/d = 10, P = 35 W, s = 30 bits/ms. From the experimental
results, conclusions are—(i) the circular wet patch is noticed in both smooth and
roughness cases, (ii) higher heat transfer occurs at stagnation points, (iii) higher heat
Time, T, Re=5000 Re=10000

S ºC
t=0
t=0.30
t=0.60
t=1.40
t=2.0
(a)
Smooth surface for Rough surface for
Re=5000 at t=1.0s P=30W, speed, S=30,
Re=5000 at t=1.0s
(b)
Fig. 2 a IR image thermal contour for smooth surface during transient cooling of hot plate,
b comparison of smooth and roughness surfaces
Fig. 3 Temperature profile z/d=10, Re=5000

along the radial direction 300
r/d
Surface Temperature, T( C)
(r/d) for smooth surface, Re 0
= 5000, z/d = 10 250 1.75
3.5
200 5.25
7
8.75
150
100
50
0
0.0 0.5 1.0 1.5
Cooling time, t(s)
flux value is noticed on roughened surface, and it may be concluded that rough-
ness surface gives higher heat transfer rate compared to smooth surface on bottom
impingement cooling.
Fig. 4 a Surface heat flux

along radial direction for
smooth surface for Re =
5000, z/d = 10, b surface
heat flux along radial
direction for roughened
surface fro Re = 5000, z/d =
10, P = 35, s = 30 bits/ms
Acknowledgements One of the author, P. Maheandera Prabu, sincerely thanks and acknowledges
to the Science and Engineering Research Bored (SERB)-DST, Government of India, for the award
of (Young Scientist award) National Post-doctoral Fellowship and Financial Support (File No:
PDF/2017/000151).
References
1. Stevens J, Webb BW (1991) Local heat transfer coefficients under an axisymmetric single-phase
liquid jet. ASME J Heat Transf 113:71–79
2. Paulraj MP, Sahu SK (2019) Conjugate heat transfer enhancement of laminar slot jets with
various nanofluids on an array of protruding hot sources using MPM approach. 76(4):1–22
3. Sharma AK, Modak M, Sahu SK (2018) The heat transfer characteristics and rewetting behavior
of hot horizontal downward facing surface by round water jet impingement. Appl Thermal Eng
138:603–617
4. Sahu SK, Das PK, Bhattacharyya S (2010) An experimental investigation on the quenching of
a hot vertical heater by water injection at high flow rate. Nucl Eng Des 240:1558–1568
5. Agrawal MK, Sahu SK (2014) An experimental study on the rewetting of hot vertical surface
by circular water jet impingement. Exp Heat Transf. https://doi.org/10.1080/08916152.2014.
973973
6. Modak M, Srinivasan S, Garg K, Chougule S, Agrawal M, Sahu SK (2015) Experimental
investigation of heat transfer characteristics of the hot surface using Al2 O3 –water nanofluid.
Chem Eng Process 91:104–113
7. Sharma AK, Modak M, Sahu SK (2016) Experimental investigation of rewetting during
quenching of hot surface by round jet impingement using Al2 O3 -water nanofluids. In: 2016
24th international conference on nuclear engineering (ICONE), Charlotte, NC, USA
8. Issa RJ (2011) Heat transfer performance of an oil jet impinging on a downward-facing stainless
steel plate. Therm Sci 15(2):397–408
9. Abdelsalam MY, Kamal MM, Aboelnasr M (2014) Flat surface heat transfer enhancement by
an impinging circular free water jet. Exp Heat Transf 27:276–295
10. Attalla M, Salem M (2015) Experimental investigation of heat transfer for a jet impinging
obliquely on a flat surface. Exp Heat Transf 28:378–391
Study on Pulsatile Blood Flow
in Cerebral Stenosed Artery
Pawan Kumar Pandey, Raghav Agrawal, Parimal Mukul,

and Malay Kumar Das
1 Introduction
Peripheral arterial diseases (PAD) such as atherosclerosis and aneurysm arise due to
adverse local hemodynamic conditions. Deposition of plaque leads to narrowing of
artery lumen and may cause blockage or stroke [1]. Consequences can be fatal and
may also manifest in long-term. Pulsatile flow through a stenosed artery has been
studied widely for different waveforms. Pulsatile nature of the blood flow cannot be
neglected in the study of the evolution of the vortex structures downstream from the
stenosis [2].
Physiological blood flow waveform varies in different parts of the body. Even
for similar artery, flow waveform along with anatomical structure, may vary from
patient to patient. Variations can also happen with age, gender, physiological health
condition and under the influence of several medications and drug abuse. Although
several cases and patient-specific simulations and experimental studies are conducted
to gauge the pulsatility effects. However, intrinsic variability in different patients
often thwarts the predictions. Cardiovascular complications are usually associated
with irregular lifestyle, junk foods and lack of physical activity. Recommended exer-
cises [3] also alter the flow waveform; thus, it is imperative to get better insight on
hemodynamic consequences of flow waveform variability. Many experiments and
computational studies have been done to study the hemodynamic patterns of blood
P. K. Pandey (B) · R. Agrawal · P. Mukul · M. K. Das

Department of Mechanical Engineering, Indian Institute of Technology Kanpur, Kanpur, India
e-mail: pawansut@iitk.ac.in
R. Agrawal
e-mail: araghav@iitk.ac.in
P. Mukul
e-mail: parimal@iitk.ac.in
M. K. Das
e-mail: mkdas@iitk.ac.in
582 P. K. Pandey et al.
flow in stenosed artery. Usually, such studies are based on some patient specific
flow waveform with single-phase blood flow model. Although some studies are also
conducted using multiple flow waveforms, still there is not much clarity on the effects
of waveform variability.
High pressure difference, wall shear stress across the constriction and recircula-
tion after stenosis are usually considered as critical hemodynamic indicators. While
pressure and wall shear stresses may cause anatomical changes or rupture, recircu-
lation or vortex structure may induce other complications due to increased particle
residence time. Spatially varying stresses may cause damage to endothelium cell
layer.
Major objective of present work is to study hemodynamics of stenosed artery
subjected to inlet flow waveform which is similar to physiologically realistic condi-
tions. The waveform is reproduced using Fourier series and simulations are repeated
for each individual Fourier term imposed over mean flow velocity. An attempt is made
to recognize the important Fourier modes which, if simulated standalone, can give
similar results or behavioral pattern to original results. Comparative analysis is based
on vortex structures’ strength, count and position.
2 Methodology
The present study is carried out by simulating blood flow in moderate and
severe stenosed cerebral arteries [4] shown in Fig. 1. Blood flow is modeled
as laminar, isothermal, incompressible, single-phase flow with constant density
(1050 kg/m3 ) and Newtonian viscosity (0.0035 kg/ms) [5, 6]. An unsteady, unstruc-
tured finite-volume method-based collocated solver is developed and used. A
waveform, mimicking the realistic physiological pulsatile flow, is imposed at the
inlet. While neumann condition is imposed at the outlet, no-slip condition is
imposed on the arterial wall.
Fig. 1 Three geometries with stenosis; diameter ratio is 40% in first and 20% in second and third
geometry
Study on Pulsatile Blood Flow in Cerebral Stenosed Artery 583
Fig. 2 Flow waveform and

its Fourier components
(imposed over waveform
mean)
2.1 Waveform and Its Fourier Components
A short time-period flow waveform, used in simulations, is shown in Fig. 2. Time

period of waveform is 0.45 seconds which is similar to exercise conditions. A 5-mode
Fourier series, Eq. (1), is used to reproduce full waveform. Flow is simulated for all
Fourier components imposed over the waveform mean. Fourier modes imposed over
waveform mean are also shown in Fig. 2.
v(t) = a0 + a1 cos(wt) + b1 sin(wt) + a2 cos(2wt)

+ b2 sin(2wt) + a3 cos(3wt) + b3 sin(3wt)
+ a4 cos(4wt) + b4 sin(4wt) + a5 cos(5wt)
+ b5 sin(5wt)
a0 = 0.4248; a1 = −0.06805; b1 = 0.1644;
a2 = −0.08076; b2 = −0.00885; a3 = −0.04299; b3 = −0.0357;
a4 = 0.01435; b4 = −0.02632; a5 = 0.01557; b5 = −0.005171; w = 14.77;
(1)
2.2 Numerical Computations
The flow in the artery is modeled using the 2D Navier–Stokes equation, see Eqs. (2)
and (3). Numerical computations are done using SIMPLE like algorithm based on
pressure smoothing.
∇.V = 0 (2)

∂V
ρ + (V.∇)V = −∇P + μ∇ 2 V (3)
∂t
Validation and Grid Independence

The solver is well validated, and for this work, we did validation by simulating
flow in channel with backward-facing step and calculating the re-attachment length.
Our results are within 1% of values reported in [7]. After the grid independence
test, grid size of 10,000 triangular elements is used for current work simulations. At
chosen mesh size, the grid sensitivity for primitive variables is reduced to around 1%.
Sensitivity is calculated by comparing the velocity profile in the lumen downstream of
stenosis. Time step size is chosen initially as 10−4 s. Results are saved and analyzed
at the time interval of 10−2 s. Each case is simulated for at least 12 cycles to get
dynamically converged results.
Simulations are done for stenotic arteryy, subjected to each waveform shown in
Fig. 2. Effect of individual Fourier components of waveform is found to be heavily
influenced by their amplitude and frequency. High-frequency Fourier terms increase
the phase decoupling of inertia and pressure of the flow [8]; this makes the flow more
susceptible to form recirculation in the artery during minimum inertia of pulsatile
cycle. Amplitude of each mode contributes in instantaneous inertia. Amplitude of
each Fourier mode is the critical factor in recirculation formation at potential location.
Fourier mode’s frequency values govern the cut-off Reynolds number values for
the recirculation formation along with recirculation length. Results in Fig. 3 show
number of recirculation varying with time for different inlet flow conditions in case1
geometry. Largest number of recirculation occurs for full waveform, but its trend is
largely captured by the first mode sine function. Relative effects of individual mode’s
amplitude are evident from incoherence between the results of first cosine mode and
full waveform. With increasing frequency, the number of recirculation diminishes.
Although in present case no vortex shedding is observed, but merger, breakup,
growth, diminish and oscillation of certain vortex do happen with varying magnitude.
All these effects are captured by tracking the centers of recirculation. Tracking of
recirculations’ centers is done by using a method based on the streamlines. Because of
multiple recirculations merging, splitting and oscillations happening, it is an arduous
task, so we identify certain prime locations for tracking. All recirculation in these
locations are either originated or merged into one big recirculation at any instant of
the pulsation cycle. Figure 4 shows these vortex cluster locations for 40% stenosis
Fig. 3 Number of
recirculation forming
downstream of stenosis.
Results are shown for case 1
geometry and waveforms as
shown in the figure legend
Fig. 4 Vortex clusters in

downstream region of
stenosis
artery (case 1 geometry) simulated with full waveform. Oscillations of recircula-

tion centers decrease with downstream direction. The first recirculation appears for
complete time-period of waveform as shown in Figs. 5 and 6. Most of the vortex
clusters activate only for the half time-period of flow waveform. Time-periods of
recirculation formation at these locations are also decreasing with distance from the
stenosis. Similarity between Figs. 5 and 6 supports the primacy of Fourier term with
largest amplitude and smallest frequency to capture major haemodynamics of flow
waveform. Multiple data points of same recirculation cluster, for same time instant,
represent splitting and merging of one big recirculation.
Oscillatory velocity of recirculation centre is calculated, and found to be much
smaller compared to inlet centreline velocity. Extent of oscillation of recirculations
Fig. 5 Variation of vorticity

with time. Results are
reported for oscillating
vortex centres found in
simulation results of 40%
stenotic artery with full
waveform imposed at the
inlet
Fig. 6 Variation of vorticity

with time. Results are
reported for oscillating
vortex centres found in
simulation results of 40%
stenotic artery with first sine
mode plus waveform mean
imposed at the inlet
is shown in terms of reattachment point, for different time instants, in Fig. 7. Odd-
positioned symbol indicates start of a recirculation and even-positioned symbol indi-
cates its end. For clarity, results are plotted at 40° phase intervals. Distance between
a pair of symbols represents the size of recirculation. It is observed that amplitude
of Fourier modes affect the oscillation of recirculations. As frequency increases and
amplitude decreases, recirculations do not change their locations anymore.
Fig. 7 Re-attachment points

at different time instants.
Shown result is for 40%
stenotic artery with full
waveform imposed at the
inlet
4 Conclusions
Simulations are done for full waveform and its Fourier components imposed at the
inlet of an artery with stenosis. Fourier components with base frequency and largest
amplitude largely govern vortex dynamics characteristics. Amplitude of individual
sine or cosine terms determine the influence of single Fourier mode over behaviour of
complete waveform. Oscillation of recirculations decreases with increasing distance
from stenosis. In case of low amplitude and high frequency, size variation of
recirculation zones decreases.
Acknowledgements We gratefully acknowledge the computing facilities provided by Indian

Institute of Technology Kanpur.
References
1. Murray CJL, Lopez AD (1996) The global burden of disease: a comprehensive assessment of
mortality and disability from disease, injuries, and risk factors in 1990 and projected to 2020.
Harvard School of Public Health, Boston, Mass, USA
2. Siouffi M, Deplano V, Pélissier R (1998) Experimental analysis of unsteady flows through a
stenosis. J Biomech 31:11
3. Myers J (2003) Cardiology patient pages: exercise and cardiovascular health. Circulation 107:e2–
e5
4. Bartlett ES, Walters TD, Symons SP, Aviv RI, Fox AJ (2008) Classification of carotid stenosis
by Millimeter CT angiography measures: effects of prevalence and gender. Am J Neuroradiol
29(9):1677–1683
5. Young DF (1968) Effects of a time-dependent stenosis on flow through a tube. J Eng Ind Trans
ASME 90:248
6. Forrester JH, Young DF (1970) Flow through a converging diverging tube and its implications
in occlusive vascular disease. J Biomech 3:297
7. Biswas G, Breuer M, Durst F (2004) Backward-facing step flows for various expansion ratios
at low and moderate reynolds numbers. J Fluids Eng 126
8. Pandey PK, Das MK (2016) Three dimensional simulation of oscillatory flow in partial porous
tube. In: 6th international and 43rd national conference on fluid mechanics and fluid power.
arXiv:2104.13165 [physics.flu-dyn]
Heat Transfer and Fluid Flow
Characteristics of a Heated Vibrating
Square Cylinder
Mohammad Athar Khan, Syed Fahad Anwer, and Saleem Anwar Khan
1 Introduction
Vortex-induced vibration has always been the topic of interest for researchers because
it finds its applications in various engineering fields such heat exchanger, mechanical
system and aerospace engineering, chemical reactors, cooling towers and extended
surface [1–4]. Circular and square cylinder are two most extensively studied bluff
body shapes. Square cylinder has fixed separation points at its fixed corners while
circular cylinder has their separation points on its continuous curvature at the down-
stream side of the body. Review paper of Sarpkaya [5] and Bearman [6] gives the
detailed description of vortex-induced vibration.
To the best of author’s knowledge, heat transfer characteristics of an elastically
mounted square cylinder are not available. Therefore, in this work, heat transfer from
square cylinder is studied under the various reduced velocities ranging from 3 to 8.
Along with the heat transfer characteristics lift and drag coefficients, coefficients of
moment and study of lock-in regime have also been conducted.
2 Methodology
The dimensionless continuity, momentum and energy equation in body fitted coordi-
nate system have been shown in Eqs. (1)–(4). While Eq. (5) shows dimensionless form
of momentum equation of freely vibrating cylinder by considering only transverse
oscillations.
M. A. Khan · S. F. Anwer (B) · S. A. Khan

Computational Aerodynamics Laboratory, Department of Mechanical Engineering, ZHCET,
AMU, Aligarh, India
e-mail: sfahadanwer@zhcet.ac.in
590 M. A. Khan et al.

∂ ∂ ∂ ∂
ξx + ηx U + ξy + ηy V =0 (1)
∂ξ ∂η ∂ξ η∂

∂U ξ
ξ ∂U
η
η ∂U ∂p ∂p 1 2
+ U − Ug + U − Ug = − ξx + ηx + ∇ U
∂τ ∂ξ ∂η ∂ξ ∂η Re
(2)

∂V ∂V η ∂V ∂p ∂p 1 2
+ U ξ − Ugξ + U − Ugη = − ξy + ηy + ∇ V
∂τ ∂ξ ∂η ∂ξ ∂η Re
(3)
∂θ ξ ∂θ ξ ∂θ 1
+ U − Ugξ + U − Ugξ = ∇2θ (4)
∂τ ∂ξ ∂η Re. Pr
d2 y 4π dy 4π 2 y 1
2
+ + 2 = CL (5)
dt Ured dt Ured 2Mred
In dimensionless vibration equation, parameters reduced mass (M red = m/ρ f D2 )

denote the mass ratio while reduced velocity (U red = U/f o D) scales the natural
frequency of the cylinder where m = mass per unit length of cylinder, ρ f = density
of fluid medium, D = diameter of cylinder, f o = natural frequency of cylinder, and U
= free stream velocity. No slip and no penetration boundary condition is applied on
cylinder surface. The temperature of cylinder is set to be unity. At outflow, convective
boundary condition for velocity is employed. Square cylinder diameter D is taken as
length scale, free stream velocity U is taken as velocity scale, and D/U is time scale.
The detailed description of mesh movement algorithm and numerical scheme was
used in the present study similar to that used in the work of Anwer et al. [7] (Fig. 1).
For the purpose of validating the solver, VIV of a square cylinder for isothermal
flows at two Reynolds number was investigated. Table 1 shows that the present results
are in good agreement with the results of Sanjay mittal [8].
In Fig. 2, amplitude of cylinder vibration for reduced velocities (3–8) has been
shown. It can be clearly observed from the figure that maximum cylinder displace-
ment is found at reduced velocity 5. At reduced velocity 3, cylinder displacement is
minimum. Figure 3 shows the vortex structure at U red = 5, and the vortex shedding
mode is found to be 2S mode of vortex shedding. For all reduced velocities (3–8),
vortex shedding mode is 2S. In Fig. 4, isotherm contour for U red = 5 also shows the
same pattern as vortex pattern. Figure 5 shows the variation of Nusselt number with
reduced velocities, and maximum Nusselt number is found for U red = 5.
Figure 6 shows the variation of peak value of coefficient of lift with different
reduced velocities and at U red = 5 peak value of C L is maximum. At U red = 5,
Heat Transfer and Fluid Flow Characteristics … 591
Fig. 1 Body fitted grid

around cylinder
4
y
-2
-4
-6
-5 0 5
x
Table 1 Comparison with

Maximum displacement at reduced mass (m* = 1)
earlier works
Re Sanjay Mittal [7, 8] Present
61 0.3032 0.345
100 0.105 0.109
Fig. 2 Variation of cylinder 0.8 Ured = 3

displacement with different Re = 60 Ured =4
reduced velocities Ured =5
0.6 Ured =6
Ured =7
Ured =8
0.4
0.2
Y
-0.2
-0.4
380 385 390 395 400

τ
Fig. 3 Vortex structure for Ured = 5

reduced velocity 5 at Re =
60 Re = 60
Fig. 4 Isotherm pattern for Ured = 5

reduced velocity 5 at Re =
Re = 60
60
Fig. 5 Variation of Nusselt 9

number with time Ured = 3
Re = 60 Ured = 4
Ured = 5
Ured = 6
Ured = 7
8.5 Ured = 8
SC
Nu
7.5
7
380 385 390 395 400
τ
displacement of cylinder and of coefficient of lift is in phase with each other. Same
result is obtained for drag coefficient as shown in Fig. 7 that drag coefficient is
maximum for U red = 5.
Figure 7 shows the ratio of oscillation frequency (f ) of cylinder to the natural
frequency of cylinder (f o ). The phenomenon of lock-in regime occurs when ratio of
Fig. 6 Variation of 1.2

coefficient of lift with
different reduced velocities Re = 60
1
0.8
CL
0.6
0.4
0.2
0
2 3 4 5 6 7 8 9
Ured
Fig. 7 Variation of drag

coefficient with different Re = 60
reduced velocities 2.2
2
CD
1.8
1.6
1.4
2 3 4 5 6 7 8 9
Ured
oscillation frequency to the natural frequency of cylinder becomes equal and cylinder
vibrates at greater amplitude at this stage. Here from Fig. 8, narrower lock-in regime
can be seen between U red 8 and 9.
Fig. 8 Response of reduced 1.5

frequency with different Re = 60
reduced velocities
f/fo0.5
0
2 3 4 5 6 7 8 9 10
Ured
4 Conclusions
Numerical simulation of vortex-induced vibration of square cylinder at Reynolds

number 60 is carried out for the range of reduced velocities 3–8. Maximum amplitude
of cylinder is found at reduced velocity 5. Nusselt number is found to be maximum
for reduced velocity 5. Vortex shedding mode for all reduced velocities included in
the study shows 2S mode of vortex shedding. A narrow lock-in regime is found in
between U red 8 and 9.
References
1. Patel CG, Sarkar S, Saha SK (2018) Mixed convective vertically upward flow past side-by-side
square cylinders at incidence. Int J Heat Mass Transf 127:927–947
2. Rastan MR, Sohankar A, Alam MdM (2017) Low-reynolds-number flow around a wall-mounted
square cylinder: flow structures and onset of vortex shedding. Phys Fluids 29(10):
3. Yang R-J, Fu L-M (2001) Thermal and flow analysis of a heated electronic component. Int J
Heat Mass Transf 44(12):2261–2275
4. Zebib A, Wo YK (1989) A two-dimensional conjugate heat transfer model for forced air cooling
of an electronic device. J Electron Packag 111(1):41–45
5. Sarpkaya T (1979) Vortex-induced oscillations. J Appl Mech 46:241–258. https://doi.org/10.
1115/1.3424537
6. Bearman P (1984) Vortex shedding from oscillating bluff bodies. Annu Rev Fluid Mech. https://
doi.org/10.1146/annurev.fluid.16.1.195
7. Anwer SF, Hasan N, Sanghi S, Mukherjee S (2009) Computation of unsteady flows with moving
boundaries using body fitted curvilinear moving grids. Comput Struct. https://doi.org/10.1016/
j.compstruc.2008.11.002
8. Sen S, Mittal S (2015) Effect of mass ratio on free vibrations of a square cylinder at low Reynolds
numbers. J Fluids Struct
Development of Solver for Vibrating
Bluff Bodies with Different Geometrical
Shapes
Md.Abu Shahzer, Mohd.Athar Khan, and Syed Fahad Anwar
1 Introduction
Vortex-induced vibrations (VIV) of structures are often found in many engineering

fields, ranging from oscillations of suspension bridges in the wind to vibrations
of sea elevators in ocean currents and from aerospace to mechanical systems. Due
to the large repeated load and high stress and strain, these vibrations can cause
fatigue type of failure of the structures. In order to reduce these failures, it is very
essential to develop techniques for the reduction of VIV. And, for this purpose,
a thorough understanding of inter-linkage between the flow and similar types of
structures must be required. The vortex shedding (oscillating flow) which takes place
when a fluid flows past a bluff body due to the flow separation at points along the
body. The periodic forces such as lift, drag, and moment are generated due to the
shedding vortices which in turn can induce vibrations in transverse, in-line, and in
the azimuthal directions. So by solving the Navier–Stokes equations subjected to
the boundary conditions, the forces and moment are calculated which further act as
forces and moment in forced vibration equations and finally response amplitudes are
calculated.
Literature shows that the study of VIV and VIR of bluff bodies is not new. Bishop
and Hassan [1], Feng [2], Griffin and Ramberg [3], Williamson and Roshko [4] and
Koopmann [5] were the first researchers that have investigated the vortex-induced
vibration of a cylinder and have demonstrated that the vortex-induced vibration is a
self-limiting motion. The investigation of 1-DOF VIV of a circular cylinder is carried
out extensively in the previous studies [6–8]. Singh and Chatterjee [9] numerically
M. Shahzer · M. Khan · S. F. Anwar (B)

Computational Aerodynamics Lab, Department of Mechancal Engineering, ZHCET, Aligarh
Muslim University, Aligarh 202002, India
e-mail: sfahadanwer@zhcet.ac.in
M. Shahzer
e-mail: md_abushahzer@zhcet.ac.in
598 Md. Abu Shahzer et al.
Fig. 1 Problem definition: boundary conditions and schematic of flow domain of a bluff body with
both transverse and rotational DOF
studied the 2-DOF VIV of a single-cylinder with different shear parameters. Ryu
and Iaccarino [10] numerically studied the vortex-induced rotations of a rigid square
cylinder at 45 ≤ Re ≤ 150. Kumar et al. [11], Rao et al. [11] and Bao et. al [12] have
observed different vortex shedding modes. Wang et al. [13] have reported the study
of VIV of an elliptic cylinder with both transverse and rotational degrees of freedom.
It can be observed from the literature review that the study of VIV incorporating the
VIR is not investigated systematically.
The focus of this work is to develop the numerical solver to compute the VIV
responses of a bluff body with both transverse and rotational degrees of freedom.
Structural properties of the cylinder such as mass and stiffness were translated into
two non-dimensional parameters: reduced mass and reduced frequency, respectively.
Figure 1 represents the problem definition in which springs and dampers are mounted
elastically on the bluff body. Figure 1 also shows the direction of flow and boundary
conditions. In the present study, the circular, square, and elliptic cylinders are taken
as bluff bodies for the numerical investigation purpose.
2 Methodology
To develop the numerical solver, a clear and distinctive formulation of the Navier–
Stokes equations and rigid body motion equations (vibrational equations) in proper
form is required which are solved by using finite difference type of discretization over
a structured body-fitted curvilinear non-orthogonal mesh. The governing equations
Development of Solver for Vibrating Bluff Bodies with Different Geometrical Shapes 599
and the numerical scheme used for the investigation of moving boundary flows are
elaborated in this section.
The flow is governed by the continuity and momentum equations for an incompress-
ible fluid. The non-dimensional continuity and Navier–Stokes equation for laminar,
unsteady, incompressible, viscous flow can be expressed as:
∇.u = 0
∂u 1 2
+ u.∇.u = −∇ p + ∇ u
∂t Re
where u and p are the velocity and pressure of the flow field. Re (= UD/ϑ) is
the Reynolds number. The equivalent diameter of the cylinder D is taken as the
characteristic length, and the free stream velocity U is the characteristic velocity.
The pressure p is non-dimensionalized by ρ f U 2 .
A structured, collocated, and body-fitted O-type grid is generated using well-
known developed techniques [14]. The doubly connected physical domain (x, y, t)
is connected with the computational domain (ξ, η, t) and mapped on the rectangular
computational domain. The body-fitted coordinates ξ (x, y) and η(x, y) are chosen as
to satisfy the Laplacian equations in the physical domain given as (Fig. 2):
∇2ξ = 0 (3)
∇2η = 0 (4)
The equations (continuity and momentum) transformed into the body-fitted

coordinates given as follows:
Continuity:

∂ ∂ ∂ ∂
ξx + ηx U + ξy + ηy V =0 (5)
∂ξ ∂η ∂ξ ∂η
X-Momentum:

∂U ξ ∂U η ∂U
+ U − Ugξ + U − Ugη
∂t ∂ξ ∂η

∂p ∂p
= − ξx + ηx
∂ξ ∂η
0.7 1
0.9
0.6
0.8
0.5 0.7
0.6
0.4
y
0.5
y
0.3 0.4
0.3
0.2
0.2
0.1
0.1
0 0
0 0.2 0.4 0.6 0 0.2 0.4 0.6 0.8
x x
(a) (b)
1
0.9
0.8
0.7
0.6
0.5
y
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8
x
(c)
Fig. 2 Enlarged view of the grid near the bluff bodies a elliptic cylinder, b circular cylinder, and
c square cylinder
2
1 ∂ U ∂ 2U ∂ 2U ∂U ∂U
+ α 2 − 2β + γ 2 + ∇2ξ + ∇2η (6)
Re ∂ξ ∂ξ ∂η ∂η ∂ξ ∂η
Y-Momentum:

∂U ξ ∂V η ∂V
+ U − Ugξ + U − Ugη
∂t ∂ξ ∂η

∂p ∂p
= − ξx + ηy
∂ξ ∂η
2
1 ∂ V ∂2V ∂2V 2 ∂V 2 ∂V
+ α 2 − 2β +γ 2 +∇ ξ +∇ η (7)
Re ∂ξ ∂ξ ∂η ∂η ∂ξ ∂η
The coefficients which are appearing in the above equations are same as calculated
by Anwer et al. [15].
The motion of a solid body vibration in fluid can be described by a general
dynamics equation of an elastically mounted body. The equation of the body
dynamics in non-dimensionalized form can be written as:
d 2 Qi 4π ξi d Q i 4π 2 1
2
+ + 2 Qi = Ci (8)
dt Ured,i dt Ured,i 2Mred
Q i is the displacement vector where i = 1, 2, 3 denotes translational displacements

x and y and angular displacement θ with respect to the mass center of the body. Q i is
non-dimensionalized by the cylinder equivalent diameter D in the current study and t
is non-dimensionalized by D/U. On the right hand side, Ci is the fluid force coefficient
in the unit length. ξi , Ured,i and Mred,i are damping ratios, reduced velocities and mass
ratio, respectively, and Ired,θ is the moment of inertia ratio. For details refer to Wang
et al. [13].
2.2 Numerical Scheme
For the purpose of capturing the unsteady physics of the flow, we employ a semi-
implicit, pressure correction type of scheme, on a non-staggered structure body-fitted
grid mesh using a finite difference type of spatial discretization. The concept of
momentum interpolation of Rhie and Chow [16] is utilized in order to avoid grid-
scale pressure oscillation that can arise owing to decoupling between the velocity and
pressure at grid points. The scheme was described by Hirsch [17] and is conceptually
similar to SMAC algorithm described by Amsden and Harlow [17]. Guided by the
works of Kim and Benson and Cheng and Armfield [18] who have demonstrated
the efficiency of SMAC scheme over SIMPLE, SIMPLEC, and PISO methods for
computing unsteady, incompressible flows, the present work uses a scheme that is
similar to SMAC. The present numerical scheme is same which is implemented
by Hasan and his co-workers [14]. The details of the scheme are spread over two
sections: time integration and spatial discretization. For details refer [15].
The numerical solver is developed using mesh movement algorithm which is
depicted in Fig. 3, is same as that of used by Anwer et al. [15].
Fig. 3 Geometric
representation of the mesh
movement algorithm
explaining different terms
For the case of circular cylinder, VIV responses are validated for cross flow with
various literature and it is found that the present results are in accordance with the
available results. At Re = 150, Mred = 2 and ξ = 0.01 the translational displacement
y is validated with the literature which is shown in Table 1.
For the case of square cylinder transverse vibration in y-direction is validated
with the literature of Sanjay Mittal [13] at various Reynolds number. The results of
maximum displacement of cylinder in y direction are in good agreement with the
available literature at Mred = 1, and ξ = 0 and it is shown in Table 2.
Table 1 Comparison of results of ymax displacement of circular cylinder with available literature
Ur = 3 4 5 6 7 8
Present study 0.069 0.572 0.556 0.486 0.403 0.081
Wang et al. [19] 0.073 0.575 0.539 0.480 0.399 0.080
Bao et al. [12] 0.066 0.568 0.550 0.477 0.394 0.083
Table 2 Validation of results

Maximum displacement at Mred = 1
of ymax displacement of
circular cylinder with Re Sen and Mittal [20] Present
available literature 61 0.3032 0.345
100 0.105 0.109
Table 3 Comparison of results of ymax (1DOF) of elliptic cylinder with available literature
Ur = 3 4 5 6 7 10
Present study 0.017 0.508 0.987 1.226 0.403 0.150
Wang et al. [13] 0.018 0.589 0.991 1.445 0.399 0.145
Table 4 Comparison of results of ymax (2DOF) of elliptic cylinder with available literature
Ur = 2 4 5 6
Present study 0.007 0.135 0.410 0.56
Wang et al. [13] 0.0076 0.162 0.369 0.533
Table 5 Comparison of results of θ max (2DOF) of elliptic cylinder with available literature
Ur = 2 4 5 6
Present study 0.008 0.415 0.520 0.589
Wang et al. [13] 0.010 0.326 0.40 0.567
Finally, for the case of an elliptic cylinder, the results are validated with the
literature Wang et al. at various fixed parameters such as Re = 150, Mred = 10,
and ξ = 0.001. The aspect ratio which is defined as the ratio of major semi-axis
to minor semi-axis [13] is taken as 2. The natural frequencies and damping ratios
in translational and rotational degree of freedom are same. The elliptic cylinder is
allowed to vibrate in transverse direction (y) and in azimuthal direction (θ ). The
maximum displacement at various reduced velocities of the cylinder in transverse
direction when the rotational motion is blocked is compared in Table 3 and when the
rotational motion is allowed, is compared in Table 4. Table 5 shows the comparison
of maximum angular displacement with the literature of Wang et al. [13]
4 Conclusions
The VIV responses are investigated for various types of cylinders (circular, square,
and elliptic) at various fixed parameters. The numerical solver is developed using
moving mesh algorithm and a special type of outflow boundary condition [14]. It can
be concluded that the developed solver gives results for the vibrational responses in
transverse as well as azimuthal directions are in good agreement with the available
published results. Thus, the solver can be used further for deep study in the field of
flow over bluff bodies and its VIV responses.
References
1. Bishop RED, Hassan AY (1964) The lift and drag forces on a circular cylinder oscillating in a
flowing fluid. Proc Royal Soc A Math Phys Eng Sci 277(1368):51–75
2. Feng CC (1968) The measurement of vortex induced effects in flow past stationary and
oscillating circular and d-section cylinders. University of British Columbia
3. Griffin OM, Ramberg SE (1974) The vortex-street wakes of vibrating cylinders. J Fluid Mech
66(3):553–576
4. Williamson CHK, Roshko A (1988) Vortex formation in the wake of an oscillating cylinder. J
Fluid Struct 2(4):355–381
5. Koopmann GH (1967) The vortex wakes of vibrating cylinders at low reynolds numbers. J
Fluid Mech 28(3):501–512
6. Bearman PW (2011) Circular cylinder wakes and vortex-induced vibrations. J Fluids Struct
27:648–658
7. Gabbai RD, Benaroya H (2005) An overview of modeling and experiments of vortex-induced
vibration of circular cylinders. J. Sound Vib 282:575–616
8. Sarpkaya T (2004) A critical review of the intrinsic nature of vortex-induced vibrations. J Fluids
Struct 19:389–447
9. Singh SP, Chatterjee D (2010) Impact of transverse shear on vortex induced vibrations of a
circular cylinder at low reynolds numbers. Comput Fluids 93:61–73
10. Ryu S, Iaccarino G (2017) Vortex-induced rotations of a rigid square cylinder at low reynolds
numbers. J Fluid Mech 813:482–507
11. Kumar S, Cantu C, Gonzalez B (2011) Flow past a rotating cylinder at low and high rotation
rates. J Fluids Eng 133:041201
12. Bao Y, Huang C, Zhou D (2012) Two-degree-of-freedom flow-induced vibrations on isolated
and tandem cylinders with varying natural frequency ratios. J Fluid Struct 35:50–75
13. Wang H, Zhai Q, Chen K (2018) Vortex-induced vibrations of an elliptic cylinder with both
transverse and rotational degrees of freedom. J Fluids Struct 84(2019):36–55
14. Hasan N, Anwer SF, Sanghi S (2005) On the outflow boundary condition for external
incompressible flows: a new approach. J Comput Phys 206:661–683
15. Anwer SF, Hasan N, Sanghi S, Mukherjee S (2009) Computation of unsteady flows with moving
boundaries using body fitted curvilinear moving grids. J Comput Struct 87:691–700
16. Rhie CM, Chow WL (1983) Numerical study of the turbulent flow past an aerofoil with trailing
edge separation. AIAA J 21:1525–1532
17. Hirsch C (1990) Numerical computation of internal and external flows. Wiley–Inter
science/Wiley
18. Cheng L, Armfield S (1995) A simplified marker and cell method for unsteady flows on
non-staggered grids. Int J Numer Methods Fluids 21:15–34
19. Wang C, Tang H, Duan F, Yu SCM (2016) Control of wakes and vortex-induced vibrations of
a single circular cylinder using synthetic jets. J Fluid Struct 60:160–179
20. Sen S, Mittal S (2015) Effect of mass ratio on free vibrations of a square cylinder at low
reynolds numbers. J Fluids Struct
21. Amsden AA, Harlow FH (1970) The SMAC method: a numerical technique for calculating
incompressible fluid flows. In: Los Alamos, Scientific Report, LA 4370
Floating Solid Waste Collection System
Using Free Vortex Flow
Shashikant S. Pawar, Glen L. Dsouza, Oyster J. D´souza,

and Alston R. Furtado
1 Introduction
Since past four to five decades due to the advent of industrial revolution, a substantial
increase in disposing off the waste materials in water bodies has taken place which
mainly includes plastic and its by-products. This has not only caused the water
pollution but also has affected the aquatic life to a large extent.
Attempts have been made at a much larger scale to collect and remove the floating
waste materials from the ocean [1]. The Ocean cleanup Foundation and the Seabin
Foundation are two such examples. However to collect the floating waste from the
surface of an ocean is not only difficult but also challenging. Interested readers may
visit the websites [2, 3] of above-mentioned projects for more details. Collecting the
floating waste from calm water bodies such as lakes, ponds and reservoirs is relatively
easier. The traditional methods used in these cases mainly require manual labour and
are cumbersome. The aim of the present work is to develop a system which will help
in collecting and removing the floating waste using the concept of free vortex flow.
Effectiveness of this system over the conventional methods lies in the fact that the
floating waste comes to a single collection point, which otherwise needs to be chased
and collected. Determination of strength of the vortex in such a system is crucial from
the design point of view. We found not many experimental studies on the free vortex
flows [4] and carried out with the aim of utilizing its potential, although these flows
are commonly seen in nature, e.g. tornadoes, whirlpools, etc.
Presently, we have restricted the application of this work to calm water bodies
since in case of wavy, turbulent ocean, as mentioned before, it is quite difficult to
sustain the free vortex flow of moderate strength.
S. S. Pawar (B) · G. L. Dsouza · O. J. D´souza · A. R. Furtado

Department of Mechanical Engineering, Don Bosco Institute of Technology, Mumbai, 70, India
606 S. S. Pawar et al.
2 Methodology
2.1 Experiments
Figure 1 shows the schematic diagram of the experimental setup. A plastic collector of
height 110 mm was used in the study. The collector was tapered with top and bottom
diameters of 168 mm and 106 mm, respectively. It was mounted on an adjustable
stand so that the immersion depth of the top edge of the collector from the free water
surface could be adjusted. The collector stand assembly was immersed in a tank of
diameter 600 mm and height 560 mm. At the bottom of the collector was connected
the suction of the 0.5 HP centrifugal pump. The discharge of the pump was sent back
to the tank through the ring of perforated flexible PVC pipe, kept at the bottom of
the tank so as to have minimum disturbance to the vortex flow. Rigid PVC pipes
were used for piping since pinching of the braided flexible pipes was observed at
high flow rates. Guide vanes (thin rectangular flaps), six in numbers, were fixed to
the collector top from outside so that the fluid entered tangentially into the collector.
The flap top edges were above the water surface for all the immersion depths. In the
final design, the guide vanes are proposed to be hinged at the collector surface on top
so as to allow size variation of the waste while entering the collector. Free surface
profile and the surface flow velocities were measured for five immersion depths, viz.
2, 3, 4, 5 and 6 cm. The results reported are from the measurements using clear water
without any floating waste; however, a filter of stainless steel wire mesh having the
inside shape of the collector, designed to segregate the solid waste from water, was
also used during the experiments, so as to have similar head loss across the filter
which would take place during the actual run of the system. Free surface profile
measurements were carried out by using a measuring instrument which was made
with traversing mechanisms in both the vertical and lateral directions.

of the setup
Floating Solid Waste Collection System Using Free Vortex Flow 607
2.2 Theoretical Background
The superimposition of a free vortex and a sink flow is considered for the analysis
since it represents the fluid flow in the present case closely [5, 6].
The stream function in the cylindrical polar co-ordinates then becomes,
Ψ= k. - C log r (1)
where k = − q/2π is the sink strength and C = /2π is the vortex strength, with
as the circulation.
Similarly, the velocity potential function is given by,
Φ= k log r + C. (2)
The radial and tangential velocity components at any point (r, P) are given by,
1 δ k
ur = = (3)
r δ r
and
u = = (4)
respectively. Applying the Bernoulli equation between the stagnation point and
any desired point (since the flow is irrotational),
p0 p u 2 + u 2
= + r (5)
ρ ρ 2
p − p0 (C 2 + k 2 )
=− (6)
ρ 2r 2
i.e. the differential pressure can be obtained from the radial and tangential velocity
components.
3 Results and Discussions
3.1 Free Surface Profile Measurements
Figure 2 shows the measured surface profiles of the vortex for different immersion
depths d, of the top edge of the collector from undisturbed water surface, measured
before starting the pump.
Fig. 2 Free surface profiles 0

for different immersion r
-6 -4 -2 -0.5 0 2 4 6
depths -1
-1.5
-2 2cm
-2.5
-3 3cm
-3.5 4cm
-4
-4.5 5cm
d -5
6cm
-5.5
-6
For immersion depths of more than 5 cm, the vortex core was shallow and the
eye diameter was smaller. This was primarily due to the fact that the momentum to
a deeper slab of water was required to be imparted. For lower depths, it was mainly
the near surface water which was being drawn towards the centre of the collector,
and hence, the vortex profile was deeper at the centre and larger in diameter. It was
also observed that for immersion depths smaller than 3 cm, the vortex was unsteady
and measurement of the profile was difficult.
3.2 Velocity Measurements
For surface velocity measurements, fine aluminium particles were sprinkled on the
free surface. They were illuminated by LED light, and the video was recorded by a
camera at the rate of 60 frames per second. Figure 3 shows the vortex flow visual-
ized by the streaks of the aluminium particles. The images were then extracted and
processed by the software ImageJ.
For each immersion depth, two to three particles were tracked for about 10 s,
and the radial and tangential velocities were obtained from their displacements
in the radial and tangential directions, respectively. This method works with the
linear approximation between the two frames [7]. However, it may be noted that
the maximum measured velocities in the experiments were not more than 10 cm/s;
i.e. the maximum distance the particles moved between the two frames was about
1.67 mm.
Figure 4 shows the variation of the tangential velocity uP with the radial distance
for immersion depth 4 cm. The tangential velocity is inversely proportional to the
radial distance, as is evident from Eq. (3). Blue discrete points show the measured
velocities while the red curve is a fit to the data. It may be noted that the discrete data
points are clustered between r = 2–4 cm and r = 6.5–8.5 cm.
The reason of the clustering becomes clear from the corresponding plot of radial
velocity, as shown in Fig. 5. Initially, the particle can be seen to move in and out
(indicated by positive and negative radial velocities) while orbiting around the centre,
Fig. 3 Visualization by fine

aluminium particles (d =
2 cm)
Fig. 4 Variation of
tangential velocity (d =
4 cm)
between the radial location r = 6.5–8.5 cm. It then starts moving towards the centre;
however, it again keeps orbiting around the centre between the radial location r =
2–4 cm, with in and out motion. The particle does not go further closer to the centre.
It shows therefore that the surface flow is quite different than the flow in the lower
layers.
Fig. 5 Variation of radial

velocity (d = 4 cm)
3.3 Pressure Calculation
Using Eq. (6), the differential pressure variation is calculated from the measured
radial and tangential velocities. Figure 6 shows the variation of the differential pres-
sure (blue curve). References also shown are the curve fits to the radial (black curve)
and the tangential (red curve) velocity data; however, they are scaled suitably. Since
the flow is irrotational, the reference point can be taken anywhere on the free surface,
where the flow velocities are negligibly small. The pressure at the reference point p0
Fig. 6 Pressure difference,

radial and tangential velocity
variation (d = 4 cm)
in the present case is therefore the atmospheric pressure. The curve shows that for
large radial distance the pressure p, approaches p0 .
4 Conclusions
Preliminary experimental results are presented from the study of free vortex flow
created in the laboratory. The aim was to utilize the features of this flow to collect
floating solid waste in calm water bodies like lakes and reservoirs. Free surface
profile and the surface velocity measurements are carried out. Although the measure-
ments show the velocity variations as expected from the theoretical considerations,
it appears that the surface flow is quite different than the flow in different layers
underneath. A CFD study is planned to explore the detailed flow features which is
expected to throw more light on the flow physics. To have similar head loss, while
conducting the experiments, a wire mesh filter was used similar to the one to be used
in the actual system for segregation of the solid waste. The results from this study
are therefore expected to provide inputs to the actual system design.
References
1. Sainte-Rose B, Lebreton L, de Lima Rego J, Kleissen F, Reisser J (2016) Multi-scale numerical

analysis of the field efficiency of an ocean plastic cleanup array. In: Part-1&2, ASME-2016,
35th international conference on ocean, offshore and arctic engineering, Busan, South Korea,
19–24 June 2016
2. www.theoceancleanup.com
3. www.seabinproject.com
4. Gupta S, Panda JP, Nandi N (2014) A model study of free vortex flow. In: Proceed-
ings of ICTACEM 2014, international conference on theoretical, applied, computational and
experimental mechanics, 29–31 December, IIT Kharagpur, India ICTACEM-2014/017
5. Batchelor GK (1967) An introduction to fluid dynamics. Cambridge University Press
6. Kundu PK, Cohen IM (2010) Fluid mechanics. Academic Press
7. Raffel M, Willert CE, Wereley ST, Compenhans J (2007) Particle image velocimetry: a practical
guide. Springer
DEM Simulation of Finite Size Particle
Impact on a Loose Granular Bed
B. Gokul and Dhiraj V. Patil
1 Introduction
Granular materials exhibit different behaviours when compared to fluid and

molecules. Due to particle–particle and particle–wall interactions, they dissipate
more energy. Granular materials have many such peculiarities. DEM is a method
proposed by Cundall and Strack [1], which is a model used to numerically examine the
particle–particle and particle–wall interactions. LIGGGHTS (stands for LAMMPS
Improved for General Granular and Granular Heat Transfer Simulations) is an open-
source DEM particle simulation software based on C++, distributed under the terms
of the GNU Public License [2]. LIGGGHTS supports MPI message passing library so
that parallel computing is possible on both shared and distributed memory machines.
Rock penetration [3], meteor impacts [4], ballistic impact [5], impact of mixer
blade on granular material [6] are some of the impact phenomena and range from
civilian to defence applications. Crater profile and its dependency on projectile energy
and size for low velocity impact [7] have also been studied.
So, the objectives are to study the influence of size, shape and material for various
velocities of the impactor on (a) the penetration depth, i.e. the distance to which the
impactor has penetrated into the granular bed, (b) total number of particles ejected
from the bed after the impact and (c) kinetic energy of the ejected particles, when
impacted on a loose granular bed.
B. Gokul (B) · D. V. Patil

Indian Institute of Technology Dharwad, Dharwad, India
e-mail: gokul.kbs1@gmail.com
D. V. Patil
e-mail: dhiraj@iitdh.ac.in
614 B. Gokul and D. V. Patil
2 Methodology
Computer simulations are performed through which physics of granular impact on the
substrate surface is investigated. DEM uses lagrangian method to study the particle–
particle interactions. Lagrangian method resolves individual particle trajectories and
numerically integrates the equations of motion.
2.1 Mathematical Modelling
In this research, the different types of forces considered to be acting on the parti-
cles are gravitational force, frictional force and contact forces. We make use of the
Hertzian contact model given by Eqs. (1) and (2).
F = Finj + Fitj (1)
Finj = kn δinj − γn vinj , Fitj = kt δit j − γt vit j (2)
where
Finj , Fitj are normal and tangential contact forces between particles i and j,
Here,
4 ∗ n
kn =E Rδi j (3)
3

5
γn = −2 β Sn m ∗ ≥ 0 (4)
6

kt = 8G ∗ Rδit j (5)

5
γt = −2 β St m ∗ ≥ 0 (6)
6
where
kn is the elastic constant for normal contact, kt is the elastic constant for tangen-
tial contact, γn is the viscoelastic damping constant for normal contact, γt is the
viscoelastic damping constant for tangential contact, vinj and vit j are normal and
tangential component of relative velocity for the particles i and j, δinj is the overlap
distance between the particles i and j, δit j is the tangential displacement vector between
the particles i and j which is truncated to satisfy a frictional yield criterion, E ∗ is the
DEM Simulation of Finite Size Particle Impact on a Loose Granular Bed 615
equivalent Young’s modulus, G ∗ is the equivalent shear modulus, R ∗ is the equivalent

radius, m ∗ is the equivalent mass of two colliding particles, μ∗ and e∗ are Poisson
ratio and coefficient of restitution, which are given by the following equations,
1 1 1
= 1 − μ21 + 1 − μ22 (3)
E E1 E2
1 1
= 2 2 − μ 1 + μ
G∗ E1 1 1

+ 2 2 − μ2 1 + μ2 (4)
1 1 1
∗
= + (5)
R R1 R2
1 1 1
∗
= + (6)
m m1 m2
where
E 1 and E 2 are Young’s modulus, μ1 and μ2 are Poisson’s ratio, R1 and R2 are
radius, m 1 and m 2 are mass of the two individual interacting particles, respectively,
and β, Sn and St are
ln(e)
β= (7)
ln (e) + π 2
2

Sn = 2E Rδinj (8)

St = 8G Rδinj (9)
2.2 Numerical Modelling
Discretization scheme: LIGGGHTS uses the velocity Verlet scheme by default as it

is second order accurate. The velocity Verlet algorithm is used for discretizing and
converting the equations of motion to algebraic equations. The discretization scheme
is shown below:
1
x(t + t) = x(t) + v(t)t + a(t)t 2 (10)
2
1
v(t + t) = v(t) + [a(t) + a(t + t)]t (11)
2
Table 1 Number of particles

Material properties Aluminium α-alumina particles
detected
particles
Young’s modulus 70.0 415.74
(GPa)
Poisson’s ratio 0.35 0.23
Restitution 0.85 0.85
coefficient
Friction coefficient 0.4 0.4
Density (kg/m3 ) 2700 3900
Input script details: A cylindrical geometry (0.25 m radius and 0.2 m height) is
used, which will be filled with particles and it is attached to another cylinder (4.5 m
radius and 4 m height) and obtained as a single domain, in order to avoid collision with
the boundary. The simulation domain consists of the cylindrical geometry, impactor
and the bed of particles. The particles and the impactor are in spherical shape. The
particle diameter and the impactor diameter(s) are 8 mm and 20, 30, 40, 50 mm,
respectively.
The walls of the geometry are taken as aluminium, the impactor as α-alumina and
the bed particles as aluminium. Table 1 shows the properties of the particles used in
the simulation.
The boundary conditions are given as non-periodic and shrink-wrapped with a
minimum value, in all the directions of the domain.
The initial conditions are obtained by allowing the bed particles to settle under
gravity, in the simulation domain. After that, the impactor is set at a height of 2 m
from the top surface of the bed particles. The gravitational force is acting on the
particles in the simulation domain. The domain is considered to be with negligible
air resistance. 80,000 aluminium particles are taken as bed particles. One α-alumina
particle is made to impact from a height of 2 m on the bed.
The penetration depth is calculated as the difference between the height to which
the particles are filled and the height at which the impactor reaches the zero velocity.
This is calculated from the data in the dump file.
In the calculation of Maximum KE of the ejected particles, the kinetic energy of
a single particle from the bed which has the maximum velocity after the impact is
taken as maximum kinetic energy.
The calculation of the total number of ejected particles is done by adding the
number of particles ejected from the bed, after the impact, having positive value of
z-component of velocity. As all the bed particles will be having some velocity (as
they do not reach zero in numerical methods), the particles with velocity greater than
0.1 m/s are taken as ejected particles.
A representation of the results of some of the test cases is shown here. The bed
particles are impacted with different initial velocities from 0 to 25 m/s. The snapshots
of the simulation of α-alumina impacting a bed of aluminium particles with 40 mm
diameter at 25 m/s are shown (Figs. 1, 2, 3 and 4).
The scatter plots for impactor initial velocity versus penetration depth, maximum
kinetic energy of the ejecta mass and total number of ejected particles are shown in
Figs. 5, 6 and 7.
Fig. 1 At t = 0.078 s for

α-alumina impacting at
initial velocity 25 m/s and
40 mm diameter

40 mm diameter

40 mm diameter
Fig. 4 At t = 0.09 s for α-alumina impacting at initial velocity 25 m/s and 40 mm diameter
Fig. 5 Penetration depth for initial impactor velocities from 0 to 25 m/s
Fig. 6 Maximum KE of ejecta mass for initial impactor velocities from 0 to 25 m/s
Fig. 7 Total number of ejected particles for initial impactor velocities from 0 to 25 m/s
4 Conclusions
From the given results, the penetration depth, maximum kinetic energy of the ejecta
mass and total number of ejected particles seem to follow a certain scale w.r.t the
impactor’s initial velocities and diameters. The scaling laws have to be identified,
and the simulation for different materials and shape of the impactor for different
angles of impact are in progress.
References
1. Cundall PA, Strack ODL (1979) A discrete numerical model for granular assemblies. Geotech-
nique 29(1):47–65
2. https://www.cfdem.com/liggghts-open-source-discrete-element-method-particle-simulation-
code. LIGGGHTS website.
3. Heuze FE (1989) An overview of projectile penetration into geological materials, with emphasis
on rocks
4. Asphaug E, Collins G, Martin J Global scale impacts. Arizona State University, Imperial College,
London, University of Bern
5. Nair RP, Rao CL (2012) Simulation of depth of penetration during ballistic impact on thick
targets using a one-dimensional discrete element model. Sadhana 37(2):261–279
6. Chandratilleke GR, Zhou YC, Yu AB, Bridgwater J (2010) Effect of blade speed on granular
flow and mixing in a cylindrical mixer. Ind Eng Chem Res 49:5467–5478
7. de Vet SJ, de Bruyn JR (2007) Shape of impact craters in granular media. PhysRevE 76:041306
Using Automated Finite Element
Framework to Analyze Offshore Grid
Turbine Blades Design
Utkarsh Thakre and Asim Tewari
1 Introduction
In a scenario where there is an ongoing hunt for renewable energy sources, harnessing
wind energy through offshore wind turbine grids has emerged as a promising
option. This can be attributed to an abundance of space and greater and consistent
wind resources. That results in a 40% higher output than onshore [1]. Design and
manufacturing of turbine blades for efficient implementation are a very complicated
job due to the operational conditions to which it will be subjected to. In this paper,
an insight is given on finite element analysis for the design of turbine blades for
offshore grids. Today, numerical simulations with complex finite element models are
being solved with highly parallelized processing. They have become a complement
to experimental analysis. The validation is still done against the experimental data
since the precision is not always a guarantee. To get highly precise results, it is
very important to implement advanced problem-specific methods with dedicated and
refined algorithms rather than using methods offered by programs with a predefined
algorithm for broad sets of problems. Besides, a solver environment that can be
highly customized for problem-specific conditions to optimize the computations
can give more room for increasing the complexity of simulations. Thus, for a
given computing power, thereby increases the reliability of results. Often the most
commercially available software does not give the freedom to implement advanced
custom procedures as mentioned above since the user never actually gets to formulate
the actual algorithm. Besides, the most available software in C++, FORTRAN, etc.
involves traditional (manual) development of the complete computational model for
the numerical solution of PDEs which is time-consuming and error-prone. However,
U. Thakre (B) · A. Tewari

Department of Mechanical Engineering, Indian Institute of Technology, Bombay, India
e-mail: 17d100018@iitb.ac.in
A. Tewari
e-mail: asim.tewari@iitb.ac.in
622 U. Thakre et al.
there exists a solution to address this problem. FEniCS, a finite element analysis
framework in Python/C++ which is aimed at creating mathematical methods and
software for automated computational and modeling. It is developed in collaboration
with researchers around the world under the FEniCS project. It even provides the
capability to implement an integrated finite element analysis framework, having all
stages of pre-processing, solving, and post-processing, all under one hood. In the
upcoming section, automated finite element implementation in the FEniCS Python
framework will be discussed.
2 Automated Finite Element Framework in Python
Generality and flexibility are the key attributes of a finite element analysis. However,
due to the inherent nature of these attributes, the automation of FEA using a
computer code becomes a challenge [2]. That is because the code must necessarily
be parameterized over the choice of variational problem and finite elements. Much
of the work in traditional FEA software must still be done manually, which is tedious
and less reliable. FEniCS stands out among the numerous finite element software by
its extensive use of automated low-level code generation capabilities.
The two primary components of FEniCS are DOLFIN and FIAT. DOLFIN is a
C++ finite element library for automated assembly of weak forms, describing meshes,
solvers, and interface to linear algebra of discrete systems. Finite element automated
tabulator (FIAT) is a library for automated generation of finite element spaces and
basis functions. The generation of low-level code for element matrix and vector
construction is implemented by the FEniCS form compiler (FFC). Another feature of
FEniCS which makes it very user-friendly is the unified form language (UFL). UFL is
a domain-specific language for the specification of variational formulations of PDEs.
UFL is embedded within the programming language Python and has built-in support
for automatic differentiation, derivation of adjoint equations, etc. [3]. A special-
purpose just-in-time compiler automatically generates the low-level code based
on the variational formulations, links it into the environment, and finally invokes
the suited solver. Hence, FEniCS efficiently combines scripting with symbolic
mathematics and code generation which are compiled with special-purpose PDE
compilers to generate a C++ script that even outperforms hand-written quadrature
script [3]. This provides both user-friendliness and efficiency. In this paper, a semi-
coupled fluid-structure interaction model is used for the simulations to analyze the
turbine blades in the FEniCS Python framework.
3 Algorithms Used for Analysis in This Paper
The core methods used for the computational model formulation determine the
efficiency of processes and reliability of results. The FEniCS framework tools allow
Using Automated Finite Element Framework .... 623
easy assimilation of custom methods with the code, besides vast control over the
solver process. The various methods used in conjunction with FEniCS functionalities
are as follows:
3.1 Multimesh Functionality
The multimesh finite element methods are formulated on two or more distinct, non-
matching and potentially over-lapping meshes. The multimesh framework allows
for the formulation of efficient finite element methods on complex and dynamic
geometries, by removing the need for costly mesh generation and regeneration.
Multimesh support in FEniCS includes special integration measures for expressing
stabilized finite element methods involving integrals on cut cells, interfaces, and
overlaps involving many meshes [4]. The multimesh functionality has been efficiently
used in the fluid-structure interaction analysis with the solution for the fluid and the
structure domain being solved in respective Eulerian and Lagrangian domain meshes.
The arbitrary Eulerian–Lagrangian (ALE) domain is where the mesh movement
equations are solved to account for updating the mesh according to the displacements
at the interface.
3.2 Efficient Mesh Refinement
The mesh reuses edge bisection to subdivide the marked triangles of a mesh. They
could be the cells in two dimensions, or the facets of tetrahedra in three-dimensions.
The subdivision is chosen to maximize the internal angles of the new triangulation.
This maintains mesh quality after multiple refinements. Edges that are marked for
refinement propagate between processes in parallel, while the refinement operation
is local, resulting in good scaling [5]. In the FSI analysis, during the subsequent
iterations, the mesh refinement algorithm is called which refines the mesh based on
the value of a parameter related to pressure and stress fields near the interface.
3.3 Function Spaces and Finite Elements
The central aspect of the finite element method is the construction of discrete
subspaces Vh out of infinite-dimensional function spaces V . The construction of
subspace (Vh ⊂ V ) can be performed by patching together local function spaces
and defined by a set of finite elements and finding the approximate solution of the
variational problem based on these function spaces. To allow the use of piecewise
polynomial function spaces, H1 (), the mathematically well-known Sobolev spaces
are used [6]. They contain functions v such that v 2 and |∇v|2 have finite integrals
over () which allows functions with discontinuous derivatives [6]. In the problems
analyzed in this paper, the Pq , (q ≥ 1) Lagrange element, a widely known family of
finite elements are used (here, Pq (T ) denotes the space of polynomials of degree up
to and including q on a domain T ⊂ R d ) [7]. Further, vector-valued or tensor-valued
Lagrange elements can also be constructed by using a Lagrange element for each
component [7].
3.4 Solution Methods Implemented for FSI Analysis
The fluid sub-problem has been modeled under incompressible and low Reynolds
number flow conditions. For a domain ⊂ R d ,(1 ≤ d ≤ 3), the Stokes equation
read:
−u + ∇ p = f in , (1)
.u = 0 in , (2)
u = 0 on ∂, (3)
where u : → R d is the velocity field, p : Ω → R is the pressure field, and

f : → R d is the source term [7]. The formulation and solution process of the
Navier–Stokes problem poses a big challenge for the scientific community even
with the recent advances in computing power as of today. In formulating numerical
simulations involving the Navier Stokes problem, several issues like instability of
the finite element problem, the iterations leading to indefinite matrices, difficulty in
finding divergence-free function space (which can satisfy the conservation equations
point-wise, etc. [7]) are encountered. For analysis in this paper, Taylor-hood family
of finite elements consisting of a continuous Pq (q ≥ 2) Lagrange element for the
velocity components and a continuous Pq−1 Lagrange element for the pressure field
are used. With their implementation, convergence is obtained within priori estimate
limits with the smallest velocity errors in comparison with other finite element
schemes [7]. The fluid problem is solved in an Eulerian coordinate system, and
the solver is based on the stress formulation of the Incremental Pressure Correction
Scheme (IPCS). The fluid velocity and the fluid pressure are discretized in space
using Taylorâe“Hood elements. The structural sub-problem is solved using the
linear elasticity formulation with the structural deformation discretized by order one
Lagrange elements with the traction boundary conditions obtained from the fluid
sub-problem.
4 Formulation and Implementation of FSI Simulation
FSI analysis of offshore wind turbine blades becomes increasingly complicated due to
torsion and rotation of the blades. These cause additional gravitational and centrifugal
forces on blades which act in a time varying manner. Apart from this, the air flow
direction is not a single directional flow. Often there are two components acting. One
is the direct flow due to the winds, and the other is the flow due to the rotational
movement of the blades accompanied by convective air flow from over the water
surface. Thus, a full-blown simulation description is out of the scope of this paper.
The simulation workflow is as follows: CAD model design using the OpenSCAD
libraries, meshing using the Gmsh program, FEniCS framework to solve the FEA
problem, and post-processing done in Paraview. Thus, a fully integrated FEA
environment as discussed earlier is realized in Python. The problem is framed in a
simpler manner where a very basic model of a blade is taken in a cuboidal chamber,
and one of the faces is marked as inlet, chosen concerning the orientation of the
blade. The opposite face is marked as an outlet, and all other faces are marked as
walls with no-slip condition. The same is also applied to the blade surface. The inlet
face is given the velocity profile as follows, combining flow due to wind and rotation
of the blade in proximity to the blade in actual conditions

u in = u k x|| + r ∗ ωx⊥ (4)
The outlet velocity has been set to zero. The fluid velocity is denoted by the u and
the fluid pressure by p. In the IPCS used here, first we compute a tentative velocity
u ∗ from the momentum equation using a fully implicit formulation of the convective
term and the previously computed pressure pn [6]. The variational formulation for
the same is as follows. Here, μ is the dynamic viscosity, and ρ is the density. Further,
q is a scalar-valued test function from the pressure space, whereas the test function
v is a vector-valued test function from the velocity space.
Also, σ (u, p) denotes the stress tensor,
σ (u, p) = 2μ (u) − p I, (5)
where (u) is the strain-rate tensor.
1
(u) = ∇u + (∇u)T . (6)
2
Variational Form I
1

u − u n /t, v +
u n · ∇u n , v + σ u n+ 2 , p n , (v)
1

+ p n n, v ∂ − μ∇u n+ 2 · n, v = f n+1 , v . (7)
∂
In the next step, the pressure pn is corrected to pn+1 with the continuity equation
based on the computed velocity u ∗ from the first step [6].
Variational form II

∇ p n+1 , ∇q = ∇ p n , ∇q − t −1 ∇ · u , q . (8)
Finally, the velocity is corrected using the corrected pressure from the previous
step [6].
Variational form III

u n+1 , v = u , v − t ∇ p n+1 − p n , v . (9)
The pressure from the fluid problem is used as a traction T boundary condition
for the structure problem.
T = p ∗ n̂ (10)
Here, n̂ is unit normal to the interface. The variational formulation for the linear
elasticity model is as follows [8]. Here, v is a vector-valued test function from the
displacement space. Also, stress Ïƒ is given by

σ (u) = λ(∇ · u)I + μ ∇u + (∇u) , (11)
Variational form I

σ : ∇v dx = f · v dx + T · v ds. (12)
∂T
following the solving processes of structure problem, the mesh gets refined selectively
on places based on the conditions on parameters discussed earlier.
5 Results and Figures
The FSI simulation results for the turbine blade are shown in Fig. 1. The results
of displacement fields provide an insight into the distortion due to flow over in the
blade. This can be extended to cyclic stresses and can be used to optimize the blade
design for fatigue and performance. As can be seen from the figure the distortion
increases as we move along the leading edge away from the connected face. The
streamlines show the development of flow over blade profile. The flow has almost
free-stream velocity (red) far away and zero velocity (blue) due to no slip BC on the
blade surface.
Fig. 1 Figure shows the flow characteristics and related displacements on the turbine blade part
6 Future Work to be Done
In its current scope the study focuses on describing the basic level implementation of
framework utilizing automated FE code for FSI analysis. The further complexities
weren’t considered like composite blade structure, fatigue type loading, turbulence,
etc. These have to be incorporated in the model for design of blades meeting the
industry level quality controls and tolerances. Future work to be done is as discussed.
6.1 Efficient Mesh Movement Algorithm
The analysis is able to take into account only one-sided coupling. That is, it can only
account for the transfer of fluid pressure as traction to the structure. The analysis
essentially considers the solid only negligibly deformable, thus avoiding mesh update
at the interface. However, in reality, the finite deformation of the structure requires
formulating and solving mesh movements equations at the interface. There are many
methods to achieve movements but many of them cannot be used stably over the
iterations. They tend to distort mesh with due time and eventually leading to a
situation where further iteration can’t be performed without re-meshing which is
a very costly operation. The current analysis implements parts of the actual ALE-
FSI framework, which will be the basis for future works. This framework allows
mesh movements using the bi-harmonic and the Laplace approach. It has been found
that the bi-harmonic model provides greater freedom in the choice of boundary
and interface conditions. In general, the bi-harmonic mesh motion model leads to
a smoother mesh (and larger deformations of the structure) compared to the mesh
motion models based on second order PDEs [9].
6.2 Error Control and Adaptivity
The error estimates can be expressed in terms of the regularity of the actual problem.
This may provide useful information about the order of convergence of the finite
element method. Error estimates can also be expressed in terms of computable
quantities like the residual or possibly the solution of the auxiliary framed dual
problem. This in applications is often necessary to control the error in certain output
functional (M : V → R) within some given tolerance e > 0 [7]. The error estimates
can also be used to implement adaptive solving. In one such approach, in the solver,
the code adapts over the iterations by selective mesh refinement. In this approach, the
solver starts off with a coarse mesh and successively refines only those cells where
the error indicator is large. For this approach to work, the previously described mesh
refinement method needs to be integrated with the solver. In an overall sense, the error
analysis can be used for two purposes, viz. to implement it as a tolerance check for
stopping iterations and for an adaptive solution to enhance stability and convergence.
6.3 Fully (Strongly) Coupled Problem Using ALE
The current formulation is an important step for realizing a solution in actual

condition. The work on the separate fluid and structure sub-problems are performed
with an effort to implement the components of the ALE-FSI framework in parts.
In the formulations, the other side of the coupling has to be achieved using the
mesh movement algorithm as discussed. Once the coupling is achieved, the adaptive
solving and error control measures can be integrated into the formulation, resulting in
a completely coupled FSI problem. A step in the direction to achieve strong coupling,
a thermo-elastic problem was solved with strong coupling between thermal and the
structural domains [8].
The results obtained for this strongly coupled problem are shown in Fig. 2. The
future work will be to implement a similar strong coupling in FSI on the same lines.
6.4 Implementing Extensive Parallel Processing
FEniCS supports high-performance, distributed parallel simulations for large-scale

applications. This is based on an efficient implementation of scalable distributed
meshes, distributed mesh refinement, parallel IO and access to high-performance
parallel linear algebra backends, such as PETSc. Simulations of elliptic problems
Fig. 2 Thermo-elasticity problem solved in strong coupling formulation. Figure shows, a stresses,
b displacements, c temperature, d T versus x variation in time, for an elastic cube that is heated at
a corner with fixed supports on adjacent faces
with over 12 billion degrees of freedom have been performed [6]. This framework
allows portability of the same code from a low-end workstation to deployment on
massive supercomputers without the need for any change. This will be a promising
extension to current capabilities, allowing us to not only simulate the actual problem
more realistically than ever, but will also open doors for further complex analysis.
7 Conclusions
Automated finite element procedures can be implemented for obtaining very precise
simulation results data to improve up on the design of the turbine blade to face
the offshore conditions. The fluid-structure interaction problem today poses a very
big challenge but also has huge application in numerous areas. The work discussed
in this paper put slight on the immense potential of the automated finite element
framework for not only FSI problems but also other complex multi-physics domains.
The FEniCS framework not only provides a highly customizable environment in
which every type of problem can be formulated but also that too can be done with
ease as the actual algorithms are nearly the same as the hand-written expressions. The
simplified formulation done in this paper has huge space to be refined and improved
by various additional algorithms as discussed previously to achieve simulation results

closely resembling reality. Besides this, even more advanced formulations like
complete monolithic continuous description of the fluid and solid domains can also be
implemented with the support of highly parallelized processing. The automated finite
element framework discussed in this paper can be easily adopted by the scientific
community around the globe, which provides not only ease and freedom for problem
formulation but also the capability and power to solve them efficiently. This also gives
them an upper handover the traditionally used closed source counter parts, since the
researchers implement their own customized formulations with proper control over
the solution process and thus contributes to the development of the environment
itself, whereas the open-source nature of environment guarantees reliability due to
being verified by the community actively.
Acknowledgements The author would like to thank National Center for Aerospace Innovation
and Research, Department of Mechanical Engineering, Indian Institute of Technology Bombay, for
their kind support and guidance.
References
1. Strong future for offshore wind power in global energy mix. https://www.siemensgamesa.com/
offshore
2. Logg A Automating the finite element method. Toyota Technological Institute at Chicago.
https://arxiv.org/pdf/1112.0433.pdf
3. Mortensen M, Langtangen HP, Wells GN A FEniCS-based programming framework for
modeling turbulent flow by the Reynolds-averaged Navierâe“Stokes equations. https://doi.
org/10.1016/j.advwatres.2011.02.013
4. Larrson S The FEniCS Project. European Consortium for Mathematics in Industries (ECMI),
Chalmers University of Technology
5. Alnaes MS, Blechta J, Hake J, Johansson A, Kehlet B, Logg A, Richardson C, Ring J, Rognes
ME, Wells GN The FEniCS Project 1.5. Simula Research Laboratory, Chalmers University of
Technology and University of Gothenburg
6. Langtangen HP, Logg A The FEniCS tutorials. www.fenicsproject.org
7. Logg A, Mardal K-A, Wells G (2011) Automated solution of differential equations by the finite
element method the FEniCS book
8. Bleyer J Numerical tours of computational mechanics using FEniCS. https://cometfenics.rea
dthedocs.io
9. Wick T Fluid-structure interactions using different mesh motion techniques. https://doi.org/10.
1016/j.compstruc.2011.02.019
Periodic Transient Performance
Prediction of the Rotary Desiccant Wheel
1 Introduction
The desiccant wheel consists of process air flow, and regeneration air flow sections
split by the clap board as shown in Fig. 1. The ambient air with high moisture content
is made to pass through the process air section to dehumidify the air by transferring
the moisture from air to desiccant pores by the process of adsorption. The adsorbed
moisture content is desorbed out of the desiccant pores in the regeneration section
by passing the regeneration air at high temperature [1].
Ge et al. [2] have developed a model coupling solid side and gas side resistances.
The simulation results show very good correlation with the experimental data over
the earlier gas resistance model. Zhai [3] modeled the effect of residual moisture in
the desiccant wheel after the regeneration process by varying the desiccant density
in a controlled manner. Gao et al. [4] extended this study to taking into account the
effect of channel shape and the desiccant layer thickness in the simulation. Narayanan
et al. [5] modeled the desiccant wheel with parallel and counter flow directions of
process air and regeneration air flow directions with GSR and GSSR models and
numerical simulations performed by Selvaraji et al. [6, 7]. The experimental setup of
air handling system with desiccant wheel and the experimental results in comparison
with simulation results are presented by Angrisani et al. [8, 9].
2 Methodology
The major gap found in the wide literature is that the simulations of rotary desiccant
wheel under transient input conditions are not investigated. Hence, the authors have
S. Muthu · N. Sekarapandian (B)

Department of Thermal and Energy, School of Mechanical Engineering, VIT, Vellore, India
e-mail: sekarapandian.n@vit.ac.in
632 S. Muthu and N. Sekarapandian
Fig. 1 Schematic of rotary desiccant wheel
focused primarily on the periodic transient input conditions and their effect on the
wheel output results.
The mathematical model of rotary desiccant wheel is based on the following major
assumptions:
(1) Matrix material is having negligible moisture adsorption capacity.
(2) The inlet air conditions are varying gradually or suddenly from low value to
high value or visa-versa.
(3) The thermo-physical properties of the dry air, dry desiccant, and matrix material
are constant, but bulk properties are the function of moisture content.
(4) The heat of adsorption is fully taken by desiccant layer.
(5) The pressure loss of the air stream in axial direction is negligible.
The governing equations are derived based on the conservation of moisture and
conservation of energy for the air and desiccant medium.
The moisture conservation in the air can be expressed as:

∂(ρα Ya ) ∂(ρα Ya ) ∂(ρα Ya )
de +ω +u = K y (Yd − Ya ) (1)
∂t ∂θ ∂z
Moisture conservation in desiccant is given by:

Periodic Transient Performance Prediction of the Rotary Desiccant Wheel 633

moisture content profile of a Yp(0,t) = 14.2 g/kg
Moisture content, kg/kg

0.03
rotary desiccant wheel with Yr(L,t) = 18.2 g/kg
Narayanan et al. [5] and 0.025 Tp(0,t) = 35° C
Brillhart [10] Tr(L,t) = 120° C
0.02 and
0.015 L = 0.2 m, Simulation
0.01
α = 180° Narayanan et al.
Experiment -
0.005 Brillhart
0
0o Adsorption 180 o Regeneratio 360 o
Wheel Rotation angle, °

∂W ∂(W ) ∂ ∂W
δρd +ω − De = K y (Yd − Ya ) (2)
∂t ∂θ ∂z ∂z
Energy conservation in the air is given as:

∂(ρa Ta ) ∂(ρa Ta ) ∂(ρa Ta )
de c pa +ω +u
∂t ∂θ ∂z

∂ ∂ Ta
− de ka = h(Td − Ta ) + c pv K y (Yd − Ya )(Td − Ta ) (3)
∂z ∂z
Energy conservation in the desiccant can be written as

∂ Td ∂(Td ) ∂ ∂ Td
c pd ρd δ +ω −δ kd = h(Ta − Td )
∂t ∂θ ∂z ∂z
+ K y (Ya − Yd )qst + c pv K y (Ya − Yd )(Ta − Td ) (4)
The nomenclatures and geometrical data adopted in the simulation are referred
from Selvaraji et al. [6, 7] The present simulation results are validated with the
numerical results Narayanan et al. [5] and experimental results of Brillhart [10] as
per the corresponding wheel parameters and operating conditions, shown in Fig. 2.
The impact of varying the design parameters is investigated by analyzing the perfor-
mance of the various designs. The parameters varied are the direction of flow of the
process air and regeneration air streams with varying the inputs in periodic transient
conditions.
During the dehumidification process, the outlet moisture content of air decreases
sharply at the beginning and reaches a minimum value and then increases gradually
until the end of the dehumidification process, whereas for regeneration, the moisture
Fig. 3 Performance of Ta(z=L) Ta(z=0) Ya(z=L) Ya(z=0)

rotary desiccant wheel with
140 35
Temperature of air, C
Moisture conent of air ,

uniform inlet air moisture
content and temperature 120 30
100 25
80 20
g/kg
60 15
40 10
20 5
0 0
5.0 5.5 6.0
Wheel Rotation, rev
content of air rises sharply at the beginning and then gradually declines. The sharp
edge in the start of regeneration air outlet moisture curve is due to effects of the
sudden transition from process section to regeneration section where air is coming
in opposite direction.
Temperature of air at the outlet drops from the regeneration temperature steadily
during de-humidification. Throughout the regeneration process, the temperature
increases, but slightly at a faster rate, the beginning is shown in Fig. 3.
The effect on output moisture and temperature of air is predicted under the periodic
transient variable conditions of input moisture contents and shown as in Fig. 4 for
various cases outlined as below in Table 1. The nominal input moisture contents for
process and regeneration air are used as 14.2 g/kg and 18.2 g/kg, respectively, as
same as Narayanan et al. [5], and the change in linear and step cases is either 50%
lower or higher than these nominal values are used.
4 Conclusions
The numerical model shows a good agreement with experimental data. The effect
on output moisture content and temperature of air are predicted under the periodic
transient variable input moisture contents. As a future scope, this simulation is to
be used to predict the performance of rotary desiccant wheel for the solar desiccant
air-conditioning system.
Fig. 4 Performance of Moisture of Process air

wheel under periodic A - Linear Change input Low - High
Moisture conent of air , g/kg

transient variable input Ta(z=L) T(z=0) Ya(z=L) Ya(z=0)
conditions (Cases A → H)
140 35
120 30
100 25
80 20
60 15
40 10
20 5
0 0
5.0 5.5 6.0
Wheel Rotation, rev
Moisture of Process air

B - Linear Change input How - Low

Ta(z=L) T(z=0) Ya(z=L) Ya(z=0)
140 35
120 30
100 25
80 20
60 15
40 10
20 5
0 0
5.0 5.5 6.0
Wheel Rotation, rev
Moisture of Process and Regeneration air

C
- Linear Change input : Low - High
140 37.5
120 32.5
100 27.5
80 22.5
60 17.5
40 12.5
20 7.5
0 2.5
5.0 5.5 6.0
Wheel Rotation, rev
Fig. 4 (continued) Moisture of Process and Regeneration air

D
- Linear Change input : High - Low

160 45
140 40
120 35
100 30
80 25
60 20
40 15
20 10
0 5
5.0 5.5 6.0
Wheel Rotation, rev
Moisture of Process air

E - step change : Low-high

150 35
25
100
15
50
5
0 -5
5.0 5.5 6.0
Wheel Rotation, rev
F Moisture of Process air

- step change : High- low

140 35
120 30
100 25
80 20
60 15
40 10
20 5
0 0
5.0 5.5 6.0
Wheel Rotation, rev
Fig. 4 (continued)
G Moisture of Process & Renegeration air
- Step change : Low-high & Low - high

160 40
140 35
120 30
100 25
80 20
60 15
40 10
20 5
0 0
5.0 5.5 6.0
Wheel Rotation, rev
H Moisture of Process & Renegeration air

- Step change : Low-high & High -Low

160 42.5
140 37.5
120 32.5
100 27.5
80 22.5
60 17.5
40 12.5
20 7.5
0 2.5
5.0 5.5 6.0
Wheel Rotation, rev
Table 1 Various cases of simulations performed

Periodic transient variable moisture content inputs conditions
Cases Process air Regeneration air
A Linear low –> High Constant
B Linear high –> Low Constant
C Linear low –> High Linear low –> High
D Linear high –> Low Linear high –> Low
E Step low –> High Constant
F Step high –> Low Constant
G Step low –> High Step low – > High
H Step low –> High Step high – > Low
References
1. Ge TS, Li Y, Wang RZ, Dai YJ (2008) A review of the mathematical models for predicting
rotary desiccant wheel. Renew Sustain Energ Rev 12:1485–1528
2. Ge TS, Ziegler F, Wang RZ (2010) A mathematical model of predicting the performance
of compound desiccant wheel (A model of compound Desiccant Wheel). Appl Therm Eng
30:1005–1015
3. Zhai C (2008) Performance modelling of desiccant wheel design and operation. Phd Thesis,
Carnegie Mellon University, Pittsburgh
4. Gao Z, Mei VC, Tomlinson JJ (2005) Theoretical analysis of dehumidification process in a
desiccant wheel. J Heat Mass Transf 41:1033–1042
5. Narayanan R, Saman WY, White SD, Goldsworthy M (2011) Comparative study of different
desiccant wheel designs. Appl Therm Eng 1–8
6. Muthu S, Talukdar P, Jain S (2015) Effect of regeneration section angle on the performance of
a rotary desiccant wheel. ASME J Therm Sci Eng Appl 8(1):011013. https://doi.org/10.1115/
1.4030966
7. Muthu S, Talukdar P, Jain S (2017) Performance enhancement of rotary desiccant wheel by
the innovative designs of multiple desiccant layers. In: Fluid mechanics and fluid power—
contemporary research, Springer India. https://doi.org/10.1007/978-81-322-2743-4_102
8. Angrisani G, Minichiello F, Roselli C, Sasso M (2012) Experimental analysis on the
dehumidification and thermal performance of a desiccant wheel. Appl Energ 92:563–572
9. Bareschino P, Diglio G, Pepe F, Angrisani G, Roselli C, Sasso M (2015) Modelling of a
rotary desiccant wheel: numerical validation of a variable properties model. Appl Therm Eng
78:640–648. https://doi.org/10.1016/j.applthermaleng.2014.11.063
10. Brillhart PL (1997) Evaluation of desiccant rotor matrices using an advanced fixed- bed test
system. Ph.D. thesis, University of Illinois, Chicago
Study of Hybrid LES/RANS Models
in OpenFOAM Applied
to Taylor–Couette Flow
Sai Guruprasad Jakkala and S. Vengadesan
1 Introduction
Counter-rotating Taylor–Couette (CRTC) flow is a classical problem in fluid

mechanics which has been studied over many years. The simple geometry and myriad
flow characteristics have championed the study of this flow. Over the years, many
experimental, numerical and analytical approaches have been used to inspect the
problem. This makes CRTC a good benchmark to test different models. There are
also extensive data sets for the given problem which helps in validating the model.
Experimental studies have been carried out by many researchers. The work by
Andereck et al. [1] helped in identifying the different regimes of flow in the CRTC
flows. Many numerical studies have been carried out for the laminar and turbu-
lent cases. Dong [2] carried out one of the most recent extensive DNS studies on the
problem. Poncet et al. [3] used large eddy simulation (LES) models to study the effec-
tiveness in predicting the turbulent quantities and the mean velocity profile. Luo et al.
[4] developed a new hybrid model by modifying the k-equation used in LES calcu-
lations. The coefficients in the k-equation were modified in the Reynolds-averaged
Navier–Stokes (RANS) region. The eddy viscosity was made a linear function of the
turbulent viscosities from the LES and the RANS region. The length scale was fixed
prior to the start of the simulation.
DNS is computationally too expensive and requires considerable amount of
computational resources. To avoid massive computational resource requirement,
turbulence modelling is pursued. Hybrid LES/RANS models are a class of turbulence
models which can leverage the advantages of LES and RANS. The unsteadiness of
the flow is captured by LES, and the near wall regions are modelled using the RANS
models. This hybridization avoids the grouping of a large number of cells in the near
S. G. Jakkala · S. Vengadesan (B)

Indian Institute of Technology Madras, Chennai 600036, India
e-mail: vengades@iitm.ac.in
640 S. G. Jakkala and S. Vengadesan
wall region. The cutback in the number of grid points helps reduce the computa-
tional time, while still capturing major flow features due to the hybrid modelling of
turbulence. The switching between RANS and LES regions can either be predefined
or be made to switch automatically based on the grid size, and the latter is used in
this work.
An overview on turbulence modelling and its procedures are given in the books by
Pope [5] and Wilcox et al. [6]. A review on the recent developments in the turbulence
modelling (RANS and hybrid models) has been done by Durbin [7]. Even though
hybrid models have been present for the past few decades, not a lot of work has been
done to exploit their advantages in many flow problems. This study aspires to show
the effectiveness of hybrid LES/RANS models to a classical flow problem. In this
current study, we apply different hybrid LES/RANS models to the CRTC flows at a
Reynolds number Rei = Reo = 4000 which had not been pursued previously to study
the model’s applicability to swirling geometry and the prospect of their extension to
problems where DNS data is not available for comparison and benchmarking. The
mean flow quantities and turbulent quantities are compared with the DNS and LES
results available in the literature. The automatic switching between LES and RANS
region in DES, DDES and IDDES is studied. The results are discussed in detail in
the respective sections.
2 Methodology
The simulations in this study are done using the open-source software OpenFOAM-
v1806 [8]. The Reynolds number is chosen based on the DNS study of Dong [2], Rei =
−Reo = 4000. The turbulence models (LES, DES, DDES and IDDES) implemented
in OpenFOAM are discussed in the following subsections.
2.1 Computational Set-Up
The computational domain is shown in Fig. 1. The domain dimensions are taken
identical to the one used in the DNS study by Dong [2].
In the present study of CRTC system, the inner and outer cylinders are rotating
in opposite directions at constant angular velocities, but the angular velocities are
different for the inner and outer walls, i , and o , respectively. The flow geometry
is characterized by the radius ratio, η = Ri /Ro , where Ri and Ro are, respectively,
the radii of the inner and outer cylinders. The annulus gap width d = (Ro − Ri ) is
taken as the length scale. Two Reynolds numbers (Re), namely inner and outer Re
numbers, are Rei and Reo defined as Rei = U i d/ν and Reo = U o d/ν, where ν is the
fluid kinematic viscosity and U i and U o are the tangential velocities on the inner and
outer cylinder walls, respectively. In all the simulations performed here, Re = Rei =
−Reo = 4000 and radius ratio was kept at η = 0.5.
Study of Hybrid LES/RANS Models in OpenFOAM Applied … 641
domain
The results obtained from the WALE LES model and the Spalart–Allmaras (SA)-
based and k-ω SST-based DES, DDES and IDDES are discussed in the following
subsections for a given mesh size. The results are compared with the DNS results of
Dong [2] to evaluate the effectiveness of the turbulence models.
The averages for the mean flow quantities are taken after stationary fluid flow is
reached for the LES and hybrid LES/RANS simulations. The quantities are averaged
along the periodic directions, θ and z. This averaging over the periodic directions
helps in achieving better accuracy results.
The mean velocity profile of the flow is the first result discussed, followed by
the Reynolds stress. Both these quantities are averaged over the periodic regions.
One of the variables corresponds to the flow feature, and the other corresponds to
a turbulence feature. This helps in validating the accuracy of the turbulence models
against DNS. The interface regions between LES and RANS are plotted to decide on
the performance of the hybrid LES/RANS models. Finally, the computational time
is compared between the various models.
3.1 Mean Quantities
The mean velocity profile (u θ ) is computed, and the results are compared with the
DNS result of Dong [2]. The LES results are comparable to the DNS in terms of
peak value and the overall trend of the profile. The hybrid LES/RANS models are
also able to predict the mean velocity profile accurately for a much smaller grid size.
The differences between the models occur in predicting the position of zero velocity.
The velocity profiles comparing the different turbulence models are shown in Figs. 2
and 3.
Comparing the two different hybrid models, we find that the SA-based models
perform better in predicting the mean velocity profile. The k-ω SST-based models
(DES and DDES) are unable to capture the bends accurately and also the zero
Fig. 2 Velocity profile for SA-based models
Fig. 3 Velocity profile for k-ω SST-based models

crossing. The k-ω SST-based IDDES performs better compared to the DES and
DDES model in predicting the mean velocity profile.
The prediction of mean velocity profiles is a key in fluid problems as it helps in
deriving quantities like wall shear. From Figs. 2 and 3, we can conclude that the
hybrid models are capable of accurately capturing the mean velocity fields.
3.2 Turbulent Quantities
The Reynolds stresses are essential to the way a turbulent flow develops and also
the turbulent structures. The DNS study by Dong [2] reports only the u r u θ Reynolds
stress. The turbulence model results are plotted for the same in Figs. 4 and 5. LES
model is able to correctly predict the peak value of Reynolds stress and also the overall
trend of the Reynolds stress profile. The SA-based hybrid LES/RANS models are
able to predict the overall trend of the Reynolds stress, but fall short in predicting
the peak value. The k-ω-based hybrid models except the IDDES are very under-
predicting. The trend is not correctly predicted in the k-ω SST-based DDES. The
error is 4.25, 4.35 and 4.64% for the SA-based DES, DDES and IDDES, respectively.
The error is 14.13, 13.95 and 5.85% for the k-ω SST-based DES, DDES and IDDES,
respectively.
Fig. 4 Reynolds stress profile for SA-based models

Fig. 5 Reynolds stress profile for k-ω-based models
3.3 Length Scales
The LES region (red) for the hybrid LES/RANS models is shown in Fig. 6. The
regions are switched automatically during the simulation based on the equations
Fig. 6 Length scale for LES-RANS switching for SA-based DES

given below. The LES region increases as we go from the DES to the DDES and the
IDDES model. In the IDDES model, the RANS region is limited to the near wall
region.
The different length scales used for DES, DDES and IDDES, respectively, are:
d̃ = min(CDES , y) (1)
d̃ = max[L RS − f d , max(L RAS − L LES , 0)] (2)
d̃ = max( f d L RAS + (1 − f d )L LES , 1e − 6) (3)
The variables in the above equations are defined below:
L RAS = y (4)
L LES = CDES (5)

C
f d = 1 − tanh Cd1 rd d2 (6)

νeff
rd = min , 10 (7)
|∇u|κ 2 y 2
The increase in the LES region helps in capturing more of the turbulence structures
in the flow. The DES and the DDES models have almost equal representation of the
LES and the RANS models. The IDDES model length scale switching helps in
achieving a larger LES region.
4 Conclusions
This study focuses on the utilization of hybrid LES/RANS models against the LES
and DNS models. CRTC model was chosen for its simplicity in geometry and
complexity in flow features.
The hybrid LES/RANS are able to predict the mean quantities with a good accu-
racy as seen in the Results section. The models are able to predict the overall trend
of the Reynolds stress values, with the correct location of peak Reynolds stress, but
not the correct peak value. The SA-based models are more accurate than k-ω SST
models for this particular flow problem.
The computational mesh required for the hybrid LES/RANS regions is nearly half
of the one used in LES. This is a significant decrease in mesh size, while still getting
results within accuracy of 4% for the Reynolds stress profile. This shows that for
industrial turbulence problems, DES could be a good alternative compared to RSM.

The computational time required for the simulations of hybrid LES/RANS models
is less compared to the LES models.
For this particular type of flow problem, the DES model is sufficient to accu-
rately capture the flow features. Comparing the SA-based and k-ω SST-based hybrid
models, SA-based models perform better in terms of accuracy. There is a 22–24%
speed-up in the hybrid models when compared to the LES WALE model. This is a
significant speed-up which necessitates the use of the hybrid models in large-scale
turbulent flows.
The results show that hybrid LES/RANS are satisfactory in predicting the turbulent
and averaged features of the flow at lower mesh sizes when compared to LES. The
extension of these models to complex geometries lies in successfully separating the
region for RANS and LES. These models are applicable in problems where RANS
models are still used to obtain results in industries.
References
1. Andereck D (1986) Flow regimes in a circular Couette system with independently rotating
cyclindes. J Fluid Mech 164:155–183
2. Dong S (2008) Turbulent flow between counter-rotating concentric cylinders: a direct numerical
simulation study. J Fluid Mech 615:371–399
3. Poncet S, Viazzo S, Oguic R (2014) Large eddy simulations of Taylor-Couette-Poiseuille flows
in a narrow-gap system. Phys Fluids 26(10)
4. Luo G, Yao Z, Shen H (2018) A new hybrid turbulence model applied to highly turbulent
Taylor-Couette flow. Phys Fluids 30(6)
5. Pope SB (2000) Turbulent flows. Cambridge University Press
6. Wilcox DC (1998) Turbulence modeling for CFD. DCW industries La Canada, CA
7. Durbin PA (2018) Some recent developments in turbulence closure modeling. Annu Rev Fluid
Mech 1–47
8. Weller HG, Tabor G, Jasak H, Fureby C (1998) A tensorial approach to computational continuum
mechanics using object-oriented techniques. Comput Phys 12(6):620
Investigation of Second Throat
Supersonic Exhaust Diffusers
for the Altitude Testing of Rocket Motors
R. Ashokkumar, Sathiyavageeswaran Sankaran, and T. Sundararajan
1 Introduction
Rocket motors designed for high-altitude operations usually employ higher area ratio
nozzles, in order to utilize the low ambient pressure and produce higher thrust., gener-
ally for upper stage rocket motors of launch vehicles. In the development phase,
such rocket motors are to be tested at ground level to evaluate their performance.
During a ground-level test with nozzle back pressure (pb ) value equal to the atmo-
spheric pressure (pa ), flow separation may take place at the divergent portion of the
nozzle. Under such conditions, it is not practically feasible to qualify a rocket engine
for a launch vehicle.
To evaluate the performance of upper stage rocket motors, the low-pressure envi-
ronment of the flight situation needs to be simulated in the ground testing installation.
Different fluid dynamic systems are being used in high-altitude test (HAT) facility to
simulate the high-altitude conditions. Experimental and numerical studies have been
carried out by several researchers since 1950 to develop fluid dynamic systems [1–5]
which can simulate the high-altitude conditions on the ground for the altitude testing
of rocket motors. In general, two types of fluid dynamic systems are in use for the
simulated altitude testing of rocket motors and they are supersonic exhaust diffusers
and ejector diffusers. Two types of supersonic exhaust diffusers are used for HAT;
one is the straight cylindrical supersonic exhaust diffuser (SED), and the other one is
R. Ashokkumar (B)
QD-CSP&SMA, SR, SDSC SHAR, ISRO, Sriharikota, India
e-mail: ashokkumar.r@shar.gov.in
S. Sankaran
SMP&ETF, SDSC SHAR, ISRO, Sriharikota, India
T. Sundararajan
Department of Mechanical Engineering, IIT Madras, Chennai, India
e-mail: tsundar@iitm.ac.in
648 R. Ashokkumar et al.
the second throat supersonic exhaust diffuser (STED). STED uses the kinetic energy
of rocket exhaust itself to maintain low nozzle back pressure through a benign shock
structure, which isolates the vacuum test chamber from the ambient. Chen et al.
[6] carried out a numerical analysis of a second throat diffuser and concluded that
the averaged Mach number on the entrance plane of the second throat is a suitable
criterion for justification of the performance of the STED system.
At present, published research literature is limited on the parametric studies related
to STED [7]. They pertain to mostly numerical simulations and that too with nozzles
of less than 50 area ratio. Advanced launch vehicles include upper stage solid rocket
motors with nozzle area ratios greater than 50. For such motors, it is essential to
generate experimental data on the starting conditions in a high-altitude test facility.
In the present study, a series of numerical simulation and cold flow tests have been
conducted to study the effects of the following diffuser geometrical parameters on
STED performance: (i) annular gap between nozzle and diffuser (yg ), (ii) second
throat contraction ratio (Ad /Ast ) and (iii) distance from nozzle exit to the start of
diffuser contraction (X c ). The main objective of the study is to develop a near optimum
STED configuration, which can start at a minimum stagnation pressure and deliver
a smooth pressure recovery.
2 Experimental and Numerical Simulation
2.1 Cold Flow Experimental Setup
The cold flow high-altitude test setup with second throat exhaust diffuser (STED) is
shown in Fig. 1. For these tests, gaseous nitrogen (γ = 1.4) is used as the driving fluid.
Cold flow tests were conducted with a small vacuum chamber of volume 0.0013 m3
(referred to as ‘cell region’). For all the tests, the STED is started against atmospheric
pressure. The diffuser is said to be started, when the flow attaches to the diffuser wall
and a series of oblique shocks followed by a normal shock are positioned in the throat
region of the diffuser. Diffuser starting can also be confirmed by the gradual pressure
recovery along the diffuser length and the steady cell pressure (less than nozzle exit
pressures) measured at the vacuum chamber. Strain gauge-type absolute pressure
transducers were used to record the nozzle inlet total pressure, vacuum chamber
pressure and diffuser wall static pressure.
2.2 Numerical Simulation of STED
Numerical simulations are carried out to provide an initial prediction for the cold
flow experiments and to have an insight on the internal shock structure during the
steady-state operations of the STED. For the computational analysis, a half plane of
Investigation of Second Throat Supersonic Exhaust Diffusers … 649
Fig. 1 Schematic of cold flow test setup
the axisymmetric geometry is considered, from the axis to the wall of the domain
consisting of small vacuum chamber, rocket motor and STED. The model is created
using the commercial software Gambit, and the governing equations of the flow are
solved using the finite volume-based commercial CFD solver, ANSYS Fluent. A fully
coupled implicit compressible flow solver with Spalart–Allmaras (SA) turbulence
model [8] has been adopted to compute the flow pattern inside the STED system. The
assumptions considered are the flow is considered to be axisymmetric, gravitational
effects are negligible, the exhaust flow from the nozzle exit is treated as an ideal
gas mixture, and flow is considered as steady. Wall boundary conditions are applied
on the motor, nozzle, vacuum chamber and diffuser walls. The rocket motor inlet is
defined as the pressure inlet, and the subsonic exit is defined as a pressure outlet.
The grid was generated using quadrilateral elements. Simulations have been carried
out until the residues fall below 10−6 for all the flow variables (i.e., convergence
criteria). The grid used for the numerical analysis has been tested for its sensitivity
on the accuracy level of the predicted results.
3.1 Effect of Annular Gap (Yg )
In the simulated altitude test setup, the rocket nozzle will be positioned inside the
diffuser inlet for the positive impingement of the exhaust jet onto the diffuser wall.
The radial gap between the nozzle outer wall and the diffuser inner wall at the nozzle
Table 1 Nozzle and diffuser parameters for annular gap variation studies
Test case Ae /At Dt (mm) Ad /At yg (mm) Ad /Ast Aso /Ast
1 100 14.5 164 19 2 4
2 100 10 164 12 2 4
Fig. 2 Vacuum chamber pressure characteristics for Case-1 and Case-2
exit plane is termed as the annular gap, and it is denoted as ‘yg ’. During the STED
unstarted mode, reverse flow into vacuum chamber occurs through this gap only.
Cold flow tests were conducted with STED configuration as listed in Table 1 for a
100 AR nozzle. For the both cases, length of the second throat diffuser (L st ) is equal
to 8 Dst and nozzle exit position X c is equal to 0.5 Dd . Tests were carried out with
discrete nozzle inlet pressure (Po ). The vacuum chamber pressures for varied nozzle
inlet pressures for Case-1 and Case-2 are presented in Fig. 2.
For Case-1, the required minimum starting pressure (Po )ds is 67.5 bar, whereas
for Case-2 it is 63.5 bar. It is inferred from the results that the annular gap (yg ) plays
a significant role in determining the minimum starting pressure (Po )st . If the annular
gap is too large, the exhaust flow expanding from the nozzle exit may not attach to
the diffuser wall and it may not protect the vacuum chamber against the ingress of
ambient atmosphere leading to a flow separation in the nozzle divergent.
3.2 Effect of Diffuser Contraction Ratio (Ad /Ast )
It is desirable to have a second throat area (Ast ) as small as possible from the point of
view of structural simplicity and HAT facility dimensions (as second throat diameter
Table 2 STED
Parameters STED-1 STED-2
configurations for the studies
Ad /At 100 134.5
Ad /Ast 4.5, 2.9, 2.5, 2.2, 2 and 2.9, 2.5, 2.2, 2, 1.7 and
1.4 1.4
Aso /Ast 4 4
L st /Dst 8 8
Xc 0.5 Dd 0.5 Dd
Yg 6 mm 6 mm
dictates the second throat diffuser length and subsonic diffuser length). Therefore,
understanding the changes in the flow phenomena due to the variation in contraction
ratio (Ad /Ast ) is essential for the design optimization of STED.
Numerical simulations and cold flow experiments were conducted for a nozzle
area ratio of 70 and 100 using STED configurations 1 and 2, respectively, as listed in
Table 2. The required minimum diffuser starting pressure (Po )ds values for STED-1
and STED-2 obtained from the experiments are presented in Fig. 3. For 70 AR nozzle,
STED-1 started for the contraction ratio (Ad /Ast ) range of 1.4–2.9. For 100 AR nozzle,
STED-1 started for the contraction ratio (Ad /Ast ) range of 1.7–2.5. STED-2 with a
contraction ratio (Ad /Ast ) of 2.5 requires a minimum starting pressure of 46.5 bar, as
against 66 bar for a contraction ratio of 1.7.
STED-1 with a contraction ratio (Ad /Ast ) of 4.5 has not at all started, even for a
pressure of 70 bar. Also, for this case (Ad /Ast = 4.5), the vacuum chamber pressure
increases with further increase in nozzle inlet pressure (Po ), which indicates choking
of the second throat. This leads to a reverse flow of exhaust gases into the vacuum
chamber, leading to a flow separation in the nozzle. Similar is the case for STED-2
Fig. 3 STED contraction ratio limits

Fig. 4 STED-1 pressure recovery for the case of Ad /Ast = 2.9 and Ad /Ast = 1.4
configuration with a contraction ratio of 2.9, which did not start even for an inlet
pressure of 70 bar, due to the too small throat of the diffuser.
STED-2 configuration with a contraction ratio of 1.4 did not start even for an inlet
pressure of 70 bar. This STED configuration requires still higher-pressure values to
start the diffuser as it is close to the straight cylindrical diffuser configuration. Wall
static pressure variations along the diffuser length (L) for STED-1 with contraction
ratio (Ad /Ast ) of 1.4 and 2.9 are shown in Fig. 4.
Test results show that for the contraction ratio (Ad /Ast ) of 2.9, recovery of pressure
occurs through a series of weak oblique shocks terminating with a weak normal shock
almost at the exit of the second throat diffuser. However, for the contraction ratio
of 1.4, pressure recovery does not occur through multiple weak shocks, but through
one or two strong oblique shocks ending with a terminal normal shock almost at the
middle of the second throat region. In this case, though the diffuser is started, the
pressure recovery process through the shock train is not in a favorable condition. In
such a case, even a small reduction in the rocket engine pressure (Po ) may result in
the shock being swallowed by the nozzle, leading to flow separation in the nozzle.
3.3 Effect of Xc
In the STED configuration, the axial distance between nozzle exits to the start of the
second throat contraction (Xc ) influences the diffuser starting pressure. Cold flow
experiments were conducted with the STED configuration as given in Table 3 for a
70 AR nozzle, to study the effect of nozzle exit position with respect to the start of
diffuser contraction (X c ).
Table 3 Diffuser configuration for X c studies

Ad /At Ad /Ast Aso /Ast L st /Dst X c /Dd yg
100 2.9 4 8 0 to 1 (in steps of 0.1) 3 mm
The minimum starting pressure (Po )ds required remains almost constant and inde-
pendent of X c in the range of 0–0.3 Dd (Fig. 5). Any further increase in X c beyond
0.3 Dd increases the minimum starting pressure required. But for X c = Dd , STED
has not started even for a pressure of 60 bar, which is almost 50% higher than the
minimum starting pressure required (Po )st for lower X c values (≤0.7 Dd ). This is
due to the occurrence of strong shock in the diffuser near the nozzle exit, leading to
a flow separation in the diffuser before the second throat region.
Numerical simulations were carried out for the diffuser configuration-B with 70
AR nozzle for X c = 0, 0.5 and 0.7 Dd , and the corresponding Mach number contours
at the minimum diffuser starting pressure (Po )ds are presented in Fig. 6. The formation
of multiple oblique shock cells with a terminal weak normal shock inside the second
throat diffuser is observed. From the numerical results, it is evident that the terminal
normal shock (TNS) location inside the second throat is significantly influenced
by the position of nozzle exit with respect to the start of contraction (X c ). For the
case of X c = 0, the normal shock terminates almost at the midpoint of the second
throat, whereas for the case of X c = 0.5 Dd , the normal shock terminates further
downstream of the second throat diffuser mid-location and for the case of X c = 0.7
Dd , it terminates close to the exit of second throat. STED with X c = 0.7 Dd is the best
Fig. 5 Diffuser starting pressure for different X c /Dd ratios

Fig. 6 Mach contour of STED with different X c values
configuration for the rugged operation of STED as in this case the terminal normal
shock is located at the favorable location, i.e., exit of second throat.
4 Conclusions
The systematic investigation has been carried out to study the effect of STED geomet-
rical parameters on its starting characteristics. From the results, it is concluded that
the diffuser starting pressure increases with increase in annular gap (yg ). In fact, the
larger annular gap would require larger expansion of the exhaust flow when it enters
the diffuser from the nozzle. Based on the minimum starting pressure required, the
near optimum diffuser contraction ratio for 70 AR nozzle with STED-1 is 2.9 and
that for a 100 AR nozzle with STED-2 is 2.5. It is apparent from these results that
there exists a limit on the maximum diffuser contraction ratio for STED, beyond
which the second throat chokes resulting in flow separation within the nozzle. An
increase in the axial distance X c beyond 0.7 Dd leads to an increase in the STED
starting pressure and also X c influences the position of the terminal normal shock.
Acknowledgements The authors thank the management of SDSC SHAR and ISRO for their
support to this R&D study, and also we thank IITM computation center for permitting us to perform
the CFD studies.
References
1. German RC, Bauer RC (1961) Effect of diffuser length on the performance of ejectors without
induced flow. Technical report, Report No. AEDC-TN-61-89
2. Goethert BH (1962) High altitude and space simulation testing. ARS J 32:872–882
3. German RC, Bauer RC, Panesci JH (1966) Methods of determining the performance of Ejector-
Diffuser systems. J Spacecraft 3:193–200
4. Annamalai K, Visvanathan K, Sriramulu V, Bhaskaran KA (1998) Evaluation of the performance
of supersonic exhaust diffuser using scaled down models. Exp Therm Fluid Sci 217–229
5. Kim H, Lee Y (1999) Numerical simulation of the supersonic flows in the second Throat Ejector-
Diffuser systems. J Therm Sci 8:214–222
6. Chen F, Liu CF, Yang JY (1994) Supersonic flow in the second throat ejector diffuser system. J
Spacecraft Rock 31:123–129
7. Sankaran S, Satyanarayana TNV, Annamalai K, Visvanathan K, Babu V, Sundararajan T (2002)
CFD analysis for simulated altitude testing of rocket motors. Canadian Aeronaut Space J 48:153–
161
8. Kumaran MR, Vivekanand PK, Sundararajan T, Kumaresan K, Manohar RD (2009) Optimiza-
tion of second throat ejectors for high altitude test facility. J Propulsion Power 25:667–706
Aerodynamic Noise from Circular
Cylinder in Low Subcritical Flow Regime
Using Large Eddy Simulations
Joemon Jacob and S. K. Bhattacharyya
1 Introduction
Noise generated from bluff bodies is one of the common problems in engineering
applications, examples being automobile mirrors, marine risers, submarines, etc.
Understanding the generation mechanism and estimation of noise is the first step
in designing better noise control mechanisms. Earlier, experimental and analytical
methods were used for flow induced noise studies. With the progress in computa-
tion facilities, computational fluid dynamics (CFD) techniques are widely used for
studying fluid flow problems. Effective numerical techniques can help in predicting
far field aerodynamic sound as well as help in designing better experiments in
future. For easiness, complex bluff bodies are approximated to circular cylinders
for numerical studies.
Etkin et al. [1] found that sound radiated from a circular cylinder is related to the
oscillating lift and drag forces acting on the cylinder. Aeroacoustics is a multi-scale
problem as there is large disparity in the scales of turbulence and acoustics. A hybrid
approach is adopted, where flow features are calculated first for the acoustic sources
and analogies are used to obtain far field pressure fluctuations. Lighthill acoustic
analogy was modified by Ffocws Williams [2] to account for the introduction of
solid surfaces in the fluid medium and is widely used in aeroacoustic studies.
One of the earliest experimental study on flow induced noise was that of Revell
et al. [3] who established relation between far field sound and drag coefficient for
various Re ranging from 4.5 × 104 to 4.5 × 105 . Experiment for rod airfoil config-
uration of Re 4.6 × 104 was done by Jacob et al. [4], which remains a benchmark
study for aerodynamic and acoustic features in that Re range. Recently Geyer et al.
J. Jacob (B) · S. K. Bhattacharyya

Department of Ocean Engineering, IIT Madras, Chennai, India
e-mail: oe15d020@smail.iitm.ac.in
S. K. Bhattacharyya
e-mail: skbh@iitm.ac.in
658 J. Jacob and S. K. Bhattacharyya
[5] did experiments to understand the influence of porous materials on flow induced
noise of cylinders for wide range of Re from 1.6 × 104 to 1 × 105 . The flow studied
in the present numerical study is that of a circular cylinder subjected to uniform flow
of Re 19,800. Circular cylinder considered here is to be of infinite length such that
the effect of free end vortices and horse shoe vortices is not affecting the flow, and
only von karman vortex streaks are significant.
Numerical techniques help us to overcome the limitations of experiments and
enable to study flow problems for a wide range of Re. One of the pioneer numerical
study for estimating sound was by Hardin et al., [6] for a circular cylinder subjected
to Re = 200. Kato et al. [7] used LES to predict far field sound from a circular
cylinder subjected to flow of Re 1 × 104 . They used a short span length of L/D =
2 (Length/Diameter) for numerical simulations to predict the far field noise from a
span length of 50 D and obtained agreeable results. A methodology was proposed to
account for the unsimulated lengths based on the coherence between surface pres-
sure fluctuations. Far field sound prediction for Re 9 × 104 was done by Orselli et al.
[8] using LES and Kato methodsm and agreeable results were obtained. Norberg
[9] reviewed a number of experiments and proposed an empirical formula for the
spanwise correlation length (), based on surface lift fluctuations. Instead of coher-
ence functions, suitability of coupling and Kato formulas to obtain far field sound
correction is explored in this study.
It is observed that experimental and numerical studies on flow induced noise are
limited to a few Reynolds numbers only, and sound radiated from low subcritical Re
flows is not thoroughly studied yet. Low sub-critical Re number flows find application
in marine risers, offshore pipelines, etc. Agreeable results for far field sound were
obtained with a reasonable grid resolution and above-mentioned correction method.
2 Analysis Methodology
Velocity and pressure fluctuations on cylinder surface, which constitute acoustic

sources, were calculated first using LES, and far field pressure fluctuations were
obtained using FWH equations. Flow feature experiments for the exact Re were not
available, and hence, various reference experiments were used for validation [8, 9].
Flow induced noise was compared with [5].
2.1 Flow Characteristics
Statistical flow parameters like coefficient of drag (C d ), lift (C l ), Strouhal number

(St), etc., were captured and compared with literature. For a circular cylinder with
projected area A, mean flow velocity of U c and fluid density of ρ, coefficient of drag
and lift can be defined as C d = F d /(0.5 ρAU 2c ) and C l = F l /(0.5 ρAU 2c ), respectively.
Strouhal number is a non-dimensional number defined as St = fD/U c , where f is
Aerodynamic Noise from Circular Cylinder in Low … 659
the frequency of vortex shedding and D the diameter of cylinder. Mean coefficient
of pressure (C p ) was plotted along the circumference of cylinder to understand the
pressure distribution on the surface
and the lee
side of the cylinder. Coefficient of
pressure can be defined as C p = P − Pref / 0.5 ρUc2 , where P is the pressure at
the location and Pref is the pressure in the mean flow. Apart from the standard flow
parameters mentioned above, wake region of cylinder was analysed to understand
the flow pattern and to ensure the accuracy of LES simulations. Flow feature in the
turbulent wake region of cylinder shall be explained with recirculation length (L r ),
which is defined as the distance between the base of the cylinder and where the
centreline mean velocity component changes it sign.
2.2 Far Field Sound
In order to estimate sound from long span bodies, numerical simulations were made
for shorter segments and corrections factors to account for unsimulated lengths were
applied onto the results of short span simulations. Conventionally, spatial coherence
lengths (L c ) are obtained from the coherence functions of surface pressure fluctua-
tions along the span length of cylinder. Coherence values thus obtained are distributed
mathematically (e.g., Gaussian) to obtain L c and further corrections. Owing to limits
in computation, lesser span length is chosen for numerical simulations. Lesser span
length and periodic boundary conditions can lead to variations in L c obtained by
the above conventional methods [10, 11]. In the present study, a simple empirical
relation from [8] was employed to obtain spanwise correlation length () and used
with Kato formulas to obtain correction factors. Normalised spanwise correlation
length (/D) is given as
−0.20
/D ≈ 2.6 × Re/2.4 × 105 (1)
where Re is the Reynolds number.
2.3 Numerical Method
Ansys FLUENT, a commercial finite volume code was used to solve the incompress-
ible flow equations and far field pressure fluctuations. The filtered Navier-Stokes
equations for incompressible flows in LES are given as
∂(u i )
=0 (2)
∂ xi

∂(u i ) ∂ u i u j 1 ∂( p̄) ∂ σi j ∂ τi j
+ =− +ν − (3)
∂t ∂x j ρ ∂ xi ∂x j ∂x j
With the velocity and pressure fluctuations from above equation, far field pressure
fluctuations are obtained using FWH equations. Volume integral term (quadrapole)
in the FWH equation is not calculated because of its negligible contribution to the
sound field in this Re regime [1].
2.4 Grid and Boundary Conditions
A rectangular computational domain of size 29 D × 17 D × 3 D (x, y, z) was

chosen for present study as shown in Fig. 1. Origin of the coordinate system is at
the base of the circular cylinder and was placed 8.5 D from the inlet of the domain.
Downstream length of computational domain extends to 20.5 D from the origin of
cylinder. Spanwise length of the cylinder in the z direction (L s ) was chosen as 3D,
based on previous LES studies and available computation capacity.
Fig. 1 Schematic of
computation domain
Fig. 2 Computational grid
Spatial discretisation was based on the characteristic scale of turbulent eddies

such as length scale, l = 0.05 D and velocity scale u = 0.2Uc [10, 12]. A non-
dimensionalised time step t* (t* = tU c /D, t is the simulation time (s)) of 0.05 was
used which resolves the vortex shedding cycle with 105 time steps. In the present LES
study, y + (y+ = yu* /υ, y is the wall normal distance, u* is the friction velocity
and υ is the kinematic viscosity of the fluid) was kept < 1 in the vicinity of cylinder
wall, and a growth ratio of 1.05 was applied in the boundary layer. 48 nodes were
used along the span, and 360 grid points were used along the circumference of the
circular cylinder. With the above-mentioned specifications, a structured mesh of 3.3
million cells was generated. Different views of the grid used are shown in Fig. 2.
A 32-CPU work station with Intel dual-core Xeon processors was used for running
the simulations and it took nearly 71 s to calculate the flow for each time step. These
time statistics were taken after the simulation reached steady statistics.
A uniform flow velocity (U c ) with zero turbulence level was used as inlet boundary
condition, and outflow boundary condition was given at downstream outlet of the
domain. At spanwise boundaries, periodic boundary condition was applied, and the
top and bottom walls were given a zero shear slip wall boundary condition.
Statistical flow parameters for Re 19,800 LES simulations were compared with exper-
iments and are shown in Table 1. Cd , Cl rms and St values were observed within the
range of sub-critical flow regime. Separation angle (θ s ) was found over estimated
than experiments, probably because of the less number of grid points along the
circumference of the cylinder. Figure 3 shows the distribution of mean C p around
cylinders. Qualitatively, the trend was predicted well, but the absolute value of C p
was observed more than the experiments, and C p was seen deviating away in the lee
side of the cylinder.
Table 1 Flow parameters comparison

Case (Re) St Cl θ s (°) Cd
2 × 104 [8] 0.194 0.47 78 1.0–1.4
1.4 × 105 [14] 0.19–0.215 1.0–1.4
13,100 [9] 0.203 0.545 87.6 1.31
19,800 (LES) 0.198 0.6 85 1.31
Fig. 3 Mean pressure distribution around circular cylinder
Apart from the conventional flow parameter studies, the accuracy of simulations
performed could be explained based on the velocity details from the turbulent wake
region of cylinder. Figure 4 shows the streamwise velocity profiles in (x, z) and (y,
z) plane of circular cylinder. In Fig. 4a, minimum velocity was seen closer to the
circular cylinder base for high Re and was seen away from cylinder base for low Re.
For Re 3900, L r was obtained as 1.57 D, and for high Re of 1.45 × 105, L r reduced
to 0.5 D [13, 14]. For Re 19,800, L r was obtained as 0.76 D, which falls in the range
mentioned above, which confirms the effective capturing of velocity features in the
wake region. In Fig. 4b, lower Re flows display larger velocity minima, which is a
corollary of the explanation for L r .
Pressure fluctuations in the far field were captured at 16.6 D, 90° above the trans-
verse axis of circular cylinder. from Eq. (1) was obtained as 4.28 D, which is
greater than the chosen span length (L s ) and less than the experiment length (L e ) of
10 D. Correction factor corresponding to = 4.28 D was calculated as 6.8 dB and
was used to obtain final sound pressure spectrum (SPL) as follows [7]:
SPL f = SPLs + 20 log(L c /L s ) + 10 log(L e /) (4)

(a) (b)
Fig. 4 Time averaged streamwise velocity profile (u) in the wake region: a In (x, z) plane, y/D =
0. b In (y, z) plane, x/D = 1.06. Re 19800, LES; Re 13100, [9]; Re 1.45 × 105 , [14, 16]; Re 3900,
[13]
where ‘f ’ and ‘s’ subscripts indicate final and simulated length of sound pressure level
(SPL), respectively. Obtained SPL spectrum compared with experiment is shown in
Fig. 5.
Sound in the region around the peak frequency (10 dB below the aeolian tone)
was taken as a parameter for comparing with experiments. Aeolian tone predicted
by present LES is 70.1 dB which is slightly higher than experiment value of 66.9
(Table 2). For engineering applications, exact value of sound is not required as sound
(dB) is generally categorised into various regimes based on a range of decibel values.
The predicted values agree well with the measured values (Fig. 5).
Fig. 5 Sound pressure level at 16.6 D

Table 2 Comparison of
Cases St Aeolian tone (dB)
aeolian tone
(Geyer et al. 2016) 0.188 66.9
LES (present) 0.193 70.1
4 Conclusions and Future Work
Aerodynamic noise radiated from a circular cylinder in the low subcritical Re was
predicted numerically using LES and FWH equations. Conclusions from the study
can be summarised as:
(1) Capturing flow fields accurately is of primary importance in the estimation of
aerodynamic sound.
(2) Moderate grid resolution on the surface and near field region of solid surface
is sufficient to obtain agreeable dipole sound characteristics.
(3) Empirical formula from Eq. (1) can be used for estimating spanwise correlation
length and can be used for obtaining correction factors for far field sound.
In this study, we have coupled from empirical formula and Kato formulas to
obtain correction factors. Kato and Seo Moon [15] proposed the correction formulas
based on L c , which is based on the coherence between surface pressure fluctuations
along the span length and its mathematical distribution. A comparison of present
method with coherence method is intended as a future work.
References
1. Etkin B, Korbacher GK, Keefe RT (1957) Acoustic radiation from a stationary cylinder in a
fluid stream (Aeolian Tones). J Acoust Soc Am 29(1):30–36
2. Ffocws Williams JE, Hawkings DL (1969) Sound generation by turbulence and surfaces in
arbitrary motion. Philos Trans R Soc A Math Phys Eng Sci 264(1151):321–342
3. Revell JD, Prydz RA, Hays AP (1978) Experimental study of aerodynamic noise vs drag
relationships for circular cylinders. AIAA J 16(9):889–897
4. Jacob MC, Boudet J, Casalino D, Michard M (2005) A rod-airfoil experiment as benchmark
for broadband noise modeling. Theor Comput Fluid Dyn 19:171–196
5. Geyer TF, Sarradj E (2016) Circular cylinders with soft porous cover for flow noise reduction.
Exp Fluids 57(3):1–16
6. Hardin JC, Lamkin SL (1984) Aeroacoustic computation of cylinder wake flow. AIAA J
22(1):51–57
7. Kato C, Iida A, Takano Y, Fujita H, Ikegawa M (1993) Numerical prediction of aerodynamic
noise radiated from low mach number turbulent wake. In: 31st AIAA, aerospace sciences
meeting and exhibit, 11–14 January 1993
8. Norberg C (2003) Fluctuating lift on a circular cylinder: review and new measurements. J
Fluids Struct 17(1):57–96
9. Abrahamsen Prsic M, Ong MC, Pettersen B, Myrhaug D (2014) Large eddy Simulations of
flow around a smooth circular cylinder in a uniform current in the subcritical flow regime.
Ocean Eng 77:61–73
10. Orselli R, Meneghini J, Saltara F (2009) Two and three-dimensional simulation of sound
generated by flow around a circular cylinder. In: 15th AIAA/CEAS aeroacoustics conference
(30th AIAA aeroacoustics conference), May 2009
11. Manoha E, Delahay C, Sagaut P, Mary I, Ben Khelil S, Guillen P (2001) Numerical predic-
tion of the unsteady flow and radiated noise from a 3D lifting airfoil. In: 7th AIAA/ICEAS
aeroacoustics conference, May 2001
12. Kim S-E (2006) Large eddy simulation of turbulent flow past a circular cylinder in subcritical
regime. In: 44th AIAA aerospace sciences meeting and exhibit, pp 1–17
13. Parnaudeau P, Carlier J, Heitz D, Lamballais E (2008) Experimental and numerical studies of
the flow over a circular cylinder at Reynolds number 3900. Phys Fluids 20(8):1–14
14. Cantwell B, Coles D (1983) An experimental study of entrainment and transport in the turbulent
near wake of a circular cylinder. J Fluid Mech 136:321–374
15. Seo JH, Moon YJ (2007) Aerodynamic noise prediction for long-span bodies. J Sound Vib
306(3–5):564–579
16. Breuer M (2000) A challenging test case for large eddy simulation: high Reynolds number
circular cylinder flow. Int J Heat Fluid Flow 21(5):648–654
Review of Experimental Test Rigs
for Pump as Turbine Applications
Rahulkumar Painter, Ashish Doshi, and Mukund Bade
1 Introduction
Energy is associated with every human being and their day-to-day life. As the
population increases, the demand will increase which requires to be satisfied by
the alternative source of energy instead of relying on conventional energy sources.
Renewable sources of energy are the best alternative to depleting energy resources.
The total installed capacity of renewable energy in India as of March 31, 2019, is
77742.12 MW, and from that, small hydropower is 4604.80 MW. The cost of low
head turbines is high per output power generation, and more research is required in
that direction of lowering the cost so that it will be beneficial for the remote location
[1]. In rural areas, usually, power generation can be done by diverting some water path
of run-of-the-river before penstock dropping [2]. A centrifugal pump when running
in the reverse direction which acts as a turbine is called PAT. Centrifugal pumps and
Francis turbines are hydraulically comparable excluding the flow control device. It
is also observed that because of hydraulic losses, the pump mode has less head and
discharge compared to turbine mode of operation [2, 3]. The PAT has various advan-
tages compared to turbines such as a large number of standard sizes, easily available
components, bulk production, and less costly and easy installation [2].
Though PAT is not yet commercial, the performance characteristics curves are not
entirely known and pump manufactures also not providing reverse characteristics of
the pump so to establish its performance characteristics is a challenging task of
research, and lots of researchers are heading toward it but still it is not clarified;
R. Painter (B) · A. Doshi · M. Bade

MED, SVNIT Surat, 395007 Surat, India
e-mail: ds17me005@med.svnit.ac.in
A. Doshi
e-mail: avd@med.svnit.ac.in
M. Bade
e-mail: bmh@med.svnit.ac.in
668 R. Painter et al.
hence, the standard test rig is the first step for that, and this paper will help to make
an acceptable test rig for PAT testing.
2 Literature Review
Derakhshan and Nourbakhsh [4] carried out experimental work with the centrifugal
pumps having different specific speed (Ns < 60 (m, m3 /s)). Four centrifugal pumps
having specific speed 14.6, 23.0, 37.7, and 55.6 (m, m3 /s) tested and found that in
PAT mode, it operates at higher head and flow rate than in pump mode with the
same rotational speed. The experimental setup consists of a feed pump, PAT, orifice
meter for discharge, barometer, and piping systems. The uncertainty in the measured
parameters was given in Table 1.
Nautiyal et al. [5] studied an experimental analysis for pumps having specific
speed 18 (m, m3 /s) with 1500 rpm.
The centrifugal pump was operated in reverse mode with the different head and
discharge values and observed that reverse mode of operation requires higher head
and flow in comparison with pump mode operation with compromising less efficiency
(8.53%). In an experimental test rig synchronous generator of 12.5 kVA was coupled
with PAT and having desired flow and head, an auxiliary pump is used with 60 m
head and 0.035 m3 /s. For controlling the excessive flow, a bypass line was connected
between the main supply pipe and tank, and the venturi meter was used for the flow
measurement.
Singh and Nestmann [6] examine the impeller rounding and studied the internal
hydraulic analysis for nine PATs.
The tested pumps having specific speed in the range of 20–94.4 rpm. The exper-
imental setup was an open-loop-type hydraulic circuit in which water is supplied
by the help of a feed pump and surge tank. The experiment result shows that modi-
fication leads to decrease system loss coefficient, whereas in exit, relative angle is
increased for all PATs. Modification of impeller rounding at blades inlet reduced
wake formation and overall efficiency was increased within the range of 1–3% band
in all regimes as BEP, part load, and overload for all PATs.
Bozorgi et al. [7] examine the axial flow pump in turbine mode with the experi-
ment and the numerical approach for Pico hydropower plants. The PAT operated at
750 rpm, and for controlling constant rotational speed, an electronic load controller
was associated. Moreover, the conventional synchronous generator was used in a test
setup. The maximum efficiency obtained by the experiment was 61% and found that
both experiment and numerical results were in good agreement.
Raman et al. [8] performed an experimental study with an end suction pump
to find performance characteristics of PAT. A synchronous AC generator of 5 kW
coupled with PAT. They used fifty pieces of bulbs, 100 kW each as ballast load for the
generator power. PVC pipe with 100 mm diameter was used in the test rig. The results
were matched with the available literature and observed acceptable agreement, but
the efficiency was less. It is also noticed by the experiment that it operates at higher
Table 1 Summary of pump as turbine experimental test rig
S. No. Author name PAT Measuring instruments Remark
H (m) Q Ns (rpm) Pressure Flow Speed Torque/output
(m3 /sec) head power
1 Ismail et al. 14 0.008 17.91 Pressure Digital flow – Torque sensor Speed controlled
[3] gage meter uncertainty uncertainty based on torque
uncertainty ±3% ±0.1% applied
±0.25%
2 Derakshshan 25 0.15 14.6–23 Pressure Orifice meter – Generator with Speed control by the
and 37.6–55.6 gage uncertainty torque arm flow control valve
Nourbakhsh uncertainty ±3.4%
[4] ±5.5%
3 Nautiyal et al. 32.8 0.0148 18 Pressure Venturi meter – Synchronous Gate valve for speed
[5] gage generator with regulator
electric resistance
4 Singh and – – 24.5–35.3 Pressure Magnetic flow Optical counts Torque sensor Speed regulation with
Nestmann [6] 36.4–39.7 transducer meter ±1% of ±1 rpm ±0.1% of full the help of a control
45.2–46.4 accuracy full scale scale valve
79.1–94.4 ±0.5%
5 Yang et al. Pressure Turbine flow – Torque meter Energy dissipation
[17] meter uncertainty uncertainty pump for speed
Review of Experimental Test Rigs for Pump as Turbine Applications
gage ±0.4%
uncertainty ±0.5% ±0.2% regulation
6 Bozorgi et al. 4 0.2 118.58 Pressure Venturi meter Optical rotor Generator with Speed
[7] gage uncertainty meter torque arm regulation–electronic
uncertainty ±5.4% uncertainty controller with ballast
±3.5% ±4.1% load power output
6 kW
(continued)
669
Table 1 (continued)
670

7 Raman et al. 22 0.0083 15.36 Pressure Digital flow Pulse counter Generator with Inverter used to
[8] gage meter meter ballast load regulate feed pump
motor frequency
8 Giosio et al. 12.25 0.222 104.3 Pressure Ultrasonic flow Optical sensor Torque Speed regulation by
[9] transducer meter accuracy sensor–digital VSD insertion of
uncertainty ±0.01 m/s over telemetry adjustable guide vane
±0.06% of the sonic path accuracy ±0.2% in pat for flow
full scale of full scale regulator
reading
9 Jain et al. [10] 20 0.0292 25.29 Pressure Electromagnetic Digital speed Generator with Speed regulating by
transmitters flow meter sensor resistive load valve and bypass line
sensitivity sensitivity sensitivity (voltage, no VFD
±0.1 psi ±0.0005 m3 /s ±1 rpm ammeter
sensitivity ±1 V
and±0.01A)
10 Wang et al. 38.6 0.0264 18.1 Pressure Turbine flow Speed sensor Electrical eddy Compared original
[11] 16.4 0.0252 36.4 transmitter meter accuracy accuracy current impeller and special
9.73 0.025 52.8 accuracy ±0.5% ±1 rpm dynamometer impeller special
±0.1% accuracy ±0.4% impeller with forward
curved blades
11 Doshi et al. – – 19.9 Pressure Electromagnetic Speed sensor Torque sensor Blade/shroud
[12] 24.6 sensors flow meter uncertainty uncertainty rounding efficiency
33.4 uncertainty uncertainty ±0.1% ±0.66% rise 1–2.5%
53.6 ±0.31% ±0.5%
(continued)
R. Painter et al.
Table 1 (continued)
12 Abazariyan 5 0.0038 27.04 Pressure Volume/time Tachometer Generator with Part load viscosity
et al. [18] gage method uncertainty torque arm increase –loss increase
uncertainty ±1 rpm uncertainty ±1% and efficiency
±0.31% decrease BEP and
overload viscosity
increase –loss
decrease
Review of Experimental Test Rigs for Pump as Turbine Applications
671
flow and head but because of more losses in PAT compared to pump mode resulted
in lowering the efficiency in turbine mode.
Giosio et al. [9] carried out experimental analysis by incorporating the inlet flow
control facility by inserting the adjustable guide vanes. They follow IEC 60193
wherever they can apply for designing test setup and calculations. They used variable
speed drive (VFD) along with the feed pump, supply tank, delivery tank, circular
draft tube, and support frame. The maximum 79% efficiency was achieved in the
experiment and compared with the predicted data.
Jain et al. [10] performed an experimental analysis to check the effect of impeller
diameter, speed, and blade tip rounding. Selected centrifugal pump specification is
given in Table 1. The design of the draft tube was with inlet cone angle 6° and height
1 m following IS 5496:1993. The highest efficiency is obtained in 10% trimmed
impeller as 76.93% at the 100 rpm rotational speed. They observed good performance
at a low specific speed and also found 3–4% increase in efficiency for the blade
rounding and trimming of impeller operation.
Ismail et al. [3] carried a test to determine the performance of the end suction
centrifugal pump in reverse mode with a range of rotational speeds (800–1400 rpm).
All the required data were measured with the sensors and stored in the data logger.
The speed of pump was controlled by influencing the braking force at the shaft. The
digital flow meter used for measuring flow rate was a propeller type with a capacity
to measure 76–760 l/min. It was noticed that torque increased as the rotational speed
decreased and achieved a maximum of 65% efficiency for the tested PAT.
Wang et al. [11] investigated the experimental study of a special type of impeller
designed with forward curve blades for PAT and achieved a good response. The main
observation between two volutes of pumps running in turbine mode was having a
similar loss pattern, but the losses occur in the draft tube of specially designed
impeller that shows lower for the same flow rate with the original impeller.
Doshi et al. [12] have done an experimental investigation on four pumps in turbine
mode having specific speed ranging from 19–54 rpm with inner blade rounding, inner
and outer shroud rounding. The developed test rig was well designed with highly
sophisticated instruments for measuring head, flow, speed, and torque. They used
variable frequency drive and PLC with SCADA software. The performance rounding
found to be beneficial as it helps in reducing losses and increase efficiency within
1–2.5% Fig. 1.
In addition to this, many researchers had experimentally investigated PAT
performance and try to predict its performance characteristics [13–16].
Different standards are available for the selection and designing of different parts
of the experimental test rig. Standard IEC 60193 (International Electrotechnical
commission) is mainly for model acceptance tests of hydraulic turbines, storage
pumps, and pump turbines. The pressure measurement at inlet and outlet, pressure
transmitter, or pressure gages with suitable range and accuracy within ±1% is recom-
mended. According to standard ISO 20456:2017, the flow meter shall be placed on
a straight pipe length of ten times diameter of pipe to the upstream and two times
diameter of pipe to the downstream which ensures fully developed flow. Though the
Review of Experimental Test Rigs for Pump as Turbine Applications 673
Fig. 1 Pump as turbine test rig [12]
design and selection criteria for the PAT draft tube are not defined yet, we can use a
standard for elbow-type draft tube (IS 5496).
3 Conclusion
Pumps are available in a wide range so that pump as turbine can be used for the small
and micro-hydroelectric power plant for energy generation because of its various
advantages as discussed earlier. In this present work, the experimental investigation
of the pump as turbine and test rig developed is reviewed, which gives an idea about
different test rig for evaluation of PAT performance. Some researchers used feed
pumps and some used an overhead tank to supply the required head and flow to PAT.
Though the various test rig designed by researchers in order to test and achieve the
complete characteristics of PAT is a step ahead for the future, but still its performance
is not predicted perfectly for the entire range. Moreover, it was observed that to have a
standardized test rig, one needs to follow available standards, and its recommendation
for different components used in an overall test rig.
References
1. Haidar A, Senan M, Noman A, Radman T (2012) Utilization of pico hydro generation in

domestic and commercial loads. Renew Sustain Energy Rev 16(1):518–524
2. Binama M, Su WT, Bin Li X, Li FC, Wei XZ, An S (2017) Investigation on pump as turbine
(PAT) technical aspects for micro hydropower schemes: a state-of-the-art review. Renew Sustain
Energy Rev 79(April):148–179
3. Ismail MA, Muzammil WK, Rahman M, Ibrahim MWK, Misran S (2017) Experimental design
and analysis of pump as turbine for microhydro system. In: IOP Conference series: materials
science and engineering, vol 217, no 1
4. Derakhshan S, Nourbakhsh A (2008) Experimental study of characteristic curves of centrifugal
pumps working as turbines in different specific speeds. Exp Therm Fluid Sci 32(3):800–807
5. Nautiyal H, Kumar A, Yadav S (2011) Experimental Investigation of centrifugal pump working
as turbine for small hydropower systems. Energy Sci Technol 1(1):79–86
6. Singh P, Nestmann F (2011) Internal hydraulic analysis of impeller rounding in centrifugal
pumps as turbines. Exp Therm Fluid Sci 35(1):121–134
7. Bozorgi A, Javidpour E, Riasi A, Nourbakhsh A (2013) Numerical and experimental study of
using axial pump as turbine in pico hydropower plants. Renew Energy 53:258–264
8. Raman N, Hussein I, Palanisamy K, Foo B (2013) An experimental investigation of pump
as turbine for micro hydro application. In: IOP Conference series: earth and environmental
science, vol 16, no 1
9. Giosio DR, Henderson AD, Walker JM, Brandner PA, Sargison JE, Gautam P (2015) Design
and performance evaluation of a pump-as-turbine micro-hydro test facility with incorporated
inlet flow control. Renew Energy
10. Jain SV, Swarnkar A, Motwani KH, Patel RN (2015) Effects of impeller diameter and rotational
speed on performance of pump running in turbine mode. Energy Convers Manag 89:808–824
11. Wang T, Wang C, Kong F, Gou Q, Yang S (2017) Theoretical, experimental, and numerical study
of special impeller used in turbine mode of centrifugal pump as turbine. Energy 130:473–485
12. Doshi A, Channiwala S, Singh P (2017) Inlet impeller rounding in pumps as turbines : an
experimental study to investigate the relative effects of blade and shroud rounding. Exp Therm
Fluid Sci 82:333–348
13. Fernández J, Blanco E, Parrondo J, Stickland MT, Scanlon TJ (2004) Performance of a
centrifugal pump running in inverse mode. Proc Inst Mech Eng Part A J Power Energy
218(4):265–271
14. Chapallaz JM, Eichenberger P, Fischer G (1992) Manual on pumps used as turbines. Deutsches
Zentrum fur Entwicklungstechnologien, vol 11
15. Singh P, Nestmann F (2010) An optimization routine on a prediction and selection model for
the turbine operation of centrifugal pumps. Exp Therm Fluid Sci 34(2):152–164
16. Suarda M (2009) Assessment performance of pumps as hydro-turbines. J Ilm Tek Mesin
CakraM 3(1):67–72
17. Yang SS, Derakhshan S, Kong FY (2012) Theoretical, numerical and experimental prediction
of pump as turbine performance. Renew Energy 48:507–513
18. Abazariyan S, Rafee R, Derakhshan S (2018) Experimental study of viscosity effects on a
pump as turbine performance. Renew Energy 127:539–547
Assessment of Struts in Gas-Dynamic
Steering of a Supersonic Nozzle
T. Thillaikumar and Mrinal Kaushik
1 Introduction
Gas-dynamic steering, commonly known as thrust vector control (TVC), is a tech-

nique used to manipulate the direction of the thrust, which eventually controls the
attitude of the vehicle. This technique has received a great attention since long due to
its practical application in developing the control systems for attitude adjustment and
performing the manoeuvres. Thrust vectoring can be accomplished using either the
mechanical or the fluid systems. A mechanical system, such as a moving pintle used
for changing the throat area of a convergent-divergent nozzle, requires a drive mech-
anism [1, 2]. The throat area can also be modified by introducing a secondary flow in
the direction perpendicular to the primary flow, resulting interactions between these
flows [3, 4]. Indeed, this flow configuration exhibits high reliability and requires
no drive mechanism; however, it needs a large storage tank to store the sufficient
quantity of fluid for injection into the primary flow. Clearly, the injection of an auxil-
iary fluid flow, acting as the thrust vector control, makes the system heavier and less
attractive for longer endurance. To overcome these difficulties, a small solid body (or
strut) is inserted into the primary flow. A strut deployed in the convergent-divergent
nozzle controls the thrust vector in the same way as the injected fluid in transverse
direction. In addition, the strut thrust vector control (STVC) is geometrically simple,
light weight and relatively cheaper than the fluid injection system.
To quantify the thrust vectoring capability of struts, the wall static pressures for
both uncontrolled and controlled nozzles were measured at the wall mounted ports.
Also, the waves prevailing in the uncontrolled and controlled configurations have
been visualized using the Schlieren technique.
T. Thillaikumar · M. Kaushik (B)

Department of Aerospace Engineering, Indian Institute of Technology, Kharagpur 721302, India
e-mail: mkaushik@aero.iitkgp.ac.in
676 T. Thillaikumar and M. Kaushik
2 Methodology
In the present study, the experiments were conducted in the jet-test facility available
at the Department of Aerospace Engineering at the Madras Institute of Technology,
Chennai, India. The compressed air from the storage tank was supplied continuously
to the settling chamber through a control valve. The nozzle inlet pressures were set
and maintained constant using the pressure regulating valves (PRVs). The study was
performed at the overexpansion, optimum-expansion and underexpansion states of
the nozzle by varying the nozzle inlet pressures as 4 bar, 6 bar and 8 bar, respectively.
The throat and exit areas of the nozzle were 113.09 mm2 and 167.41 mm2 , respec-
tively. The semi-divergence angle of the convergent-divergent nozzle was 1.517°.
The area ratio of the nozzle was 1.48 with corresponding exit Mach number as 1.84.
The convergent-divergent nozzle and struts were made of stainless steel.
To quantify the effect of strut on gas-dynamic steering of the nozzle, three different
strut heights (1.5, 2.5 and 3.5 mm) of triangular-shape were used. The strut was
deployed at the mid-section of the divergent portion of the nozzle. The static pressures
were measured at different ports mounted along the nozzle wall. These pressure ports
were connected to a 16-channel intelligent pressure scanner. The pressure scanner
(PSI 9116) had a capability of measuring the pressure up to 17 bar. The user-friendly
software provided by the manufacturer has been used to interface the pressure scanner
with a computer that acquired the data and displays the reading from all the 16
channels, simultaneously on the monitor. The jet structure and shock pattern in the
jet field were visualized using the Schlieren technique.
3.1 Wall Static Pressure Variation
To quantify the thrust vector control efficiency of strut, the variation of wall static
pressure was plotted against the axial distance. The wall static pressure (p) was made
non-dimensional with the stagnation pressure (p0 ), and the axial distance was made
non-dimensional with the nozzle length (L). The non-dimensional static pressure
(p/p0 ) variation for the lower and upper surfaces of the nozzles is shown in Figs. 1,
2 and 3.
Figure 1a shows the wall static pressure distribution over the lower surface of the
supersonic nozzle operating at overexpanded conditions. It can be seen that the flow
starts accelerating soon after entering the convergent-section and attains the choking
at the throat and further accelerates in the divergent-section, up to x/L = 0.64 for
both uncontrolled and strut-controlled nozzles. At x/L = 0.64, the static pressure falls
below than the ambient pressure, which produces a local suction effect. However, at
the subsequent port location (x/L = 0.78), a significant rise in static pressures for the
struts of heights 2.5 and 3.5 mm is observed. In fact, when a strut is introduced in
Assessment of Struts in Gas-Dynamic Steering of a Supersonic … 677
Fig. 1 Wall static pressure variation for a lower surface, b upper surface (overexpansion)
the divergent portion of the nozzle, it produces a strong bow-shock which eventually
increases the static pressure downstream. Impingement of bow-shock on the nozzle
wall essentially leads to the boundary layer detachment. In addition, with increase of
strut height a rise in static pressure was observed in all the cases. At the farthest port
location corresponding to 0.84 L, an initial drop in static pressure is shown which
subsequently rises to freestream value for both the struts of heights 2.5 and 3.5 mm.
This gradual rise in pressure indicates the reattachment of the flow to the wall. Here,
an insignificant rise in the pressure due to 1.5 mm strut indicates the generation of a
relatively weak shock.
The wall static pressure variation at the upper surface of supersonic nozzle under
overexpanded conditions is shown in Fig. 1b. For both uncontrolled and controlled
configurations, the flow is found to be accelerating up to x/L = 0.78; with a rapid
increase in static pressure for the strut of height 3.5 mm. This is because of the
Fig. 2 Wall static pressure variation for a lower surface, b upper surface (correct-expansion)
boundary layer detachment caused by the bow-shock generated at the strut and
extends till the upper surface of the nozzle. In far field, the pressure gradually rises
to freestream value for all the nozzle configurations.
Similar behaviour in wall pressure distribution was observed for correctly-
expanded and underexpanded nozzles. At the lower surface of supersonic nozzle,
a rapid increase in static pressure was found at x/L = 0.78 for both 2.5 and 3.5 mm
strut cases (Figs. 2a and 3a). Thereafter, the flow accelerates towards the nozzle exit
(x/L = 0.98), where the same value of static pressure was observed for both correctly
expanded and underexpanded nozzles (Figs. 2b and 3b). Thus, the strut-controlled
nozzle under overexpanded conditions experiences a larger jet deflection.
Fig. 3 Wall static pressure variation for a lower surface, b upper surface (underexpansion)
3.2 Schlieren Flow Visualization
The Schlieren flow visualization images for uncontrolled and controlled nozzles at
overexpanded condition are shown in Fig. 4. For an overexpanded jet, since the nozzle
exit pressure is less than the ambient pressure and thus an oblique shock formation
at the nozzle exit can be seen in Fig. 4a. This oblique shock increases the nozzle
exit pressure to the ambient pressure. Further, the left-running and right-running
oblique shocks cross each other at the jet centerline and get reflected from the jet
boundary as the expansion waves. These expansion waves again cross each other
and get reflected from the outer jet boundary as compression waves. This process
continues and leads to the formation of a periodic shock-cell structure in the jet.
When a strut is introduced in the nozzle, the jet gets deflected downward as shown
(a) Uncontrolled nozzle
(b) Nozzle with 1.5 mm strut
(c) Nozzle with 2.5 mm strut
(d) Nozzle with 3.5 mm strut

Fig. 4 Schlieren images of the uncontrolled and strut-controlled nozzles (overexpansion)
in Fig. 4b–d. The jet deflection angles of 3.3°, 3.5° and 3.6° are obtained with the
strut of heights 1.5 mm, 2.5 mm and 3.5 mm, respectively.
For an underexpanded jet, the nozzle exit pressure is higher than the ambient
pressure and hence the expansion wave forms at nozzle exit (Fig. 5). These waves
cross each other at the jet centerline and get reflected from the jet boundary as
compression waves. These compression waves again cross each other at the jet axis
and get reflected as expansion waves. This process continues and generates a periodic
shock-cell structure (Fig. 5a). However, these shock-cell structures are modified with
Fig. 5 Schlieren images of

the uncontrolled and
strut-controlled nozzles
(underexpansion)
(a) Uncontrolled nozzle
(b) Nozzle with 1.5 mm strut
(c) Nozzle with 2.5 mm strut
(d) Nozzle with 3.5 mm strut
the insertion of struts (Fig. 5b–d). In addition, the number of shock-cells in the jet
are also reduced in the strut-controlled configurations. However, the strut-controlled
configurations in this case experience no jet deflection, which is in contrast to the
controlled nozzles operating at overexpanded conditions. Therefore, it can be said
that the deployment of a strut in the supersonic nozzle at underexpanded conditions
has no significant impact on thrust vectoring.
4 Conclusions
In this study, efficacy of a strut in controlling the thrust vector of a supersonic nozzle
has been experimentally investigated. The height of strut was varied as 1.5, 2.5 and
3.5 mm. The strut is found to be efficient in deflecting jet flow field at overexpanded
state of the nozzle. A maximum of about 3.6° of jet deflection is achieved with the
strut of height 3.5 mm. The superior performance of the longest strut is due to the
generation of strong bow-shock, which eventually increases the pressure imbalance at

the nozzle exit. The qualitative analysis of the Schlieren images supports the findings
of wall pressure distribution. It has been observed that the strut of height 3.5 mm is
more efficient in deflecting the jet at all levels of expansion prevailing at the nozzle
exit.
Acknowledgements The authors deeply acknowledge the Madras Institute of Technology,

Chennai, India, for providing the supersonic jet-test facility to carry out the experiments.
References
1. Sung H-G, Jeong K, Heo J (2017) Performance characteristics of a pintle nozzle using the
conformal sliding mesh technique. Aerosp Sci Technol 61:85–94
2. Lee J, Park B, Yoon W (2013) Parametric Investigation of the pintle perturbed conical nozzle
flows. Aerosp Sci Technol 26:268–279
3. Sung H, Heo J (2012) Fluidic thrust vector control of supersonic jet using co-flow injection. J
Propuls Power 28:858–861
4. Yagle PJ, Miller DN, Ginn KB, Hamstra JW (2001) Demonstration of fluidic throat skewing for
thrust vectoring in structurally fixed nozzles. J Eng Gas Turbines Power 123:502–507
Implementing Canonical CFD
Techniques to Determine the Effect
of Optimised Aerodynamic Coefficients
on the Performance of a Car for a Given
Race Track
K. Jayesh Balaji, K. Koushik Sundaram, S. P. Hareesh, S. Arun Sundar,
and R. Naveen
1 Introduction
Formula type cars are regulated by strict rules. These cars are aimed at maximising
performance in terms of speed and handling. The performance of these cars depends
on the engine performance, tire performance, aerodynamics, etc [1]. Adding an aero
package for the car provides an increased down force and thus improves the cornering
speeds at the cost of some additional drag. All the components of the car which
come into contact with air produce drag and down force. The major components
which can be designed to improve aerodynamic performance include a pair of front
wings, rear wing, and a diffuser [1]. It is important to strike a balance between the
down force and drag to obtain an optimised performance of the car [2]. The primary
goal while designing aero package for a car will always maintain proper handling
stability, which is ensured by the relative position of the centre of pressure with the
centre of gravity, and this is validated by aeromapping concept [3]. The selection of
optimum aerodynamic coefficients depends on the track. Tracks with long straights
and less corners require low-drag configuration and tracks with more corners require
high down force configurations [4]. Based on the obtained C l , C d values, the wing
profile is designed [5–7]. The performance of the designed wing is simulated by
CFD, taking into considerations of all the proper concepts to be employed for better
results. This paper provides the procedure and the concepts to be implemented during
CFD analysis of the aerodynamic package of the car. By following the given set of
procedures, a team might be able to select an appropriate aerodynamic package for
their car and also perform a better implemented CFD analysis.
K. Jayesh Balaji (B) · K. Koushik Sundaram · S. P. Hareesh · S. Arun Sundar · R. Naveen

PSG Institute of Technology and Applied Research, Avinashi Road, Neelambur, Coimbatore,
Tamilnadu 641062, India
684 K. Jayesh Balaji et al.
2 Methodology
As mentioned earlier, the selection of optimum aerodynamic coefficients depends

on the track. This is done by assigning weights to parameters of track like straight
line speed and cornering speed. From this weighted parameter method, coefficients
are selected for a given track. Artificial intelligence method can also be imple-
mented using algorithms like genetic algorithms [8]. After selection of optimum
coefficients, proper CFD has to be employed, so that the simulated performance is
efficient and yields accurate results. Techniques to be followed in the design proce-
dure include elimination of flow separation, surface refinement, layer refinement,
turbulence modelling and wheel modelling. Elimination of flow separation is done
by the design of multi-element wings [9]. The gap between main wing and flap in a
multi-element wing feeds in new energy flow and eliminates flow separation. Surface
refinement is done to increase mesh nodes near surfaces. Layer refinement is done
for better mesh node positions along flow where separation might happen. Flow over
aerofoils may become turbulent, which increases viscous drag and overall drag. This
effect is substituted by the use of turbulence modelling [10]. K-omega SST model
in Simscale software is used for turbulence modelling. Wheel modelling is done for
the consideration of wheels as moving elements compared to wheels being static,
thus gives accurate results. This is done by use of multiple reference frame (MRF)
concept in meshing. Thus by employing these concepts, use of CFD is effective.
The amount of drag produced is based on the following formula:

1
Drag force = ∗ ρ ∗ cd ∗ A ∗ v2 (1)
2
where
ρ Density of air in kg/m3
Cd Coefficient of drag
A Frontal area in m2
V Velocity of air in m/s
Similarly, the amount of down force produced is based on the following formula:

1
Down force = ∗ ρ ∗ cl ∗ A ∗ v2 (2)
2
Implementing Canonical CFD Techniques … 685
where
Cl Coefficient of lift (down force)
For example, a given rear wing design having particular values of cd and cl , the
drag and down forces are directly proportional to the velocity of travel of the car.
The down force for assumed rear wing at 40 m/s is found to be 360 N and at 80 m/s
is found to be 1480 N. Similarly, the drag at 40 m/s and 80 m/s was found to be 90 N
and 310 N, respectively. These are obtained from the Simscale software.
The airflow across aerofoil depends on the Navier Stokes equation.

dv
ρ = −∇ρ + ρ · g + μ∇ 2 v (3)
dt
It is also based on Bernoulli equation.
Energy per unit volume before = Energy per unit volume after. (4)
Formula Student vehicles are mostly designed for high down force configurations,
since skidpad and autocross tests have significant weightage in the competition. For
scoring high points in these competitions, high cornering speeds are required, and so,
high down force configurations are used. For this research, a sample Formula Student
vehicle is considered, which has a brake horsepower of 94 and weight of 250 kg.
The drag coefficient without aero package was 0.6, and down force coefficient was
−0.01. The track considered was Buddh International circuit. After optimisation of
the coefficients for this track, the drag coefficient of the car was 2.14 and 1.23.
These are found using Simscale software.
The meshing of the car is done by employing the above-mentioned concepts. The
meshing is explained in Fig. 1.
Thus by these process, it was ensured that a right number of mesh nodes are
present at appropriate positions.
The boundary condition for CFD simulation included the following procedure,
the inlet boundary was fixed to have the velocity of vehicle. In this research, the
velocity is considered as 20 m/s. The outlet boundary was fixed to have atmospheric
pressure. The upper and lower boundaries were fixed to be walls, so that they do
not produce viscous drag. As explained earlier, turbulence model was selected as
K-omega SST. The wheel modelling was done, and its effect is shown in Fig. 2.
From this, it is inferred that all parts of the car are at 0 velocity, since only the
airflow has velocity of 20 m/s. But wheel alone has velocity of flow, equal to 20 m/s,
highlighted by different colour, and thus, it is inferred that wheels have velocity.
Thus, all explained concepts are used during meshing.
Fig. 1 Meshing process explanation
Fig. 2 Wheel modelling using MRF
The CFD was done in Simscale software. The results of the simulation were post
processed in Paraview software. Notable results found are discussed below.
Three different slices along the width of the car were taken to study velocity and
pressure at different zones. First, the velocity and pressure near driver are studied.
These are shown in Figs. 3 and 4.
Fig. 3 Pressure distribution at driver section
Fig. 4 Velocity distribution at driver section
These figures imply that pressure at top of wings is high compared to bottom of
wings, thus down force is generated, and this is further proved by high velocities
below wings and low velocities above it. These also explain the drag produced by
the driver.
The pressure and velocities in the tire section explain the drag produced by flow
over tires. These are shown in Figs. 5 and 6.
The pressure produced on top of the car and at the bottom is viewed separately
for better understanding of the difference of pressure created, and this is shown in
Figs. 7 and 8.
Fig. 5 Pressure distribution at tire section
Fig. 6 Velocity distribution at tire section
The colour pattern for the above figures is scaled such that violet indicates high
pressure and yellow indicates low pressure. Thus, these figures explain the pressure
difference over and under the car.
The principle of multi-element wing is considered for front and rear wings in this
research. The influence when used cumulatively is explained in Fig. 9.
Fig. 7 Pressure distribution over car
Fig. 8 Pressure distribution under car
4 Conclusions
The steps done in this paper portray an efficient method for the design and simulation
of the designed aerodynamic package. Lap time simulation for the Formula Student
car is done with OptimumLap software with and without the installation of aero-
dynamic package, which was designed and simulated from the method highlighted
Fig. 9 Coefficient of pressure distribution for multi-element front wing
above. The results showed the reduction of lap times after installation of aerody-
namic package. Moreover, higher speed is achieved in skidpad and autocross tests.
Thus, installation of aerodynamic wings for the Formula car proved to be effective
(Fig. 10).
The blue line represents car without aerodynamic package and orange represents
after aerodynamic package. Although blue line has more top speed, due to less drag,
more acceleration, and braking is done on corners since traction is not sufficient.
But orange line does very good corners and thus does a lower lap time of 137.38 s,
compared to 139.54 s, without aerodynamic package. The maximum velocity at a
skidpad event which has 50 m radius, before installation is 97.8 kmph, but after
installation is 138.1 kmph, because of huge increase in C l [11]. But, the main factor
to be considered is the aerodynamic stability of the vehicle.
Fig. 10 Speed v/s distance graph for FS car without and with aerodynamic package
References
1. Toet W (2013) Aerodynamics and aerodynamic research in Formula 1. Aeronautical J

117(1187)
2. Agathangelou B, Gascoyne M (1998) Aerodynamic design considerations of a formula 1 Racing
Car. SAE Technical paper Series-980399
3. Craig C, Passmore MA (2014) Methodology for the design of an aerodynamic package for a
formula SAE vehicle. 2014-01-0596
4. Wordley S, Saunders J (2006) Aerodynamics for formula SAE: initial design and performance
prediction. SAE World Congress Detroit, Michigan
5. Prasad KS, Krishna V, Ashok Kumar BB (2016) Aerofoil profile analysis and design
optimisation. J Aerospace Eng Technol 3(2)
6. Ziemkiewicz D (2016) Simple analytic equation for airfoil shape description
7. Vadgama TN, Patel A, Thakkar D, Vala J (2015) Structural analysis of formula one racing car.
Int J Adv Res Eng Sci Technol 2(5)
8. Bhatnagar UR (2014) Formula 1 race car performance improvement by optimization of the
aerodynamic relationship between the front and rear wings
9. Zhang X, Zerihan J (2003) Aerodynamics of a double-element wing in ground effect. AIAA J
41(6)
10. Rehnberg S, Börjesson L, Svensson R, Rice J (2012) Race car aerodynamics—the design
process of an aerodynamic package for the 2012 Chalmers formula SAE Car. 2013-01-0797
11. Prasanth A, Biswal S, Gupta A, Barodawala A (2016) Complete design and optimization of
the aerodynamics of a FSAE car using solid works ANSYS and XFLR5. In: Proceedings of
the world congress on engineering 2016, vol II, WCE 2016, June 29–July 1, 2016
Flow Visualization and Computational
Studies on Unsteadiness Due to 3-D
Shock-Induced Separation of Flow Over
a Square Protrusion
K. Ramachandra, S. Bhardwaj, and R. Sriram
1 Introduction
A shock present in a supersonic flow is essentially unsteady in nature due to

the phenomenon of shock boundary layer interaction (SBLI). This particular
phenomenon has been extensively researched for over 50 years. One of the key
objective in SBLI was to understand the reason for oscillation of the shock wave.
Shock oscillations have a wide spectrum of frequencies. The high frequency is related
to the upstream turbulence in the boundary layer while the lower frequency which is
one or two orders of magnitude lower has been a topic of debate. Some researchers
believed upstream turbulence as the reason while the others speculated the down-
stream separation region as the source of low frequency unsteadiness until Piponniau
et al. [1] proposed shear layer entrainment-recharge mechanism for the dynamics of
shock. Humble et al. [2] in their study correlated high frequency oscillation with the
incoming boundary layer scales. Later, in another study, Souverein et al. [3] found
out that oscillations are better correlated with incoming turbulent boundary layer for
weakly separated flow, i.e. L/δ x ≈ 2.2, where L is the separation length while δ x is
the local turbulent boundary layer thickness. A lower correlation was observed for
strongly separated flow, i.e. L/δ x ≈ 4.2. Therefore, a consensus was attained that the
upstream turbulence leads to low frequency oscillation in case of weakly separated
while downstream mechanism is the cause in strongly separated flows.
K. Ramachandra (B)
Department of Mechanical Engineering, PSG College of Technology, Coimbatore, India
e-mail: ammurama2@gmail.com
S. Bhardwaj · R. Sriram
Department of Aerospace Engineering, IIT Madras, Chennai, India
e-mail: sourabhbhardwaj12@gmail.com
R. Sriram
e-mail: r.sriram@iitm.ac.in
694 K. Ramachandra et al.
In the present study, a square-faced protrusion is introduced in the supersonic flow

and dynamics of the 3-D shock formed are studied experimentally, complemented
with some computational results. One of the key objectives of such interactions is to
identify the mean separation line. Sufficient study had been done on the methods for
investigating three-dimensional flow structures such as the surface oil flow technique
by several researchers and were compiled by R. L. Maltby. Theoretical study along
with several experiments in low-speed and high-speed wind tunnels were carried
out to optimize the parameters for surface oil flow visualization by Squire et al. [4].
The ability of the surface oil flow technique to reveal surface topologies in a 3-D
flow was discussed in detail for the flow over Microvortex generators in a study by
Lu [5]. In our study, the surface oil flow visualization is employed for identifying
the mean separation line which would give an approximate region to further do
pressure measurements using fast-response transducers to study the unsteadiness of
the oscillating shock.
2 Methodology
2.1.1 Test Section and Test Model Details
The experiments were performed in the blowdown supersonic wind tunnel facility at
IIT Madras with a maximum gauge pressure of 12 bar in the settling chamber. The
inlet of the convergent-divergent nozzle is of a square cross section with side 100 mm.
The outlet dimensions of the nozzle are 100 mm × 120 mm. The test section used
is made of aluminium with the same rectangular cross section as the nozzle exit and
has a length of 407 mm. The dimensions of the side glass windows used are 320 mm
× 140 mm. Supersonic flow was attained in the test section with a free stream Mach
number (M ∞ ) of 2.97. The uncertainty in the free stream Mach number was 8.33%.
The test model is a cube with side 15 mm (which is twice the local boundary layer
thickness obtained using computations). The protrusion was placed at the centre of
the base (in spanwise direction) at a distance of 167 mm from the nozzle exit (Fig. 1).
2.1.2 Details of Schlieren
A light source with a slit (for divergence) is used for Schlieren flow visualization. A
concave mirror M1 with a focal length of 2 m is placed at a distance of 2 m from
the slit which makes the diverging rays parallel. The parallel beam is made to pass
through the test section glass window. The rays are then converged using another
concave mirror M2 with a focal length of 1 m. At the focus of mirror M2, a knife
edge is placed to block a portion of the beam from the mirror. A convex lens with a
Flow Visualization and Computational Studies … 695
Fig. 1 Schematic of the test section with protrusion
focal length of 53 cm is placed to converge the rays back to the camera. A high-speed
Photron camera was used to record the images with a frame rate of 30,000 fps. The
Photron Fastcam Viewer software was used to vary the frame rate, shutter speed and
record the images.
2.1.3 Details of Surface Oil Flow Visualization
The surface oil flow visualization was performed using an oil, a powdered pigment
or a dye and a dispersing agent. Since there is a limitation in the running time for
high speed tunnels, the viscosity of the oil (μoil ) to be used was decided based on
the relation formulated using experiments done with oils of varying viscosity with
different running times in [4] and is given by:

μoil
t = (36, 000 ± 12, 000) (1)
qc f
where q denotes the dynamic head and c f denotes the local skin friction coefficient.
Since the wind tunnel has sufficiently long runtime (of more than 30 s), engine
oil SAE40 was chosen. Titanium dioxide (TiO2 ) powder which is white coloured
was used as the pigment. Oleic acid was used as the dispersing agent. Optimization
of pigment to oil proportion was done, and the best proportion for the conditions
prevailed was found to be 4.5 parts of TiO2 to 3 parts of SAE40 with few drops of
oleic acid. Random distribution of drops of the mixture was found to give better flow
streaks than the ordered distribution.
The 3-D model similar to the experimental arrangement was created. The square-
faced protrusion of height 15 mm was placed at a distance of 95 mm from the nozzle
exit where the local boundary layer thickness (δ x ) is 7.5 mm. The geometry is meshed
in ICEM CFD. The grid size is 0.9 mm with inflation layer near the bottom wall to
capture the boundary layer accurately. This resulted in total number of nodes of 5
million.
Three-dimensional simulations were performed using ANSYS Fluent software.
Density-based transient simulations were performed. K-omega SST turbulence
model with compressibility effects was used. Sutherland’s viscosity model (using
three-coefficient method) was used. No-slip boundary condition was given for the
bottom and top walls, whereas translational-type periodic boundary condition was
used for the side walls. The inlet total gauge pressure was given as 5.5 bar with total
temperature as 300 K. The outlet gauge pressure was given as −87 kPa. Roe-FDS
flux formulation with explicit transient formulation was used to capture the shock
oscillations. The time stepping of 7.28 × 10−7 s was done for a better temporal
resolution of the flow. When the residuals dropped to the order of 10−3 , convergence
of solution was considered to be achieved. All the simulations were done in Virgo
Supercluster, at HPCE, IIT Madras.
3.1 Experimental Results
Schlieren flow visualization was done for the Mach 2.97 flow past the square protru-
sion to capture the flow structure near the model. Figure 2 shows the most upstream
and downstream positions of the separation shock obtained in the flow.
The Schlieren images show clearly the boundary layer separation due to shock–
boundary layer interaction, and the shock pattern formed due to the presence of
protrusion which is twice higher than the incoming turbulent boundary layer thick-
ness. The images also show the presence of shock oscillations with the corresponding
frequency of 1100 Hz, measured by taking snapshots of the flow field captured by a
(a) (b)
Fig. 2 Schlieren images showing the extreme positions of the separation shock (a most upstream
position and b most downstream position)
high-speed camera. The reasons for this low frequency oscillations which are due to a
combination of downstream mechanism and upstream mechanism can be understood
well by the review work done by Clemens and Narayanaswamy [6].
A higher temporal resolution taken for Schlieren images allowed us understand
the dynamics of the shock such as frequency and maximum and minimum excursions
of the shock, but an overall three-dimensional picture of the shock structures could
not be obtained. This motivated us to perform the oil-flow visualization to get the
mean separation line due to the shock formed.
The surface oil flow visualization was first carried out for Mach 2.97 flow over a
2-D un-swept 26° compression ramp with height as 15 mm (2δ x ) to optimize the oil-
pigment proportions to get mean separation line for the flow over a two-dimensional
model which has been well studied in the past years. The average separation distance
was clearly observed at a distance of 14 mm from the leading edge of the ramp (i.e.
x sep /h = 0.933). The surface oil flow visualization was then carried out for the Mach
2.97 flow past the square protrusion with height 15 mm (2δ x ) to capture the three
dimensionality of the separate shock oscillations.
Figure 3 clearly represents the three-dimensional nature of the unsteady separation
shock. Since the square protrusion is a bluff body to the supersonic flow, the shock is
detached and curved (bow shock). The detached separation shock is near normal at
Fig. 3 Surface oil flow

visualization of Mach 2.97
flow over a square protrusion
(a) (b)
Fig. 4 Mach number contours in the sideview midplane showing the extreme positions of the
separation shock (a most upstream position and b most downstream position)
the base and hence is a strong shock interacting with the boundary layer resulting in a
large separation distance. Figure also shows the curved mean separation line, where
oil is accumulated due to the shear stress being zero which is slightly downstream
of the separation shock foot.
The reverse flow patterns are visible in the separated bubble region. These are the
characteristics of the horseshoe vortices surrounding the protrusion that is caused
due to the spiralling separating surfaces as discussed by Babinsky and Harvey [7].
The curved shock ends with weak boundary layer interactions at the side walls. The
surface flow visualization also shows the highly symmetric nature of the streaks
which is indicative of the fact that the shock pattern is symmetric in the spanwise
direction. The average separation distance of the oscillating shock at the centre in
the spanwise direction was found to be 35 mm (i.e. x sep /h = 2.33).
3.2 Computational Results and Comparison
The flow simulations were performed for Mach 3 flow over a square protrusion of
side 15 mm (2δ x ). The square protrusion is a 3-D case of forward-facing step (2-D)
whose flow has been studied by Murugan and Govardhan [8]. Figure 4 shows the
Mach number contours of the side view (midplane in spanwise direction) obtained
for the most upstream and downstream location of the separation shock. The shock
pattern in the contour resembles closely to the Schlieren images obtained in the
experiment.
The x/h for the Mach number contour (shock position) was found to be 2.735 and
3.2 for the downstream and upstream locations of the shock, respectively. This gives
the average x/h on the spanwise centred plane to be 2.95 based on the Mach number
contour.
Figure 5 is the x-component velocity contour taken at a spanwise (x-z) plane
closer to the base wall (y/h = 0.25) to depict the average separation location (where
u = 0). The 3-D shock pattern which is much similar to the oil flow streak pattern
was observed. The mean x sep /h (from Fig. 5) through computations was found to be
2.7 which is much closer to the experimentally obtained result. It is also noted that
Fig. 5 u velocity contour for

the average shock position
for the topview plane at
y/h = 0.25
the extent of separation reduces in the spanwise direction as we go away from the
protrusion since shock becomes weaker away from the protrusion.
The amplitude of shock oscillation was found to be 0.5 δ x from computations
which is significant when compared to the separation length. The frequency of oscil-
lations observed was 1056 Hz which is higher compared to the 2-D cases that had
been studied extensively in the existing literature. Also, the frequency obtained using
computations is very close to the frequency calculated using the Schlieren images.
Having obtained the mean separation location of the separation shock for the
supersonic flow over a square protrusion using Surface oil flow visualization and
having studied the nature and the position of shock pattern through Schlieren images,
the pressure sensors can thus be placed at strategic locations both in streamwise and
spanwise directions to perform time series analysis of unsteady pressure, and thus,
correlations can be developed to study the dynamics of the shock oscillations.
4 Conclusions
The shock-induced flow separation due to square protrusion is characterized using

surface oil flow visualization technique, complemented by numerical simulations.
The unsteadiness of the separation shock has been investigated using time-resolved
Schlieren. The mean separation distance for the flow was observed to be x sep /h = 2.33
using experiment which was close to the computational result of x sep /h = 2.7. With
those results obtained, the optimal positions of the pressure sensors are to be identified
to measure the time variation of unsteady pressure and thus obtain correlations to
study the unsteadiness of the 3-D shock oscillations further.
Acknowledgements We thank IIT Madras for the permission to conduct experiments, HPCE
supercluster facility for carrying out computations and Hemanth Chandravamsi, MTech student
at IITM, for his great support in experiments.
References
1. Piponniau S, Dussauge JP, Debieve JF, Dupont P (2009) A simple model for low frequency
unsteadiness in shock-induced separation. J Fluid Mech 629:87–108
2. Humble RA, Elsinga GE, Scarano F, van Oudheusden BW (2009) Three dimensional
instantaneous structure of a shock wave/turbulent boundary layer interaction. J Fluid Mech
622
3. Souverein LJ, Dupont P, Debiève JF, Oudheusden BWV, Scarano F (2009) Effect of interaction
strength on unsteadiness in shock-wave-induced separations. AIAA J 48(7):1480–1493
4. Squire LC, Maltby RL, Keating RFA, Stanbrook A (1962) Flow visualization in wind tunnels
using indicators. Compiled by Maltby RL. AGARDograph 70:39–74
5. Lu FK (2010) Surface oil flow visualization. Eur Phys J Spec Top 182:51–63
6. Clemens NT, Narayanaswamy V (2014) Low frequency unsteadiness of shock wave/turbulent
boundary layer interactions. Annu Rev Fluid Mech 46(1):469–492
7. Babinsky H, Harvey JK (2011) Shock wave—boundary-layer interactions. Cambridge Univer-
sity Press, USA
8. Murugan JN, Govardhan RN (2016) Study of shock wave/turbulent boundary layer interaction
over a forward-facing step. J Fluid Mech 807:258–302
Performance Analysis of Two-Stage
Evaporative Cooler: A Review
Shubha Deep Paul, Sheetal Kumar Jain, Ghanshyam Das Agrawal,

and Rohit Misra
Nomenclature
Tdi Inlet Dry bulb temperature

Tdo Outlet Dry bulb temperature
Twi Inlet Wet bulb temperature
ma Mass flow rate of air
hi Enthalpy (kJ/kg) of entering air
ho Enthalpy (kJ/kg) of leaving air
Tdb Dry bulb temperature
Abbreviation
IEC Indirect evaporative cooler

DEC Direct evaporative cooler
EER Energy efficiency ratio
S. D. Paul (B) · S. K. Jain · G. Das Agrawal

MNIT Jaipur, Jaipur, Rajasthan, India
R. Misra
Government Engineering College Ajmer, Ajmer, Rajasthan, India
702 S. D. Paul et al.
Greek symbols
1 Effectiveness
η Efficiency
1 Introduction
Evaporative cooling systems have attracted much attention as an alternative to the

conventional air conditioning system because it helps to save refrigerated power
costs. Evaporative cooling technology is more sufficient and eco-friendly technology.
Therefore, in the hot and dry tropical regions, the evaporative cooling technology can
be used for saving a large amount of energy. Evaporative cooling is not only limited
to thermal comfort, but also an effective method of storage of foods and vegetables
of moderate respiration rate.
Evaporative cooling is used in many other applications, such as
• Animal housing facility cooling
• Electronics and optic fiber equipment cooling
• Turbine engine air intake cooling
• Exterior spot cooling
• Greenhouse, laundries, and manufacturing process cooling
• Power plant evaporative cooling towers.
Benefits of using evaporative cooling in its growing climatic ranges are as follows:
• Lower operating and power cost
• Mostly applicable for rural region
• Maintenance cost is much lower
• Ventilating effect is also great if needed
• Better air distribution without ducts
• Usually lower outdoor noise levels
• Evaporative cooling units remove moist of the air contaminants, such as dust, dirt,
bacteria, soluble gases, and nutrients
• Applicable in the domestic and industrial application in wide-scale
• It gives high effectiveness.
2 Direct Evaporative Cooler
The direct evaporative cooling (DEC) system uses a cellulose media with the water
contact surface where the air is passed through it at a uniform rate. The principle
underlying direct evaporative cooling is the easy conversion of sensible to latent heat.
Non-saturated air is cooled by exposure to free and colder water, both thermally
Performance Analysis of Two-Stage Evaporative Cooler: A Review 703
Fig. 1 Direct evaporative cooling system a Schematic diagram b and psychometric process [9]
isolated from other influences as shown in Fig. 1. Some of the air’s sensible heat
transfers to the water and becomes latent heat by evaporating some of the water.
The latent heat follows the water vapor and diffuses into the air. That exchange of
sensible for latent heat tends to progress until the air is saturated and air and water
temperatures and vapor pressures equalize. It is called adiabatic saturation because
no external heat is involved; saturation is approached purely by conversion of the
air’s existing sensible heat.
2.1 Limitations
• Due to the increase in relative humidity, which may cause flu-like illness when
coming in direct exposure to human
• Evaporative cooling pads required a continuous water supply
• It is only suitable for dry and hot climate.
3 Indirect Evaporative Cooler
Indirect evaporative cooling (IEC) cools air by the evaporation of water not contacting
it and so does not increase the cooled air’s moisture content. Indirect evaporative
cooling involves two streams of air passing through coolers simultaneously, but not
contacting each other. As usually identified:
• Primary Air: the air being cooled that will later become supply air for cooling
rooms.
• Secondary Air: the air that contacts and helps evaporate the water that cools the
heat exchanger surfaces that cool the primary air. The secondary air usually comes
from outdoors and is discharged there again after use.
The hearts of all indirect evaporative coolers are the heat-exchange surfaces that
separate the primary air from the secondary air, and the water the latter is evaporating.
Those surfaces absorb heat from the primary air and transfer it to moist secondary
air, which carries it away and discards it outside. The surfaces may be metal or plastic
plates, sheets, or tubes or even revolving wheels of porous metal. They must conduct
heat readily, keep the two air streams apart, and resist corrosion [1].
Limitation
• Effectiveness is less
4 Two-Stage Evaporative Cooler (Indirect/Direct)
To avoid the high relative humidity of direct evaporative cooler and by considering
the low effectiveness of indirect evaporative cooler, the combined of both the system
is called two-stage evaporative cooler as shown in Fig. 2. The principle of a two-stage
evaporative cooler is to reduce the wet-bulb temperature of outdoor air before entering
the DEC unit. The heat-exchange process reduces the dry bulb temperature of the
air stream without changing its humidity. After that, the air stream is introduced into
the DEC unit. As a result, the air temperature approaches the wet-bulb temperature
of the pre-cooled air. This temperature is lower than the wet-bulb temperature of the
ambient air.
Fig. 2 Two-stage evaporative cooling system a Schematic diagram b and psychometric process
[10]
4.1 Thermal Performance
Thermal performance analysis of two-stage evaporative cooler is obtained using

below-mentioned equations.
(a) Adiabatic effectiveness (ε) of two-stage evaporative cooler: It is the ratio
of actual drop in dry bulb temperature of the air to the maximum possible
temperature drop or wet-bulb depression.
Tdi− Tdo
ε=
Tdi− Twi
(b) Cooling capacity of two-stage evaporative cooler:
Cooling capacity = m a × (h i − h o )
(c) Efficiency of evaporative pad (ï):
Tdb air entering pad − Tdb air leaving pad

η=
Tdb air entering pad − Twb air entering pad
5 Literature Survey on a Two-Stage Evaporative Cooler
Many researchers had performed the worked on two-stage evaporative cooler.

Heidarinejad et al. [2] used two air simulator to simulate outdoor design condi-
tion of different cities in primary and secondary airstream, and results show that
effectiveness of IEC stage varies over range of 55–61% and effectiveness of Indi-
rect/Direct stage varies over a range of 108–111% under various outdoor condition.
They compared the power consumption of Indirect/Direct evaporative cooler with
mechanical vapor compression system. It was found that more than 60% power
saving could be obtained by the Indirect/Direct evaporative cooler.
Jain [3] developed and tested the two-stage evaporative cooler by using plate type
heat exchanger as IEC unit and humidifier, which is stuffed with wooden shave as
DEC unit. The result shows that the effectiveness varies from 110 to 120%, and
relative humidity of room air was observed from 50 to 75% against the 15–40%
of ambient air. Kim and Jeong [4] calculated the cooling performance of a 100%
outdoor air system integrated with two-stage evaporative cooler. It was observed that
effectiveness of two stage varies from 83.4 to 97.8% in an intermediate season where
outdoor air temperature varies from 20 to 25 °C range and humidity vary from 20 to
80%.
Jaafarian and Kazemian [5] developed a mathematical model of two-stage evapo-
rative cooling, which consists of two direct evaporative coolers and a heat exchanger.
It was noticed that the overall effectiveness of system is more sensitive to evaporative
pad efficiency than that of overall heat transfer coefficient.
El-Dessouky et al. [6] constructed a two-stage evaporative cooler and tested in
Kuwait environment during the summer season when dry bulb temperature is higher
than 45 °C. They showed that the efficiency of IEC unit, DEC unit, and IEC/DEC
unit is in range of 20–40%, 63–93%, and 90–120%, respectively.
Sharma and Darokar [7] designed two-stage evaporative cooler and provided two
air simulator to simulate outdoor design condition in primary and secondary air
stream. They showed that the effectiveness of IEC and IEC/DEC varies between
55 and 61% and 108–111%, respectively. They also showed that the average water
consumption of two-stage evaporative cooler is 55% more than the direct evaporative
cooler, and power consumption was 33% of mechanical vapor cooling system.
Mohammed [8] developed an experimental setup on two-stage evaporative cooler.
He used heat exchanger as IEC unit and 15 cm thick pad as DEC unit. It was observed
that the IEC effectiveness varies from 55 and 65% and IEC/DEC effectiveness over
a range of 90–110%. It was also seen that the average water consumption of two-
stage evaporative cooling system was 40% more than the direct evaporative cooling
system.
Al-Juwayhel et al. [11] experimentally investigated four different arrangements of
an evaporative cooler and found that the IEC/DEC had the highest EER than the other
three arrangements. It was also found that the DEC had the lowest effectiveness.
Alklaibi [12] experimentally investigated the internal two-stage evaporative cooler
and compare its thermal performance with direct evaporative cooler. Results showed
that the efficiency of internal two-stage evaporative cooler fluctuates around 0.70,
while the efficiency of direct evaporative cooler fluctuates around 0.66. The results
also showed that the efficiency of direct evaporative cooler increases by 12% and the
internal evaporative cooler increases only by 5% when fan speed switches from high
to low.
6 Conclusions
From the above review paper, following conclusions are made:

• Two-stage evaporative cooler consumes less power than conventional air condi-
tioner.
• Effectiveness of two-stage evaporative cooler is more than another cooling system.
• The EER of two-stage evaporative cooler has best value.
• Pad thickness and material, fan speed, and water circulation are the main factor
that affects the effectiveness of evaporative cooler.
• Effectiveness is also depending on outdoor temperature and relative humidity.
References
1. Watt JR (1986) Evaporative air conditioning handbook. Chapman and Hall

2. Heidarinejad G, Bozorgmehr M, Delfani S, Esmaeelian J (2009) Experimental investigation
of two-stage indirect/direct evaporative cooling system in various climatic conditions. Build
Environ 44:2073–2079. https://doi.org/10.1016/j.buildenv.2009.02.017
3. Ã DJ (2007) Development and testing of two-stage evaporative cooler. 42:2549–2554 (2007).
https://doi.org/10.1016/j.buildenv.2006.07.034
4. Kim MH, Jeong JW (2013) Cooling performance of a 100% outdoor air system integrated with
indirect and direct evaporative coolers. Energy 52:245–257. https://doi.org/10.1016/j.energy.
2013.02.008
5. Mohammad J, Ehsan KM (2017) Performance analysis of a two-stage evaporative cooler.
403:1–8. https://doi.org/10.1051/meca/2017013
6. El-Dessouky H, Ettouney H, Al-Zeefari A (2004) Performance analysis of two-stage evapora-
tive coolers. Chem Eng J 102:255–266. https://doi.org/10.1016/j.cej.2004.01.036
7. Sharma A, Darokar H (2018) Two stage indirect/direct evaporative cooling. 41–46
8. Kareem Mohammed A (2013) Experimental performance of two-stage evaporating cooling
system. Sch J Eng Technol 1:122–127
9. Mujahid Rafique M, Gandhidasan P, Rehman S, Al-Hadhrami LM (2015) A review on desiccant
based evaporative cooling systems. Renew Sustain Energy Rev 45:145–159. https://doi.org/10.
1016/j.rser.2015.01.051
10. Amer O, Boukhanouf R, Ibrahim HG (2015) A review of evaporative cooling technologies. Int
J Environ Sci Dev 6:111–117. https://doi.org/10.7763/ijesd.2015.v6.571
11. Al-Juwayhel F, El-Dessouky H, Ettouney H, Al-Qattan M (2004) Experimental evaluation of
one, two, and three stage evaporative cooling systems. Heat Transf Eng 25:72–86. https://doi.
org/10.1080/01457630490486292
12. Alklaibi AM (2015) Experimental and theoretical investigation of internal two-stage evapo-
rative cooler. Energy Convers Manag 95:140–148. https://doi.org/10.1016/j.enconman.2015.
02.035
A Review on Numerous Solar Drying
Techniques
Dilip Kumar, Sheetal Kumar Jain, Ghanshyam Das Agrawal,

and Rohit Misra
1 Introduction
Preservation of fruits, vegetables and food is essential for keeping them for a long time
without further deterioration in the quality of the product. Several process technolo-
gies have been employed on industrial scale to preserve food products; the major ones
are canning, freezing and dehydration. Solar drying is best technology to preserve
food. Drying means that it reduces the moisture content in perishable items to improve
their shelf life as well as quality of the products. It is a process of moisture removal
by the phenomenon of simultaneous mass and heat transfer. The product can be dried
in the solar dryer in the drying air temperature varying between 45 and 60 °C which
is more suitable drying air temperature range for many agriculture products.
1.1 Advantages of Solar Dryer
• Temperature developed by solar air heater in dryer is more than ambient air
temperature and movement of air increases the drying rate.
• Products are placed inside the dryer so it is protected from dust, birds, insects,
animals and rain.
• Higher temperature reduces the chances of spoilage of products by micro-
organisms.
• Higher drying rate improves the productivity at low cost.
• The dryer is waterproof; therefore, there is no need to be moved during raining.
• Economically cheap way to get the good quality products.
• Use of renewable and sustainable energy reduces impact on environment.
• Easily and economically availability of construction material of dryer.
D. Kumar (B) · S. K. Jain · G. Das Agrawal · R. Misra

710 D. Kumar et al.
• Less payback periods.

• It is zero energy cost system.
• Reduce post-harvest losses.
• Safe and hygienic.
• Improve financial condition of farmers.
• Product is uniform in colour, texture, appearance and taste.
1.2 Limitations of Solar Dryer
• Dilute nature of solar radiation.

• Sometime quality of goods is not obtained as required.
• Efficiency is less.
• Need of thermal energy storage system for continuous drying.
2 Drying Principle and Mechanism of Drying
Drying principle:—Drying is basically a phenomenon of removal of liquid by evap-

oration from a solid. Mechanical methods for separating a liquid from a solid are not
generally considered drying. The main energy consumption takes place in conver-
sion of water present in product to vapour (2228 kJ/kg at 101.3 kPa.). The water may
be contained in the solid in various forms like free moisture or bound form which
directly affects the drying rate.
Mechanism of drying:—Drying basically comprises of two fundamental and
simultaneous processes: (i) heat is transferred to evaporate liquid, and (ii) mass is
transferred as a liquid or vapour within the solid and as a vapour from the surface.
The factors governing the rates of these processes determine the drying rate. The
various dryers may utilize heat transfer by convection, conduction, radiation or a
combination of these.
3 Classification of Solar Dryer
3.1 Direct Solar Dryer
It is a type of dryer in which product is placed in the dryer to absorb direct sun
radiation. The phenomenon happens in this is natural convection so it is also called
natural convection cabinet dryer. The main issue with this dyer is that direct sun
radiation reduces the quality of the product. It consists of a drying chamber covered
A Review on Numerous Solar Drying Techniques 711
with a transparent cover made of glass or plastic. The drying chamber is usually a
shallow, insulated box with air holes in it to allow air to enter and exit from the dryer.
3.2 Indirect Solar Dryer
In this dryer, first, air is heated in the solar air heater; then, this warm air is passed
through the dryer cabinet to dry product. In comparison with direct solar dryer, the
product obtained by indirect solar is better with good quality and texture. After solar
air heater, warm air is passing through dryer, and it is vent out through chimney
for moisture removal. In indirect type of solar drying systems, a better control over
drying is achieved. An efficiency of solar dryer is improved by integration of thermal
energy storage material.
3.3 Forced Convection and Natural Convection Solar Dryer
Forced convection—In this dryer, the movement of air is created by the external
means like fan and blower. The fan or blower is placed at solar collector inlet. It is
normally referred to as active dyer. It is also improving the drying efficiency of the
dryer and reduces the drying time. The product obtained from this dryer is of good
quality.
3.4 Natural Convection
In this dryer, natural movement of air takes place due to density difference of air. It is
also called passive dryer. The quality of product obtained from this dryer is not good
as natural convection dryer. Figure 1 represents the various types of solar dryer.
4 Review on Various Drying Techniques
Bhardwaj et al. [1] performed experiments on indirect forced convection solar dryer
integrated having sensible heat storage material (SHSM) and phase change material
(PCM). The experiment was performed at meteorological environment of Himalayas
(latitude 30.91 °N). The experimental study was conducted on Valeriana Jatamansi
(a medicinal herb). The drier with heat storage material (SHSM and PCM) enables
to maintain consistent air temperature inside the drier. The Valeriana Jatamansi was
dried from initial moisture content 89% to the final moisture content 9% (wet basis).
Drying time for reduction of moisture content to 9% was 120 h in comparison with
712 D. Kumar et al.
Fig. 1 Direct evaporative cooling system a schematic diagram b and psychometric process
216 h and 336 h for without heat storage material and shaded drying, respectively.
The maximum temperature inside drying chamber was recorded to be 60 °C. Drying
started at 9 am and continued until 9% moisture content in the product was achieved.
The average energy and exergy efficiency of the solar collector without use of SHSM
found to be 9.8% and 0.14%, respectively, whereas by using SHSM, these values
improved to 26.10% and 0.81%, respectively.
El-Sebaii and Shalaby [2] designed and fabricated an indirect-mode forced
convection solar dryer. The thermal performance of the solar dryer under Tanta
(latitude, 30o 47 N and longitude, 31 o E) prevailing weather conditions was experi-
mentally investigated. The experimental investigation was done on thymus and mint.
It consists of V-grooved plate solar air heater. The experiment was conducted during
summer month of 2009. The initial moisture content of thymus and mint was 95%
and 85% (wet basis), respectively, at initial temperature of 29 °C. The sample size
was taken 2 kg for mint and 4 kg for thymus. The average solar radiation and ambient
temperature during experiment were 592 W/m2 and 32.9 °C, respectively. The drying
air temperature inside the dryer during experiment was 39–54°C for both mint and
thymus. Time taken for final moisture content (11 ± 0.5%) for both thymus and mint
34 and 5 h, respectively.
Gupta et al. [3] designed and fabricated a solar dryer for application in agro-
products. It consists solar collector with baffles. The location of experiment was
Nagpur, India. The average dimension of solar dryer was (0.5 × 0.47 × 0.45 m). The
average solar radiation (daily) on horizontal surface was 657 W/m2 . The minimum
and maximum temperature for drying was 35 °C and 65°C, respectively.
Sandali et al. [4] used passive solar dryer with the help of geothermal water heat
exchanger. It consists double level tubular heat exchanger which contains geothermal
water. The experiment was conducted at Lenreza laboratory, University of Ouargla,
Algeria. The temperature of circulating water was 70 °C. The experiment was started
at 9 am on the first day and finished at 9 am in the second day. The maximum solar
radiation was observed to be 643 W/m2 . The axial air velocity was varying between
0.05 and 0.15 m/s. The dryer with heat exchanger is produced lowest and highest
temperature of 46 °C and 58 °C, respectively.
Badaoui et al. [5] experimentally studied the drying of tomato pomace waste
by solar greenhouse dryer. The experiment was conducted at Algeria. The drying
temperature of the product with the help of this dryer varies between 40 and 58 °C.
This dryer helps to municipal corporation to control waste produce by food processing
industries. It took 5 h to get the final product. The effective diffusivity was varied
between 3.2 E-9 and 4.7 E-10 m2 /s with activation energy 75.6 kJ/mol. The drying
was carried out on 10/7/2017 in Bouismail. The test was beginning at 9 am and end at
2 pm. The maximum solar radiation during experiment was achieved to be 809 W/m2 .
The pomace drying was obtained to be from initial moisture content 88.5% to the
final moisture content of 10.3%.
Mewa et al. [6] experimentally investigated the beef drying kinetics in a solar
tunnel dryer. The solar radiation, ambient temperature, relative humidity and
air velocity during experiment were observed to be 476.3–1000 W/m2 , 38.9 °C
(maximum) and 21.3 °C (minimum), 48–69.5%, 0.02–0.18 m/s, respectively. The
experimental results were fitted to different models from and found that most suit-
able model for beef drying was page model. The effective moisture diffusivity was
varied between 2.282 E-10 and 2.536 E-10 m2 /s. The experiment was held at Isiolo
county in Kenya. The experiment was conducted in months of August and September
2017. The drying duration for all samples was 11 h. The sample size for experiment
was 20 kg, and it was dried to 7.5 kg. The sample moisture content was reduced to
2.32–9.56% (dwb) in 11 h in solar tunnel dryer as compared to 24.76% (dwb) in a
similar sample size with open sun drying.
Essalhi et al. [7] dried the grapes by passive solar dryer and open sun drying
method. It was designed at solar energy and environment laboratory of Mohammed
V University, Rabat Morocco. It consists storage tank to maintain temperature of
grapes higher than ambient temperature during night and off sunshine hours. The
714 D. Kumar et al.
moisture content of grapes was reduced from 79.8% (w.b.) to 20.2% (w.b.) in 120
and 201 h for indirect solar dryer and open sun drying, respectively. Drying data
fitted to 10 thin layer drying models. Through these models, it was obtained that
Midilli et al. model is found to best for showing drying behaviour of grapes in both
indirect and open sun drying. The effective moisture diffusivity of grapes for sun
drying and indirect solar drying when applying fick’s diffusion model were observed
to be 2.34 E-11 and 4.08E-11 m2 /s. The quantity of grapes for indirect soar dryer
was (300.84 ± 0.01) gm and for open sun drying (300.42 ± 0.01) gm. The maximum
solar radiation and ambient temperature was obtained to be 938 W/m2 and 32 °C
during experiment.
Rabha et al. [8] comparative studied the drying of ghost chilli pepper in a forced
convection solar tunnel dryer. Two double pass solar air heater with semi-continuous
type tunnel dryer and shell and tube heat exchanger has been used. The experiment
was conducted at Guwahati, Assam (latitude 26.18o ). The sample size for experiment
was taken to be 9 kg of ghost chilli pepper. The experimental data was fitted to
11 thin layer drying models. The page and modified page model were best fitted
for the open sun drying. The drying experiment was carried out at every day for
8 h. The experiment duration was 29th September 2015 to 3rd October 2015. The
ambient temperature, solar radiation and relative humidity during experiment were
noticed to be 29–37 °C, 166 W/m2 −1011 W/m2 with an average of 641 W/m2 and
57.4%–85.5%, respectively. The drying air temperature was varying between 44 and
66 °C with an average of 57 °C. The reduction of moisture content of sample from
589.6%(db) to 12%(db) in 123 h and 193 h for solar dryer and open sun drying.
5 Conclusions
An extensive review has been outlined of the multiple models, design information
and operational principles of the broad range of practical models of solar energy
drying devices. It is possible to identify two major groups of solar dryers, i.e. passive
or natural solar dryers, and active or forced solar dryers for convection. Some easy-
to-build and user-friendly dryers have been provided that can be adequately used
in small-scale plants or in remote rural villages. In rural regions, these small-cost
foods drying techniques can be easily used to decrease spoilage and enhance product
quality, leading in general hygiene for processing.
References
1. Bhardwaj AK, Kumar R, Chauhan R (2019) Experimental investigation of the performance

of a novel solar dryer for drying medicinal plants in Western Himalayan region. Sol Energy
177:395–407. https://doi.org/10.1016/j.solener.2018.11.007
2. El-Sebaii AA, Shalaby SM (2013) Experimental investigation of an indirect-mode forced

convection solar dryer for drying thymus and mint. Energy Convers Manag 74:109–116. https://
doi.org/10.1016/j.enconman.2013.05.006
3. Gupta PM, Das AS, Barai RC, Pusadkar SC, Pawar VG (2017) Design and construction of solar
dryer for drying agricultural products. Int Res J Eng Technol 4. e-ISSN: 2395-56
4. Sandali M, Boubekri A, Mennouche D, Gherraf N (2019) Improvement of a direct solar dryer
performance using a geothermal water heat exchanger as supplementary energetic supply. An
experimental investigation and simulation study. Renew Energy 135, 186–196. https://doi.org/
10.1016/j.renene.2018.11.086
5. Badaoui O, Hanini S, Djebli A, Haddad B, Benhamou A (2019) Experimental and modelling
study of tomato pomace waste drying in a new solar greenhouse: evaluation of new drying
models. Renew Energy 133:144–155. https://doi.org/10.1016/j.renene.2018.10.020
6. Mewa EA, Okoth MW, Kunyanga CN, Rugiri MN (2019) Experimental evaluation of beef drying
kinetics in a solar tunnel dryer. Renew Energy 139:235–241. https://doi.org/10.1016/j.renene.
2019.02.067
7. Essalhi H, Benchrifa M, Tadili R, Bargach MN (2018) Experimental and theoretical analysis
of drying grapes under an indirect solar dryer and in open sun. Innov Food Sci Emerg Technol
49:58–64. https://doi.org/10.1016/j.ifset.2018.08.002
8. Rabha DK, Muthukumar P, Somayaji C (2017) Experimental investigation of thin layer drying
kinetics of ghost chilli pepper (Capsicum Chinense Jacq.) dried in a forced convection solar
tunnel dryer. Renew Energy 105:583–589. https://doi.org/10.1016/j.renene.2016.12.091
Effect of Air Co-flow on Flickering
Motion of LPG Laminar Diffusion Jet
Flame
P. Satyanarayana Raju, Guguloth Mahesh Nayak, and S. Balusamy
1 Introduction
Diffusion flames are observed in most of the combustion appliances which makes it
mandatory to have a clear understanding of the flame structure. Chen et al. [1] reported
two different kinds of vortical structure in the diffusion flames, inner vortices in the
inside luminous flame and the other outside the flame. The modified K-H (Kelvin–
Helmholtz) instabilities on either side of the flame surface responsible to form an
outer cortical structure due to buoyancy. These vortices are periodic in nature and
influence the flow structure to oscillate. These phenomena generally occur in laminar
premixed and diffusion flame [1–7]. The stability criteria for laminar unsteady flame
are reported by Chen et al. [1]. The flame oscillation frequency is in the range of 10–
20 Hz. It relies only slightly on the experimental conditions such as exit velocity of
fuel jet, tube diameter and fuel type [2]. It revealed that the frequency of oscillations
is slightly sensitive to the Froude number (Fr = u 2 /gd, where ‘u’ is jet exit velocity,
‘g’ is gravitational acceleration, and ‘d’ is tube diameter). The non-dimensional
relationship in between the Froude number and Strouhal number (St = f d/u, where
‘f ’ is flickering frequency) is reported for scaling the oscillation frequency [2].
A closer understanding of outer vortical structure interaction with the flame
surface in the unsteady jet diffusion flames can be brought by analysing the impacts
of air co-flow on the flame dynamics. Experimental studies of the impacts of air
P. Satyanarayana Raju
Department of Mechanical Engineering, National Institute of Technology, Surathkal, Karnataka,
India
G. M. Nayak · S. Balusamy (B)
Department of Mechanical and Aerospace Engineering, Indian Institute of Technology,
Hyderabad, Telangana, India
e-mail: saravananb@iith.ac.in
G. M. Nayak
e-mail: me15m17p000001@iith.ac.in
718 P. Satyanarayana Raju et al.
co-flow on the laminar diffusion, premixed, and partially premixed flames are previ-
ously reported [3–6]. It revealed that full suppression of vortical structure in the
flickering diffusion flame was noticed by increasing the velocity of air-co-flow to the
exit velocity of fuel jet [3, 4]. In pre-mixed flame, a decrease in the temperature vari-
ation was observed with air co-flow, indicating a lessening in oscillation amplitude
with a rise in velocity of air co-flow [5]. The effects of air co-flow and equivalence
ratio on the flickering behaviour of a partially premixed methane–air mixture flame
were studied under different test conditions, such as fuel velocity to air co-flow
velocity and burner diameter [6]. It is reported that critical co-flow velocity ratio for
suppressing flickering is independent of the equivalence ratio. The flickering energy
analysed with POD analysis under the influence of co-flow [5, 6]. Furthermore, Katta
et al. studied the effects of soot on flickering of flame [7]. Both experimental and
numerical calculations produce a steady flame when the soot generated is more but
starts producing a dynamic flame when the formation of soot is restricted.
The objective of the present study is to investigate the influence of co-flow air
on low sooting LPG diffusion flame. The oscillation frequency, flame height and,
percentage dominant mode energy as a function of the co-flow velocity of air were
measured for fixed nozzle diameter and exit velocity of LPG. This study was accom-
plished by fast Fourier transform (FFT) analysis of photomultiplier tube (PMT) time
signals, image-processing technique and proper orthogonal decomposition (POD)
analysis.
2 Experimental Set-up
The experimental set-up, shown in Fig. 1, consists of three major components: fuel
flow and co-flow air system, high-speed camera and photomultiplier set-up. LPG
gas is supplied to a burner nozzle of 8 mm inner diameter from a compressed gas
cylinder, and co-flow air is supplied to a circular hollow pipe of 55 mm inner diameter
Fig. 1 Schematic of experimental set-up

Effect of Air Co-flow on Flickering Motion of LPG Laminar … 719
placed coaxially to the burner pipe. Honeycomb-shaped structures having 2-mm-

mesh sieve are situated inside the 55-mm-diameter pipe to have uniform supply
of air co-flow. The flow rates of LPG and co-flow air are measured by mass flow
meters. The experiments are conducted for different combinations of the Reynolds
number of fuel (Re150, Re200, Re250) and air co-flow velocity. The exit velocity of
LPG gas jet is approximately 0.0745 m/s with Reynolds number 150. The co-flow
air velocities are in the range of 0.05–0.25 m/s. The evolution of flame structure
is captured using a monochromatic high-speed Phantom VEO 710L camera. The
images are obtained at rates of up to 5000 frames per second, and spatial resolution
is 680 * 1280 with 8 bits. The flickering frequency and oscillation spectrum are
obtained using a photomultiplier set-up. The photomultiplier set-up is discussed in
detail in the paper [3]. In this work, POD is used to find out the percentage of dominant
mode energies. Complete information about the computational methodology of POD
analysis is discussed in the literature [8].
The experiments are carried out in a different combination of fuel Reynolds number
and co-flow velocities. In this section, the Reynolds number of the diffusion flame
is fixed to 150 (diameter of burner nozzle is 8 mm). Initially, images were captured
without co-flow, and it gradually increases to 0.20 m/s through a step of 0.05 m/s.
A complete cyclic sequence of high-speed images of LPG-air diffusion flame in
the absence of co-flow air is shown in Fig. 2a. A frequent and repeatable oscillation
is observed in this flame owing to the movement of hot gases and the recurrent
interaction of vortices in the flame caused by Kelvin–Helmholtz instability with
ambient air. The time elapsed between two images Fig. 2a is 11.16 ms, which indicates
the flickering frequency about 12.8 Hz. There is waxing of the flame along the axis
initially, and then, there is waning of flame to reach its initial state, thereby exhibiting
the cyclic order. This cycle can be attributed to the repeated interaction of vortices in
the flame with air present in its vicinity. Flame–air interaction leads to the improved
fuel–air mixture at a certain time and accordingly changes the rate of combustion.
This causes a momentary discontinuity in the flame. However, when the co-flow
starts to flow, it is found that the discontinuity of the flame tends to appear quicker
than before and the flame shows a decrease in its lateral size and increase in its
longitudinal size. For higher velocities of air co-flow, flame bulge and necking start
arise well downstream of the flame resulting in no separation at the flame tip. The
outer toroidal vortices tend to travel symmetrically along the centreline of flame as
co-flow starts, as shown in Fig. 2. The time intervals between two successive images
in the Fig. 2b–e are 10.46 ms, 9.9 ms, 9.4 ms, and 9.07 ms, respectively.
There is a highly noticeable difference in the shape of the flame when the co-flow
velocity is 0 m/s to that of 0.2 m/s (Fig. 2a–e). The flame appears to have gained flame
height when co-flow changes from 0 to 0.2 m/s. The flame also has reduced its size
laterally. Puff formation which could be seen distinctly when the co-flow velocity at
Fig. 2 High-speed flame

images
(a)
(b)
(c)
(d)
(e)
0 m/s vanished. The cyclic variations/oscillations of the flame are on verge of seizure
when co-flow velocity reaches 0.2 m/s. There is no momentary discontinuity of the
flame during the combustion cycle when the flame becomes completely smooth. This
phenomenon happens as the co-flow of air can drive away completely those toroidal
vortices which were responsible for the alteration of flame structure from the vicinity
of the luminous flame. The FFT analysis on PMT time signals is performed with no
co-flow, and the fundamental frequency of flame is found to be around 12.8 Hz. With
an increment in the co-flow, the fundamental frequency tends to move to a higher

value with a reduction in its oscillation amplitude as shown in Fig. 3a, b.
Fig. 3 Frequency spectra, obtained from PMT signal

Initially, the flame’s flickering frequency increases with an increase in velocity

of co-flow. Gradation in the flickering frequency of the flame decreases as the co-
flow air velocity is increased to 0.2 m/s. Once the co-flow velocity reaches 0.2 m/s,
the frequency of the flame becomes constant indicating that the flame structure has
attained a stable state.
The frequency of the flame for different Reynolds numbers (fixed burner diameter
= 8 mm and different gas-jet velocity) with varying co-flow shown in Fig. 4. It clearly
depicts that the flame flickering frequency is increased with increase in co-flow
velocity irrespective of the Reynolds number. But as the co-flow reaches 0.2 m/s,
the frequency of the flame for different Reynolds number converges to the same
frequency. This indicates that the stable flame structure is reached at a fixed frequency.
Figure 5 shows the mean flame height with the co-flow velocity at Reynolds number
150. Note that, measurement of the mean flame height also includes disconnected
portion of the flame. The mean flame height increases with increase in velocity of air
co-flow (refer Fig. 5). The flame images from the Fig. 2a–e clearly shows the evidence
of increment in the mean flame height with varying air co-flow. The present results
indicate that the flame is thin at the moment of necking owing to elongation of flame
in the direction of air co-flow, that is induced by the shift of big sized outer vortices
in the longitudinal direction of flame with co-flow. Figure 6 shows the fluctuating
energy of dominant POD modes of the LPG flame at different co-flow velocities,
which are obtained by 4096 instantaneous images.
The similar phenomenon is observed in the fluctuating energy of dominant POD
modes with varying co-flow. Initially, without co-flow the dominant mode energy
shows 27.37%. As co-flow starts, an increment in the fluctuating energy is observed.
Fig. 4 Peak frequency of LPG flame at three Reynolds numbers (Re = 150, 200, and 250) with
co-flow
Fig. 5 Mean flame height of LPG flame with co-flow
Fig. 6 Dominant POD mode energy with co-flow
When the velocity of air co-flow is increased, the velocity difference between fuel
flow and co-flow also increases and so the flow shear termed, thus, the dominant
mode energy. It can be seen in the graph that there is a sudden hike in the percentage
of dominant mode energy when co-flow increased from 0.15 to 0.2 m/s. This is due
to increment in the co-flow velocity from 0.15 to 0.2 m/s which results in pushing
away the outer toroidal vortices from the vicinity of the flame resulting in a reduction
of oscillations and increment in the effect of flow shear term due to an increase in
velocity difference. When the co-flow is increased beyond 0.25 m/s, the percentage
of dominant mode energy remains constant as the oscillations are suppressed.
4 Conclusions
Effects of co-flow air on flickering motions of laminar LPG gas-jet diffusion flames
were experimentally examined by evaluating the flame height, flickering frequency,
and percentage of dominant mode energy. This study was accomplished by fast
Fourier transform analysis on PMT time signals, image processing technique, and
POD analysis. The following results were obtained:
1. It is found that increasing co-flow velocity will increase oscillation frequency.
This phenomenon happens as the co-flow of air can drive away completely
buoyancy induced vortices which were responsible for the alteration of flame
structure from the vicinity of the luminous flame.
2. It is confirmed that the flame flickering frequency is not a function of fuel flow
rate, but it is increasing with co-flow air in a linear trend.
3. The most prominent observation is that, when the velocity of co-flow increases
up to a certain value, flame height becomes constant. It must be noted that the
percentage of dominant mode energy follows a similar trend as that of flame
height when the co-flow velocity of air is increased.
Acknowledgements The authors would like to thank Dr. Arumuga Perumal D, NIT Surathkal for
his valuable support during this study.
References
1. Chen LD, Seaba JP, Roquemore WM, Goss LP (1989) Buoyant diffusion flames. In: Symposium
(International) on combustion, vol 22, no 1, pp 677–684. Elsevier
2. Hamins A, Yang JC, Kashiwagi T (1992) An experimental investigation of the pulsation
frequency of flames. In: Symposium (international) on combustion, vol 24, no 1, pp 1695–1702.
Elsevier
3. Darabkhani HG, Zhang Y (2010) Stabilisation mechanism of a flickering methane diffusion
flame with co-flow of air. Eng Lett 18(4):369
4. Fujisawa N, Matsumoto Y, Yamagata T (2016) Influence of co-flow on flickering diffusion flame.
Flow Turbul Combust 97(3):931–950
5. Fujisawa N, Abe T, Yamagata T, Tomidokoro H (2014) Flickering characteristics and temperature
field of premixed methane/air flame under the influence of co-flow. Energ Convers Manag
78:374–385
6. Fujisawa N, Okuda T (2018) Effects of co-flow and equivalence ratio on flickering in partially
premixed flame. Int J Heat Mass Transf 121:1089–1098
7. Katta VR, Roquemore WM, Menon A, Lee SY, Santoro RJ, Litzinger TA (2009) Impact of soot
on flame flicker. Proc Combust Inst 32(1):1343–1350
8. Feng LH, Wang JJ, Pan C (2011) Proper orthogonal decomposition analysis of vortex dynamics
of a circular cylinder under synthetic jet control. Phys Fluids 23(1):
Natural Convection Through High
Porosity Metal Foams—A Numerical
Study
Banjara Kotresha, Prakash H. Jadhav, and N. Gnanasekaran
1 Introduction
The metal foams are considered as a good candidate for heat transfer enhancement
in thermal systems because of their unique mechanical structural properties like low
density, high surface area density, ability to increase turbulence, etc. Phanikumar
and Mahajan [1] studied gravity-induced flow through highly porous metal foams
both by experiments and numerical method. They concluded that the aluminium
metal foam with combination of air gives the heat transfer enhancement factor of
4 and local thermal non-equilibrium (LTNE) is a good model for predicting heat
transfer through the interfaces of the metal involved. Zhao et al. [2] carried out
experimental and numerical studies on natural convection through metal foams and
concluded that the heat transfer in buoyancy-induced flow is very significant due to
the interconnected open cells and high porosity of the metal foams. They also reported
that the metal foams increase conductivity by 50%. Baez and Nicolas [3] numerically
performed the natural convection heat transfer and flow through porous medium in
a rectangular cavity for different inclinations. The study concluded that the global
heat transfer increases as Rayleigh number increases. Saeid [4] studied the natural
convection through porous structure with finite wall thickness. The problem is solved
as a conjugate heat transfer since it involves the thickness of the wall. The study
revealed that as thickness increases the Nusselt number decreases and circulation
region decreases. At lower Rayleigh number, the conduction heat transfer dominates
both through wall thickness and porous medium. Piller and Stalio [5] numerically
B. Kotresha
Department of Mechanical Engineering, Government Engineering College, Haveri 581110,
Karnataka, India
P. H. Jadhav · N. Gnanasekaran (B)
Department of Mechanical Engineering, National Institute of Technology Karnataka, Surathkal
575025, Karnataka, India
e-mail: gnanasekaran@nitk.edu.in
728 B. Kotresha et al.
studied the buoyancy-driven heat transfer and flow through metal foams partly filled
in inclined parallel plate channel. They concluded that the natural convection region
is independent of channel inclination when the non-Darcian effects are negligible.
Abbas et al. [6] carried out a natural laminar convection study in porous media. For
constant Darcy number, the convection increases with increase of Grashoff number.
Barbieri et al. [7] experimentally studied the effect of thickness of metal foam on
the heat transfer characteristics by natural convection. They reported that the heat
transfer increases with increase of thickness of the metal foam and also developed
correlation for Nusselt number in terms of Rayleigh number, Darcy number and
shape function. Sheikholeslami [8] carried out the natural convection of magnetic
nanofluid through porous medium numerically and concluded that with the increase
of Hartmann number the temperature gradient decreases.
Based on the above literature survey, the authors found that there is a gap in
the study of buoyancy-driven convection through the metal foams attached with the
vertical plates. Therefore, this paper explores the possibility of a numerical study on
a two-dimensional conjugate heat transfer from aluminium metal foam attached to
the vertical channel.
2 Methodology
2.1 Experimental and Numerical Domain
The experiment is conducted on an aluminium plate-heater assembly for validating

the buoyancy-driven CFD model. The test plate assembly used in the experiments
consists of two aluminium plates of dimensions 150 × 250 × 3 (all in mm) with a flat
heater sandwiched between them. The flat heater is prepared by winding a Nichrome
wire over a mica sheet and the same is electrically insulated from the unexposed
side of the test plates by using mica sheets and the power is supplied to the heater
with the help of DC power source. The aluminium plates and heater are held tightly
together by using eight screws and nuts. The heater plate assembly is hanged by using
wires on the supporting frame and is placed at the centre of the plywood box. The
data acquisition system NI 9213 along with LABVIEW software is used to store the
temperature data. Five K-type thermocouples are used for recording the temperature
on the aluminium plate and are placed at different locations of aluminium plate and
one thermocouple is used for noting down the ambient temperature. The schematic
diagram along with experimental setup is shown in Fig. 1. In experiment, the steady-
state condition is confirmed when the variation of temperature of the aluminium plate
is observed to be 0.2 °C for time duration of 10 min.
Natural Convection Through High Porosity Metal Foams … 729
Fig. 1 Experimental setup a schematic view b, c photographic view
2.2 Numerical Problem Domain
The problem considered for the numerical investigation consists of a two aluminium
plates in which the heater is sandwiched between them. The aluminium metal foams
of four different pore densities of 10, 20, 30 and 45 are attached on either side of the
aluminium plate. The size of the aluminium plates is 250 × 150 × 3 and of metal
foams is 250 × 150 × 10 (all are in mm). The schematic diagram of the problem
domain considered for the present study is shown in Fig. 2.
2.3 Computational Domain
From Fig. 2, it is clear that the physical geometry is symmetrical about the Y-axis;
hence, a two-dimensional computational domain is chosen which consists of one
aluminium plate, heater, metal foam and extended domain for the present numerical
simulation. The heater is assigned with heat flux boundary condition, the symmetrical
part of the extended domain is defined with symmetry and the extended domain is
Fig. 2 Schematic of
problem domain (1) heater,
(2) aluminium plate, (3)
metal foam
defined with zero pressure (pressure outlet) boundary condition. Since the total heat
given to the heater is equally divided on both sides of the aluminium plates, only
half of the heat input is considered in the numerical simulation. The computational
domain with boundary conditions assigned is shown in Fig. 3.
Two-dimensional numerical simulations are carried out by using commercially avail-

able ANSYS FLUENT 15.0 [9]. The heat transfer from the aluminium plate to the
surrounding air takes place due to buoyancy forces caused because of tempera-
ture difference. To capture the buoyancy-driven flow, Boussinesq approximation is
considered in the present study. The metal foam region is considered as a homoge-
neous porous medium with Darcy extended Forchheimer model to capture the flow.
The heat transfer through the metal foam to the adjacent and surrounding air is solved
by using local thermal non-equilibrium model (LTNE).
domain and boundary
conditions
The pressure drop across the porous media is modelled by adding the following
source term to the momentum conservation equation which involves the inertia and
viscous effects because of the form drag coefficient and permeability of the porous
metal foam.

3
3
1
Si = − Dμν j + C ρ|ν|ν J (1)
j=1 j
2
In order to validate the present methodology, the experiment is carried out on an

aluminium plate for different heat inputs and the corresponding steady-state temper-
ature of the aluminium plate is recorded. The experimental results obtained are then
compared with the numerical results obtained for the same aluminium plate without
the metal foam using the CFD software and the comparison is shown in Fig. 4.
From the plot it is clear that the results obtained in the present simulations for the
buoyancy-driven flow are in good agreement with the experimental results.
The interfacial heat transfer coefficient and interfacial surface area are calculated
based on the correlations given by Calmidi and Mahajan [10]. The properties of the
metal foams used for the present study are taken from Kamath et al. [11] and are
listed in Table 1.
Fig. 4 Comparison of excess temperature obtained in the present study with experiments
Table 1 Properties of metal foams considered in the present study [11]

PPI Porosity (ε) Permeability Form drag Ligament dia Pore dia d p Surface area
(K) 107 , m2 coefficient (c) d f (mm) (mm) density asf
(m−1 )
10 0.95 2.480 94.98 0.45 4.95 360.6
20 0.90 2.177 208.82 0.45 3.15 960.7
30 0.92 1.644 148.97 0.22 2.32 936.4
45 0.90 0.420 397.01 0.18 1.65 1671.7
A detailed grid sensitivity analysis is carried out in the numerical examination to

select an optimum number of grids for the computations. The grid sensitivity study
is carried out on an aluminium plate for three different grids of 26,950, 46,200 and
70,300 for a heat input of 20 W. The excess temperature obtained is almost same for
all the three grids studied, so 46,200 grids are selected to save computational time
Fig. 5 Nusselt number

variation for different
Rayleigh numbers
and space. The effect of domain dependency is also carried out while selecting the
extended domain for the computations.
The variation of the Nusselt number for all the metal foams for different Rayleigh
number is shown in Fig. 5. It is clear from the plot that the Nusselt number decreases
as PPI increases at a particular Rayleigh number. This indicates that the heat transfer
rate decreases as PPI increases because the flow resistance increases as metal foam
PPI increases. The 30 PPI metal foam shows higher heat transfer compared to 20 PPI
metal foam because the porosity of 30 PPI metal foam is higher than 20 PPI metal
foam. The results obtained in the present study follows the similar trend of results
obtained by Phanikumar and Mahajan [1]. It is also noticed that the Nusselt number
increases with increase in Rayleigh number for all PPI metal foam.
The enhancement ratio for the presence of metal foam is calculated based on
Eq. (2).
(Nu − Nu0 )
Enhancement ratio E = (2)
Nu0
where N u0 is the Nusselt number for aluminium plate and N u is Nusselt number for
metal foam based on the thermal conductivity of air for the purpose of comparison.
The enhancement ratio for all the metal foams with respect to Rayleigh number
is shown in Fig. 6. For particular metal foam PPI, the enhancement ratio remains
the same. The presence of metal foams with air as working fluid shows an average
enhancement of 4–6 times compared to the without metal foam case. Phanikumar and
Fig. 6 Enhancement ratio for all metal foams
Mahajn [1] reported that air-metal foam combination gives an average of enhance-
ment ratio of 4.0. The results obtained for natural convection in the present study are
in good agreement with the results available in the literature [1].
4 Conclusions
A two-dimensional natural convection study is carried out on a metal foam attached to

the aluminium plate numerically using commercial available software FLUENT. The
problem considered consists of two aluminium plates in which heater is sandwiched
between them and the metal foams of different pore densities are attached on either
side of the aluminium plate to enhance the heat transfer. The buoyancy-driven flow
is predicted by using Boussineq approximation and the flow through the metal foam
is predicted using Darcy extended Forchheimer model. The heat transfer through the
metal foam is calculated using LTNE model. The experiment is carried out on an
aluminium plate to verify the methodology used in the CFD model and an excellent
agreement between the experiment and present simulation is achieved. The heat
transfer through the metal foam decreases as PPI increases at a particular Rayleigh
number, but 30 PPI metal foam shows higher heat transfer rate compared to 20 PPI
metal foam since the porosity of the 30 PPI metal foam is higher than 20 PPI metal
foam. The metal foams considered in the present study enhances the heat transfer by
an average of 4–6 times than the without metal foam case.
References
1. Phanikumar MS, Mahajan RL (2002) Non-Darcy natural convection in high porosity metal
foams. Int J Heat Mass Transf 45:3781–3793
2. Zhao CY, Lu TJ, Hodson HP (2005) Natural convection in metal foams with open cells. Int J
Heat Mass Transf 48:2452–2463
3. Baez E, Nicolas A (2007) Natural convection fluid flow and heat transfer in porous media. J
Mech Mater Struct 2:1571–1584
4. Saeid NH (2007) Conjugate natural convection in a porous enclosure: effect of conduction in
one of the vertical walls. Int J Therm Sci 46:531–539
5. Piller M, Stalio E (2012) Numerical investigation of natural convection in inclined parallel-plate
channels partly filled with metal foams. Int J Heat Mass Transf 55:6506–6513
6. Abbas AH, Messaoud H, Saada D, Abdennacer B (2015) Numerical study of laminar natural
convection in porous media: Darcy Brinkman Forcheimer model. Energ Procedia 74:77–86
7. Barbieri M, Ilio GD, Patane F, Bella G Experimental investigation on buoyancy induced
convection in aluminum metal foams. Int J Refrig. https://doi.org/10.1016/j.ijrefrig.2016.
12.019
8. Sheikholeslmi M (2017) Numerical simulation of magnetic nanofluid natural convection in
porous media. Phys Lett A 381:494–503
9. ANSYS Fluent (2017). Available from http://www.ansys.com/Products/Fluids/ANSYS-Fluent
10. Calmidi V, Mahajan R (2000) Forced convection in high porosity metal foams. ASME J Heat
Transf 122:557–565
11. Kamath PM, Balaji C, Venkateshan SP (2011) Experimental investigation of flow assisted
mixed convection in high porosity foams in vertical channels. Int J Heat Mass Transf 54:5231–
5241
Optimization of Engine Operating
Parameters to Meet Global Emission
Norms and CAFE Norms
N. Sekarapandian and Selvaraji Muthu
1 Introduction
Emission control methodologies for diesel and gasoline vehicles are assessed by
Sanchez et al. [1]. Gasoline engine emission control is based primarily on precise
air-fuel control and catalytic after-treatment. These emission control technologies
have reached a significant level of maturity up to the most stringent existing stan-
dards. Nitrogen oxides (NOx ) and particulate matter (PM) emission control from
diesel engines is far more complex and requires the implementation of relatively
new technologies involving air management, fuel injection control, after-treatment
and system integration. The implementation of new technologies for diesel engine
emissions control has a significant impact compared with gasoline engine emissions
control. Emission norms to comply for Gasoline vehicle is given in Table 1. In-
cylinder control and after treatment technologies to meet emission norms for gasoline
vehicles are presented in Tables 2 and 3, respectively. Similarly emission norms to
comply for Diesel vehicle is given in Table 4. In-cylinder control and after treatment
technologies to meet emission norms for Diesel vehicles are presented in Tables 5
and 6, respectively. The particulate matter is controlled by number (PN) for Euro 6
and which requires the use of wall-flow DPF for this PM control. At the same time
in-cylinder control measures require continuous research, design and development in
combustion, including different optimum mode of fuel injection strategies at higher
fuel injection pressures and the variable geometry turbocharger (VGT) required to
deliver optimum amounts of fuel and air at specific operational conditions of the
engine. The emission levels of NOx are reduced by 66% from Euro 5, and requiring
the use of NOx after-treatment devices in addition to in-cylinder measures like cooled
EGR. Then LNTs (Lean NOx Traps) have shown very good NOx reduction perfor-
mance and durability. On the other hand, SCR, while offering also very good NOx
N. Sekarapandian (B) · S. Muthu

Department of Thermal and Energy, School of Mechanical Engineering, VIT, Vellore, India
738 N. Sekarapandian and S. Muthu
Table 1 Emission norms to comply for gasoline vehicles

Technology matrix Norms Emission norms to comply (gasoline vehicle)
EU1 EU2 EU3 EU4 EU5 EU6
Fuel Sulphur ppm 2000 500 150 50 10 10
Emission Gases CO2 g/km – – – – 130 95
CO g/km 2.72 2.2 2.3 1 1 1
THC g/km – 0.2 0.1 0.1 0.1
NMHC g/km – – – 0.068 0.068
NOx mg/km 490 – 150 80 60 60
HC + NOx g/km 0.97 0.5 – – – –
PM mg/km – – – 5 5
PM (Nos) #/km – – – – 6 × 1011
Table 2 In-cylinder emission control technologies for gasoline vehicles

Technology matrix Emission norms to comply (gasoline vehicle)
Old EU1 EU2 EU3 EU4 EU5 EU6
In-cylinder control Electromechanical yes
distributors
Electronic ignition yes
Multipoint fuel injection yes yes yes yes
(MPFI) (electronic
controls for fuel injection
and ignition spark timing)
Gasoline direct injection yes yes yes yes yes
(GDI)
Combustion yes yes
improvements through
engine calibration
Incremental yes yes
improvements in air-fuel
management
reduction performance, it offers more flexibility for fuel economy and reduction of
CO2 emissions. Most of manufacturers will likely choose the NOx after-treatment
technology between EGR, LNT and SCR, based on a combination of fuel economy,
cost, reliability, and consumer acceptance, as explained in detail by Sanchez et al.
[1].
Optimization of Engine Operating Parameters to Meet … 739
Table 3 After-treatment emission control technologies for gasoline vehicles

Technology matrix Emission norms to comply (gasoline
vehicle)
Old EU1 EU2 EU3 EU4 EU5 EU6
After treatment systems Three way catalytic yes yes yes yes
convertor
Single O2 sensor yes yes yes yes
Under floor (UF) yes yes
catalyst
Closed coupled catalyst yes yes
(CC) for cold start
Exhaust gas yes yes yes yes
recirculation (EGR)
More responsive heated yes yes
oxygen sensors (HO2 S)
Secondary oxygen yes yes
sensors after the
catalyst
Universal wide range yes yes yes
oxygen sensors
Table 4 Emission norms to comply for diesel vehicles

Technology matrix Norms Emission norms to comply (diesel vehicle)
Fuel Sulphur ppm 2000 500 350 50 10 10
Emission gases CO g/km 2.72 1 0.64 0.5 0.5 0.5
THC g/km – – – – – –
NMHC g/km – – – – – –
NOx mg/km – – 500 250 180 80
HC + NOx g/km 0.97 0.7 0.56 0.3 0.23 0.17
PM mg/km 140 80 50 25 5 5
PM (Nos) #/km – – – – 6 × 1011 6 × 1011
2 Methodology
Increasing compression ratio (CR) is one of the most proven method to improve
engine efficiency, but the CR of practical spark ignition engines is limited by knock
and spark retard at high loads. An advanced CR mechanism could improve efficiency
by using higher CR at low loads, and lower at high loads, as explored by Karthikeya
et al. [2] as shown in Fig. 1 and Rafael et al. [3].
Table 5 In-cylinder emission control technologies for diesel vehicles

Technology matrix Emission norms to comply (diesel vehicle)
In-cylinder Mechanical rotary pump fuel yes
control injection systems
Electro-mechanical cam-controlled yes yes
fuel injection
Rotary fuel injection system yes yes
Indirect fuel injection yes yes
Common-rail or unit-injector yes yes yes yes
systems
Electronic fuel timing yes yes
Naturally aspirated (NA) yes
Turbocharging Without yes yes yes
inter-cooling
With inter-cooling yes
Variable geometry yes yes yes yes
turbocharger
(VGT)
Electronic assistance for fuel yes yes yes yes yes
metering
Variable fuel timing and metering yes yes yes yes yes
strategies
Injection pressure. Bar <900 1300 1600 1900 2100
Combustion improvements through yes yes
engine calibration
Incremental improvements in yes yes yes yes
air-fuel management
Exhaust gas Mechanically yes
recirculation activated EGR
(EGR) circuits
Cooled EGR yes
Cooled, yes
electronically
controlled and
solenoid-operated
EGR.
Cooled EGR with yes
a DC motor
actuator
(continued)
Table 5 (continued)
Cooled yes yes
high-pressure
EGR
Table 6 After-treatment emission control technologies for diesel vehicles

After-treatment DPF DPF with yes yes
systems active
regeneration.
Wall-flow DPF yes
Diesel oxidation yes yes yes yes
catalysts DOC
Lean-NOx Traps (LNT) yes
technology (<3.0L)
Selective catalytic yes
reduction (SCR)
technology (>3.0 L)
Fig. 1 In-cylinder pressure 30.0

versus crank angle at BP 1.0
25.0
In-Cylinder Pressure
different boost pressures and BP 1.25

swirl ratios BP 1.5
20.0
BP 1.75
( MPa)
15.0 BP 2.0
10.0
5.0
0.0
680.0 700.0 720.0 740.0 760.0 780.0 800.0
Crank Angle (deg)
3 Component Technologies
High thermal conductivity of the light alloy material, aluminum pistons have excel-
lent heat distribution. This reduces the temperatures at the very highly stressed rim of
the combustion bowl. The combination of optimized piston cooling to allow power
densities of around 100 kW/L and maximum peak cylinder pressures (PCP) of over
200 bar for aluminum pistons as well. For such applications, the use of a bronze
bi-metal bushing to reinforce the pin boss is recommended, if necessary.
Steel pistons can generally be lower in compression height (CH) than aluminum
pistons, which produce weight savings in the entire engine. Steel pistons demonstrate
advantages with regard to frictional loss [5–7] and thermodynamics in comparison
with aluminum pistons. When using cast-iron crankcases, a greater frictional loss
advantage can be expected with steel pistons than with aluminum crankcases. Overall
fuel savings of 3–5% have been observed in engine testing as shown in Fig. 2.
Special type of U-flex oil rings (as shown in Fig. 3) are used to achieve technical
alternative solution for highly deformed engine blocks and for Particle emission
reduction (SI) and also for high loaded engines with EGR (HSD) explained by
Nocera [8] and Tomanik [9, 10] and in combination with variable compression ratio
engines by Shelby et al. [11] and methodologies to Achieve Highest CO2 emission
reduction are presented with advanced technologies by Selvaraji et al. [12, 13].
There are advanced variants of these U-Flex rings with Land wear protection (Laser
hardening/GNS/DLC) [10]. The following are certain unique technology advantages
of U-Flex Oil Rings: (a) Excellent oil sealing and releasing properties to reduce the oil
consumption; (b) Excellent conformability ideal for application in highly distorted
Fig. 2 Benefit of steel

piston compared to
aluminium piston [4]
Fig. 3 U-Flex oil ring
bores; (c) Low sensitivity to wear ideal for application with low viscosity oil and
high amount of soot.
4 Conclusions
System, component and test methodologies were discussed for evaluating the
fuel consumption impact when changing engine components design. The devel-
oped methodology reproduces the vehicle emission cycles in engine dynamometer,
allowing fuel consumption tests with higher accuracy and overall lower time for
development that the vehicle chassis homologation tests. With use of available low
friction and light weight product solutions, fuel consumption and CO2 emissions can
be reduced by 3–5%.
References
1. Sanchez P, Bandivadekar A, German J (2012) Estimated cost of emission reduction technologies

for light-duty. In: The international council on clean transportation (ICCT), Washington DC
2. Karthikeya Sharma T, Amba Prasad Rao G, Madhu Murthy K (2015) Effective reduction of in-
cylinder peak pressures in homogeneous charge compression ignition engine—a computational
study. Elsevier Alexandria Eng J 54(3):373–382
3. Bruno R, Dezotti R, Mordente P (2017) Mahle Metal Leve S.A.—Power cell unit (PCU) for
modern HDD engines. SAE Technical paper 2017-36-0301. https://www.sae.org/publications/
technical-papers/content/2017-36-0301/
4. Spangenberg S, Hettich T, Lazzara M, Schreer K, Pistons for passenger car diesel engines—
aluminum or steel
5. IBler W, Eberhard Kopf-MAHLE GmbH: modern DI diesel pistons-development and tech-

nologies
6. Fernandez M, Tomanik E, Carlini R (2013) Comparison of the potential to reduce fuel
consumption by engine energy losses mapping. SAE paper 2013-36-0221
7. Ferrarese A et al (2010) Piston ring tribological challenges on the next generation of flex-fuel
engines. SAE Int J Eng 3(2):85
8. Nocera E (2013) Reducing mechanical losses for fuel targets on the inovar-auto. In: 11th
symposium SAE of powertrain, Sorocaba, Brazil
9. Tomanik E et al (2000) Reduced friction power cell components. Paper SAE 2000-01-3321
10. Tomanik E, Ferrarese A (2006) Low friction ring pack for gasoline engines. ASME paper ICEF
1566
11. Shelby M, Leone T, Byrd K, Wong F (2017) Fuel economy potential of variable compression
ratio for light duty vehicles. SAE Int J Eng 10(3)
12. Selvaraji M, Pramod K, Sathishbabu S, Rajkumar M, Li Q, Shaobo Zeng and Chai H (2017)
Advanced piston technologies for gasoline and diesel engine applications to meet EU6 emission
norms. In: Chakrabarti A, Chakrabarti D (eds) Research into design for communities, vol 1,
Smart Innovation, Systems and Technologies 65. © Springer Nature Singapore Pvt. Ltd.
13. M Selvaraji, K Pramod (2019) Methodologies to achieve highest CO2 emission reduction in
automotive systems to meet global CAFE/CAFC Norms. In: Chakrabarti A (ed) Research into
design for a connected world, smart innovation, systems and technologies, vol 134. © Springer
Nature Singapore Pvt. Ltd.
Enhancement of Heat Transfer Capacity
of Heat Pipe Heat Sink
V. Saravanan, K. Rakshith, Vishakh, Nithesh Bhaskar, and Nataraj Badiger
1 Introduction
In recent years, developments in electronics field have led to high-speed electronic

equipment. The effects of these developments may be commonly seen in computers.
The heat dissipation from functional elements can no longer be efficiently done with
the conventional approaches like passive heat sinks. Heat pipes are two-phase flow
heat transfer devices where processes of liquid to vapour and vice versa circulate
between evaporator to condenser with high effective thermal conductivity. Heat pipe
technology has found increasing applications in enhancing the thermal performance
of heat exchangers in microelectronics, energy savings in heating, ventilating, and
air conditioning (HVAC) systems.
Solomon et al. [1] developed analytical expression to predict the thermal conduc-
tivity of a heat pipe based on the heat transport limit equations. The thermal conduc-
tivity of heat pipe obtained from the heat transport equations was compared with
the lumped thermal resistance network model. The maximum thermal conductivity
obtained using the analytical model showed reasonable agreement with the experi-
mental thermal conductivity. Maydink et al. [2] developed loop heat pipe for high
heat transfer capacity using ammonia. Study was conducted for horizontal orien-
tation. At a cooling temperature of 20 °C, a maximum heat load of 1700 W was
achieved at a vapour temperature of 62 °C. Lin et al. [3] has studied thermal char-
acteristics of long heat pipe using ammonia which is capable of dissipating 350 W
at temperatures from 40 to 60 °C. Jasvanth et al. [4] developed loop heat pipe using
ammonia with evaporator length 2.5 m and inside diameter 4.57 mm. A maximum
heat dissipation capacity of about 600 W was achieved at a vapour temperature
of 40.3 °C for 5 °C condenser temperature. Kcuck [5] studied the effects of three
V. Saravanan (B)
Department of Mechanical Engineering, IIT Kharagpur, Kharagpur, India
K. Rakshith · Vishakh · N. Bhaskar · N. Badiger
Department of Mechanical Engineering, BNMIT, Bangalore, India
746 V. Saravanan et al.
different working fluids, namely water, ammonia and mercury on performance of

heat pipe. Two commonly used wick structures were also selected to demonstrate
the wick structure effect on the heat transfer capacity of the heat pipe such as wire
screen mesh and groove mesh. The study shows that it is preferable to use wire screen
meshes with two layers of 24 meshes/inch. The performance of a heat pipe using a
thermal resistance networking model was studied by Seo et al. [6], and the effect of
the heat pipe length was also analysed. One-dimensional numerical simulations were
conducted to verify the operating limits, and it was found that the capillary limit was
the dominant limit that determined the heat transport capacity of the heat pipe. The
effect of operating temperature and the fill ratios for heat pipes was presented by
Mozumder and Chowdhury [7], The study reveals that the saturated boiling temper-
ature, latent heat of vaporisation, evaporator surface temperature, working fluids and
their fill ratios play a vital role in determining the heat transport capacity of the heat
pipe. Faghri and Harley [8] studied the transient effects of heat pipes.
From the literature survey, it is observed that a majority of study is conducted on
influence of wick properties, working fluid, heat pipe length and the filling ratios on
the performance of the heat pipe. A thorough evaluation of the ideal temperature range
of the capillary limit needs to be made, for it being the major factor in determining
the thermal performance of a HPHS at lower temperatures. There also lies a scope
to design a HPHS with a pre-determined thermal dissipation capacity to meet the
global demand. In the present work, a heat pipe heat sink is designed and tested to
dissipate a pre-determined thermal dissipation capacity.
2 Methodology
The experimental setup, schematically represented in Fig. 1 consists of the heat pipe
heat sink (HPHS), heating element, thermocouples, data logger, power supply and
Fig. 1 Schematic representation of experimental setup

Enhancement of Heat Transfer Capacity of Heat Pipe Heat Sink 747
the external experiment chamber. To record the temperature data, a set of three K-type
thermocouples, placed along the length of random pipes are used in conjunction with
a data logger. All thermocouples are equally spaced along the length of the HPHS
and mounted using Kapton tape. The data logger is set to log data at a sampling rate
of 1000 ms throughout the start-up, steady state and shut down of the HPHS as well
as the conventional heat sink.
The heat pipe heat sink assembly consists of common manifold made of copper
which acts as base of the HPHS where heat pipes are attached to it. This common
manifold is considered to be a common evaporator section of dimensions 51 mm ×
51 mm × 10 mm. The manifold contains a channel of 6 mm diameter drilled through
it to attach vacuum and charging port. The top face of the manifold consists of three
holes along the centre of 9.52 mm diameter up to a depth of 8 mm. The heat pipes are
placed inside these holes as shown in Fig. 2. Copper tubes of 10 mm outer diameter,
0.8 mm thickness and 80 mm length are used. The stainless-steel wick is inserted
into the copper tubes after which they are capped at one end. Stainless-steel mesh
is used which has as the wick material (9.9 × 10−10 m2 permeability and 44.08%
porosity) with a wire diameter of 0.210 mm and a wire opening of 0.415 mm. The
charging port is used to aid the vacuuming and evacuation of the PHS to a desired
vacuum pressure.
Fig. 2 Heat pipe heat sink

test setup
All dimensions
are in mm
2.2 Theoretical Modelling
Analytical computation was performed for two different working fluids, namely
methanol and water at 60 °C to evaluate heat dissipation rate using following Eq.
(1). The thermos physical properties of water and methanol are considered at 60 °C
2σ 16μv L eff Q μl L eff Q

+ ρl gL t ≥ 2 + + ρl gdv (1)
rc 2rv Av ρv h f g K Al ρl h f g
Wire mesh specification: Wire diameter dw = 0.210 mm, wire opening w =

0.415 mm
Porosity,
w2
ε= (2)
(dw + w)2
Permeability,
dw2 ε3
K = (3)
122(1 − ε)2
Capillary radius,
dw + w
rc = (4)
2
Effective length,
Le + Lc
L eff = + La (5)
2
Le, Lc and La are evaporator, condenser and effective length, respectively.
Number of turns of mesh N = 3 turns
Mesh thickness,
tw = 2dw N (6)
Area of the vapour section Av = 2.2902 × 10−5 m2.

Area of the liquid section Al = 2.6362 × 10−5 m2.
Total length of the pipe Lt = 80 × 10−3 m.
Heat dissipated, Q methanol ≤ 110 W, Q water ≤ 306.6 W. Since water has a
better heat transport capacity when compared to methanol at the given temperature,
it is chosen as the working fluid for the HPHS.
In the present work, the thermal performance of HPHS is studied for five different
heat inputs of 100, 150, 200, 250 and 300 W at a fill ratio of 60, 80 and 100%. The
working fluid used for the experimental analysis is distilled water. The performance
of the HPHS is studied based on temperature vs time graph, which is obtained from
the experimental work. The difference in temperature across the axial length of the
heat pipe is measured to verify the proper working condition at the steady state.
Figure 3 represents the temperature distribution across the length of the HPHS at
100 W for a fill ratio of 60%. The start-up, steady-state and shut down characteristics
are depicted in the graph. Thermocouples are placed along the axial length of the heat
pipe at lengths of 10 mm, 40 mm and 70 mm from the base. These are represented
by bottom, middle and top positions in the graph. The following setup remains the
same throughout all the test cases. The graph clearly shows that the HPHS is able
to dissipate the heat that is generated by the nichrome resistance heater, which is
in contact with the base of the HPHS. It is observed that there is very minimal
temperature difference across the axial length of the heat pipe during the steady state
which proves the fact that the HPHS is working under the capillary regime in the
given conditions. Once the temperature goes beyond 90 °C, there is small variation in
the temperature difference along the length of the heat pipe which may be attributed
to other parameters like boiling limit coming into picture.
Fig. 3 Temperature distribution of HPHS at 100 W

Figure 4 represents the temperature distribution across the length of the HPHS at
200 W at a fill ratio of 100%. The start-up, steady-state and shut down characteristics
are depicted in the graph. The graph obtained shows that the HPHS is able dissipate
the heat that is generated by the heater up to a certain point in the steady-state region.
There is very minimal temperature difference across the length of the heat pipe until it
reaches 100 °C in the steady-state region. After the temperature goes beyond 100 °C,
there is significant variation in the temperature difference along the length of the heat
pipe which may be caused due to the nucleate boiling of water that takes place due to
the boiling limit and the shear force exerted by the vapour on the liquid at the wick
surface that hinders the heat transport capacity.
Figure 5 represents the temperature distribution across the length of the HPHS
at 250 W at a fill ratio of 100%. The graph obtained shows that there is significant
temperature difference between the evaporator and the condenser section of the HPHS
right at the beginning of the steady state. This trend continues throughout the steady
state with the evaporator and the adiabatic section temperatures remaining close
to each other, whereas the condenser section temperature lags by a small amount.
After the temperature goes beyond 130 °C, the adiabatic section temperature goes
beyond that of the evaporator section, hinting to the fact that the HPHS breakdown
has occurred.
4 Conclusions
A heat pipe heat sink has been designed and developed for electronic cooling appli-
cation, and its thermal performance has been studied. The following conclusions can
be drawn from the present study: (i) The present HPHS array is able to dissipate
heat up to 250 W without undergoing any breakdown. (ii) There is a significant
increase in the heat transport capacity of the HPHS on increasing the fill ratios up
to a certain working temperature limit. (iii) The working fluid plays a significant
role in determining the transport capacity in the working temperature considered
with water being the best, as it has the highest latent heat among all other fluids that
are suitable for operating conditions between 30 and 140 °C. (iv) Some of the major
design parameters that needs to be considered while developing a HPHS are the wick
structure, and the diameter of the vapour section of the heat pipe as both of them
have a significant role is determining the heat transport capacity of the HPHS.
Figure 6 represents the temperature distribution across the length of the HPHS
at 300 W at a fill ratio of 100%. The graph obtained shows that although there is
very minimal temperature difference along the length of the heat pipe throughout
the analysis, the adiabatic section temperature is higher than that of the evaporator
section over the entire steady-state region. This shows that the HPHS is not working
as theoretically predicted at 300 W, and it breaks down at high heat inputs beyond
250 W.
References
1. Solomon AB, Sekar M, Yang SH (2016) Analytical expression for thermal conductivity of heat
pipe. Appl Therm Eng 100:462–467
2. Maydanik Y, Pastukhov V, Chernysheva M (2018) Investigation of a loop heat pipe with a high
heat transfer capacity. Appl Therm Eng 1052–1061
3. Lin G, Li N, Bai L, Wen D (2010) Experimental investigation of a dual compensation chamber
loop heat pipe. Int J Heat Mass Transf 53:3231–3240
4. Jasvanth VS, Adoni A, Jaikumar V, Ambiraja A (2017) Design and testing of an ammonia loop
heat pipe. Appl Therm Eng 111:1655–1663
5. Kcuck S (2007) A comparative investigation of heat transfer capacity limits of heat pipes.
Graduate School of Natural and Applies Sciences
6. Seo YM, Park YG, Ha MY (2011) Effect of variation in length of the conventional heat pipe on
the thermal performance. In: 13th International conference on heat transfer
7. Mozumder AK, Chowdhury MSH, Akon AF (2011) Characteristics of heat transfer for heat pipe
and its correlation. Int Sch Res Netw
8. Faghri A, Harley C (1994) Transient lumped heat pipe analyses. Heat Recovery Syst CHP
14(4):351–363
The Forced Convective Heat Transfer
from an Isoflux Sphere for Liquid Metals
Dipjyoti Nath and B. Hema Sundar Raju
1 Introduction
The study of forced convective heat transfer for liquid metals from an isoflux sphere
is important due to industrial and engineering applications. In most of the studies,
researchers have considered water, air and molten salts as the working fluid. However,
in many practical problems such as Solar Central Receiver Systems, liquid metals
are proposed as the working fluid because of the most excellent thermophysical
properties, specifically higher heat capacity, higher conductivity, lower density and
higher boiling temperature (Lorezin and Abanades [1]). The flow of liquid metals
from the sphere is utilized in nuclear and rocket fuel systems, in which the shape of
the fuel particles is close to spherical droplets (Hsu [2] and Witte [3]).
The problem of forced convection in sphere surface has been studied by Dhole
et al. [4] with influences of isoflux and isothermal conditions by employing finite
volume method (FVM) for Pr ≥ 0.7. Ramteke and kishore [5] have studied the role
of linear slip velocity and isoflux thermal condition in Newtonian fluids across a
sphere by employing semi-implicit marker and cell (SMAC) for Pr ≥ 1. The forced
convective heat transfer from a cylinder with the effect of isoflux and isothermal
boundary conditions has been studied by Bharti et al. [6], Paramane and Sharma [7]
and Sufyan et al. [8] by using FVM for Pr ≥ 0.7.
The above-mentioned numerical schemes applied by the researchers are mainly
second-order accurate. For improving the accuracy of the scheme, Sekhar et al. [9],
Sekhar and Raju [10], Nath et al. [11], Raju et al. [12] and Nath and Raju [13]
attempted the problems of Navier–Stokes equation and energy equation in spherical
coordinates by employing SGHOCS which is fourth-order accurate. The present
work investigates the role of various boundary conditions, namely isothermal and
D. Nath · B. H. S. Raju (B)

Department of Mathematics, National Institute of Technology Silchar, Silchar 788010, Assam,
India
e-mail: drbhsraju@gmail.com
754 D. Nath and B. H. S. Raju
isoflux on forced convective heat transfer past a sphere for liquid metals (Pr 1) by
using SGHOCS.
2 Problem Formulation
Consider the steady, laminar and incompressible motion of Newtonian fluid past a
sphere of radius d/2 with a uniform stream velocity and temperature. The stream
function, vorticity and energy equations together with velocity components after
applying the transformation r = eξ are listed below.
Components of velocity:
e−2ξ ∂ψ e−2ξ ∂ψ
vr = , vθ = − (1)
sin θ ∂θ sin θ ∂ξ
Stream function equation:
∂ 2ψ ∂ψ ∂ 2ψ ∂ψ
− + − cot θ = −e3ξ ωsinθ (2)
∂ξ 2 ∂ξ ∂θ 2 ∂θ
Vorticity equation:
∂ 2 ω ∂ω ∂ 2 ω ∂ω
+ + 2 + cot θ − ω csc2 θ
∂ξ 2 ∂ξ ∂θ ∂θ

Re −ξ ∂ψ ∂ω ∂ψ ∂ω ∂ψ ∂ψ
= e − + ω cot θ − ω (3)
2 sin θ ∂θ ∂ξ ∂ξ ∂θ ∂ξ ∂θ
Energy equation:
∂ 2 ∂ ∂ 2 ∂
+ + + cot θ
∂ξ 2 ∂ξ ∂θ 2 ∂θ

RePr −ξ ∂ψ ∂ ∂ψ ∂
= e − (4)
2 sin θ ∂θ ∂ξ ∂ξ ∂θ
The following boundary conditions are used to solve the Eqs. (2)–(4).
∂ψ 1 ∂ 2 ψ ∂
ψ= = 0, ω = − , = −1 at ξ = 0
∂ξ sin θ ∂ξ 2 ∂ξ
1
ψ = e2ξ sin2 θ, ω = = 0 at ξ → ∞
2
∂
ψ = 0, ω = 0, = 0 at θ = 0, θ = π
∂θ
The local and average Nusselt numbers on the surface of sphere are evaluated
from temperature as follows.
The Forced Convective Heat Transfer from an Isoflux Sphere … 755

2
Nu = ,0 ≤ θ ≤ π
ξ =0
π
1 2
Nu = sin θ dθ.
2 ξ =0
0
The governing Eqs. (2)–(4) together with boundary conditions are solved by utilizing
SGHOCS. The details of discretization of conservation equations and boundary
conditions are elaborately explained in Sekhar et al. [9], Sekhar and Raju [10], Nath
et al. [11], Raju et al. [12] and Nath and Raju [13].
2.1.1 Validation of Results
The N u values calculated from the SGHOCS are validated with the correlated values
of Dhole et al. [4] in Table 2 at Pr = 10 for various values of Re together with absolute
percentage difference. It can be noticed from Table 2 that the present results vary
with the maximum percentage of 1.1474% and indicates that current results are in
concurrent with the correlated values of Dhole et al. [4].
The numerical computations are conducted for various parameters: Reynolds number
in the range 1 ≤ Re ≤ 200 and Prandtl number in case of liquid sodium (Pr = 0.004),
liquid gallium alloy (Pr= 0.0208) and liquid lithium (Pr = 0.065).
To show the grid independence of present scheme, numerical experiments are
conducted in distinct grids, namely 81× 81, 101× 101, 121× 121, 141× 141, 161×
161, 181 × 181, 201 × 201 at Re = 200 and Pr = 0.0208 for both thermal boundary
conditions. The N u values attained from the SGHOCS in the above referred grids
are furnished in Table 1. The values of N u referred in Table 1 are almost identical
for two grids, namely 181 × 181 and 201 × 201. Consequently, for all numerical
experiments, 181 × 181 grid has been chosen as optimum grid size in the current
work.
To show the order convergence of the SGHOCS, the decay of divided differences
(1st order) of N u values reported in Table 1 in regard to h is displayed in Fig. 1
together with reference line of fourth order. The slopes of the results for both boundary
Table 1 Results of grid independence analysis of fourth-order accurate N u values in isothermal

and isoflux case for Re = 200, Pr = 0.0208. In bracket, the values denote step size (h) of the grid
Grid Re = 200 (isoflux) Re = 200 (isothermal)
81 × 81 3.34450889 3.16454530
(0.054775)
101 × 101 3.35019445 3.16893220
(0.043820)
121 × 121 3.35212421 3.17041349
(0.036517)
141 × 141 3.35293651 3.17103529
(0.031300)
161 × 161 3.35332465 3.17133307
(0.027388)
181 × 181 3.35352802 3.17148972
(0.024344)
201 × 201 3.35372853 3.17174220
(0.021910)
Table 2 Comparison of N u
Re Dhole et al. [4] Present % Diff
values calculated from
SGHOCS with the literature 1 3.4948 3.4754 0.5551
values in case of isoflux 10 7.2216 7.2326 0.1523
boundary condition at distinct
40 12.3494 12.4911 1.1474
Re and Pr = 10 with absolute
percentage error 100 18.3337 18.4782 0.7882
200 25.2408 25.1357 0.4164

Fig. 1 Variation of d N u /dh with refer to the step size (h) for Pr = 0.0208 and Re = 200
(isothermal and isoflux case)
Fig. 2 Angular variation of Nu in isothermal and isoflux case for distinct values of Pr and Re =
200
conditions are in line with fourth-order accurate reference line. It shows that as grid
refines (h → 0), the present results approach to fourth-order accuracy.
Figure 2 shows the Nu variation on the sphere surface at Re = 200 and Pr =
0.004, 0.0208, 0.065 in case of isothermal and isoflux thermal boundary conditions.
The Nu patterns are qualitatively equivalent in both isoflux and isothermal boundary
cases. It is noticed from Fig. 2 that at θ = π, the Nu values are higher in isothermal
case as compared to isoflux, whereas the opposite effect is shown at rear stagnation
point. Moreover, the difference between local Nusselt number values of isoflux and
isothermal cases are increasing with Pr.
The variation of N u with different Re at Pr = 0.004, 0.0208, 0.065 in isoflux
and isothermal cases is shown in Fig. 3. It is shown from Fig. 3 that for lower Pr
= 0.004 and all the values of Re, there is no much difference between N u values
for isoflux and isothermal cases. The similar result is shown for small values of Re
and Pr = 0.0208 and 0.065. Further, it can be observed that the difference between
N u values of isoflux and isothermal cases is increasing for Pr = 0.0208, 0.065 and
higher values of Re. Moreover, the heat transfer rate is more in isoflux case when
compared to isothermal case.
Fig. 3 N u values calculated from SGHOCS for different values of Re and Pr
4 Conclusions
The role of isoflux and isothermal boundary conditions has been investigated in this
work for the problem of forced convective heat transfer from a sphere for liquid
metals by applying SGHOCS. The main points which can be concluded from the
current numerical investigations are highlighted as follows.
(1) As grid refines, the present results approach to fourth-order accuracy.
(2) At θ = π, the Nu values are higher for isothermal case in compare to isoflux
case, while opposite results are seen at θ = 0 and also, the difference between
Nu values for isoflux and isothermal is increasing, as Pr increases.
(3) For Pr = 0.0208, 0.065 and higher values of Re, the difference between average
Nusselt number values of isothermal and isoflux cases is increasing.
(4) The heat transfer rate for isoflux boundary condition is more as compared to
isothermal boundary condition.
Acknowledgements The authors would like to thank the TEQIP-III NIT Silchar for funding
supports.
References
1. Lorezin N, Abanades A (2016) A review on the application of liquid metals as heat transfer
fluid in concentrated solar power technologies. Int J Hydrogen Ener 41:6990–6995
2. Hsu CJ (1967) Analytical study of heat transfer to liquid metals flowing along a row of spheres.
Int J Heat Mass Transf 10:1693–1708
3. Witte LC (1968) An experimental study of forced-convection heat transfer from a sphere to
liquid sodium. ASME J Heat Transf 90:9–12
4. Dhole SD, Chhabra RP, Eswaran V (2006) A numerical study on the forced convection heat
transfer from an isothermal and isoflux sphere in the steady symmetric flow regime. Int J Heat
Mass Transf 49:984–994
5. Ramteke RR, Kishore N (2017) Effects of uniform heat flux and velocity-slip conditions at
interface on heat transfer phenomena of smooth spheres in newtonian fluids. J Heat Transf 139:
6. Bharti RP, Chhabra RP, Eswaran V (2007) A numerical study of the steady forced convection
heat transfer from an unconfined circular cylinder. Heat and Mass Transf 43:639–648
7. Paramane SB, Sharma A (2010) Heat and fluid flow across a rotating cylinder dissipating
uniform heat flux in 2D laminar flow regime. Int J Heat Mass Transf 53
8. Sufyan M, Sufyan M, Nadeem AS (2015) Free stream flow and forced convection heat transfer
across rotating circular cylinder in steady regime: effects of rotation, prandtl number and
thermal boundary condition. J Mech Sci Tech 29:1781–1797
9. Sekhar TVS, Raju BHS, Murthy PVSN (2016) Higher order compact scheme for laminar
natural convective heat transfer from a sphere. Appl Math Modelling 40:2039–2055
10. Sekhar TVS, Raju BHS (2012) An efficient higher order compact scheme to capture heat
transfer solutions in spherical geometry. Comp Phy Commun 183:2337–2345
11. Nath D, Pati S, Raju BHS (2019) Analysis of mixed convection past a heated sphere. Proc Inst
Mech Eng Part E: J Pro Mech Eng 233(2019):601–616
12. Raju BHS, Nath D, Pati S (2018) Effect of prandtl number on thermo-fluidic transport
characteristics for mixed convection past a sphere. Int Commun Heat Mass Transf 98:191–199
13. Nath D, Raju BHS (2019) Effect of isoflux thermal boundary condition on mixed convective
heat transfer from a sphere for liquid metals. Int J Amb Ener. https://doi.org/10.1080/01430750.
2019.1636881
Numerical Analysis of Coupled Heat
and Mass Transfer in Solid Desiccant
for Stationary Hybrid Desiccants Beds
1 Introduction
Adsorption in a desiccant cooling cycles and desiccant air dehumidification systems

are used as good alternative methods for the vapor compression cooling systems. It
gives the provision to use the waste heat from various heat releasing applications
and as well as low grade energy sources like solar energy as a regeneration energy
source for desorption phase in the desiccant air dehumidification systems and many
evaporative cooling systems. The fixed desiccant bed is filled with homogenous
desiccant particles as shown in Fig. 1. The moist wet air enters in the inlet and dry
hot air leaves from the outlet during the adsorption process. To reduce the effect
of heat of adsorption and in order to improve the bed performance, Mujumdar [1]
studied the operation of a composite packed desiccant bed dehumidifier in which
the normal particles were mixed with inert high specific heat particles. Chang et al.
[2] studied the performance of a matrix material, wherein silica gel particles were
applied on a stainless steel substrates.
2 Problem Description and Governing Equations
The operation of desiccant packed bed during adsorption and desorption processes
includes both heat and mass two transfer processes. The first is mass transfer, referred
to as adsorption process. During the adsorption process, the adsorbate molecules are
transferred from the bulk gas flow to the desiccant surface due to adsorbate vapor
S. Muthu (B) · N. Sekarapandian

Department of Thermal and Energy, School of Mechanical Engineering, VIT Vellore, Vellore,
India
N. Sekarapandian
Fig. 1 Schematic of
stationary desiccant bed
pressure difference, and the mass transfer direction is inversed during the desorption
process, and the desiccant bed gets regenerated. As the moisture vapor pressure on
the surface of the desiccant particle is highly influenced by the desiccant temperature
and the moisture content on the desiccant, the heat of adsorption is dependent on the
mass transfer rate, which is influenced by the moisture vapor pressure on the desiccant
surface. Hence, this problem of adsorption and desorption in the fixed desiccant bed
has to be treated as a coupled transient heat and mass transfer phenomena.
The mathematical model of fixed desiccant bed is based on the following major
assumptions:
1. Bed wall material is having negligible moisture adsorption capacity.
2. The inlet air conditions are uniform. The thermo-physical properties of the dry
air and dry desiccant material are constant, but bulk properties are function of
moisture content.
3. The heat of adsorption is fully taken by desiccant layer.
4. The adsorption heat of composite silica gel is similar to that of conventional
silica gel.
Numerical Analysis of Coupled Heat and Mass Transfer … 763
The governing equations are derived based on the conservation of moisture and
conservation of energy for the air and desiccant medium in the fixed bed.
The moisture conservation in the air is given by
∂W km a
= (ωa − ωs ) (1)
∂t (1 − ε)ρs
Moisture conservation in desiccant is given by:
∂ωa G a ∂ωa km a
=− + (ωs − ωa ) (2)
∂t ερa ∂ y ερa
Energy conservation in the air is given by
∂θa G a ∂θa kh a
= + (θa − θs )
∂t ρa ε ∂ y ρa Ca ε
C v km a
+ (ωa − ωs )(θa − θs ) (3)
Ca ρa ε
Energy conservation in the desiccant is given by:
∂θs λe f f ∂ 2 θs Φ km a
= + (ωa − ωs )
∂t (1 − ε)ρs cs ∂ y 2 cs (1 − ε)ρs
kh a
+ (θa − θs ) (4)
(1 − ε)ρs cs
Nomenclatures are referred from Ramzy et al. [3].
The governing equations aforementioned in Sect. 2 are discretized using the finite
volume method with fully implicit scheme for time discretization, and the first and
second order spatial derivatives discretized using upwind and central differences
schemes, respectively. The system of linear algebraic equations so obtained is solved
iteratively using Gauss Siedel method. The numerical solutions produced using the
developed solver are validated against the theoretical and experimental results of [4,
5] subjected to the same bed geometrical parameters and operating conditions.
4 Motivation
The effect of particle diameter on the pressure drop in packed bed given in Ramzy
[3] is reproduced in Fig. 2. This figure depicts that, when the particle diameter and
the flow velocity are changed simultaneously, a wide range of pressure drop in the
bed is obtained. Hence, a logarithmic scale has been used for the pressure drop
axis. At a specified particle diameter, it can be observed that with the increase in
flow velocity, the pressure drop in the fixed desiccant bed increases. The impact of
varying the design parameters is investigated by analyzing the performance of the
various designs. The parameters varied are the direction of flow of the process air
and regeneration air streams with varying the inputs in periodic transient conditions.
During the dehumidification process in the bed, the outlet moisture content of air
decreases sharply at the beginning and reaches a minimum value and then increases
gradually until the end of the dehumidification process; whereas for regeneration in
the bed, the moisture content of air rises sharply at the beginning and then gradually
declines. The sharp edge in the start of regeneration air outlet moisture curve is
due to effects of the sudden transition from process section to regeneration section
Fig. 2 Pressure drop of fixed desiccant bed for various particle sizes [3]
where air is coming in opposite direction in the fixed bed. The packed desiccant
bed geometrical dimensions and four operating conditions for adsorption and two
operating conditions for regeneration are referred from Pesaran [4, 5].
In fixed desiccant bed, temperature of air at the outlet drops from the regeneration
temperature steadily during dehumidification. Throughout the regeneration process,
the temperature increases, but slightly at a faster rate, the beginning in the bed is
shown in Fig. 3. The effect on output moisture and temperature of air are predicted
under the periodic transient variable conditions of input moisture contents and shown
as Fig. 3.
6 Future Work
The author has come up with innovative hybrid stationary desiccant beds with
multiple types of desiccants to get the benefit of both Silicagel and molecular sieves
for the drying and dhimmification applications, as shown in Fig. 4. The parametric
simulation to predict the performance improvements of hybrid beds in comparison
with conventional mono-desiccant beds is to be presented, subsequently.
7 Conclusions
The numerical model shows a good agreement with experimental data. The effect
on output moisture content and temperature of air is predicted under the periodic
transient variable input moisture contents. As a future scope, this simulation is to
be used to predict the performance of hybrid desiccant beds for the drying and
dhimmification applications.
Fig. 3 Performance of fixed desiccant bed under different operating conditions (cases 1 → 6)
Fig. 4 Schematic of innovative designs of hybrid Stationary vertical packed desiccant beds
References
1. Majumdar P (1998) Heat and mass transfer in composite desiccant pore structure for
dehumidification. Sol Energy 62:1–10
2. Chang K-S, Chen M-T, Chung T-W (2005) Effects of the thickness and particle size of silica
gel on the heat and mass transfer performance of a silica gel-coated bed for air-conditioning
adsorption systems. Appl Therm Eng 25:2330–2340
3. Ramzy KA, Kadoli R, Babu TA (2012) Performance studies on the desiccant packed bed with
varying particle size distribution along the bed. Int J Refrig 35:663–675. https://doi.org/10.1016/
j.ijrefrig.2011.11.016
4. Pesaran AA, Mills AF (1987) Moisture transport in silica gel packed beds. I—theoretical study,
Int J Heat Mass Transfer 30(6):1037–49
5. Pesaran AA, Mills AF (1987) Moisture transport in silica gel packed beds-II. Experimental
study. Int J Heat Mass Transfer 30(6):1051–60
Computational Analysis of Indoor Air
Distribution Assessment on Under-floor
Air-Conditioning System
Lakshmanan Ramesh Krishnan and Lachireddi Gangadhara Kiran Kumar
Nomenclature
ρ Density
μ Viscosity
k Thermal conductivity
h Sensible enthalpy
cp Specific heat
PMV Predicted Mean Vote
PPD Percentage of People Dissatisfied
ACH Air Change per Hour
1 Introduction
The target of room air distribution was to maintain the appropriate combination of
temperature, velocity and humidity in the occupied zone of conditioned space. Poor
indoor environment condition affects the health, productivity and performance of
occupants in offices and classrooms. Computational fluid dynamics (CFD) tool was
used to predict the capability of UFAD in maintaining indoor environment and air
distribution. CFD has been commonly used in the design and optimal operation of
the indoor environment control in HVAC system [1]. The airflow pattern from floor
to ceiling has been applied to HVAC and developed as a different approach of venti-
lation, i.e., under-floor air distribution (UFAD) system. The major benefits of UFAD
system are supply of fresh air directly to the occupant zone and also reduction in
L. Ramesh Krishnan · L. Gangadhara Kiran Kumar (B)

CFD Lab, Mechanical Engineering Department, National Institute of Technology Calicut,
Kozhikode, Kerala, India
e-mail: ganga@nitc.ac.in
770 L. Ramesh Krishnan and L. Gangadhara Kiran Kumar
contaminant level. In UFAD system, the conditioned air passes through the under-
floor plenum and is supplied to the space by diffusers [2]. Three different angles of
supply conditions were tested in the UFAD office building. The 2D numerical simu-
lation of airflow, temperature distribution, relative humidity, contaminant concentra-
tion and thermal comfort indices of UFAD and overhead air distribution (OHAD)
system was evaluated, while the numerical results indicated that the UFAD system
saves 20–30% of energy consumption than OHAD [3]. An isothermal impinging jet
flow was investigated, and the turbulence model was predicted by comparing the
experiment results. The predicted results suggested that the RNG k − ε model was
considerably better than the SST k − ω model [4]. The investigation of different
ventilation systems was carried out numerically and was reported that the thermal
comfort and IAQ were within the acceptable level of both UFAD-improved desk
displacement ventilation (DDV+) and individual UFAD system. The energy savings
were compared with mixing ventilation system and reported that the savings are
16.6% for UFAD-DDV+ and 15.1% more for UFAD [5]. In the indoor environ-
ment, the thermal comfort strongly depends on the function of operating controls in
HVAC system. The detailed thermal comfort reviews were assessed for uniform and
non-uniform environment [6]. The objective of this study is to use the 3D compu-
tational domain model and to simulate the indoor air distribution of the room. CFD
simulations are examined for two cases, namely case A with 15ACH and case B with
10ACH, of UFAD system. Temperature distribution and airflow pattern are presented
in this paper.
2 Modeling and Simulation
The computational domain of the UFAD model is developed for a typical

computational-based research laboratory in university with the dimension of 5 m
(length) × 5 m (width) × 3 m (height) as shown in Fig. 1. The exhaust vent (1 m ×
0.25 m) is placed in the north wall and 2.1 m above the floor. The seven inlet personal
comfort diffusers are located in the floor. The different supply swirl diffusers are used
such as circular diffuser (diameter = 0.254 m) and rectangular diffuser (0.25 m ×
0.25 m), respectively.
The seated occupants with computer table and PC monitor, and CPU are positioned
nearer to the south and north walls. The internal heat sources are listed in Table 1.
The minimum ventilation rate of the computer laboratory should be 5 l/s per occu-
pant and 0.6 l/s.m2 according to ASHRAE standard 62.1. Simulations are performed
with identical cooling load, supply air temperatures and humidity ratio for both cases
A and B. The detailed boundary conditions considered for simulation are listed in
Table 2.
To attain the temperature and velocity distribution of thermal flow field, Navier–
Stokes equations with Boussinesq approximation are considered. The following
governing equations of mass, momentum and energy conservation are considered
for 3D steady state, incompressible and turbulent flow [5].
Computational Analysis of Indoor Air Distribution Assessment … 771
Fig. 1 Schematic
illustration of computational
model. Note: D1, D2, D3,
D4—circular swirl diffuser;
D5, D6, D7—rectangular
swirl diffuser; L1, L2, L3,
L4—ceiling lighting;
EV—exhaust vent; d—door;
W1, W2—windows; O1, O2,
O3, O4, O5, O6,
O7—occupants with PC
CPU; PC—monitor,
computer table; NW—north
wall; SW—south wall;
EW—east wall; and
WW—west wall
Table 1 List of internal heat

Heat source Load (W) No. s Total load (W)
gains
Occupants 130 7 910
Lighting 28 4 112
PC—CPU 90 7 630
PC—monitor 10 7 70
Total 1722 W
Table 2 Boundary
Boundary Conditions
conditions
Supply air flow rate Case A—15 ACH Case B—10 ACH
290 1/s 200 1/s
Supply velocity, round 0.8 m/s 0.65 m/s 0.56 m/s 0.45 m/s
rectangular
Supply air temperature 18 °C and 11 g/kg
and humidity ratio
Exhaust vent (outlet) Pressure outlet
Wall temperature Ceiling—27 °C, Floor—21 °C
NW—33 °C, SW—31 °C,
EW—32 °C, WW—32 °C
Continuity equation,
∇.(ρv) = 0 (1)
Momentum equation,
∂
(ρv) + ∇.(ρvv) = −∇ p + ∇.τ + ρ g (2)
∂t
Energy equation,
∂
(ρh) + ∇.(ρhv) = ∇.[(k + kt )∇T ] + Sh (3)
∂t
C pμt
kt = (4)
Prt
kt is the turbulent transport conductivity, and S h is the source term. The above
equations are solved by using ANSYS Fluent® based on the finite volume method.
The selected RNG k − E turbulence model has been widely used in the swirl diffuser
airflow, also for prediction of indoor airflow in different applications.
The 3D model computational domain is discretized by using unstructured tetra-
hedral mesh. A second-order discretization method is used to solve all the equations.
By using grid independence study, the suitable mesh cell size is identified based on
the accuracy and the computation time. For identification of grid independency, mesh
cells are tested from 7.9 to 24.78 lakhs as shown in Fig. 2. The mesh independence
is realized at 18.23 lakhs cells for the prediction of vertical air temperature profile.
Fig. 2 Mesh study

The 3D computational domain was simulated and investigated the local thermal
discomfort around the occupants from ankle to head, temperature and velocity distri-
bution at Y = 1.1 m above the floor and predicted mean vote (PMV)–percentage of
people dissatisfied (PPD) for each occupant.
Table 3 shows the numerical temperature gradient between head and feet level of
seated occupants for both the cases. The 4 poles are located in the 4 sides of each
occupant at a distance of 0.1 m. The mean temperature value of 4 poles at an ankle
(0.1 m) and head level (1.1 m) was tabulated clearly. As per the standard ISO7730,
the vertical temperature gradient does not exceed more than 3 °C; if it exceeds,
the occupant may feel local thermal discomfort. The activity levels of computational
laboratory are reading or writing or typing (i.e., 1.1 met), and the thermal insulation of
clothing level is 0.61clo. The theoretical calculation of minimum and maximum oper-
ative temperature range (25.46–26.68 °C) was found for the corresponding activity
level and clothing [7].
From case A, the vertical temperature difference has exceeded more than 3 °C for
occupant 1, occupant 3, occupant 4 and other occupants. T is within the acceptable
limit. Some occupants’ (1, 2, 3, 4 and 5) head level temperature range is attained
outside the comfortable operative temperature. From case B, the numerical temper-
ature of occupant’s head level is almost 2 °C higher than the acceptable level. The
case B supply airflow rate is not suitable to attain the comfortable environment in
the laboratory.
Figures 3 and 4 represent the contour of temperature distribution at 1.1m horizontal
plane from the floor. As shown in Fig. 3, the observed temperature range between
north wall to occupants and south wall to occupants is about 26 to 29 °C. The midplane
of the room has maintained the temperatures from 24 to 26 °C. Moreover, the heat
interaction between occupants and wall is higher when compared to the midplane
(Z = 2.5 m) and that results in temperatures more than 26 °C. Figure 4 shows the
temperature distribution of case B supply conditions. The obtained simulation results
Table 3 Local thermal discomfort for each occupant

Position Occupant 1 Occupant 2 Occupant 3 Occupant 4
Case A Case B Case A Case B Case A Case B Case A Case B
Tankle (0.1m) °C 24.09 25.43 25.93 27.95 23.81 25.69 23.81 25.78
Thead (1.1m) 27.29 29.49 28.77 29.07 28.18 28.36 26.94 28.52
T °C 3.2 4.057 2.84 1.145 4.37 2.67 3.13 2.74
Position Occupant 5 Occupant 6 Occupant 7
Case A Case B Case A Case B Case A Case B
Tankle (0.1m) °C 26.66 28.205 24.6 24.55 23.58 25.18
Thead (1.1m) °C 29.02 29.72 26.32 27.75 26.39 28.15
T °C 2.36 1.515 1.72 3.2 2.81 2.97
Fig. 3 Temperature distribution at Y = 1.1 m—case A
Fig. 4 Temperature distribution at Y = 1.1 m—case B
of temperature are around 27.36–29.23 °C. It feels uncomfortable to the occupants.

Figures 5 and 6 represent the vector plot of velocity distribution at Y = 1.1 m above
the floor. From Fig. 5, it is clear that the inlet of the diffusers supplies the conditioned
air in 4 different directions at the angle of 35°. The air flows under the computer table,
CPU and occupants, and it absorbs heat from the heat source. The velocity around
the occupants is much lower (<0.06 m/s) than the mid of the plane. The velocity
range at this level is maintained less than 0.2 m/s and is within the acceptable level.
As shown in Fig. 6, the velocity range (>0.25 m/s) of occupant 4 to occupant 7
at this plane is higher than the acceptable range. Due to this condition, the occupant
Fig. 5 Velocity distribution at Y = 1.1 m—case A
Fig. 6 Velocity distribution at Y = 1.1 m—case B

Table 4 Thermal sensation scale and PMV-PPD [7]
Table 5 PMV-PPD indexes of each occupant

Cases Occupant 1 Occupant 2 Occupant 3 Occupant 4
PMV PPD % PMV PPD % PMV PPD % PMV PPD %
Case A 0.74 17 1.15 33 1.05 28 0.62 13
Case B 1.22 36 1.33 42 1.17 34 1.22 36
Cases Occupant 5 Occupant 6 Occupant 7
PMV PPD % PMV PPD % PMV PPD %
Case A 1.31 41 0.47 10 0.5 10
Case B 1.6 56 0.98 25 1.11 31
feels draft risk. The occupant 1 to occupant 3 experience the velocity less than 0.2 m/s
at face level, and it falls under the acceptable condition.
The classic steady-state method of Fanger model was used to predict the indoor
thermal comfort for numerical simulations. This was the first model and widely used
to determine the comfort level. Based on the Fanger model, the thermal comfort
indices of PMV and PPD were developed. The variables of occupant activity,
clothing, mean radiant, dry bulb temperature, relative humidity and air velocity are
used to evaluate PMV. PMV index is a thermal sensation scale based on seven-point
scale as shown in Table 4, and PPD is the percentage of person dissatisfied, i.e., based
on the PMV.
For both simulation case studies, the PMV and PPD are given in Table 5. The
PMV-PPD is evaluated around the whole body of individual occupant. In case A, the
occupants 2, 3 and 5 attained more than 20% of PPD and other occupants experienced
less than 20%. In case B, all the occupants experienced more than 20% of PPD and
they had a maximum risk of discomfort as compared with case A.
4 Conclusion
The UFAD system in computational laboratory for two different cases of ACH
at constant heat load was analyzed computationally. The temperature distribution,
airflow movement and local thermal discomfort from head to ankle level were carried
out by using ANSYS Fluent® . As per the discussion of results, case B seems to have
more number of occupants with a risk of local thermal discomfort and vertical temper-
ature gradient found to be more than 3 °C when compared with case A. Based on the
simulation results, the thermal comfort indices of PMV-PPD around each occupant
had been found using the Fanger model. For case A, the PMV-PPD was perceived
within the acceptable level for a greater number of occupants than case B. This inves-
tigation clearly states the comfort conditions and temperature gradient around the
individual occupant for identical internal heat gains. Further study is needed for the
parametric analysis of supply conditions and to identify the effectiveness of various
diffuser types in the UFAD system.
References
1. Chen Q, Srebric J (2002) A procedure for verification, validation, and reporting of indoor
environment CFD analyses. HVAC&R Res 8:201–216
2. Zhang K, Zhang X, Li S, Jin X (2014) Review of underfloor air distribution technology. Energy
Build 85:180–186
3. Ho SH, Rosario L, Rahman M (2011) Comparison of underfloor and overhead air distribution
systems in an office environment. Build Environ 46:1415–1427
4. Chen HJ, Moshfegh B, Cehlin M (2012) Numerical investigation of the flow behavior of an
isothermal impinging jet in a room. Build Environ 49:154–166
5. Rahmati B, Heidarian A, Jadidi AM (2018) Investigation in performance of a hybrid under-
floor air distribution with improved desk displacement ventilation system in a small office. Appl
Therm Eng 138:861–872
6. Enescu D (2017) A review of thermal comfort models and indicators for indoor environments.
Renew Sustain Energy Rev 79:1353–1379
7. ASHRAE Fundamentals handbook, (2009)
Review and Performance Evaluation
of Artificially Roughened Solar Air
Heaters
Ayush Jain, Sheetal Kumar Jain, Ghanshyam Das Agrawal, and Rohit Misra
1 Introduction
Sun is the main source of all forms of energy. The demand of energy to our society is
growing, and the conventional forms of energies are on the verge of extinction. So,
to meet the demands of the society, there is a need to look for an alternative source
of energy, and the answer to this is solar energy. Solar air collectors have very low
thermal efficiency because of low convective heat transfer coefficient between the
air and the absorber plate. The use of artificial roughness on a surface of the solar air
heater is an efficacious method to enhance heat transfer to fluid flowing in the duct.
Artificial roughness in the form of wires and in various other arrangements has been
used to create turbulence near the wall or to disturb the boundary layer. Air heating is
an important source of solar thermal applications, used for space heating and process
heating such as desalination, crop drying, laundry, and other drying processes. The
solar energy not only reduces the operational cost of the system but also a clean form
of energy so as to curb the pollution.
A. Jain (B) · S. K. Jain · G. Das Agrawal

R. Misra
Government Engineering College Ajmer, Ajmer, Rajasthan, India
780 A. Jain et al.
Fig. 1 Solar air heater
1.1 Solar Air Heater
A conventional solar air heater is essentially a flat plate collector with absorber plate,
a transparent cover system at the top, and insulation at the bottom and on the sides
as shown in Fig. 1. The whole assembly is encased in a sheet metal container. The
working fluid is air, though the passage for its flow varies according to the type of
air heater.
1.2 Artificial Roughness
All materials have naturally rough surfaces as no surface can be considered as smooth
at microscopic level. However, this roughness in the plate has only negligible effect
on the performance of the solar air heater, so it can be considered as convective
heat transfer process. It has been observed that the presence of artificial roughness
elements of optimal dimension can lead to a considerable change in the fluid flow and
heat transfer. The roughness of different shapes and size can be applied over surfaces
that are exposed to fluid stream in order to create the turbulence in the fluid flow
Review and Performance Evaluation of Artificially Roughened … 781
pattern. The popular roughness geometries include grit, ribs, dimples, protrusions,
wings, blockages, etc. [1–4].
1.3 Classification of Solar Air Heater Based on Without

Thermal Energy Storage
• Non-porous-type solar air heater: In non-porous type, air stream does not flow
through below the absorber plate, but air may flow above and/or behind the plate.
The non-porous type with air passage below the absorber is most commonly used.
A plate parallel to the absorber plate is provided in between the absorber and the
insulation, thus forming a passage of high aspect ratio. In another variety of non-
porous type air heater, the absorber plate is cooled by air stream flowing on both
sides of the plate.
• Porous-type solar air heater: The second type of air heaters has porous absorber
which may include slit and expanded metal, overlapped glass plat absorber, and
transpired honeycomb.
1.4 Solar Air Heater Advantages
• The need to transfer heat from working fluids to another fluid is eliminated as air
is being used directly as the working substance. The system is compact and less
complicated.
• Corrosion is a great problem in solar water heater, and this problem is not
experienced in solar air heaters.
• Leakage of air from the duct does not create any problem.
• Freezing of working fluid virtually does not exist.
• The pressure inside the collector does not become very high.
• Thus, air heater can be designed using cheaper as well as lesser amount of material,
and it is simpler to use than the solar water heaters.
1.5 Solar Air Heater Disadvantages
• The poor heat transfer properties of air.

• The need for handling large volume of air due to its low density.
• Air cannot be used as a storage fluid because of its low thermal capacity.
• In the absence of proper design, the cost of solar air heaters can be very high.
782 A. Jain et al.
1.6 Solar Air Heater Applications
The applicability of the solar air heater depends on various factors like high efficiency,
low fabrication cost, low installation, and operational cost and some other specific
factors regarding specific uses. Extensive work in solar air heaters has been done.
Various geometries have been proposed, and their theoretical investigation is carried
out. But it needs commercial exploitation.
This review focuses to show the efforts of the researchers working on SAH and
find the ways to present it by means of robust applications to use and thus to increase
the performance for consideration in the design and development in present context.
The researchers have used many geometries to enhance the performance of the solar
air heater and develop the empirical correlations.
2 Roughness Geometries Used in Solar Air Heater
Patel and Lanjewar [5] conducted an experimental and numerical investigation on

novel V-rib roughened SAH and observed that the enhancement in Nu was observed
to be 1.55–2.26 times to that of smooth plate and an increase in friction factor was
observed as 2.63–3.40 times to that of smooth plate. Thermo-hydraulic performance
parameter at p/e = 10 was found to be maximum, and it varies from 1.14 to 1.59.
Varun et al. [6] conducted an experiment to investigate the thermal performance of
solar air heater having roughness elements as a combination of inclined and transverse
ribs on the absorber plate. The best results obtained for the roughened collector with
the geometry having the relative roughness pitch of 8.
Deo et al. [7] conducted an experimental and CFD analysis of solar air heater
duct roughened with multiple broken transverse ribs. The results obtained show that
the maximum increase of Nusselt number is 3.34 times the smooth plate while the
thermo-hydraulic performance is observed as 2.45 times the smooth plate corre-
sponding to the P/e value of 12, e/Dh value of 0.044, and angle of attack value of
60° at Reynolds number of 12,000, whereas maximum increase in friction factor is
observed as 3.38 times than the smooth plate corresponding to the P/e value of 12,
e/Dh value of 0.044, and angle of attack value of 80° at Reynolds number of 12,000.
Singh et al. [8] conducted an experimental and CFD analysis of solar air heater duct
roughened with multiple broken transverse ribs. The observed results show that the
maximum thermal efficiency was 44.26% for a simple solar air heater, for square
wave-shaped ribs, it is found as 71.46%, and for the multiple broken transverse
ribs, it is found as 72.25%. Kumar et al. [9] conducted experimental investigation
and developed a correlation for heat transfer and friction factor for solar air heater
duct artificially roughened with ‘S’ shape ribs, and maximum deviation for devel-
oped correlations for Nusselt number and friction factor was observed to be ±10.8%
and ±10%, respectively. Komolafe et al. [10] conducted an experimental investi-
gation and thermal analysis of solar air heater having rectangular rib roughness on
Table 1 Various ribs roughened geometries and results

S. No. Author Year Geometry Input parameters Results
1 Patel and 2019 Novel V-rib P/e = 6–14, e/Dh = Nu/Nus =
Lanjewar [5] 0.043 1.55–2.26
g/e = 4, d/w = 0.65 f/fs = 2.63–3.40
α = 60°, p’/p = THPP = 1.14–1.59
0.4, r/e = 4
g’/e = 1, Ng = 3
ng = 4, Re =
4000–14,500
2 Varun et al. [6] 2008 Inclined and Re = 2000–14,000 The roughened
transverse rib P/e = 3–8 collector with
e/Dh = 0.030 absorber plate
having relative
roughness pitch of
8 gives the best
performance
3 Deo et al. [7] 2016 Multigap V-down W/H = 12, Re = Nu/Nus = 3.34
ribs combined 4000–12,000 f/fs = 3.38
with staggered P/e = 4–14 THPP = 2.45
ribs e/Dh =
0.026–0.057
α = 40–80°, g/e =
1
w/e = 4.5, p/P =
0.65
n=2
4 Singh et al. [8] 2019 Multiple broken P/e = 10 Maximum thermal
transverse ribs e/Dh = 0.043, W/w enhancement of
=7 2.50 times with
Re = 3000–18,000 corresponding
pumping power
penalty of 3.92
times is obtained
for square
wave-shaped ribs,
whereas in case of
multiple broken
ribs, the
enhancements are
3.24 and 3.85
times, respectively
5 Kumar et al. 2017 S-shaped ribs W/H = 12, P/e = Nu/Nus = 4.64
[9] 4–16 f/fs = 2.71
e/Dh = THPP = 1.57
0.022–0.054
α = 30–75°, W/w
= 1–4
Re = 2400–20,000
(continued)
784 A. Jain et al.
Table 1 (continued)
S. No. Author Year Geometry Input parameters Results
6 Komolafe et al. 2019 Rectangular rib I = 827.87 W/m2 ηthermal =
[10] Ta = 33.77 °C 14.0–56.5%
Min. Temp. = 21
Max. Temp. =
127 °C
7 Jain and 2018 V-rib with P’/P = 0.65, g/e = Nu/Nus = 2.30
Lanjewar [11] symmetrical gap 4 f/fs = 3.13
and staggered rib w/e = 4, α = 60° THPP = 2.45
geometry Ng = 3, P/e =
10–16
Re = 3000–14,000
8 Patel and 2018 Multiple discrete P/e = 12, e/Dh = Nu/Nus = 2.27
Lanjewar [12] V patterns 0.045 f/fs = 4.05
combined with α = 60°, p°/P = THPP = 1.55
staggered ribs 0.65
g/e = 1, w/e =
2–5.5
Ng = 4, Re =
3000–12,000
9 Kumar and 2018 Twisted rib Re = 3500–21,000 Nu/Nus = 2.46
Layek [13] roughness P/e = 6–10, α = f/fs = 1.78
30–90°
twist ratio (y/e) =
3–7
10 Singh and 2018 Square wave P/e = 4–30 Nu/Nus = 2.14
Singh [14] profiled Re = 3000–15,000 f/fs = 3.55
transverse ribs e/Dh = 0.043 THPP = 1.43
the absorber plate and calculated thermal efficiency value ranged between 14.0 and
56.5%. The simulated minimum and maximum temperature of the solar air heater
were 21 and 127 °C, respectively. Jain and Lanjewar [11] observed an increase in
Nusselt number in V-rib with symmetrical gap and staggered rib of 2.30 times the
smooth duct while the increase in friction factor is 3.13 times as compared to the
smooth duct for novel v-rib geometry.
Patel and Lanjewar [12] conducted an experimental analysis for enhancement of
heat transfer in multiple discrete V patterns combined with staggered ribs on solar air
heater and observed that Nusselt number ratio comes out in between 1.94 and 2.27
and the friction factor ratio is in between 2.46 and 3.6 over the range of Reynolds
number. Kumar and Layek [13] developed a correlation for Nusselt number and
friction factor of solar air heater having twisted rib roughness. Singh and Singh
[14] have done a CFD analysis of solar air heater duct having square wave profiled
transverse ribs as roughness elements. The maximum value of Nusselt number and
friction factor was found to be 2.14 and 3.55 times the smooth duct. Table 1 shows
the geometries investigated in recent past along with their results.
3 Conclusions
This paper provides a thorough overview of the various artificial roughness that
many scientists have used to improve heat transfer. This paper discusses the impact
of different rib geometric parameters such as rib height and pitch on heat thermal
performance of SAH.
References
1. Jain SK, Agrawal GD., Misra R (2019) A detailed review on various V-shaped ribs roughened
solar air heater. Heat Mass Transf. https://doi.org/10.1007/s00231-019-02656-4
2. Jain SK, Agrawal GD, Misra R (2019) Heat transfer augmentation using multiple gaps in arc-
shaped ribs roughened solar air heater : an experimental study. Energy sources, part a recover.
Util Environ Eff 1–12. https://doi.org/10.1080/15567036.2019.1607945
3. Jain SK, Misra R, Verma P, Rathore S, Jamuwa DK, Agrawal GD (2019) Performance investi-
gation of triangular solar air heater duct having broken inclined roughness using computational
fluid dynamics. J Sol Energy Eng 1. https://doi.org/10.1115/1.4043751
4. Jain SK, Agrawal GD, Mishra R (2017) Review on thermal performance enhancement of Solar
air heater using artificial roughness. In: Proceedings of 2017 IEEE international conference on
technological advancements in power and energy: exploring energy solutions for an intelligent
power grid, tap energy 2017. IEEE
5. Singh Patel S, Lanjewar A (2019) Experimental and numerical investigation of solar air heater
with novel V-rib geometry. J Energy Storage 21:750–764. https://doi.org/10.1016/j.est.2019.
01.016
6. Saini RP, Singal SK (2008) Investigation of thermal performance of solar air heater having
roughness elements as a combination of inclined and transverse ribs on the absorber plate.
Renew Energy 33:1398–1405. https://doi.org/10.1016/j.renene.2007.07.013
7. Deo NS, Chander S, Saini JS (2016) Performance analysis of solar air heater duct roughened
with multigap V-down ribs combined with staggered ribs. Renew Energy 91:484–500. https://
doi.org/10.1016/j.renene.2016.01.067
8. Singh I, Vardhan S, Singh S, Singh A (2019) Experimental and CFD analysis of solar air
heater duct roughened with multiple broken transverse ribs: a comparative study. Sol Energy
188:519–532. https://doi.org/10.1016/j.solener.2019.06.022
9. Kumar K, Prajapati DR, Samir S (2017) Heat transfer and friction factor correlations develop-
ment for solar air heater duct artificially roughened with ‘S’ shape ribs. Exp Therm Fluid Sci
82:249–261. https://doi.org/10.1016/j.expthermflusci.2016.11.012
10. Komolafe CA, Oluwaleye IO, Awogbemi O, Osueke CO (2019) Experimental investigation
and thermal analysis of solar air heater having rectangular rib roughness on the absorber plate.
Case Stud Therm Eng 14: https://doi.org/10.1016/j.csite.2019.100442
11. Jain PK, Lanjewar A (2019) Overview of V-RIB geometries in solar air heater and performance
evaluation of a new V-RIB geometry. Renew Energy 77–90. https://doi.org/10.1016/j.renene.
2018.10.001
12. Patel SS, Lanjewar A (2018) Experimental analysis for augmentation of heat transfer in multiple
discrete V-patterns combined with staggered ribs solar air heater. Renew Energy Focus 25:31–
39. https://doi.org/10.1016/j.ref.2018.03.003
13. Kumar A, Layek A (2019) Nusselt number and friction factor correlation of solar air heater
having twisted-rib roughness on absorber plate. Renew Energy 130:687–699. https://doi.org/
10.1016/j.renene.2018.06.076
14. Singh I, Singh S (2018) CFD analysis of solar air heater duct having square wave profiled
transverse ribs as roughness elements. Sol Energy 162:442–453. https://doi.org/10.1016/j.sol
ener.2018.01.019
Experimental Study of Gas-Solid
Fluidized Bed Hydrodynamics Using
Optical Fiber Probe
Mona Mary Varghese, Narasimha Mangadoddy, and Teja Reddy Vakamalla
1 Introduction
Gas-solid fluidized beds have numerous applications in the recent growing indus-
trial world. The main applications of these include catalytic cracking, gasification
of coal, granulation, pyrolysis of food waste, etc. The distinct properties include
high heat and mass transfer, ideal mixing, uniform temperature gradient and a large
area of contact. The bed remains in the packed condition unless the drag force is
balanced by the gravitational force. Characterizing the hydrodynamics will help in
optimizing the design of a fluidized bed reactor. To determine the solid holdup and
velocities in the fluidized bed, an optical fiber probe is one of the potential tools that
can be employed [1]. The non-invasive techniques such as tomography are ineffec-
tive in case of industrial operating conditions, i.e., at high gas holdups and solids
concentration, and fail to work efficiently in turbulent flows [2]. Taofeeq [3] used
an optical fiber probe to measure the time series solid holdup and solid velocity.
The obtained results were compared against the gamma-ray computed tomography
measurements for solid holdup and radioactive Particle tracking for solid velocity. Xu
et al. [4] reviewed different calibration methods used for the optical fiber probe and
proposed a new calibration technique. The newly proposed calibration technique was
capable of calibrating the optical fiber probe at the full range from zero to packed bed
concentration. Hydrodynamics of turbulent fluidized bed using optical fiber probe
was studied by Ellis et al. [5]. He found that with an increase in bed diameter, the
M. M. Varghese (B) · T. R. Vakamalla

Department of Chemical Engineering, NIT Calicut, Kattangal, Kerala, India
e-mail: mona_p180010ch@nitc.ac.in
T. R. Vakamalla
e-mail: teja@nitc.ac.in
N. Mangadoddy
Department of Chemical Engineering, IIT Hyderabad, Kandi, Telangana, India
e-mail: narasimha@iith.ac.in
788 M. M. Varghese et al.
transition velocity (U c ) decreases. Using optical probes, void fraction and velocity
profiles were obtained and compared with different column diameters. The compar-
ison indicated that the effect of the superficial velocity was higher than the column
diameter in radial profiles of time-averaged mean voidage. From the literature, it
is observed that a detailed parametric study of the effect operating conditions on
gas/solid flow dynamics is absent.
In the last 2 decades, computational fluid dynamics (CFD) become an essen-
tial tool for obtaining internal fluid dynamics of gas-solid fluidized beds [6]. The
main computational approaches used in the case of gas-solid fluidized beds are the
Eulerian–Eulerian and Eulerian–Lagrangian approaches. In the earlier case, both
the phases are treated as an interpenetrating continuum while the case of Eulerian–
Lagrangian, the gas phase is treated as a continuum and the particles as discrete.
Taghipour et al. [6] conducted an experimental and computational study on a 2D
gas-solid fluidized bed with spherical glass beads of 250–300 µm. The adopted drag
correlations (Syamlal O’Brien, the Gidaspow and Wen & Yu) showed a reasonable
agreement for mean pressure drop, bed expansion, and qualitative glass-solid flow
pattern. Bakshi et al. [7] carried out Eulerian–Eulerian simulation and concluded
that the specularity coefficient in the range of 0.01–0.3 predicted bubble diameters
are in good agreement with experimental values. It is concluded that the Gidaspow
model predicted well for the homogeneous system while Syamlal O‘Brien model
predictions were superior for higher gas velocities.
The current work aims to measure the hydrodynamic characteristics of fluidized
beds such as mean solid, gas holdup and solid axial velocity for different operating
conditions with optical fiber probe. A total of 3 static bed height and 3 flow rate
conditions is varied with Geldart B particles. Further, it is also planned to conduct
a CFD study for predicting the internal flow dynamics. Cross-validation of CFD
predicted flow properties against in-house measured experimental measurements is
also included in the present work.
2 Methodology
The experimental setup consisted of a cylindrical Plexiglas fluidized bed with 0.105 m
inner diameter and 1.02 m height with a thickness of 0.005 m (Fig. 1). A porous
plate with 63 holes of 0.003 m diameter is used to distribute the air to the column.
Spherical glass beads with a density of 2500 kg/m3 and size range 400–600 µm are
considered to conduct the experimental studies. The particles are filled up to a bed
height ranging from 0.10 m to a maximum of 0.21 m with a solid volume fraction
of 0.6. The flow rates are varied between 110 and 150 lpm. A U-tube manometer
is utilized for measuring the pressure drop. PV6 optical fiber probe (Institute of
Process Engineering, Chinese Academy of Sciences) is used to measure the solid
Experimental Study of Gas-Solid Fluidized Bed Hydrodynamics … 789
Fig. 1 Fluidized bed test rig with PV6 optical fiber probe and data acquisition system
holdup and velocity at 3 different axial positions and four radial positions at each
axial position. All the experiments are repeated a minimum of three times, and error
bars are provided to show the accuracy level of the experiments. Unless it is specified,
the experimental measurements at each point are taken for a minimum of 10 s after
the starting of fluidization.
The gas-solid flow in the fluidized bed is simulated using the Eulerian multi-fluid
approach. The finite volume method is used for solving the governing equations
of mass and momentum. The kinetic theory of granular flow, which considered
the conservation of solid fluctuation energy, is used for the closure of solid stress
terms. The pressure outlet boundary condition is specified at the top and no-slip
Table 1 Details of variables

Variables Values
used in simulation studies
Inlet flow rate, lpm 150
Particle diameter, µm 450
Particle density, kg m−3 2500
The angle of internal friction, º 28
Time step, seconds 10−5
Simulation time, seconds 30
Restitution coefficient 0.95
Specularity coefficient 0.6
Coefficient of lift 0.25
Frictional stress model Syamlal O’Brien
Discretization Third-order
for the gas phase and partial slip with a specularity coefficient of 0.6 is used for
particles at the walls. Pressure and velocity are coupled by the SIMPLE scheme. At
maximum, 20 iterations per time step are used, and the residual of 10−5 is specified
for the convergence criteria. The parameters considered for the CFD simulation of
the fluidized bed are provided in Table 1. The 2D simulations are conducted with
ANSYS’s Fluent 19.1.
The hydrodynamic parameters analyzed in this experimental study includes mean

solid holdup and axial velocity of particles at different operating conditions. The solid
holdup measurement is essential to analyze the operation taking place in a fluidized
bed reactor as it greatly influences the mass transfer and heat transfer properties in
a fluidized bed. Figure 2a shows solid holdup radial profiles (r/R = 0 represents
the center and r/R = 1 represents the wall) obtained for different flow rates. The
displayed results are plotted at an axial position of 8 cm from the bottom for a 21 cm
static bed height (BH) of the column. A raise in the solid holdup can be witnessed by
moving away from the center at all the flow rates. This can be true since most of the
gas trying to escape from the center, thereby pushes the solids away from the center
toward the wall. With the increase in flow rates from 110 to 150 lpm, a decrease in
the solid holdup of 5.9% at the center and 7.6% at the wall is observed. Since the rise
in gas flow rate increases the air entering into the fluidized bed, a reduction in the
solid holdup can be expected. Analysis of variation of radial solid holdup for a fixed
axial position (8 cm) and flow rate (110 lpm) revealed that solid holdup decreases
with a reduction in static bed height (Fig. 2b). The fixed axial position taken for
plotting is in the top portion for a static bed height of 10 cm. Whereas the same
point is positioned in the middle for a static bed height of 15 cm, and it is further
Fig. 2 Radial profile of

solid holdup a. for 21 cm
static bed height at 8 cm
axial position for different
flow rates b. at 8 cm axial
position and 110 lpm for
different static bed heights c.
for 21 cm static bed heights
and 130 lpm flow rate at
different axial positions
moved to a down position for 21 cm bed height. In the case of 10 cm bed height, the
presence of the gas phase is more near to the top portion; hence, a low solid holdup
is observed. Figure 2c depicts the radial solid holdup variation with axial positions
for a fixed bed height of 21 cm and 130 lpm flow rate. A decrease in solid holdup is
observed with the axial position. This can be true since most of the tiny gas bubbles
form near the bottom positions trying to coalesce as moving upward the fluidized
bed. This can improve the gas holdup with the axial position; therefore, a reduction
in the solid holdup is witnessed with the axial position.
3.1 Axial Velocity
The properties exhibited by a fluidized bed mainly depend upon the velocity of
particles, i.e., the movement of particles inside the bed. In general, the velocity is
positive near the center and approaches negative value toward the wall. The negative
velocities near the wall are due to the particles back mixing. The measurements are
taken up to the radial position of r/R = 0.75 because of experimental difficulties
associated with the probe. Figure 3 shows the radial profile of particles velocity at
8 cm axial position for different flow rates. A decrease in particle velocity is observed
toward the wall. With increasing the flow rate from 110 to 150 lpm, a raise in particle
velocity is observed. It can be true since the increment in air velocity causes a large
fluctuation of solid particles. (Figure 3a). It is clear from Fig. 3b that axial velocity
decreases as we increase the static bed height of the column. With the increase in bed
height from 10 cm to 21 cm, higher velocity is needed for carrying the particles to the
top. It is noticed from Fig. 3c that the radial profile of axial velocity increases with
the axial position. This can be explained by the increase in bubble size moving from
the bottom to the top. A rise in the axial velocity enhances the mixing of particles
inside the bed.
The corresponding gas superficial velocity associated with 150 lpm is introduced
through the sparger at the bottom of the bed. At time t = 0 s, the bed is packed with
Fig. 3 Radial profile of

axial velocity a. for 21 cm
static bed height at 8 cm
axial position for different
flow rates b. at 8 cm axial
position and 110 lpm for
different static bed heights c.
for 21 cm static bed heights
and 130 lpm at different
axial positions
particles up to a height of 21 cm. The gas injected results in the formation of bubbles
as shown in Fig. 4. It is clear that the increase in velocity results in the expansion of
the bed and after 1.4 s, complete mixing of particles is observed. The time-averaged
values are obtained after taking the average for 16 s after 10 s simulation. Figure 5
shows the comparison of simulated and experimental measurements for the radial
solid holdup. Close predictions are observed near the wall and the predictions started
deviating as moving toward the wall. A large deviation from the experimental data
is observed near the wall. This can be because of the 2D model used for simulation.
As to capture the realistic phenomena inside the bed, a 3D model has to be used.
Fig. 4 Simulated volume fraction of solid contours of 2D fluidized bed (for an inlet flow rate of
150 lpm)
Fig. 5 Comparison of CFD

predicted radial profile of
solid holdup with
experimental measurement
for 21 cm static bed height
and 150 lpm flow rate at
8 cm axial position
4 Conclusions
In the present work, an experimental study is performed to understand the gas-solid

fluidized bed hydrodynamics using an optical fiber probe. This measurement allows
the determination of internal solid holdup and axial particle velocity. As the flow
rate increases from 110 to 150 lpm, a reduction of 11.87% gas holdup is noted. It is
found at a specific position of 8 cm from the bottom. The solid holdup increases with
bed height. A reduction of 9.16% in solid holdup is noticed with the increase in axial
position as a result of an increase in bubble size because of coalescence from bottom
to top. Further, an increase in axial velocity of the particle is noted with flow rate. High
velocities are observed near the center compared to wall positions. Additionally, CFD
is used to show the dynamic behavior of bubble formation in the fluidized bed. Also,
the CFD predicted solid holdup is cross-validated against experimental measurement,
and close predictions are observed near the center compared to wall position.
Acknowledgements The authors would like to thank NIT Calicut for funding and IIT Hyderabad
for providing experimental facilities.
References
1. Liu J, Grace JR, Bi X (2003) Novel multifunctional optical-fiber probe: I. Development and
validation. 49:1405–1420
2. Li X, Yang C, Yang S, Li G (2012) Fiber-optical sensors: basics and applications in multiphase
reactors. Sensors (Switzerland) 12:12519–12544
3. Taofeeq HM (2017) Impact of vertical internals on the hydrodynamics and heat transfer
coefficient in a gas-solid fluidized bed
4. Xu G, Liang C, Chen X, Liu D, Xu P, Shen L et al (2013) Investigation on dynamic calibration
for an optical-fiber solids concentration probe in gas-solid two-phase flows. 9201–9222
5. Ellis N, Bi HT, Lim CJ, Grace JR (2004) Hydrodynamics of turbulent fluidized beds of different
diameters. Powder Technol 141(1–2):124–136
6. Taghipour F, Ellis N, Wong C (2005) Experimental and computational study of gas-solid fluidized
bed hydrodynamics. Chem Eng Sci 60:6857–6867
7. Bakshi A, Altantzis C, Bates RB, Ghoniem AF (2015) Eulerian–Eulerian simulation of dense
solid-gas cylindrical fluidized beds: impact of wall boundary condition and drag model on
fluidization. Powder Technol 277:47–62
Advection Stabilization Using
Lower-Order Scheme Blending: A Case
Study of Rayleigh–Taylor Instability
Rahul Jha, Ashish Arote, and Jyotirmay Banerjee
1 Introduction
Solving Navier–Stokes equation and arriving at an acceptable solution become chal-

lenging due to its nonlinearity. This nonlinearity is posed by the advection term
(∇.(ρ −→
v v)) which needs to be discretized properly to avoid numerical errors. The
use of an advection scheme for such discretization exceeding the first-order accuracy
tends to create numerical dispersion near the discontinuity. Figure 1 shows a typical
single-shock problem where the scalar at different magnitudes enters from the left
and bottom faces of the domain creating discontinuity at the diagonal. It can be
seen from Fig. 1 that the schemes such as Lax–Wendroff (LW (O(2))) and Quadratic
Upstream Interpolation for Convective Kinematics (QUICK O(3)) are able to capture
the discontinuity with reasonable accuracy but create dispersion (oscillations) near
the shock breaking the monotonicity of the solution due to their nonlinearity.
The nonlinear scheme that can preserve monotonicity in the solution can be
achieved by two ways: (i) blending two or more schemes and (ii) using flux limiters.
The latter is known as total variation diminishing (TVD) scheme [1]. The former
method is discussed and analysed in the present work to understand the numerical
behaviour of the blended schemes.
The present work tries to analyse the effect of the blending factor on the inter-
face topology of RT instability and the numerical stability of the two new schemes
resulting from the blending. The use of two blends, namely B1 (first-order upwind
(FOU) and LW) and B2 (FOU and QUICK), is used to demonstrate this effect.
R. Jha · A. Arote (B) · J. Banerjee

SV National Institute of Technology, Surat, India
796 R. Jha et al.
Fig. 1 Scalar magnitude plotted along X at the mid of Y
The one-fluid model is used for the present work to perform the two-phase flow
simulations. In doing so, it is presumed that both the fluids hold continuum approx-
imation, and based on its scales, the interface between the two fluids is supposed to
be sharp.
The incompressible Navier–Stokes equations govern the flow field, and in their non-
dimensional form, it can be written as
∂u j
=0
∂x j
∂u i ∂u j u i 1 ∂p μ∗ 1
+ =− ∗ + ∗ ∇ 2ui − (1)
∂t ∂x j ρ ∂ xi ρ Re Fr 2
where i and j represent the non-repeating and repeating indices, while u, p,ρ ∗ , μ∗ ,
Re and Fr denote the non-dimensional velocity vector, pressure, density, viscosity,
Reynolds number and Froude number, respectively.
Advection Stabilization Using Lower-Order Scheme Blending… 797
The scaling used for Eq. (1) is as follows
x ∗ y ∗ t u p
x∗ = y = t =√ u∗ = √ p∗ =
L L L/g gL ρl gL
ρ μ ρl L g
3/2 1/2
ρ∗ = μ∗ = Re =
ρl μl μl
where the terms with superscript * denote the non-dimensional form of the variable.
The terms with subscript l represent the liquid properties.
The set of equations are discretized based on finite volume method and are solved
using semi-explicit method. The temporal and diffusion terms are discretized using
forward Euler and central difference method, respectively. The pressure Poisson
equation is solved using Gauss–Seidel successive over-relaxation solver with residue
achieved up to 1 × 10−6 . The advection terms here are discretized using the blending
of higher-order scheme and lower-order scheme. Thus, the two blended schemes B1
and B2 are used to obtain the scalar face value (φ f ) as follows

B1 : φ f = ψ φ f FOU + (1 − ψ) φ f LW

B2 : φ f = ψ φ f FOU
+ (1 − ψ) φ f QUICK
Here,

φ f FOU = φu
Co
φ f LW = φu + (φu − φuu )
2
3 3 1
φ f QUICK = φu + φd − φuu
4 8 8
where subscripts u, uu and d denote upstream, far upstream and downstream nodes,
respectively. Co represents the cell face courant number. The blending factor ψ lies
between 0 and 1.
2.2 Interface Tracking
The interface between the two fluids is tracked using volume of fluid (VOF) method
[2]. The interface is reconstructed using case-wise piecewise linear interface calcula-
tion (PLIC) method [3, 4]. The solution of VOF Eq. (2) is performed using operator
split method which provides the second-order accuracy in space.
∂ϕ ∂u j ϕ
+ =0 (2)
∂t ∂x j
798 R. Jha et al.
The use of operator split method results in the dilatation term which causes over-
shoots and undershoots in the volume fraction values. These oscillations are treated
using truncations.
The fluid properties in the mixed cells are obtained by interpolation using volume
fraction (ϕ) as follows
ρ ∗ = ϕ + (1 − ϕ)ρl /ρg
μ∗ = ϕ + (1 − ϕ)μl /μg
2.3 Simulation Setup
The domain of dimensions 0.5 × 4 with the boundary conditions used is depicted in
Fig. 2. The grid resolution of 52 × 416 is used after performing the grid convergence
study which resulted in an asymptotic ratio of 0.98 that lies in the convergence region.
The left face of the domain is treated as the symmetry, while the right face is
treated as the free-slip boundary. The top and bottom boundaries are dealt as a wall.
domain and boundary
conditions
The initial interface perturbation is given by the relation y =

−0.01L cos(2π x/L). The density ratio ρl /ρg = 3 while the viscosity of both
fluids is treated to be same. The time-stepping is adaptive with average time step
taken to be 0.001.
The present section deals with the comparative study between blended schemes
B1 and B2. This comparison is based on the simulations performed for Rayleigh–
Taylor instability. The results of the blended schemes are compared with simulations
performed by [5] for Re = 1000. As Re is high, the flow can be termed as advection
dominant and can be said to be good test case for the comparison.
The section also pays attention towards the stability of the higher-order scheme,
i.e. LW and QUICK, when blended with FOU.
The validation of the present model is done using B2 scheme (ψ = 0.6) with
results obtained in [5]. The maximum error within the results is observed to be 4.8%.
Figure 3 shows the flow topologies for the present simulations which agree well with
[5]. Also, the fluid front measured with time as shown in Fig. 4 shows considerable
agreement with the literature.
Fig. 3 Flow topologies of RT instability at t = 0, 1, 1.5, 2, 2.5 with scheme B2 using ψ = 0.6
800 R. Jha et al.
Fig. 4 Front positions for gas and liquid with respect to time compared with [5]
3.1 Blended Scheme B1
As seen in Fig. 1, the pure LW scheme shows erratic dispersion error near the discon-
tinuity. Hence, blending FOU with LW will dampen the oscillations and improve the
accuracy of the solution.
Figure 5 shows the effect of blending factor on the flow topology, where the
solution is seen to get distorted at lower ψ values due to spurious velocities. At
higher ψ values, the solution starts showing dissipation error. The behaviour of
the blending can be understood through Fig. 6 where normalized variable diagram
(NVD) is plotted for the scheme. It shows that as ψ increases the solution becomes
more dissipative as it approaches FOU, while solution becomes more dispersive
when ψ decreases as it approaches LW. It is also seen that to achieve accurate results
by preserving monotonicity the ψ value should be greater than 0.5. But, it should
not go beyond 0.8 as shock-capturing capabilities are lost. Hence, the optimal value
obtained here is 0.6.
Fig. 5 Flow topologies of RT instability using B1 at t = 2.5 with varying blending factor ψ
3.2 Blended Scheme B2
Similar to scheme B1, the scheme B2 from Figs. 7 and 8 shows the same charac-
teristics and optimal value of ψ. It is interesting to observe here that QUICK in
its pure form is third-order accurate scheme, whereas LW is second-order accurate
scheme, but still after blending, LW shows similar results. Thus, scheme B1 that
is obtained from second-order accurate scheme when stabilized by the lower-order
scheme performs equivalent to stabilized third-order accurate scheme.
802 R. Jha et al.
Fig. 6 Normalized variable diagram for scheme B1 with varying blending factor ψ
3.3 Von Neumann Stability Analysis
The one-dimensional advection dominant flow field is decomposed into its Fourier
components g(k)eikm x , and the behaviour of each component is observed when
used after discretization of governing Eq. (1). Here, g(km ) is the amplification factor
which is a function of wave number k m . Here, g(km ) needs to be ≤1 for the scheme
to be stable.
For pure unsteady advection case, the Fourier decomposition for the blended
schemes can be given as
B1 : g n+1 eikm x = g n eikm x + ψCo(g n eikm−1 x − g n eikm x )

Co n ikm−1 x
+ (1 − ψ) (g e − g n eikm+1 x )
2

Co2 n ikm+1 x
+ (g e +g en ikm−1 x
− 2g e )
n ikm x
2
B2 : g n+1 eikm x = g n eikm x + ψCo(g n eikm−1 x − g n eikm x )

7 n ikm−1 x 1 n ikm−2 x 9 n ikm x 3 n ikm+1 x
+ (1 − ψ)Co g e + g e − g e − g e
8 8 8 8
Fig. 7 Flow topologies of RT instability using B2 at t = 2.5 with varying blending factor ψ
After solving the above equations and finding the magnitude of complex vector
g(km ), we get
B1 : |g| = 1 − 2ψCo(1 − cos(km x))(1 − Co)

− (1 − ψ)Co2 1 − sin2 (km x)(Co − 1)2
B2 : |g| = 1 − 2ψCo(1 − cos(km x))(1 − Co)

⎡
2 ⎤
Co2 1 − 0.5 cos(km x) + 0.25 sin2 (km x)
⎢ 2 ⎥
− (1 − ψ)⎣ 5 ⎦
+ sin(km x) + 0.25 sin(km x) cos(km x)
4
804 R. Jha et al.
Fig. 8 Normalized variable diagram for scheme B2 with varying blending factor ψ
Thus, the numerical stability for both schemes has a limiting equation in the form
of Co ≤ 1. It is interesting to view the stability of scheme when used in their pure
form. FOU and LW both need to have Co ≤ 1 to be stable; thus, B1 too has the same
characteristic. On the other hand, QUICK has a limiting case of Co ≤ min(2, 0) for
long wavelength case which was improved to Co ≤ 1 in scheme B2.
4 Conclusions
The present work demonstrates the effect of blending parameter on the flow topology
and the numerical stability of the higher-order scheme. It is observed that the blending
factor needs to be more than 0.5 to have accurate and stable results. Also, the blended
schemes display better numerical stability as compared to the pure higher-order
schemes.
Acknowledgements The authors would like to acknowledge the financial assistance provided by
Sardar Vallabhbhai National Institute of Technology, Surat, towards the development of Advanced
Fluid Dynamics Laboratory where this work has been carried out.
References
1. Harten A (1983) High resolution schemes for hyperbolic conservation laws. J Comput Phys
49(2):357–393
2. Hirt C, Nichols B (1981) Volume of fluid (vof) method for the dynamics of free boundaries. J
Comput Phys 39(1):201–225
3. Rudman M (1997) Volume-tracking methods for interfacial flow calculations. Int J Numer Meth
Fluids 24(7):671–691
4. Saincher S, Banerjee J (2015) A redistribution-based volume-preserving PLIC-VOF technique.
Numerical Heat Transfer, Part B: Fundamentals 67(4):338–362
5. Guermond JL, Quartapelle L (2000) A projection FEM for variable density incompressible
flows. J Comput Phys 165:167–188
Modeling and Simulation
of an Electro-Hydraulic System Using
Fuzzy Logic Approach
A. Tony Thomas, S. K. Thangarasu, and T. Sowmithra
1 Introduction
Hydraulic systems are widely used in industry due to their capabilities of providing
large driving force or torques, higher speed of response with fast motion, and possible
speed reversals and continuous operation. Many industrial applications such as
suspension of electro-hydraulic servo system, fly by wire system of aircraft, sheep
steering gear system, and numerical machine tools use hydraulic system [1]. Electro-
hydraulic system combines together the versatile and precision available from elec-
trical technique of measurement and signal processing with the superior performance
which high pressure hydraulic mechanism can provide when moving heavy loads
and applying large forces. Modern industry is looking for flexible solutions that will
be able to provide some new characteristics of hydraulic systems, such as the ability
of controlled motion, the possibility for continuous control of the required values,
simple data transfer and signal processing, the possibility of monitoring and process
visualization, etc. The rapid developments in microelectronics in recent years have
reduced the cost of computer equipment to a level acceptable for industrial appli-
cations, which has enabled the implementation of sophisticated control strategies in
practice. In hydraulic system, proportional and servo valves are used to control the
velocity or force of an actuator. But in conventional hydraulic systems, there exist
fluid leakages, and subsequently, there occurs nonlinearities in the output. To solve
this issue, a proper controller must be installed in the system for accurate positioning
of the actuators and for obtaining linearity. Fuzzy model is developed to overcome
the nonlinearities existing in the system. A simplified mathematical model is derived,
and to reduce the complexity, leakage and friction between piston and cylinder are
neglected. In this study, fuzzy model is successfully developed for the hydraulic
system with proportional valve, and accurate position of the actuators is obtained for
A. Tony Thomas (B) · S. K. Thangarasu · T. Sowmithra

Department of Mechatronics Engineering, Kongu Engineering College, Erode, Tamil Nadu
638052, India
808 A. Tony Thomas et al.
the corresponding change in input pressure and flow rate. The entire fuzzy system
is validated by comparing its output to the experimental output through correlation
chart. Both the output readings are found to be highly correlated, and this control
model can be certainly used in the system for improving the efficiency.
2 Literature Survey
Accurate models for the electro-hydraulic components have been constructed using
the characterization data provided by manufacturers and measurements obtained in
the laboratory with a custom build electro-hydraulic manipulator was proposed. In
this case, a commonly available general-purpose simulation package, simulink has
been used to solve the nonlinear differential equations [2]. Fuzzy logic was applied
to control a hydraulic servo system for position control, and the effect of internal
leakage on the control performance was investigated [3]. A simulation of the model of
the hydraulic cylinder with vertical load and this model has been introduced within
the MATLAB simulink environment in order to obtain the theoretical simulation
program of the electro-hydraulic system operation. This model has been used to
investigate the performance of the system when works under different operating
conditions [4]. The square—root term in the orifice flow equations was presented.
This term is essential a nonlinear input gain, which varies as a function of chamber
pressure in a cylinder attached to the valve. Non-smooth nonlinearities arise from
geometric imperfections of the valve and its spool [1]. Convenient control strategies
are applied to achieve the accurate and precise position control of the hydraulic
system. The conventional control methods including PD and PID are insufficient
to overcome the nonlinearities in the hydraulic system to achieve the accurate and
precise position control. Fuzzy logic is used to minimize the uncertainties in the
hydraulic actuation system, and it produces notable result. So, combination of both
PID and fuzzy logic results in the hybrid fuzzy-PID controller with coupled rules
(HFPIDCR) can give a better performance.
3 System Description
Hydraulic system consists of cylinders, direction control valve, and flow control
valve. The hydraulic press experimental setup used for developing fuzzy logic model
is shown in Fig. 1, and the description of the components present in the press is shown
in Table 1.
The hydraulic circuit of the system is shown in Fig. 2.
Modeling and Simulation of an Electro-Hydraulic System … 809
Fig. 1 Electro-hydraulic
system
Table 1 Components and its manufacturers list

S.No Description Make Qty
1 Hydraulic cylinder Hydrosmith 1
100 mm × 50 mm × 200 stroke
2 Direction control valve Yuken 1 No
DSG-01-3C2-D24-N1-50
3 Proportional flow control valve Sun hydraulics 1
FPCH-XCN
4 Pressure relief valve Yuken 1
BSG-03-3C3-D24-N1-4680
5 Pressure line filter Bhagwathi 1 No
6 In-line check valve Yuken 1
CIT-03-5-2080
7 Vane pump Yuken 1
PVR1T-8-FRA-28
8 Electric motor Monark/CGL 1
3 hp × 1440 rpm; 50 Hz
Foot mounting
4 Mathematical Modeling of the Components
4.1 Modeling of Valves and Actuators
The input to the system is the movement of the valve x i . This allows the flow of
oil into the cylinder of Q m3 \s which makes the cylinder move a distance x o [5].
Fig. 2 Circuit diagram of hydraulic system
Assuming constant supply pressure, the flow rate is directly proportional to the valve
position,
Q = kv · xi (1)
where k v is the valve constant and examining its unit, it would be m2 /s and area of
the piston is m2 .
The velocity of the actuator is v = dx 0 /dt, and this is related to the flow and the
piston area by the law of continuing that
Q = kv · xi = A · dx0 /dt (2)
Changing to a function of s this becomes,
kv · xi = A(s) · x0 (3)
G(s) = x0 /xi (s) = 1/(A/kv )s (4)
G(s) = 1/T S where T = A/kv (5)
By using the values of piston and cylinder area, it was found that G(s) = 1/0.2945 s.
4.2 Modeling of Direction and Flow Control Valve
A spool valve used for metering purposes controls flow rate by throttling. Each port
in a valve that is partially closed by a land on the spool becomes a control throttle.
The rate of flow of fluid through such a valve depends on the spool displacement
from the null position “x” and on the pressure upstream and downstream of the valve.
One way of representing the flow rate “q” through a valve is
q = k q · x − k c · pm (6)
where
q Flow rate
kq Flow gain
x Spool displacement from null position
kc Pressure flow coefficient
pm Pressure drop
The equation is then transformed to
dq/dt = kq · x(s)−kc d pm /dt (7)
dq/dt = x(s) · [kq −kc · pm /x(s)] (8)
Q 0 /x(S) = kq −kc · pm (s)/x(s) (9)
From the equation,
kq = 6.7π d1 ps1\2
kc = 6.7π d1 x ps1\2 /2 ps
Substituting the values of Pm and x(s) obtained from the observed experimental
values,
For the directional control valve,
Q 0 /x(S) = 8.89135−(7.2086s/8)
For the flow control valve,
Q 0 /x(S) = 8.89135−(7.9690s/8)
The above analysis predicts that the flow gain k q can be treated as constant for a
particular valve and supply pressure but the pressure flow coefficient k c varies with
the valve opening x. In reality, spool lands never exactly match the annular ports
in the valve body. Actual test results with a constant pressure drop across the valve
ports show variations, particularly near the central or null position of the spool.
5 Experimentation
In order to obtain the fuzzy model for obtaining the accurate position, the flow rate
through the valve is measured through experiment. The position of the cylinder is
measured by a Linear Variable Displacement Transducer (LVDT). The flow rate is
measured by a flow transmitter. The stroke length is limited by proximity sensors.
Since the hydraulic system experimental setup is equipped with flow transmitter, the
flow rate is obtained for proper change in input pressure. The relationship between
the input and output is obtained from the modeling of the system.
The maximum range of pressure for which the hydraulic system reacts is from 10
to 30 bar. The experiment is carried out by changing the input pressure to the propor-
tional valve. The corresponding change in the position of the actuator is obtained
through the setup. The values obtained for change in input pressure are given in
following Table 2.
From the table, it is evident that the leakage from the system increases when
pressure increases. Thus, through the experiment for small change in input pressure,
there is no change in output flow rate. In order to obtain the position directly, the fuzzy
model is developed, thus, reducing nonlinearities of the system. From the range of
output measured through the experiments, the measurements needed to develop the
fuzzy model are selected based on maximum change in output that is obtained. For
the selected output flow rate, the corresponding position of the actuator is calculated.
6 Development of Fuzzy Model
The proposed model was developed using the FIS Editor graphical user interface
in the fuzzy logic tool box within the framework of MATLab V7.6. The developed
MISO fuzzy model for proportional hydraulic system for estimating the accurate
position is shown in the below Fig. 3.
Here, pressure at the proportional valve and the corresponding flow rate to the
solenoid for actuating the proportional valve spool are considered as input. The
accurate position of the actuator for the corresponding pressure and current to the
flow controlled valves are considered as output.
The observed experimental data are normalized for improving the performance of
the model. The input and output variables are quantified by using linguistic terms. For
this hydraulic model, input variables include pressure, and flow rate are expressed
into seven fuzzy sets, namely negative very big, negative big, negative small, zero,
positive small, positive big, and positive very big. The generated fuzzy membership
Table 2 Experimental data obtained through the setup

Supply pressure (Bar) Flow during extension Time taken for System pressure drop
stroke (Lpm) extension stroke (Sec) (Bar)
10 8.0 10.15 8.3
11 8.5 9.72 8.9
12 9.1 8.94 9.5
13 9.6 8.78 10.6
14 9.7 8.10 11.9
15 10.2 7.40 12.7
16 10.7 7.25 13.2
17 11.2 6.90 14.1
18 11.7 6.44 14.4
19 11.2 6.91 14.4
20 11.2 6.57 15.8
21 11.6 6.30 13.3
22 11.7 6.13 13.7
23 10.6 6.53 13.8
24 11.3 5.75 13.8
25 11.6 6.12 13.7
26 11.3 6.10 13.7
27 11.3 6.15 13.6
28 11.3 5.97 13.6
29 11.2 5.93 13.5
30 11.7 5.84 13.5
function for the input variable pressure, input flow rate, and output position is given
in Figs. 4, 5 and 6, respectively.
The next stage of the fuzzy logic is to construct the IF-THEN rules to represent
the relationship between input and output variables based on the linguistic terms.
The rules entered in rule editor window are shown in Fig. 7.
The last stage of fuzzy model is defuzzification process. In this model, center
of area method is used for defuzzification. The defuzzification viewer of the fuzzy
model is shown in Fig. 8.
For small change in the range of input pressure to the proportional valve and
flow rate, the fuzzy model predicts an output which is not obtained directly from the
experiment. The experimental and predicted values of position of the actuator are
given in Table 3.
The correlation chart for the output is obtained from the table. The error between
the actual and the predicted value is calculated and given in Table 4.
Fig. 3 Two input and one output fuzzy model
In this design and analysis, fuzzy logic model developed to predict the position of the
actuator of the proportional hydraulic system. Figure 9 shows the deviation between
the obtained values and the fuzzy predicted values for the position of the actuator
correspondingly. Theoretically, the position of the actuator should vary proportion-
ally and linearly with respect to change in input pressure. Results from the developed
fuzzy model show that uncertainties get minimized at higher pressure, and it also
increases proportionally.
The results from Fig. 9 show that developed fuzzy model is sufficient for predicting
the above-mentioned parameter. Figure 10 shows the deviation between predicted and
obtained valves, and it is observed that the uncertainties in the output are minimized
from the fuzzy predicted value.
The accuracy of the proposed model is expressed through various statistics. The
statistical values are indicated that the predicted results of the proposed model are
good correlations with the experimentally obtained data. The correlation chart for
position of the actuator is shown in Fig. 11, respectively. The developed fuzzy model
shows a correlation factor of an R = 0.9952 for position of the actuator which is
Fig. 4 Membership function for input variable pressure
shown in Fig. 11. Thus, the correlation of 0.99 concludes that the developed fuzzy
model is best suited for the proportional hydraulic system.
8 Conclusion
A fuzzy model has been successfully developed for the proportional hydraulic
system. Accurate position of the actuators is obtained with the corresponding change
in input pressure and flow rate. The entire system is validated by comparing the exper-
imental and fuzzy model readings by plotting the values in the correlation chart, and
the error value is determined. It is found to be the novel model and certainly replaces
the existing system. The predicted and experimental values of the fuzzy model are
highly correlated with the correlation factor of R = 0.9952. This reveals that the
fuzzy model is useful to test the operation of the hydraulic system with the use of
limited number of experimental data.
Fig. 5 Membership function for flow rate

Fig. 6 Membership function for position

Fig. 7 Rule editor of the proposed model

Fig. 8 Output viewer of the fuzzy model
Table 3 Experimental and fuzzy predicted values of velocities

Pressure (Bar) Flow rate (Lpm) (Experimental) position (Fuzzy) position
(mm) (mm)
10 7.5 88 89
12 8.5 91 91
14 8.9 102 103
16 10.2 108 109
18 11.2 127 127
20 10.6 122 124
22 10.6 131 131
24 11.3 133 128
26 11.2 132 133
28 11.7 135 135
30 11.7 136 136
Table 4 Error value for position of the actuator

Supply pressure (bar) Experimental displacement Fuzzy predicted displacement Error %
value (mm) value (mm)
10 88 90.6 −0.0295
12 91 91 0
14 102 100 0.0196
16 108 106 0.0185
18 113 111 0.0176
20 122 123 0.0081
22 127 127 0
24 133 133 0
26 132 133 −0.0075
28 135 135 0
30 136 136 0
Fig. 9 Input pressure versus position

Fig. 10 Flow rate versus position
Fig. 11 Correlation chart for position of the actuator
References
1. Margolis DL, Hennings C (2011) Stability of hydraulic motion control systems. J Dyn Syst
Measur Control 119(4):605–613
2. Li ZD, Corke PI, Gurgenci HZ (2008) Modeling and simulation of an electro-hydraulic mining
manipulator. In: Proceedings of the IEEE international conference on robotics and automation,
vol 2, pp 1663–1668
3. Kalyoncu M, Haydim M (2009) Mathematical modeling and fuzzy logic based position control
of an electrohydraulic servosystem with internal leakage. Mechatronics 19(6):847–858
4. Menshawy TM, Moghazy MA, Lotfy AH (2011) Investigation of dynamic performance of an

electro-hydraulic proportional system. In: 13th international conference on aerospace sciences
and aviation technology, vol 18, no 7, 950–956
5. Qiu H, Zhang Q, Reid JF (2001) Fuzzy control of electrohydraulic steering systems for
agricultural vehicles. Trans ASAE 44(6):1397–1402
Experimental Investigation
of Aerodynamic Coefficients of NACA
4412 Aerofoil
Vikas Sharma, Sushanta Dutta, K. Shashi Kumar Naidu,

and M. Ajay Kumar
1 Introduction
Aerofoil plays an important role in the aviation and energy sectors. Their profile and
angle of attack (AOA) determine their performance characteristics. Thus, by selecting
an optimum AOA, desired lift or drag could be obtained. An aerofoil is basically
defined by its cross section and is a 2D shape. In presence of free stream velocity, an
aerofoil experiences aerodynamic drag and lift forces. Thus, by analysing the shape
and AOA of an aerofoil, its performance characteristics could be maximised for
practical applications. Aerofoils are used in aeroplane wings, propellers and turbine
blades, etc. The study of an aerofoil is an important aspect in successful modeling of
micro-aerial vehicles [1], which is an important research field these days. The shape
of a typical asymmetric aerofoil is shown in Fig. 1.
Lift force is the force experienced by the body in a direction normal to the free
stream flow direction. Lift is generated due to unsymmetrical flow behaviour over
a body. Asymmetry in the flow structure above and below the aerofoil results in a
pressure difference which causes lift [3]. The drag force is the force experienced by
the body due to the fluid shear on its surface. Drag force is calculated parallel to
the flow. The AOA plays a vital role in determining the magnitude of drag and lift
force. A straight line joining the leading and trailing edge of an aerofoil is called
as chord line. Its length is chord length (c). The AOA is the angle this chord line
makes with the free stream of the flow. Camber line is a line which divides aerofoil
into two symmetric parts as shown in Fig. 1. The AOA at which lift produced by the
aerofoil is maximum is called as stall angle for that particular aerofoil. After stall
flow separation occurs and lift decreases.
The main objective of this investigation is to analyse the flow over an aerofoil
body and to calculate the performance coefficients at Re = 92,500. In the present
V. Sharma · S. Dutta (B) · K. Shashi Kumar Naidu · M. Ajay Kumar

Mechanical and Industrial Engineering Department, IIT Roorkee, Roorkee, Uttarakhand, India
824 V. Sharma et al.
Fig. 1 Schematic diagram of asymmetric aerofoil [2]
work, wind tunnel experimentation of a NACA 4412 aerofoil is done to find lift and
drag coefficient at different AOA.
2 Methodology
2.1 Aerofoil Nomenclature and Structure
National Advisory Committee for Aeronautics (NACA) develops aerofoil shapes

for aircraft wings. The geometric parameters and complete cross section of any
aerofoil could be generated using the aerofoil series number [4]. Among the various
aerofoils, the asymmetry aerofoil NACA 4412 is analysed as it is widely used in
various aircraft. Four-digit series of NACA aerofoil represents crucial information
regarding the aerofoil shape. 150 mm is the chord length used in this analysis along
with 60 cm span length. For NACA 4412:
• First number shows maximum camber in % of chord length (0.04 × c = 6 mm)
• Second number shows maximum camber’s location along chord line in 1/10 of
cord (0.4 × c = 60 mm)
• Last two digits gives the maximum thickness in % of chord (0.12 × c = 18 mm).
2.2 Fabrication of Aerofoil
Initial CAD profile of NACA-4412 is designed and is 3D printed for performing

experimental analysis. The length of the aerofoil is taken as 60 cm, and its cord
length is 0.15 m. The aerofoil is fabricated in four equal parts with proper design
inputs for easy joining. Poly lactic acid (PLA) material is used for 3D printing.
All the parts fabricated are joined using industrial grade adhesive. The surface is
Experimental Investigation of Aerodynamic Coefficients … 825
Fig. 2 3D printed aerofoil with hollow provision for pressure taps
finished using sandpaper, and spray painting is done to attain a better surface finish.
The aerofoil has 14 openings on its surface, seven on the top and others on the
bottom surface as seen in Fig. 2. Hollow sections are made through cross section
to aid drawing pressure port pipes through aerofoil. Pressure taps or pressure ports
are required to measure the values of pressure above and below the surface of the
aerofoil. First pressure port is 2 cm from the leading edge. The pressure values are
used to evaluate the lift coefficient.
8 mm holes are done on the surface to serve as pressure ports. Silicone pipes
are used for joining the pressure ports to the manometer and pipes came out of the
aerofoil through the hollow section as shown in Fig. 3. It should be done carefully
such that the pipes do not get blocked. Adhesive clay is used to fill the pressure ports,
and further, a small hole is pinched in the clay for pressure measurement.
2.3 Wind Tunnel Experimental Setup
Wind tunnel is used to carry out the experimental analysis of the fabricated aerofoil. In
this analysis, the aerofoil is analysed for a given Reynolds number. The wind tunnel
Fig. 3 Pipes connecting

pressure ports to the
micro-manometer
used is an open circuit flow wind tunnel. The velocity in the tunnel is controlled
using a AC drive unit. It generates flow using a fan powered by a motor which
creates suction in the tunnel. The free stream velocity fluctuation in the tunnel is
very small (0.08%) over the velocity range studied. The air is inlet from contraction
zone, and flow straightener is used to make flow uniform in the test section. The
wind tunnel is shown in Fig. 4. The dimension of the cross section of test section is
60 × 60 cm. The analysis is done at velocity 9 m/s which correspond to Reynolds
number 92,500 using chord length of the aerofoil. The AOA is varied from 0° to 18°
at an interval of 3°.
For performing wind tunnel experiments, Pitot static tube and digital manometer are
used. Pitot static tube is a standard velocity-measuring instrument and it measures the
air velocity with great accuracy. It consists of two hollow tubes to measure stagnation
and static pressure. L shaped Pitot is used having a dimension of 70 and 13 cm in
length, and outer diameter is 8 mm. The velocity (V ) in Pitot tube is calculated using
the following formula where ρ is the density of air and P is the difference between
Fig. 4 Wind tunnel used for aerofoil testing
static and dynamic pressure.

V = 2P/ρ (1)
Digital manometer is connected with the Pitot static tube to measure the velocity in
the tunnel as well as over the aerofoil. Pressure taps from the Pitot tube is connected to
the high-quality differential pressure micro-manometer (Furness Control company,
model FC012) with pre-calibrated velocity output and a digital display. It is a capaci-
tance type differential transducer and could measure pressure up to 1.99 mm of water
with a resolution of 0.01 mm of water.
3.1 Lift Calculations
Asymmetric aerofoils are used for generation of more lift. Due to asymmetric shape,
the air velocities above and below the aerofoil surface are different which affects the
pressure distribution over the aerofoil. Literature reveals that lift force on an aerofoil
increases with an increase in AOA till a critical angle called as Stall angle of the
aerofoil. Beyond this angle, the aerofoil is said to have stalled which decreases the
lift force. The coefficient of pressure (C p ) is calculated using the equation given
below. p is the pressure on surface of the aerofoil, and V is the free stream velocity
of flow [5].
p − pref
Cp = (2)
1
2
ρV 2
The lift force on an airfoil of span b and chord c is calculated in presence of free
stream velocity V. The lift force expressed as a dimensionless lift coefficient (C L )
Fig. 5 Forces acting over an aerofoil [5]
per unit span length.
FL
CL = (3)
1
2
ρV 2 bc
where lift force (FL ) on the aerofoil is calculated using the pressure distribution
obtained by the pressure ports and integrating it over the chord length (Fig. 5).
For small angles, |sin α| |cos α|. Therefore, the lift force becomes:

FL ≈ pgl − pgu cos αb dx (4)
where pg is the gauge pressure. Pressures on the upper and lower surfaces are denoted
by subscripts u and l, respectively. The final equation used for C L in this analysis is
[5]:
x 1 x
1
C L = cos α ∫ C pl d − ∫ C pu d (5)
0 c 0 c
To determine lift coefficient, the integral needs to be solved. So, the values of C pl
and C pu are found using pressure ports and are integrated using MATLAB [6].
From Fig. 6, it could be clearly seen that apart from the leading edge, most part
of the lower surface is on higher pressure as compared to the upper surface which is
the reason for lift generation.
3.2 Drag Calculation
The drag coefficient of the wing is determined by the momentum balance method as
discussed by Schlichting [7]. This method required two velocity profiles, one profile
of free stream and second at the location in the wake where static pressure does
Fig. 6 C p values plot for upper and lower aerofoil surface at 0°
not vary across the direction of flow. Data of the velocities in the wake region are
obtained using the Pitot tube installed in the wake region at a distance 10 × c and is
used for further calculations. Momentum principle states that the drag force over a
body in the fluid flow is equal to the loss or reduction in the linear momentum of the
flow in the free stream direction. When air passes over the aerofoil, its momentum is
reduced due to shear induced be the aerofoil surface and that momentum loss is equal
to drag force. V is the velocity measured at the wake zone in a plane perpendicular
to the air stream. V0 is free stream velocity, and dl is a differential length vertically
across the wake, perpendicular to the span direction of the wing.

2
Cd = V (V0 − V )dl (6)
cV02
For discrete data of experiments this, equation could be approximated by [7]:
Cd = Vi (V0 − Vi )dl (7)

cV02
Thus, the coefficient of drag for the aerofoil by the momentum method is calcu-
lated by inserting the wake velocity and the free stream velocity distributions in the
above equation.
Table 1 Drag and lift

Angle of attack Coefficient of drag Coefficient of lift
coefficient from the
(degrees) CD CL
experiment
0 0.034 0.3209
3 0.0245 0.3291
6 0.0250 0.3130
9 0.0261 0.2976
12 0.0322 0.2790
15 0.0481 0.3604
18 0.0614 0.2472
Fig. 7 Variation of drag and 0.5

lift coefficient with angle of Drag Coefficient
attack Lift Coefficient
0.4
Coefficient Value
0.3
0.2
0.1
0
0 3 6 9 12 15 18
Angle of Attack
3.3 Coefficients for Different AOA
From the results shown in Table 1 and Fig. 7, it could be clearly seen that there
is steep descent in lift coefficient values from 15°. Thus, 15° is considered as stall
angle. Drag coefficient is almost similar for small angle of attack and increases for
higher AOA. As the AOA increases, the lower surface of the aerofoil comes close
to the ground which affects the pressure distribution. As a result, the high pressure
coefficient is there on the lower surface of the entire aerofoil, which eventually results
in higher lift coefficient. For a higher AOA, upper surface causes an adverse pressure
gradient which results in decay of kinetic energy. The decay in kinetic energy leads
to an enlarged wake region which results in increase of drag at a higher AOA.
4 Conclusions
The main objective of this investigation is to analyze the flow over a NACA 4412
aerofoil and to calculate the performance coefficients at Re = 92,500. Fabrication
and experimentation of an asymmetric aerofoil are done for AOA varying from 0°
to 18°. Pressure and velocity distributions are used to calculate the lift and drag
coefficient on the aerofoil. From the results, it could be clearly seen that apart from
the leading edge, most part of the lower surface of an aerofoil has higher pressure. For
a higher AOA, the upper surface causes an adverse pressure gradient which results
in the decay of kinetic energy and increases drag. The stall angle at which lift is
maximized is found to be 15°. The study could be helpful in selecting the optimum
angle of attack based on the required application and finding stall angle.
References
1. Somashekar V (2014) A computational investigation of unsteady aerodynamics of insect-

inspired fixed wing micro aerial vehicle’s 2D airfoil. 2014(504049) (Hindawi Publishing
Corporation)
2. Aerofoil nomenclature. https://en.wikipedia.org/wiki/Airfoil#/media/File:Wing_profile_nom
enclature.svg
3. Llorente E, Gorostidi A, Jacobs M, Timmer A, Munduate X, Pires O (2014) Wind tunnel tests
of wind turbine airfoils at high Reynolds numbers. J Phys Conf Ser 524:
4. NACA Aerofoil profile generator. http://airfoiltools.com/airfoil/naca4digit
5. Tavoularis S Lab manual. Department of Mechanical Engineering, University of Ottawa. http://
by.genie.uottawa.ca/~mcg3341/LabManual.pdf
6. MATLAB Release (2018) The math works Inc., Natick, Massachusetts, United States
7. Schlichting H (1979) Boundary layer theory, 7th edn. McGraw Hill
Solar Energy Simulation of Fresnel Lens
Concentrated System for Thermal
Electric Generator
S. Babu, R. Sriram, S. Gopikrishnan, and A. Praveen
1 Introduction
Fossil fuel is a general term for buried combustible geologic deposits of organic
materials, formed from decayed plants and animals that have been converted to
crude oil, coal, natural gas, or heavy oils by exposure to heat and pressure in the
earth’s crust over hundreds of millions of years. The total primary energy uses of
Inhdia are coal (44%), petroleum and other liquids (23%), biomass and waste (23%),
natural gas (6%), nuclear (1%), hydroelectric (2%) and Other renewable (1%). About
80% of India’s electricity generation is from fossil fuels. The burning of fossil fuels
by humans is the largest source of emissions of carbon dioxide, which is one of the
greenhouse gases that contribute to global warming [1].
Solar Energy: Solar power is usable energy generated from the sun in the form of
electric or thermal energy. Solar energy is captured in a variety of ways, solar thermal
systems, solar photovoltaic (PV) and concentrating solar power (CSP) systems, and
the most common of which is with photovoltaic solar panels that convert the sun’s
rays into usable electricity. Even though solar energy is a renewable source, it is not a
clean source, because one of the major ways of tapping solar energy is by using the PV
panels which is a PN junction diode made up of single-crystal silicon or amorphous
silicon. These materials are generally manufactured by various techniques like the
Czochralski process, float zone process, zone refining or by cutting wafer process;
out of these, Czochralski is the imminent one, and this process of producing solar
PV panels is not eco-friendly, and they arrive with serious environmental issues. So,
there is a need for another efficient or prominent way of tapping this solar energy
which is by using the solar thermoelectric generators [2].
A solar collector is an object that is used to collect energy from the sun, which
it does by absorbing solar radiation and converting it into heat or electricity. The
parabolic trough reflector is one of the majorly used solar thermal energy collectors
S. Babu (B) · R. Sriram · S. Gopikrishnan · A. Praveen

834 S. Babu et al.
designed to capture the sun’s direct solar radiation over a large surface area and focus,
or more generally “concentrate it” onto a small focal point area on the trough [3]. Even
though it offers good concentrating solar energy at the TEG system, the maintenance,
overall weight and the cost of this conventional trough collector is too high. So, we
need a simple and lightweight collector to collect the solar radiation. To rectify this
problem, a Fresnel lens collector is used. A Fresnel lens replaces the curved surface
of a conventional optical lens with a series of concentric grooves. These contours
act as individual refracting surfaces, bending parallel light rays to a common focal
length [4]. As a result, a Fresnel lens, while physically narrow in profile, is capable
of focusing light similar to a conventional optical lens but has several advantages
over its thicker counterpart. One of the most common applications for a Fresnel lens
is the collection of solar light, which is considered very nearly parallel (an infinite-
conjugate system). Using a Fresnel lens for light collection is ideal for concentrating
light onto a photovoltaic cell or to heat a surface [5]. There are two main types of
Fresnel lens: imaging and non-imaging. Imaging Fresnel lenses use segments with
curved cross sections and produce sharp images, while non-imaging lenses have
segments with flat cross sections and do not produce sharp images. As the number of
segments increases, the two types of the lens become more similar to each other [6].
The major application of this Fresnel lens is in imaging, photography, illumination,
solar power, etc., Fresnel lenses can concentrate sunlight onto solar cells with a ratio
of almost 500:1. This allows the active solar-cell surface to be reduced, lowering cost
and allowing the use of more efficient cells that would otherwise be too expensive
[7].
2 Thermoelectric Generators
It is a solid-state device that either converts heat directly into electricity or transform
electrical energy into thermal power for heating or cooling. Such devices are based
on thermoelectric effects involving interactions between the flow of heat and of
electricity through solid bodies. A thermo-electrical generator consists of a heat
source, which is kept at a high temperature and a heat sink, which is maintained at
a temperature less than the heat source.
3 Background of the Work
Initially a Fresnel lens of the specifications (Table 1) was chosen. The lens was
purchased and framed; with the help of concrete tubes, the lens was raised from
ground horizontally at a height of 200 cm (as shown in Fig. 1) with respect to its
focal length.
Optical pyranometer along with a beaker (containing water) was used in order
to know the variation of focal point with respect to time, and Fig. 2 was obtained.
Solar Energy Simulation of Fresnel Lens Concentrated System … 835
Table 1 Specifications of
Focal length 200
Fresnel lens [2]
Length 140
Width 109
Spot size (dia.) 8
Fig. 1 TEG experimental setup with Fresnel lens
Fig. 2 Two Fresnel lens system

836 S. Babu et al.
The figure shows that the focal point moves along a path following a semicircle of
radius 200 cm. Thus by experiment, it was found that solar tracking was essential
for efficient solar energy utilisation. However, says that conventional solar tracking
consumes power and might be economically inefficient [8]. Thus, an efficient Fresnel
lens system had been developed in this paper which is more economical and simple
compared to the previous systems [9].
4 Proposed Fresnel Lens System
(a) Initial conditions: The Fresnel lens specification (Table 1) was kept constant.
Lead Telluride was chosen as the TEG module due to its high melting point
and cost effectiveness.
(b) Boundary Conditions: The latitude of the place where the solar system must
be considered for inclination about north-south direction, i.e. the latitude of
Coimbatore is 11.01° N, so the lens must be tilted about 15 + 11.01 = 26.01°
during summer and 11.01 − 15 = −3.99° during winter [10].
The Target material was considered to be placed in line with the optic centre of
the Fresnel lens.
Model 1: In order to do away with the traditional power consuming solar tracking
setup, initially two lens system was considered as shown in Fig. 2.
The above model was designed with consideration that the concentrated spot
would lie always in a specified area (common area of the two semicircles) throughout
the solar day. The design was imported into Trace pro software, and the inclination
was varied in order to obtain efficient concentration of solar power in the common
area throughout the day. The lens was considered to be inclined at 15° to the horizontal
as shown in Fig. 2. Model 2: Another type of setup such as the three-lens setup was
proposed as shown in Fig. 3.
The model was analysed in Trace pro software, and the optic centre distance
between the left and right was changed with respect to change in inclination of
lenses, and the amount of radiation incident on the target with respect to radiation
incident on the lens was noted, and the inclination and distance between the lenses
at optimal point were recorded. 20 target materials each of 4 * 4 cm2 was arranged
in a line about 200 cm from each lens. Thus, the final proposed model is shown in
Fig. 3.
The optic centre distance was set finally as 300 cm. The left and right Fresnel lens
were inclined at about 10° towards right and left, respectively.
Fig. 3 Final design with 25

TEG modules
Model 1: Considering the inclination angle and the optic centre distances between
the lenses, simulation results stated that the system concentrates effectively during
before and after noon, however, the system does not concentrate during the noon
(hours of peak intensity) due to the significant gap between the lens. Hence, the
system has proved to be inefficient during the significant noon time, thereby this
model was rejected.
Model 2: The amount of radiation incident at target material with respect to
radiation incident at Fresnel lens during AM period was found out by setting light
Fig. 4 Trace of light beam during AM

838 S. Babu et al.
Fig. 5 Trace of light beam during PM
source in respective manner (Fig. 4). Similar method was used during noon and PM
(Fig. 5).
The following calculation was done in order to obtain payback period of the
proposed design similar to the calculation done in the work [8]:
Radiation input to Fresnel lens = 280 W Net radiation

incident on target = 246.829 W
Efficiency of Fresnel lens = 88.15%
Std. Solar irradiance at Coimbatore = 5.86 kWh/m2 /day
Irradiance input at inclined lens = 9.0244 kWh/day

Heat energy I/p to target material due to inclined lens
= (0.881)(9.024) = 7.95 kWh/day
Efficiency of central lens during AM/PM = 44% Heat
power input due to central lens : 0.44 ∗ 5.86 = 2.5784 kWh/day
Heat input at target:
7.95 + 2.5784 = 10.5284 kWh/day Energy generated

during the period : 0.1 ∗ 10.5284
= 1.05284 kWh/day(11)Revenue per day
= 8 ∗ 1.05284 = Rs.8.42272
Total cost of the setup was Rs. 22,000
Payback period = 22,000/8.422 = 7.15 years
The payback period in the proposed system is considerably less than that of
conventional solar system. Hence, the proposed three lens concentrated system can
be considered as a candidate for an efficient solar thermal system.
6 Conclusion
The critical parameters of a Fresnel concentrated system were identified, and suit-
able decisions and values were chosen for them. An efficient tracking concentrated
system was designed using Trace pro software, and necessary calculations were
done to obtain the payback period which was found to be less than the conventional
concentrated system stating that the concentrated system designed is efficient.
References
1. Kibria A et al Fossil fuel share in the energy mix and economic growth
2. Kasaeian A et al (2018) A review on parabolic trough/fresnel based photovoltaic thermal
systems. Renew Sustain Energy Rev 91:193–204 (Department of Renewable Energies)
3. Xu Y, Wang L, Zhang X et al (2008) Experimental study on solar cell arrays of trough
concentrating solar system, Pantoa Ti Hsueh Pao. Chin J Semicond 29:2421–2426
4. Kumar V, Shrivastava RL, Untawale SP (2015) Fresnel lens: a promising alternative of reflectors
in concentrated solar power. Renew Sustain Energy Rev 44:376–390
5. Zheng H, Feng C, Su Y et al (2014) Design and experimental analysis of a cylindrical compound
Fresnel solar concentrator. Sol Energy 107(9):26–37
6. Li P, Cai L, Zhai P et al (2010) Design of concentration solar thermoelectric generator. J
Electron Mater 39:1522–1530. https://doi.org/10.1007/s11664-010-1279-0
7. Khan F, Baek SH, Kim JH (2014) Intensity dependency of photovoltaic cell parameters under
high illumination conditions: an analysis. Appl Energy 133(6):356–362
8. Zhai H, Dai Y, Wu J (2007) Investigation of concentrating solar photovoltaic/thermal system
performance based on Fresnel lens. J Eng Thermophys 28(5):725–728
9. Seme S, Srpčič G, Kavšek D et al (2017) Dual-axis photovoltaic tracking system. Design and
experimental investigation. Energy 139
10. MaMiller DC, Kurtz SR (2011) Durability of Fresnel lenses: a review specific to the
concentrating photovoltaic application. Sol Energy Mater Sol Cells 95
Flow Visualization of Two-Dimensional
Bodies Using Hele-Shaw Apparatus
Mayank Sharma, Shikhar Jaiswal, Sumit Singh Rajput, Shreya Tripathi,

Subiksha C, and G. Vinayagamurthy
1 Introduction
In fluid dynamics, visualization plays a vital role in understanding the flow physics
in and around any structures or models. The motion of the fluid around the body can
be studied by various methods such as smoke visualization in wind tunnels, water
channels and computational fluid dynamics (CFD), and the recent discoveries have
extended to the particle image velocimetry (PIV) and DPIV. The PIV and DPIV
are expensive and require high skill for operating, whereas Hele-Shaw apparatus is
economical for academic illustrations.
In this study, the models having application in three different domains have been
used such as aerofoils for streamline bodies, trucks in automotive sectors and build-
ings in architecture. The streamline bodies are very essential as a lifting surface in
aerodynamics point of view, and thus, an aerofoil is considered. Heavy goods vehi-
cles (HGVs) play an important role in daily domestic goods transportation within
India and the whole world. A lot of fuel is used for removing the effect of the viscous
and inertial forces resulting in drag. Body designs were tried for reduction in drag
without rising lift so as to reach an optimal condition. Altaf et al. concluded that as
much as 65% of fuel is consumed to overcome the aerodynamic drag encountered
by buses and HGVs in long-haul journeys [3]. Similarly, Bradley indicated that the
aerodynamic drag contributes approximately 21% of energy loss when a 36 tons
heavy goods vehicle is traveling at 105 km/h [4]. Hsu and Davis deduced that an
M. Sharma · S. Jaiswal · S. S. Rajput · S. Tripathi · S. C · G. Vinayagamurthy (B)

School of Mechanical Engineering and Center for Innovation and Product Development, VIT,
Chennai, India
e-mail: vinayagamurthy.g@vit.ac.in
S. S. Rajput
e-mail: sumitsingh.rajput2018@vitstudent.ac.in
S. C
e-mail: subiksha.c2018@vitstudent.ac.in
842 M. Sharma et al.
annual fuel cost saving of US $10,000 could be achieved if the aerodynamic drag
acting on a heavy vehicle is reduced by 40% [5]. Similarly, Bradley also anticipated
that a 20% aerodynamic drag reduction on a heavy goods vehicle could lead to 4%
of fuel saving during high-speed operations. The flow over large span roof structure
is receiving more concern nowadays by civil, structural and wind engineers due to
its aesthetic look and maximum space occupancy. But advanced and more complex
curved roof structural design results in certain aerodynamic complexities such as
vortices and wake regions. Similar work was established numerically for building
structures [6, 7] which gives fair idea about the cause at ideal or theoretical level but
lacks to depict practical applications. Importance of flow visualization was discussed
in earlier studies [1, 2]. There is a scarcity of published work on flow visualization
using Hele-Shaw apparatus used in this work, even though studies on air foils were
done in similar conditions and experimental setup earlier [8, 9]. Similar visualization
around domed shape structures and tall buildings using smoke wire technique and
CFD has been published [10, 11]. Flow behavior and parameters such as stagnation,
separation and wake formation will vary based on the geometry. In this study, the
considered models are tested for various angle of attack (or wind incidence angles)
using the Hele-Shaw apparatus. These results were compared with various other
theoretical studies and show similarities in flow features.
2 Flow Visualization Using Hele-Shaw
Hele-Shaw experimental setup helps in visualization of flow around models by gener-

ating streamlines that depict the boundary layer, stagnation, separation, wake and
vortices. A tank is used to supply dye for depiction of streamlines around the model
mounted in the testing area of the apparatus as shown in Fig. 1. The Reynolds number
is varied by varying the flow rate. The influence of the flow on the model is captured
using a camera kept vertically down. The images and videos obtained are processed
to study the flow parameters. The following explains the flow features over various
models.
2.1 Flow Visualization Around an Aerofoil
NACA 2414 was considered for the flow visualization. Flow of air was depicted as
water using Hele-Shaw apparatus at a Reynolds number of 4555 which is almost
same for the airfoil at turbulent wind conditions at three different angle of attacks
such as 0°, +15° and −15° and is as shown in Fig. 2.
Flow Visualization of Two-Dimensional Bodies Using … 843
Fig. 1 Hele-Shaw apparatus setup used for flow visualization
2.2 Flow Visualization Around a Truck
The models for side views of trucks with, or without deflector and top view
were scaled down to 1:100 compared to actual truck size, and laser cut out of
acrylic sheets. The side view models created were of length 0.076 m and height
of 0.025 m, and thickness is 0.003 m. The top view model was of length 0.08 m and
height 0.025 m and thickness of 0.003 m. Compared to actual truck of side view (7.6
× 2.5) m opertating in air, the downsized models were visualized using water flowing
over them. The flow was visualized for side view under different water incident angle
depicting uphill and downhill and top view depicting different yaw angles such as
0°, +10° and −10° as shown in Fig. 3.
2.3 Flow Visualization Around a Building Structure
The building structure was scaled down to 1:1600 dimensions and 3D printed using
a plastic material. The original structure is of height 22.34 m, the span of the roof is
being 98.15 m, and the mid-section and side view of the building structure were taken
for experimentation. The building structures created have a thickness of 0.003 m, side
view is being 0.028 m in height and 0.093 m in length, and mid-section has 0.032 m
height and 0.093 m length. The flow was visualized for different flow parameters as
shown in Fig. 4.
Fig. 2 Flow visualization over an aerofoil. a α = −15°, separation distance x/L = 0.818, b α =
0°, separation distance x/L = 0.467, c α = 15°, separation distance x/L = 0.953
From the images, the following conclusions can be made

The stagnation point is found on the leading edge, two separation points are seen
on the upper boundary of the aerofoil, and their positions are also determined. Due
(c) α=15°, Separation distance x/L=0.953
Fig. 2 (continued)
to camber, there is a shift in the lower separation point. Length of the wake vortices
is highly influenced by the angle of attack.
For the truck, the separation point is seen on the top of the cabin. After the
separation from the cabin, the flow moves along the boundary of the trailer and gets
separated at the end. Vortex formation can be seen at the back of the trailer before
the wake region and in the gap between the cabin and the trailer.
When the vehicle moves downhill, the separation point shifts forward along the
boundary of the cabin. The stagnation point shifts downwards in the previous case.
After the separation, the flow reattaches to the trailer at the topmost edge and flows
along the boundary of the trailer. Vortex is formed at the back of the trailer and in
between the cabin and the trailer, although in this case the vortex formation is less
than in the case where the truck is linear.
When the vehicle moves uphill, the stagnation point shifts upwards along the
boundary as compared to the case when the truck is linear. The separation occurs on
the front side of the truck, and then, the flow reattaches to the topmost edge of the
(a) Truck α=10o uphill, Separation distance x/L=

0.862, re-attachment distance x/L=0.672
(b) Truck α=0o , Separation distance x/L= 0.862, re-

attachment distance x/L=0.672
Fig. 3 Flow visualization over a truck model (different angles and views), a truck α = 10° uphill,
separation distance x/L = 0.862, re-attachment distance x/L = 0.672, b truck α = 0°, separation
distance x/L = 0.862, re-attachment distance x/L = 0.672, c truck α = −10° downhill, separation
distance x/L = 0.841, re-attachment distance x/L = 0.652, d truck with deflector α = 0°, separation
distance x/L = 0.757, e truck with deflector α = −10° downhill, separation distance x/L = 0.75,
f truck with deflector α =+10° uphill, separation distance x/L = 0.758, g top view α = −10° right
turn, separation distance x/L = 0.909, h top view α = 0°, separation distance x/L = 0.92, i top view
α = 10° left turn, separation distance x/L = 1, re-attachment distance x/L = 0.213
(c) Truck α=-10o downhill, Separation distance x/L=

0.841, re-attachment distance x/L=0.652
(d) Truck with deflector α=0o, Separation distance

x/L= 0.757
Fig. 3 (continued)
(e) Truck with deflector α=-10o downhill, Separation

distance x/L= 0.75
(f) Truck with deflector α=+10o uphill, Separation

distance x/L= 0.758
Fig. 3 (continued)
(g) Top view α=-10o right turn, Separation distance

x/L= 0.909
(h) Top view α=0o, Separation distance x/L= 0.92

Fig. 3 (continued)
trailer which then flows along the boundary till the end of the trailer. Wake region
forms at the back of the trailer but has comparatively lesser area than the other two
cases. Vortex formation is only at the back of the trailer, and no vortex formation is
observed between the cabin and trailer.
(i) Top view α=10o left turn, Separation distance

x/L= 1, Re-attachment distance x/L=0.213
Fig. 3 (continued)
Fig. 4 Flow visualization over a building structure
In the case of truck with deflector, the stagnation point is on the front of the cabin.
Due to the deflector, there is no separation of the flow, and it flows smoothly along
the boundary without getting stagnated at the top edge of the trailer. Vortex formation
can be seen at the back of the trailer and very less in between the cabin and the trailer
which is due to the addition of the deflector.
When the truck with deflector moves uphill, the stagnation is at the front of the
cabin. Separation occurs at the topmost point of the deflector, and the flow does not
reattach with the trailer. Vortex formation is at the back and very minimal in between
the cabin and the trailer.
When the truck with deflector moves downhill, the stagnation is at the front of
the cabin. Separation occurs at the topmost point of the deflector. After this, the flow
gets stagnated at the top edge of the trailer and then flows along the boundary of the
trailer. Vortex forms at the back and in between cabin and trailer due to the stagnation
on the edge on the trailer.
For the case of top view of truck, the stagnation point is at the front of the cabin.
Separation can be seen on both the sides of the cabin after which the flow flows along
the boundary of the trailer. Vortex forms at the back of the trailer and also in between
the cabin and trailer.
For all the cases discussed, wake region is formed at the back of the trailer.
When the vehicle moves uphill, the stagnation point shifts downwards along the
boundary of the upper edge of cabin. Separation occurs at the top corner of the cabin
or from the left corner. Re-attachment can be seen at left side of the trailer end. Wake
region can be seen from the left corner of the cabin itself. Vortex is formed at the
right side of the trailer end and also in between the cabin and the trailer. It can be
seen that in this case the vortex formation is more than in the previous case.
When the truck moves downhill, stagnation point shifts upwards along the
boundary of the upper edge of cabin. Separation occurs at right side of the cabin
after which the flow does not reattach to the trailer. Wake area formed is compara-
tively larger than the other two cases. No vortex formation observed at the back of
the trailer, while it is more in between the cabin and trailer with respect to the other
two cases. For building structure, the stagnation is at the beginning of the structure.
Separation occurs somewhere in between the maximum elevated parts of the struc-
ture. Vortex formation can be seen at the lowest elevation of the structure or at the
mid-section of the structure. The flow reattaches at the other half of the structure and
flows along the boundary. Wake region forms at the back.
4 Conclusions
A significant amount of work has been done on flow visualization over building struc-
ture using Hele-Shaw apparatus, and it can be concluded that Hele-Shaw apparatus is
a favorable tool for flow visualization over 2D structures. Hele-Shaw apparatus can
give fair predictions about relative pressure zones around the structure. With accu-
rate scaling and dimensioning of structures, we can get a close relationship of flow
parameters between experimental and actual structures. Utilizing Hele-Shaw appa-
ratus for flow study around building structures gives rise to different geometrical
enhancements for engineers at a very low cost relative to other techniques. Being an
economical instrument, it does not require very high skill set for the user.
Acknowledgements We would like to acknowledge Dr. G. Vinayagamurthy without supervision

of whose these works would not have been possible. Also, we would like to praise our institute
Vellore Institute of Technology (VIT) Chennai campus for letting us use their labs and assets for
completion of this work.
References
1. Ahmed H, Mohamed A, Ahmed E (Sept 2015) Flow visualization using smoke wire technique
and laser beam. In: 8th Ankara International Aerospace Conference, METU, Ankara Turkey,
pp 10–12
2. Carbonaro M (1994) Flow visualization, measurement techniques in fluid dynamics an
introduction
3. Altaf A, Omar AA, Asrar W (2014) Passive drag reduction of square back road vehicles. J
Wind Eng Ind Aerodyn 134:30–43
4. Bradley R (2000) Technology road map for the 21 st century truck program. Technical Report
21 CT-001 United States Department of Energy, Washington DC, United States
5. Hsu F-H, Davis RL (2010) Drag reduction of tractor-trailers using optimized add-on devices.
J Fluids Eng 132(8):084504 (6 pages)
6. Rajasekarababu KB, Vinayagamurthy G (2019) Experimental and computational simulation
of an open terrain wind flow around a setback building using hybrid turbulence models. J Appl
Fluid Mech 12(1):145–154
7. Rajasekarababu KB, Vinayagamurthy G, Selvi Rajan S (2019) Experimental and computational
investigation of outdoor wind flow around a setback building. Build Simul 12:891–904
8. Sivadas V, Rajagopalan R, Rajesh S, Suddapally S, Muralidharan S, Venkatesan G (2015) Flow
visualization around airfoils. 10 th Pacific, 2015
9. Shah H, Kitaba J, Mathew S, Lim C (2014) Experimental flow visualization over a two-
dimensional airfoil at low reynolds number. https://doi.org/10.1049/cp.2014.1069
10. Yaghoubi MA (1991) Air flow patterns around domed roof buildings. Renewable Energy 1(3–
4):345–350
11. Rajasekarababu KB, Vinayagamurthy G (2020) CFD validation of wind pressure distribution on
a tall building under the influence of upstream terrain. Prog Computat Fluid Dyn Int J 20(5):284–
298
Experimental Investigations on Heat
Transfer Enhancement in Double Pipe
Heat Exchanger Using PT-SCA
and PTT-SCA Twisted Insert Profile
A. P. Sivasubramaniam, K. Mayilsamy, and P. Murugesan
1 Introduction
All processing industries that require heat energy use heat exchanging devices. Over
the years it was found that there is lot of scope for energy saving and size reduction
to make them compact through research works. As these heat exchanging devices
involve convective transfer, enlarging the heat transferring surface has become one of
the main criteria. This can be noticed from the continuous rise in literature available
world-wide on heat transfer enhancing devices, a growing number of patents and
also hundreds of companies are marketing thermal products ranging from enhanced
tubes to complete set of thermal systems with integrated improved technology. The
considerations for the saving of energy, materials, and space in the design along with
cost incentives have led to producing heat exchanger equipment with more efficiency.
2 Literature Review
Dittus and Boelter [1] experimented using nano-fluids in heat exchanger and found
that it improves the heat transfer rate and also reduces fuel consumption and weight.
To further enhance the heat transfer process in the tube a twisted ribbon inclusion
A. P. Sivasubramaniam (B)
Department of Mechanical Engineering, Paavai Engineering College, Namakkal, Tamil Nadu,
India
K. Mayilsamy
Department of Mechanical Engineering, PSG College of Technology, Coimbatore, Tamil Nadu,
India
P. Murugesan
Department of Mechanical Engineering, K.S.R Institute for Engineering and Technology,
Namakkal, Tamil Nadu, India
854 A. P. Sivasubramaniam et al.
was used JFE/1991 [2]. The editorial states that uncertainty in the measurements of
fluid pressure, temperature, etc. may lead to error in the final outcome/efficiency of
experimental methods ANSI/AMSE [3] provides procedural standards to conduct
experiments in thermal devices. Moffat [4] showed that uncertainty could be calcu-
lated using either the single-sample or multiple-sample methodology, which are
conventional methods generally followed to calculate error. Manglik and Bergles
[5] studied correlation between friction factor and Nusselt number for laminar flow
using twisted tape as inserts for heat exchange between water and ethylene glycol. A
swirl parameter was estimated to obtain the inter-relation between centrifugal force,
convective inertia force, and viscous force during operation.
Cengel [6] has stated methods for heat transfer calculations and input variables that
could help improve heat transfer in various modes. Eiamsa-Ard et al. [7] examined
the influence of twisted tapes of different geometries (variation in twist ratio and
free space ratio) on heat transfer and friction characteristics of a DPHE. It was noted
that increase in twist ratio increased the heat transfer coefficient whereas increase in
space ratio increased both friction factor and heat transfer coefficient.
Eiamsa-Ard et al. [8] investigated the impact of twin-counter twisted tapes and
twin co-twisted tapes on heat transfer characteristics of a heat exchanger. The twin-
counter twist tapes produced higher heat energy transfer compared to twin co-twist
which also provided considerable improvement.
Murugesan et al. [9] conducted experiments by inserting twisted tapes with square
cuts in the fluid carrying tubes of a DPHE. The heat transfer enhancement was
considerably high compared to the friction factor which increased to a moderate
level.
Murugesan et al. [10] used twisted tapes and twisted tape containing wire nails as
insertion into DPHE to conduct experiments and compare the heat transfer enhance-
ment. The twist ratios employed were: 2, 4.4, and 6. The insert of the twisted tape
containing wire nails performed better (higher Nusselt number, friction factor, and
thermal enhancement factor) due to swirling effect produced by twists and the extra
turbulence created by the nails.
3 Objectives
The literature study reveals that experimentations have been carried out using a
variety of inserts which include plain tape, twisted tape, and making small cuts
of various profiles (square cuts, step-cut arc, etc.) to be fitted longitudinally in the
plain tube of the heat exchanger for thermal enhancement. Though experimental
studies are large in number, numerical works are very few on topics related to heat
exchanger performance analysis. Hence, the objective of the present work is to carry
out numerical investigations on heat energy transfer improvement in turbulent zones
of the tube inserted with PT-SCA of 165 mm pitch and PTT-SCA of twist ratio y
= 7 for which trial test runs are conducted in an experimental setup. Heat transfer
Experimental Investigations on Heat Transfer Enhancement … 855
as well as friction factor characteristics of the heat exchanger after insertion of the
above inserts would be investigated separately for comparison.
4 Methodology and Experimental Setup
4.1 Plain and Twisted Tape Inserts
The inserts (PT-SCA and PTT-SCA) are fitted into the copper tube to enhance the
temperature in the heat exchanger. The plain tape inserts have plain tape without
any. The inserts (PT-SCA and PTT-SCA) are fitted into the copper tube to enhance
the temperature in the heat exchanger. The plain tape inserts have plain tape without
any geometry on the tape and plain tape step-cut arc has a steeped cut arc profile on
the plain tape with a pitch of 165 mm and a length of 2000 mm. The tape inserted
into the plain tube disrupts the fluid flow creating turbulence force that influences
the heat transfer rate to increase.
4.2 Experimental Setup
The experimental setup consists of two tubes (Copper and Galvanized Iron), rotame-
ters, temperature indicators, controllers, flow control valve, voltmeter, ammeter, ON
and OFF switches, digital pressure gauge, storage tank, etc. The schematic illus-
tration of the experimental setup is given in Fig. 1. The outer side of steel tube is
insulated by glass wool to avoid heat loss to the atmosphere. Heater in the molded
tank-1 is used to heat the fluid (water) and the temperature of the inlet hot water is
maintained at 53 ± 1 °C through the temperature controller. The hot water is pumped
up to the annulus between dual pipes from the tank to the outer pipe. The hot water
is circulated from tank-1 and the cold water circulated from tank-2. Both hot and
cold water outlet pipes are coupled to the individual rotameter of 20 LPM range. The
temperature of cold fluid is maintained at the range of 30 ± 1 °C through ice cubes.
These ice cubes are stored separately in tank-3. If the cold fluid temperature
increases, ice cubes are exchanged into the cold water tank (tank-2). Thermocouple
is utilized for the measurement of the temperatures of both the tanks (T7, T8), inlet
(T3, T4, T5), and outlet (T1, T2, T6) temperatures of hot and cold water. Pressure
drop occurring in the tubes is measured by using a digital manometer which is
coupled with copper tube outlet. Thus all the required readings could be observed
simultaneously for a given flow rate under turbulent flow using the experimental
setup.
Fig. 1 Schematic diagram for counter-flow double tube heat exchanger
5 Result and Discussion
The performance characteristics such as heat transfer, Re (Reynolds number), friction

factor, and thermal enhancement factor of the double pipe heat exchanger fitted with
plain tape step-cut arc and plain twisted tape step-cut arc are discussed in this section.
The results of the experiments conducted using plain and twisted tapes with inserts
(PT-SCA and PTT-SCA) in the range of Re 2000 ≤ 14,000 are conducted as per
the procedure followed by and ANSI/ASME [2] and Moffat [4] standard. Various
suspicions associated with the experimental investigations are applied in the error
analysis.
5.1 Plain Tube Data Verification
Equations. 1, 2, and 3 show the experimental data for turbulent flow which are found
to be similar to forced convection correlations for plain tube of Dittus–Boelter (1930)
equation and Gnielinski (1976) equation (Cengel 2008). The plain tube correlation
with a discrepancy of 20, 20 and 6% for Nusselt number is shown in Fig. 2.
Fig. 2 Data verification of

Nusselt number for plain
tube under turbulent flow
0.3
Nu = 0.023Re0.8 Pr (1)
f
8 (Re
− 1000)Pr
Nu = 0.5 2 (2)
1 + 12.7 8f Pr3 − 1
0.33
Nu = 0.0126Re0.879 Pr (3)
The variation of friction factor obtained for the plain tube calculated from exper-
imental results is shown in Fig. 3. The data obtain attained are compared with that
of Blasius Equation, Petukhov (1970) Equation (Cengel 2008), and the plain tube
correlations are shown in Eqs. 4, 5, and 6. It yields the deviation of 16.25, 13.5 and
18%, respectively, for friction factor.
f = 0.316Re−0.25 (4)
f = (0.790 ln Re − 1.64)−2 (5)
Fig. 3 Data verification of

friction factor for plain tube
under turbulent flow
f = 0.999Re−0.356 (6)
5.2 Heat Transfer Effect of PT-SCA
A comparison of the data obtained from the investigational results of the tube fitted
with plain tape step-cut arc with the plain tube is made and the results are confirmed
using the correlations of available data [inlet test section turbulent flow experimental
data]. The experiment was carried out with a small cut on the plain tape which
increases the turbulence in hot water.
From the results, it was revealed that plain tape step-cut arc contributes to create
additional disturbance in the tube. The Nusselt number for the plain tape step-cut arc
was found to be 1.288 times greater than that of plain tube.
The correlations between the Nusselt number (Eq. 7) and friction factors (Eq. 8)
are also developed for the tube fitted with plain tape step-cut arc from the experimental
outcomes.
0.33
NuSCA = 0.068Re0.721 Pr (7)
f SCA = 0.999Re−0.2516 (8)
The predicted values are in agreement with the experimentally obtained data for
Nusselt number and friction factor within variations of 6% and 5%, respectively.
5.3 Effect of Friction Factor on PT-SCA
From observations it is understood that friction factor for tube fitted with plain tape
step-cut arc is greater than plain tube but there is decrease in Reynolds number
for a specified profile-cut. However, the friction factor rises due to the increasing
disturbance offered by the insert for a given Reynolds number. It is also seen that
the friction factor for plain tape step-cut arc is about 2.537 times greater than that of
plain tube for a given specified profile-cut.
5.4 Comparative Evaluation of PT-SCA with PTT-SCA
The experimental results of plain tape step-cut arc (165 mm of pitch) are evaluated in
comparison with the plain twisted tape step-cut arc (twist ratio y = 7). Plain twisted
Fig. 4 Evaluation of Nu
versus Re for PTT-SCA with
PT-SCA and plain tube
under turbulent flow
tape step-cut arc raises intensity of turbulence on the wall of the inner tube, and hence
it shows comparatively higher heat transfer in the test section. Figure 4 provides the
plot of Nusselt number versus Reynolds number for all the test cases. The plot reveals
that PTT-SCA’s results are 1.623 and 2.092 times better than plain tube and PT-SCA.
The friction factor versus Reynolds number plot for Plain tube, PT-SCA and
PTT-SCA, is presented in Fig. 5.
It has been noted that the friction factor decreases constantly with increase in
Reynolds number and the friction factor for PT-SCA. This is significantly larger than
that of PTT-SCA due to stronger disturbance in the tube. From the range of study of
this investigation the following findings could be made: The mean friction factor for
the PTT-SCA is 1.171 times higher than that of the plain tube and 2.971 times higher
than that of the PT-SCA. But PTT-SCA is found to have a higher pressure drop than
the plain tube as well as the PT-SCA.
Fig. 5 Evaluation of friction

factor versus Re for
PTT-SCA with PT-SCA and
plain tube under turbulent
flow
6 Conclusions
Experimental trials are carried out using DPHE setup with PT-SCA and PTT-SCA
inserts fitted in the flow tube to compare the heat transfer performance parameters
with that of plain tube. The study yielded following main results:
The Nusselt number and friction factor of the PTT-SCA insert are 1.623 and 1.171
times higher than the PT-SCA insert for entire turbulent flow.
Nusselt number increased and friction factor decreased when the PTT-SCA was
introduced. It generates high swirl velocity and increases forced convective currents
which in turn considerably enhances the heat transfer coefficient.
References
1. Dittus FW, Boelter LMK (1930) Heat transfer in automobile radiators of tubular type, vol 2.
University of California at Berkeley, Publications in Engineering, pp 443–461
2. J Fluids Eng Sept-1991 Editorial on Experimental Uncertainty (internet link:https://journaltool.
asme.org/Templates/JFEExperUncertainty.pdf)
3. ANSI/AMSE (1986) Measurements uncertainty, PTC-19, 1-1985
4. Moffat RJ (1988) Describing the uncertainties in experimental results. Exp Therm Fluid Sci
1:3–17
5. Manglik RK, Bergles AE (1993) Heat transfer and pressure drop correlations for twisted-tape
inserts in isothermal tubes: part I: laminar flows. Trans ASME J Heat Transf 115:881–889
6. Cengel YA (2002) Heat and mass transfer, 5th edn. McGraw-Hill (chapter 8)
7. Eiamsa-Ard S, Thianpong C, Promvonge P (2006) Experimental investigation of heat transfer
and flow friction in a circular tube fitted with regularly spaced twisted tape elements. Int Comm
Heat Mass Transf 33:1225–1233
8. Eiamsa-Ard S, Thianpong C, Eiamsa-Ard P (2010) Turbulent heat transfer enhancement by
counter/co-swirling flow in a tube fitted with twin twisted tapes. Exp Therm Fluid Sci 34:53–62
9. Murugesan P, Mayilsamy K, Suresh S (2010) Turbulent heat transfer and pressure drop in the
tube in tube fitted with square-cut twisted tape. Fluid Flow Transp Phenom 14:609–617
10. Murugesan P, Mayilsamy K, Suresh S (2010) Heat transfer and friction factor studies in a
circular tube fitted with twisted tape consisting of wire-nails. Chin J Chem Eng 18(6):1038–
1042
Characteristics of Upstream Flow
in Vertical Inline Pump
P. Srikar Sai Tej and Stephen Christopher
1 Introduction
Flow in a pump is complex, which cannot be obtained easily without special instru-
mentation. In recent scenario, particle image velocimetry (PIV) is extensively used
in turbomachines application as a part of research. Apart from PIV, computational
fluid dynamics (CFD) techniques are well established in almost all sectors of appli-
cations to predict the flow nature (internal or external flows). Around 30% of pump
is converted as closed coupled vertical inline pump, because of space constraints.
Due to this advantage, it is very widely used in all commercial complex, apartments,
multi-storeyed buildings and so on.
In general the nature of upstream flow for the pump plays a greater role. Many
literatures [1–3] reported that losses in suction pipe of pump should be less. Short
bends should be avoided or else to be replaced with long radius bend. Larger suction
diameter is required to avoid high velocity there so pressure reduction could be
protected. Tapered suction pipe is meant to carry over, the entrapped air bubbles at
the upstream of pump. It was suggested for uniform velocity at the inlet of impeller for
better performance [4, 5]. Mikhailov and Kuzmenko [6] determined the characteristic
of pump and the head losses for varying cross sections at the pump intake. Van Esch
[7] experimentally showed the distribution of suction velocity profile for mixed flow
pump along with forces on the impeller was found to be non-uniform. It is concluded
that the entry flow is critical for the performance of pump. To obtain the flow physics
at the upstream of vertical inline pump, a CFD commercial tool called ANSYS CFX
14.5 is used.
P. Srikar Sai Tej · S. Christopher (B)

Department of Mechanical Engineering, Vel Tech Rangarajan Dr. Sagunthala R&D Institute of
Science and Technology, Chennai 600062, Tamil Nadu, India
862 P. Srikar Sai Tej and S. Christopher
Fig. 1 Surface mesh of vertical inline pump
2 Methodology
2.1 Numerical Analysis
Flow analysis is performed on vertical inline pump (specific speed 36) at unsteady
state conditions from 50 to 150% of Qn in the increment of 10% with respect to
nominal flow rate (Qn). The temperature is considered as isothermal and hence conti-
nuity and momentum equations are solved. The vertical inline pump was modelled
using Unigraphics and meshed using ICEM CFD by consideration of required Yplus.
Figure. 1 shows the surface mesh of vertical inline pump.
Grid independent studies were carried out with respect to total head developed by
the pump for different mesh size. The details of mesh are reported in numerical flow
prediction of inlet pipe by Christopher et al. [8]. Table 1 shows the boundary and set
up conditions used for simulation of vertical inline pump.
Two-dimensional drawing of vertical inline pump is shown in Fig. 2. Inlet pipe
of vertical line pump is divided into various sections from 1 to 7 in x/l ratio.
2.2 Parameters Definition
Whenever the flow is turbulent, the energy associated with it is nothing but turbulence
kinetic energy (k). It is the mean kinetic energy per unit mass in flow. It is expressed
as [9],
Characteristics of Upstream Flow in Vertical Inline Pump 863
Table 1 Boundary and setup

Parameters Values/Description
conditions
Flow analysis Unsteady
Time step 1.72 × 10−4 s
(Impeller rotation by 3°)
Inlet Total pressure
Outlet Mass flow rate
Interface between inlet pipe and Frozen rotor
impeller
Interface between impeller and Frozen rotor
casing
Turbulence model SST
Reference pressure 101.325 kPa
Nominal flow rate, Qn 0.014 m3 /s
Flow range 50–150% of Qn
Convergence residual 1 × 10−4
Speed, n 2910 rpm
1 Upper curve
2
x/l=0
3
4 5 76 7
1 2 x/l=1
3 x/l=0.5 Lower curve

6
4 5
Inlet pipe
Fig. 2 Two-dimensional drawing of vertical inline pump
1 2
(u ) + (v )2 + (w )2
k= (1)
2

where the turbulent velocity u component in x direction is the difference between
the instantaneous (u) and the average velocity (ū), which is given by
u = u − ū (2)
Turbulent kinetic energy is produced and as a result of fluid shear, friction in

flow passage or the effect of buoyancy or external force. Velocity coefficient (cv ) is
defined
as the ratio of velocity at (x, y, z) to the velocity at Sect. 1 for nominal flow
rate vs,Q n .
vx,y,z
cv = (3)
vs,Q n
The meridional velocity (cm ) is the velocity which is perpendicular to the flow
passage in meridional view of the impeller, which is responsible for flow rate through
pump.
The performance test was conducted in an open circuit from zero to maximum flow
rate at constant speed. The suction and delivery pressures were recorded through
pressure transmitter. During performance test, the speed was controlled with the
help of variable frequency drive. Flow rate through test pump was measured using
electronic flow meter with an accuracy of ±0.2%. The measurement uncertainty
for total head is ±2%, input power to the pump is ±1.4%, and efficiency is ±2%.
Performance validation between experimental and computational work were reported
with respect to total head developed by the pump and it was found to be perfect.
Details about it are elaborated by Christopher et al. [8].
Figure 3 shows the distribution of turbulent kinetic energy in inlet pipe for various
flow conditions from 0.5 to 0.9 Qn. When the flow rate is 0.5 Qn, the turbulent kinetic
energy increases from 0.4 x/l till it reaches the outlet of inlet pipe. Similar trend is
followed for 0.6 Qn, whereas for 0.7 Qn, energy changes after 0.6 x/l and increases
Fig. 3 Distribution of 2.00

turbulent kinetic energy in
inlet pipe (0.5–0.9 Qn) 0.5Qn
1.50 0.6Qn
0.7Qn
k in J/kg
0.8Qn
1.00 0.9Qn
0.50
0.00
0.0 0.2 0.4 0.6 0.8 1.0
x/l
till the outlet of inlet pipe. But for 0.8 and 0.9 Qn, it is almost constant till 0.9 x/l
with slight increase at the exit. This infers that turbulent intensity is more when flow
is around 0.5 to 0.7 Qn due to back flow from the impeller (Fig. 4).
Similar kind of plot as shown in Fig. 4 is used for flow from Qn to 1.5 Qn. The
trend is same for all flow rates which are above nominal flow. It is inferred that the
trend of turbulent kinetic energy is different for flow rate less than nominal flow
when compared with flow above nominal. The magnitude of kinetic energy due to
turbulence is less as compared with flow below nominal flow (Refer Fig. 3).
The turbulent kinetic energy distribution at the interface (area) of inlet pipe and
impeller is shown in Fig. 5. This plot confirms the same conclusion as discussed
above. Turbulence intensity is high around 0.5 Qn and decreases gradually towards
maximum flow rate. The unique observation is that, even at nominal flow rate, the
level of turbulence is not minimal. This concludes, there is a scope for better design
of inlet pipe for vertical inline pump.
Fig. 4 Distribution of 0.15

turbulent kinetic energy in
inlet pipe (Qn to 1.5 Qn)
0.12 Qn 1.1Qn
1.2Qn 1.3Qn
k in J/kg
0.09
1.4Qn 1.5Qn
0.06
0.03
0.00
0.0 0.2 0.4 0.6 0.8 1.0
x/l
Fig. 5 Turbulent kinetic 4

energy at the interface of
inlet pipe and impeller
3
k in J/kg
0
0.4 0.6 0.8 1 1.2 1.4 1.6
Q/Qn
Figure 6 shows the contours of velocity as coefficient, which depicts the flow
physics at 50, 100, and 150% of Qn. The top of inlet pipe, flow recirculation is
(a) 50% Qn
(b) 100% Qn
(c) 150% Qn
Fig. 6 Contours of velocity coefficient in inlet pipe

observed at 50% Qn. At nominal flow, the flow restriction is found at the interface
of inlet pipe and impeller.
Significant separation and return flow was observed at the upstream of impeller
was reported by Posa et al. [10] at reduced flow rates. Similar trend is observed
in the inlet pipe of vertical inline pump. In general, the flow velocity should be
within the 1.5 to 3 m/s at the upstream of impeller [2]. If it exceeds, the local static
pressure falls below the vapour pressure at that prevailing temperature of liquid,
which leads to cavitation in pump. The velocity distribution in meridional view is
shown in Fig. 7 for three flow rates (50, 100, and 150% of Qn). The intensity of
velocity is high at the upstream of impeller, when flow is below nominal flow rate.
At the leading edge of the impeller vanes, the meridional velocity is high as compared
with other flow rates. The uniformity of flow pattern is not found at nominal or design
flow rate.
At nominal flow rate, the magnitude of turbulent kinetic energy is not less as
compared to 150% Qn (refer Fig. 5). This ensures that design of inlet pipe is not
optimum as well as the return flow from the impeller is not zero.
4 Conclusions
Flow characteristics or physics at the inlet of any pump determines the energy of
pump at outlet, even if impeller and casing are design to the right working conditions.
This paper enumerates the flow based on turbulent kinetic energy distribution in the
inlet pipe of vertical inline pump. Turbulent kinetic energy is found to be high at off
design condition and reduces towards the nominal flow and further. This gives the
confirmation that turbulent flow consuming more energy than laminar flow. In reality,
it is impossible to avoid turbulent but it can be controlled to some extent without
deteriorating the performance of pump by various design optimization of inlet pipe.
Fig. 7 Contours of
meridional velocity
(a) 50% Qn
(b) 100% Qn
(c) 150% Qn
Acknowledgements The author would like to recognize Professor Shouqi Yuan, and National
Research Center of Pumps, Jiangsu University, China for providing the necessary facilities for
research studies.
References
1. Lazarkiewicz S, Troskolanski A (1965) Impeller pumps. Pergamon, Oxford, UK

2. Stepanoff AJ (1957) Centrifugal and axial flow pumps, 2nd edn. John Wiley, New York
3. Torii D, Nagahara T, Okihara T (2013) Suppression of the secondary flow in a suction channel
of a large centrifugal pump. In: IOP conference series: materials science and engineering, vol
52, 032005. https://doi.org/10.1088/1757-899X/52/3/032005
4. Badowski HR (1970) Inducers for centrifugal pumps. Worthington Canada Ltd., Internal Report
5. Valle JD, Braisted DM, Brennen CE (1992) The effects of inlet flow modification on cavitating
inducer performance. ASME J Turbomach 114(1992):360–365. https://doi.org/10.1115/1.292
9152
6. Mikhailov IE, Kuzmenko AI (1985) Effect of the shape of the transition section on head losses
in intakes of a pumped-storage station. Hydrotechnical Constr 19:652–660. https://doi.org/10.
1007/BF01435434
7. Van Esch BPM (2009) Performance and radial loading of a mixed-flow pump under non-
uniform suction flow. ASME J Fluids Eng 131(2009):051101-1–051101-7. https://doi.org/10.
1115/1.3089539
8. Christopher S, Yuan S, Pei J, Cheng GX (2017) Numerical flow prediction in inlet pipe of
vertical inline pump. J Fluids Eng 140(2017):051201-1–051201-10
9. Dewan A (2011) Tackling turbulent flows in engineering. Springer-Verlag, Berlin Heidelberg.
https://doi.org/10.1007/978-3-642-14767-8_2
10. Posa A, Lippolis A, Balaras E (2016) Investigation of separation phenomena in a radial pump
at reduced flow rate by large-eddy simulation. ASME J Fluids Eng 138:121101–1–121101-13.
https://doi.org/10.1115/1.4033843
Design Optimisation of Split Case
Centrifugal Pump
Manoj M. Ugale, Raju N. Patil, and Vishal V. Chavan
1 Introduction
Computational fluid dynamics (CFD) analysis is being increasingly used in the design
of centrifugal pumps. With the help of CFD approach, the complex internal flows
in water pump impellers and pump housings, which are not fully understood until
now, can be well predicted, to fasten the pump design procedure. Thus, CFD is an
important tool for pump designers.
The use of CFD tools in turbo machinery industry is quite extensive today. Many
tasks can numerically be solved much faster and at low cost than by experiments.
However, the highly complex flow in turbo machinery raises a query of the most
appropriate technique for modelling the pump system. CFD analysis is very useful for
predicting pump performance at various mass-flow rates. For designers, prediction
of the operating characteristics curve is most important. All theoretical methods
for prediction of pump performance simply give a value, but designer is unable to
determine the core cause for the poor performance. Due to the advancement of CFD
platform, one can get the head, efficiency values as well as observe actual behaviour
of the fluid in the pump. Recent advances in computing power, together with powerful
hardware, graphics and interactive 3D manipulation of geometric models have made
the process of building a CFD model and analysing results with much less work effort,
reducing time and, hence the cost. Todays’ advanced solvers contain algorithms
which empower robust solutions of the flow field in realistic time. As an outcome
of these factors, computational fluid dynamics is now a proven industrial design
M. M. Ugale (B) · R. N. Patil · V. V. Chavan

Wilo Mather and Platt Pumps Pvt. Ltd., Pune, India
e-mail: manoj.ugale@wilo.com
R. N. Patil
e-mail: raju.patil@wilo.com
V. V. Chavan
e-mail: vishal.chavan@wilo.com
872 M. M. Ugale et al.
tool, which helps reduce design time and improve the processes throughout the
engineering domain. CFD provides a cost-effective and nearby accurate substitute
to physical model testing with variations on the simulation being performed rapidly
offering obvious advantages. From available literature, it is found that most former
research, particularly research based on numerical approaches, has emphasis on the
design state of pumps. Limited work of study about the off-design performance of
pumps is found. Centrifugal pumps are extensively used in many applications, so
the pump system is also required to operate over a widespread flow range in most of
the applications. Numerical simulation of centrifugal pumps is not an easy task due
to the usual CFD difficulties: turbulence, separation, boundary layer phenomenon,
etc. While there are also particular problems such as complex geometry of the blade
profile, high number of cells are required with unstructured grids. So structured
grids being used can give better convergence in faster time than the unstructured
ones. Energy transfer is produced primarily by the centrifugal force in the impeller.
These force source terms must be included in the equations of the moving zone. CFD
has proved to be a very useful tool in the analysis of turbomachinery, both in design
and the performance prediction.
The purpose of the current work is to demonstrate a numerical study of split case
centrifugal pump taking into consideration the whole pump 3-D geometry including
impeller side rooms. Various iterations of suction elbow design are also performed.
This study has been performed with the commercial software package ANSYS
CFX using the mesher ANSYS ICEM CFD. The software code uses finite volume
method and solves the turbulence equations with capability to handle structured and
unstructured grids.
2 Methodology
The incompressible flow through the rotating impeller is solved with constant rota-
tional speed. 3-D incompressible Navier-Stokes equations are solved to analyse the
flow in centrifugal pump. Turbulence is modelled with SST turbulence model [1].
Mass Conservation Equation
∂ρ ∂
+ ρU j = 0 (1)
∂t ∂x j
The principle of mass conservation establishes that the mass of a fluid volume (a
volume that always contains the same fluid particles) is constant. In this equation,
U j represented the three-dimensional velocity vector components of the flow. If the
flow is assumed steady,
Design Optimisation of Split Case Centrifugal Pump 873
∂ρ
=0
∂t
and then the equation is reduced to
∂U j
=0 (2)
∂x j
Momentum Conservation Equation

The conservation equation for momentum ρU i can be formulated as
∂ ∂ ∂p ∂τi j
(ρUi ) + (ρUi U j ) = − − + ρ fi (3)
∂t ∂ xi ∂ xi ∂x j
The three terms on right-hand side of above equation represented the x i compo-
nents of all forces due to the pressure P, the viscous stress tensor τ ij stress tensor is
given by

∂Ui ∂U j ∂Uk ∂Ui ∂U j
τi j = −μb δi j + + −μ + (4)
∂ xi ∂x j ∂ xκ ∂x j ∂ xi
where μb = 2/3μ is the bulk viscosity, μ is the dynamic viscosity and δ ij represents
the delta (δ ij = 1 if i = j and δ ij = 0 for i = j).
2.2 Grid Generation
It has been seen that the mesh has a great influence on the solver convergence and
solution of every CFD simulation, as it is important to adopt high-quality mesh
standards to ensure the convergence and accuracy of the numerical simulation. The
model geometry was built in UGNx and meshed with ANSYS ICEM CFD. The grid
used in pump housing was 2 million tetra prism and suction volute was 2 million
tetra prism, while that in impeller was 1.6 million hex [2].
2.3 Domain Physics
Initialization in Ansys CFX is done by providing initial guess values to solve the
governing equation so that the flow field variables can be solved by iteration towards
the solution (Table 1).
Table 1 Domain physics

Domain—DOM_IMP
Type Fluid
Materials
Water
Fluid definition Material library
Morphology Continuous fluid
Settings
Buoyancy model Non buoyant
Domain motion Rotating
Angular velocity rpm
Axis definition Coordinate axis
Rotation axis Coord 0.3
Reference pressure atm
Transitional turbulence Fully turbulent
Domain—DOM_PH
Type Fluid
Materials
Water
Fluid definition Material library
Morphology Continuous fluid
Settings
Buoyancy model Non buoyant
Domain motion Stationary
Reference pressure atm
Transitional turbulence Fully turbulent
All the test and experiment work was carried out at test bed in Wilo Mather and
Platt Pumps Pvt Ltd. Pune. Piping setup for this pump was successfully developed
and assembly of pump into the test bed system was without any leakage issue.
Experiments were carried out using water as a working fluid (Fig. 1).
Operating frequency for test was 50 Hz. Setup was operated on different operating
conditions and the effects were observed. In-house software was used to take readings
while the test. Experimental results at various operating conditions are shown in
Tables 2, 3 and 4.
Fig. 1 Experimental setup
Table 2 Experimental
Flow (m3 /h) Head (m) Efficiency (%)
results: old
4800 64.68 74.5
6400 60.72 85.9
7200 57.68 86.6
8000 53.51 87.1
8800 49.52 86.8
Table 3 Experimental
results: new optimised
suction elbow geometry 4800 64.56 76.2
6400 60.55 87.7
7200 57.56 88.5
8000 53.42 89.2
8800 49.48 88.9
Table 4 Design
Impeller diameter 900 mm
specifications
Impeller eye diameter 530 mm
Number of blades 7
Impeller width 170 mm
Volute width 275 mm
The simulations were performed using ANSYS CFX solver. Table 5 shows the CFD
results obtained from CFX post for the new optimised suction elbow geometry.
Iterative elbow design and full pump analysis has been carried out and the results
are obtained for the pump with optimised suction elbow geometry shown in Fig. 2.
The results are taken only when the convergence is obtained for the solution.
Flow vectors shown in Fig. 3 with optimised elbow geometry show no presence
of recirculation in the pump domain.
Pressure contours shown in Fig. 4 show that stagnation point is matching on the
volute tongue.
Table 5 CFD results: new

optimised suction elbow
geometry 4800 66.0 79.0
6400 62.8 87.8
7200 60.6 89.9
8000 57.7 91.0
8800 53.4 89.9
Fig. 2 Pump wet surfaces with optimised suction elbow

Fig. 3 Vector plot at XY plane
Fig. 4 Pressure contour at XY plane
4 Conclusions
Experiments are carried out at various flow rates as specified in Tables 2 and 3. The
best efficiency point for the pump is seen at 8000 m3 /h both in measurement and
CFD analysis. The cases when analysed using ANSYS CFX software [3, 4] and the
results obtained are satisfying at and near the best efficiency point. This proves the
accuracy of CFD and experimental approaches near best efficiency point with mere
1.8% difference.
The accuracy of the solution can be further increased by applying roughness to
the walls in the CFD simulation. The use of double volute has further contributed to
efficiency drop. It can be deselected for the further analysis to see improvement in the
operating performance of the pump. The use of new suction elbow with optimised
sections has increased the efficiency of pump by 2.1%.
References
1. Gülich JF Centrifugal pumps, 3rd edn. Springer Heidelberg, Dordrecht London New York
2. Liseikin VD Grid generation methods, 2nd edn. Springer Heidelberg, Dordrecht London New
York
3. Webpage: https://www.ansys.com/products/fluids/ansys-cfx
4. Guide for verification and validation of computational fluid dynamics solutions. AIAA Guide
G-077-1998. www.aiaa.org
Effect of Reynolds Number on Rotational
Behavior of Low Speed Wind Turbines
P. Viswanathan, T. Prabu, V. Vijai Kaarthi, and J. Archana
1 Introduction
Small turbines are of a limited variety of designs due to cost and performance
constraints. The most common design is a stall regulated, variable speed, horizontal
axis, fixed pitch three-blade, direct drive permanent magnet machine. Blade pitch
control would be difficult to justify economically, so the blades are given a fixed
pitch, and optimized for power production at the rated speed [1]. This results in
poorer performance at lower speeds than could be achieved by a turbine with active
pitch control. Ultimately, the rotational speed of turbine (N) is a function of self-
inertia (I), moment due to loads (bearings) on shaft (M) and wind velocity (vin ). The
motivation for the present research is that three-dimensional numerical simulations
on the rotational behavior of low speed wind turbines are not investigated to a greater
extent by fellow researchers. The dynamic rotational speeds of the turbine and wake
characteristics for the proposed design are established in this paper.
2 Problem Statement
Most of the pioneering works in the field of wind—CFD involve the determination of
airfoil characteristics [2]. It is redundant to hence complicate the geometry of turbine
blades with these sections. The proposed design is henceforth supposedly modeled
with a rectangular section.
P. Viswanathan (B) · T. Prabu

V. Vijai Kaarthi · J. Archana
ME Energy Engineering, Department of Mechanical Engineering, PSG College of Technology,
Coimbatore, India
880 P. Viswanathan et al.
A wedged profile is linearly built between the chord and tip sections to outline
the blade. The blades are riveted to the hub normally (i.e., angle of attack is zero).
Numerical evaluation of the dynamic behavior of wind turbine to the inlet flow
conditions is presented in this paper which consequently fills the research gap [4].
3 Computational Domain and Meshing
3.1 Computational Domain
The blade is modeled with chord at root, c1 = 50 mm and at tip, c2 = 25 mm.

The thickness of blade is also varied linearly from root, t 1 = 6 mm to tip, t 2 =
3 mm. The length of the blades are maintained at L = 400 mm. The hub is a circular
disk for mounting of blades, with diameter, Dhub = 250 mm. The dimensions are
parametrically analogous to compact low speed commercial wind turbines. A passive
cylindrical domain of diameter, D1 = 1200 mm, is surmounting the turbine. It rotates
along with the turbine and hence captures the wake dynamics.
The domain is offset normally from the blade to its frontal area at about Dof =
100 mm and to the rear area at about Dor = 1000 mm for an exhaustive understanding
of flow paths (streamlines). An outer cushion domain is provided in rectangular
form which actively interacts with the passive inner domain, and supposedly arrests
the wake effects outside of the inner rotating domain. The enclosure is set to the
dimensions of X = ± 100 mm, Y = ± 500 mm and Z = ± 100 mm. This
helps in tapering the scope of wake studies behind turbine. The geometry for the
computational domain is exhibited in Fig. 1.
Fig. 1 Geometry for the

computational domain
Inner
Domain
Outer
Domain
Blade
Effect of Reynolds Number on Rotational Behavior of Low Speed … 881
3.2 Meshing and Refinement
Commercial software ANSYS ICEM CFD has been used to generate an unstruc-
tured tetrahedral volume mesh for Fluent. Fine refinements like curvature—prox-
imity capture and improved orthogonal quality of mesh system are carried out. The
mesh element size is set at 0.05 m, which effectively works with the system after
fair experience. The number of nodes and elements in the geometry is 54840 and
0.279 M, respectively.
4 Conservation Equations and Boundary Conditions
4.1 Conservation Equations
The model is set to solve for a pressure based solver—being an in-compressible

flow with density changes less than 5% (or) Ma < 0.3. A transient state simu-
lation is executed to account for writing up of the motion history as set of co-
ordinates in variant time steps. The turbulence around the turbine blades can be
effectively captured by k-ε model because of its robust nature. Realizable model is
preferred for boundary layers under strong adverse pressure gradients, separation
and re-circulation. Continuity equation and three-dimensional Navier–Stokes equa-
tions define the velocity and pressure at progressive time steps in every nodal point.
Non-dimensionalized form of k-ε equations is shown here that is associated with the
turbulence model. The model constants were set to default Cμ = 0.09, σk = 1.00,
σε = 1.30, C1ε = 1.44, C2ε = 1.92. The 6 DOF Solver is used in a reduced form
with only 1 degree of freedom rotation about the center of gravity in (+Y axis) of the
rigid body [3, 5].
A uniform velocity profile is considered at the inlet. Constant magnitude and normal
to the boundary—flow condition is imposed at the inlet (vin ). Pressure outlet has
been applied at the outlet (operating pressure = 101325 Pa). Blade is supposedly, the
rigid body actively rotating about +Y axis. The inner domain is passively interacting
(rigid body) with blades to capture flow dynamics. The outer cushion is defined to
be a deforming body. The boundary conditions are imposed on inlet and outlet of
the non-rotating outer domain. The mass properties of rigid body are defined for 6
DOF Solver as inputs. The material of turbine as a whole is taken as American Oak
Wood. The mass of body is M body = 0.3445 kg. The moment of inertia about Y- axis
is fed as I yy = 0.017089 kg m2 .
5 Numerical Procedure
Finite volume-based solver is used in simultaneous solving of conservation equations

in ANSYS Fluent. Pressure and velocity coupling is achieved using Semi-Implicit
Method for Pressure Linked Equations (SIMPLE) algorithm. QUICK discretiza-
tion is used as it provides better accuracy than second order scheme for rotating
and swirling flows as is the case. Flow courant number remains below 1.0 for all
computations. The convergence criteria is set at 10–3 for pressure and momentum
convergence. The solution method is set as implicit scheme with time step of 0.1 s
for a flow period of 90 s. This time provided is sufficient enough for the turbine to
reach its steady state.
6.1 Rotational Behavior of the Turbine
The speed of rotation of turbine is found to be increasing with the wind speeds.
The simulation is run for fixed pitch—angle of attack. The results of simulation
are tabulated angular displacements in progressive time steps. RPM of turbine is
evaluated by applying the following equations:

θ2 − θ1 π
ω= × (1)
t 180

60 × ω
RPM = (2)
2×π
It is inferred from this graph Fig. 2 that at this lower speeds of rotation of turbine,
effective conversion of wind power is not possible. The numerical procedure is vali-
dated by performing large eddy simulation (LES) model against the k-ε turbulence
model for the computational domain for corresponding wind speeds. The percentage
error reduces from 25.4% to 12.57% corresponding to 3 and 15 m/s, respectively.
Adverse swirl behavior is almost captured by the k-ε model and is thus validated to
reasonable accuracy.
The importance of optimal pitching of blade and design of airfoil with a hindsight
to the material selection is the future scope of research with regard to this paper.
This approach helps designer in fixing the cut-in speed, the wind speed about which
system behaves sensible and cut-out speed, which is not operable/ extreme condition
for a low speed wind turbine. It dictates the safe operating condition of the turbine
with regard to its strength (dynamic loading).
Fig. 2 Variation of turbine

speed with wind velocity
6.2 Wake Dynamics
The rotating action of wind turbine imposes a reaction force on the flow field (stream-
lines) as dictated. A local zone of re-circulation is created behind the hub and the
wake effects are felt across the inner rotating domain. The effect of wake is directly
a function of inlet wind speed. The growth of re-circulation zone is dicey because of
adverse pressure gradient creation.
The length of re-circulation zone (RCZ) as established in Fig. 3 is an appropriate
characteristic in determining turbine placements. For a wind speed of 3 m/s, the
length of RCZ is 0.4 m. The strength and length of this swirl is directly a function
of inlet wind speed (Fig. 4).
The velocity contour shows the effect of obstacle (turbine) in flow path. The
stagnation of a wind at hub creates a local eddy formation prior to turbine. The
air molecules trapped inside this eddies and in the RCZ create a drop in efficiency
of turbine. It is significant to design the turbine to operate at the optimum point
possessing best aerodynamic characteristics with reference to this wake behavior.
7 Conclusion
The effect of inlet wind speeds (i.e., a function of Reynolds number) on the rotational
behavior of wind turbine is successfully investigated in this paper. The significance
of the research is to derive a scope for numerical analysis in wind turbine blade
Fig. 3 Formation of re-circulation zone for V in = 3 m/s
Fig. 4 Contour plot of velocity in the flow domain for V in = 3 m/s

design. Moreover, the wake characteristics of turbines can be evaluated and prior
development in the field of turbine technology can be ascertained.
References
1. Gitano-Briggs H (2012) Low speed wind turbine design. Advances in wind power, In-Tech
Open, 20 Nov 2012
2. Bakırcı M, Yılmaz S (2018) Theoretical and computational investigations of the optimal tip-
speed ratio of horizontal-axis wind turbines. Eng Sci Techno Int J, 21(6):1128–1142, 2215–0986
3. Direct numerical simulation of flow around a surface-mounted finite square cylinder at low
Reynolds numbers (2017) Physics of fluids, vol 29. American Institute of Physics, pp 045101
4. Selig MS, Guglielmo JJ (1995) Summary of low speed airfoil data. Pennsylvania State University,
vol 1. SoarTech Publications
5. Nigam PK, Tenguria N, Pradhan MK (2017) Analysis of horizontal axis wind turbine using
CFD. IJEST 9:46–60
Aerodynamic Performance of a Micro
Wind Turbine Blade with S-1223 Airfoil
Ascribable the Bionic Bumps on Leading
Edge
T. Prabu, P. Viswanathan, V. Vijai Kaarthi, and J. Archana
1 Introduction
Wind energy has become the savior for many countries [Scandinavian] in today’s
environment. The clean and green feature has become the top priority to harvest any
energy source available in the world. The installed capacity of wind energy systems
has grown to 591 GW at the end of 2018. According to the statistics of International
Energy Agency [1], the total electricity generation in the world was 24,255 TWh in
2015, and the energy produced from wind was 838 TWh, which is about 3.5% of
the total electricity generation in the world. The additive power generation depends
upon the availability of high wind. According to IEC 61,400, wind sites are sorted
from Class I to Class IV based on the wind speed prevalent to that particular area.
The availability of high windy Class I (10 m/s) and moderate windy Class II sites
(8.5 m/s) is minimal.
The Indian subcontinent predominantly has land sites belonging to Class III
(7.5 m/s) and Class IV (6 m/s) winds. Practically, these areas have less than 5 m/s
of annual average wind speed—due to the atmospheric boundary layer effect, and a
wind power potential of less than 200 W/m2 at 50 m hub-height from mean sea level,
according to the National Institute of Wind Energy, India. Most of these sites are
densely populated with urban dwellings. The wind power available in these urban
environment is less. But, a proper technology transfer to the field will reduce the
load on conventional power generation and thence the atmospheric pollution. Micro-
or Mini turbines are the systems that are best suitable for this concern. The power
produced by these systems can be used directly for domestic purposes or for other
captive applications. Micro-turbines operating in these environments were found to
T. Prabu · P. Viswanathan (B) · V. Vijai Kaarthi · J. Archana

e-mail: vichu.cbe22@gmail.com
888 T. Prabu et al.
be very efficient. The blades of the micro-turbines should possess very good aero-
dynamic performance characteristics such as high lift coefficient, low drag coeffi-
cient, attached boundary layer flow, and delayed stall at very low Reynolds number.
According to Fuglsang and Maesn [2], the airfoils are to be specifically designed for
the wind conditions pertaining at particular sites. National Renewable Energy Labo-
ratory [NREL] and the Risφ labs of Delft Technological University had significantly
contributed to the development of thick and thin airfoils for wind energy systems.
Further, to enhance the performance of the blades, vortex generators, bionic bumps,
slats, etc., were incorporated on the blades (Lin [3]; Quackenbush et al. [4]; Von
Stillfried [5]). The intent of this research work is to examine the aerodynamic perfor-
mance characteristics of a thin airfoil S-1223—blade extruded normally against the
bio-inspired blade incorporated with the bumps analogous to the fins of humpback
whale on the leading edge of the profile and to dissect the feasibility of using these
blades for the micro-turbines of the urban environment. The urban environment poses
a greater challenge for harnessing the wind energy and has a lot of constraints. In the
Indian context, in an urbanized area, the availability of space for the wind turbine
structures will be very less or will not be available. The prospect of mounting these
structures on the roof top is critical over the force and noise variables. The houses built
with tilted tops must withstand the revolutionary forces generated by the dynamic
rotating turbine system. The diameter of the turbine is restricted to 2 m, the acoustic
noise produced should not disturb the people living around, and more importantly,
the wind available over this environment is less than 5 m/s. Micro-turbines are the
best systems suitable for this type of environment, which can produce a competent
power less than 1000 W, suitable for lighting purposes and other mild applications.
The micro-turbines designed for these environments should have lower cut-in speed,
produce high torque at lower wind speeds, and generate the rated power at the avail-
able wind speed. To achieve the stated design requirements, the blades and in turn
the airfoils of the turbine should have high lift coefficient, delayed stall, and flat
operating characteristics for wide range of wind speeds.
2 Bio-Mimicry
Bio-mimicry is a design process, where patterns are inspired from nature to provide
striking solutions to the geometrical problems. The design solutions developed are
based on the evolutionary upgradation in biological systems through pure mathe-
matics. The adaptation of nature’s methodology has made revolutionary improve-
ments in the traditional systems [6]. The design process aims to examine the poten-
tial improvements of integrating the bio-inspired geometrical modifications to the
blade design process for performance enhancement of the wind turbine system.
The morphological features of the humpback whale flippers were considered and
mimicked onto the wind turbine blades to suit the case. The sharp and high-speed
banking turns executed by the humpback whale are favoured by the high lift to drag
Aerodynamic Performance of a Micro Wind Turbine Blade … 889
characteristics by the combination of the tubercles and the high aspect ratio of its flip-
pers. These tubercles(bumps) are of sinusoidal pattern that are present on the leading
edge of the flippers. This feature of the humpback whale is introduced into the design
of wind turbine blades, which delays the stall and increases the lift considerably.
2.1 Bionic Blades
Carreira Pedro and Kobayashi [7] had conducted a study on using the pattern of
the scalloped flipper onto the leading edge for the unmanned air vehicles (UAV) and
compared it against the baseline smooth edged blade. The low Reynolds number influ-
ences the type of separation in the flipper, and the aerodynamic performances were
improved two-fold. Carija et al. [8] investigated the aerodynamic effects of waving
bumps on NACA 0012 blade section and found that the bumped blade provided
greater advantage in lift to drag ratio over the angle of attack (AoA) ranging from 0°
to 20°.
NACA 2412 airfoil with leading edge bumps was investigated by Malipeddi et al.
[9] and found that, for AoA greater than 16°, the increase in lift and drag is 48%
and 46%, respectively. Gawad [10] studied the flow field of NACA 0012 airfoil with
spherical tubercles in the leading edge and had found that the tubercles influence
the coefficient of lift for different angle of attacks, delays or even remove the stall
within the actual operating range of AoA. Karthikeyan et al. [11] studied the effects
of leading tubercles on laminar flow separation of bubble for NACA 4415 airfoil. The
experimental results demonstrated that the laminar bubble, which was present in the
baseline airfoil at 6° angle of attack, was not present in the modified airfoil with the
leading edge tubercles. Besides, the point of flow separation had moved from 10%
of chord for the baseline airfoil to 60% downstream for the modified airfoil. Corsini
et al. [12] numerically analyzed the profiles of NACA 4415 and NACA 0015, without
and with the sinusoidal leading edge. A 30% gain in lift was predicted post stall and
observed a stabilizing effect on the suction side of the blade corresponding to the
peak of the sinusoidal leading edge. The authors also observed that the separation
occurred only at the trailing edge in span-wise cross sections, which corresponds to
leading edge troughs. Viswanathan et al. [13] numerically analyzed and compared
NACA 0012 blade with and without bumps and predicted that the blade with the
bumped leading edge generated 24% more lift than the normal blade for increased
AoA ranging up to 25°.
2.2 Airfoil Geometry
The urban wind systems are usually small-level or micro-turbines producing power
in the order of 1 kW and has a shorter span for the arm of blades. The building and
other high rise structures in the urban obstruct the wind flow path and reduce the
890 T. Prabu et al.
Fig. 1 S-1223 Airfoil, Ref: airfoil plotter
effective wind speed further. The environment forces a low speed wind condition
over the blades of small/micro wind turbines. Thin section airfoils are aptly consid-
ered to progress. The mass of the blade with thin airfoils will also be less, which
requires a less starting torque and is competent to thick airfoil blade sections. Wind
turbine blades with S-1223 airfoil with the bionic modification are considered for
the evaluation of suitability for micro-turbines. S-1223 is a thin airfoil, designed by
M Selig, by combining the favorable effects of both a concave pressure recovery and
aft loading to achieve maximum lift at a Reynolds number of 200,000. The S-1223
airfoil produces a maximum lift coefficient of 2.11 at an angle of attack of 16° [14].
Figure 1 illustrates the chord-wise distribution of S1223 airfoil.
The bionic modification is carried out by incorporating the bumps onto the leading
edge of the blades. The bumps are generated by simultaneous flaring and dwindling
of the chord by 10%, alternatively along the leading edge of the baseline blade. The
pitch of sinusoidal wave is considered as 30% of chord. The amplitude and the wave
length are chosen based on an experiment on leading edge protuberances on finite
span wings by Custodio et al. [15]. The authors, in their work, had observed that the
protuberance produced 10% greater lift coefficient than the baseline blade span with
NACA 634 -021 airfoil. In the modified blade geometry, the span corresponding to a
single wavelength is equal across all the segments of the blade. The surface area of the
normal baseline blade and the bionically modified, bumped blade are maintained the
same for a better comparison of aerodynamic characteristics. Figure 2a and b shows
the 2D span model of the baseline and the modified blade sections, respectively. The
geometry and the specifications of the blades considered for the numerical simulation
are given in Table 1.
(a) Baseline blade section (b) Modified blade section
Fig. 2 2D span-wise blade sections

Table 1 Comparison between the baseline and the humpback modified blade sections of S-1223
S. No. Description Baseline blade model Modified blade model
1 Chord (mm) 100 Max-110
Min-90
2 Span (mm) 100 100
3 Surface area (m2 ) 0.021 0.021
4 No of elements 565268 535248
5 No of nodes 597860 556860
3 Numerical Evaluation of Blade Spans
ANSYS Fluent is preferably one of the widely used platforms to perform numer-
ical computations, is used for this research work, and has the capability to
model flows of compressible/incompressible nature throughout the flow domain,
laminar/turbulent flow phenomena, two-dimensional/three-dimensional flow fields,
and steady/unsteady flow conditions. The compatibility of this software aids in fast
processing across computer systems. Fluent is a commercial software, which solves
the Reynolds averaged Navier–Stokes Equations (RANS); Mellen et al. [16], Beri
and Yao [17], Eleni et al. [18], and Gharali and Johnson [19], etc., had effectively
used this software to solve their flow problems over airfoils.
The RANS two equation model ‘k-ω turbulence model’ (Eq. 1–5) [20] is selected
to solve the flow domain. The two equation model attempts to predict turbulence by
the two partial difference equations for the variables, ‘k’ and ‘ω’, k being the turbulent
kinetic energy and ω being the specific rate of dissipation of the turbulent kinetic
energy into internal thermal energy. The k-ω model incorporates modifications for
the effects of low Reynolds number, compressibility, and the shear flow dissipation.
The model also predicts free shear flow spreading rates that are in close agreement
with measurements for far wakes, mixing layers and plane, round and radial jets and
is thus applicable to wall bounded flows and free shear flows, Eleni et al. [18].

∂(ρk) δ ρu j k δ ρk δk
+ = P − β ∗ ρωk + μ + σk (1)
δt δx j δx j ω δx j

∂(ρω) ∂ ρu j ω γω δ ρk δω
+ = P − βρω +2
μ + σk
∂t ∂x j k δx j ω δx j
ρσd ∂k ∂ω
+ (2)
ω ∂x j ∂x j
where
∂u i
P = τi j (3)
∂x j
892 T. Prabu et al.

2 ∂uk 2
τi j = μi 2Si j − δi j − ρkδi j (4)
3 ∂ xk 3

2 ∂u i ∂u j
Si j = + (5)
3 ∂x j ∂ xi
The solution methods are set as simple, least square cell-based, and second-order
upwind momentum. The number of iterations was set for 1000 steps, and if the
solution had not converged, the iterations were incremented by another 1000 steps.
The analysis is carried for an angle of attack ranging from −5° to 30° with an
increment of 5° and 2° closer to the peak values. The analysis is performed with an
inlet velocity of 5 m/s and a Reynolds number of 1 × 105 [21].
The performance of the modified blade sections was understood from the geometry
of the airfoil, the pressure distribution over the chord, and the non-dimensionalized
parameters such as coefficient of lift, coefficient of drag, and pitching moment, plotted
against various angle of attack (Eq. 6–7). The non-dimensional parameters are easy
to interpret and provide a clarity on comparison, Sadraey [22]—Aircraft Design.
L
Cl = (6)
1
2
ρC V∞
2
D
Cd = (7)
1
2
ρC V∞
2
m
Cm = (8)
1
2
ρC V∞
2
From the plot, lift coefficient (C l ) versus angle of attack (α), seven critical features
describe the airfoil performances, namely (1) the stall angle (α s ); (2) maximum lift
coefficient (C lmax ); (3) zero lift angle of attack (α o ); (4) ideal lift coefficient (C li ); (5)
angle of attack corresponding ideal lift coefficient (α Cli ); (6) lift coefficient at zero
angle of attack (C lo ); (7) lift curve slope (C lα ). From the plot, drag coefficient (C d )
to lift coefficient (C l ), the minimum drag coefficient (C dmin ), and the corresponding
minimum lift coefficient (C lmin ) can be observed. The optimum angle of attack (α),
which is an important feature of the airfoil, can also be observed from the plot C l /C d
versus angle of attack (α).
3.2 Coefficient of Lift and Drag
Figures 3 and 4 illustrate the numerically computed coefficient of lift and drag against
various angles of attack for both baseline and modified blade span. From the coef-
ficient of lift plot, observations were made that the coefficient of lift produced is
in similar pattern for both baseline and modified blade spans. The lift coefficient
is positive at zero degree AoA for both the blade sections. The coefficient of lift
increases linearly, up to 10° AoA and stalls for further increasing AoA due to which
the generation of lift reduces.
The maximum lift produced by normal baseline blade is 1.98 at 10° AoA; the
modified blade produced a lift coefficient of 2.2 for the same angle of attack, which
is 11.11% higher than the normal baseline blade. Stalling happens successively in
baseline blade just after achieving the maximum lift coefficient. This imposes the
critical nature in operation of the normal blade profile. Whereas, the stall is prolonged
up to 15° for the bionic-modified blade. This delay in stall will be helpful for the
blade to operate in wide wind speed variations.
Fig. 3 Comparison of lift

coefficient against angle of
attack
Fig. 4 Comparison of drag

coefficient against angle of
attack
894 T. Prabu et al.
The drag plot gives a similar observation, and the drag coefficient increases
linearly up to the maximum lift coefficient angle of 10°. The normal baseline blade
has a lower slope up to 20 ° and increases rapidly, and the drag coefficient of the
modified blade increases gradually with higher angle of attack from 10°. Both the
blade sections have the same drag coefficient at the maximum lift producing angle
of attack, and hence, it is established that the total power output from the modified
blade will be comparatively higher because of the higher lift coefficient for the same
drag coefficient.
4 Numerical Evaluation of 3-D Rotor Blades
The numerical evaluation of 2-D blade span has produced positive results, signifying
the fact that the suitability of the airfoil and the bionic mimicking can be further
examined as a 3-D rotor. The performance of a 3-D rotor blade can be analyzed and
compared with another by a non-dimensional parameter called the Coefficient of
Torque ‘CQ ’ which is given in Eq. 9, and the power produced by the rotors can be
computed from torque ‘Q’ as given in Eq. 10.
Q
CQ = (9)
ρn 2 B 5
Power, P = 2π n Q (10)
A 3-bladed, horizontal axis rotor is designed to produce a rated power of 1000 W

from a 5 m/s wind velocity. The design is reduced to 1: 10 scale for numerical and
experimental analysis. The diameter of the actual wind turbine rotor is found to be
2300 mm and that of the model is 230 mm. The rotor blades with the normal baseline
leading edge and that with the bumps on the leading edge are illustrated in Fig. 5.
Numerical analysis is carried out by constructing a local disk over the turbine
rotor. The fluid domain surrounding the local disk is made as a cubical domain to
resemble a wind tunnel test section. The dimensions of the cubical domain are 10
times that of the rotor diameter on all sides. The local disk and the domain are meshed
with unstructured coarse mesh. The turbulent kinetic energy and dissipation model
with second-order upwind scheme are used to initialize the solution.
The Courant number is used to compute the time step, initially with 100 steps
to converge. Computed results of the torque coefficients for normal baseline and
the bumped rotors are plotted in Fig. 6. The torque coefficient of the bumped rotor
varies from 4 to 3.5 and produced an average torque coefficient of 3.7. The torque
coefficient for the normal baseline rotor varies from 3.5 to 3 and averages at 3.1. The
modified rotor has produced 20% higher average torque than the normal baseline
rotor. The higher torque coefficient aids in achieving considerably a higher power
in the wind turbine system. From the above analysis, it is evident that the bumps
(a) Rotor with normal leading edge (b) Rotor with bumped leading edge
Fig. 5 Rotor with a normal and b bumped leading edge
Fig. 6 Toque coefficients of normal and modified rotors
of the humpback whale will enhance the performance of the wind turbine rotors
significantly.
5 Experimental Evaluation of the Rotors
The numerical evaluations has shown that the performance of the bionic-modified
blade section and rotor gives positive increments, but this has to be validated through
the on-field conditions. The on-field tests will reveal the real characteristics of the
airfoil, blades, and the turbine rotor as a whole. The other way is to conduct the
896 T. Prabu et al.
Fig. 7 Fabricated models of

normal and bumped
modified rotors
tests in a simulated environment, for which wind tunnels are widely used. Tangler
and Somers [23], Timmer and Van Rooij [24], Selig and Guglielmo [14], etc., have
used wind tunnels to conduct experimental study on airfoils and wings. An open
wind tunnel with comparatively lower turbulence levels is used to validate the rotor
results.
The tunnel used for the experimental validation is an induced draught type with
an axial fan on the divergent end, driving out the air. A D.C. motor of 5 kW is used
to power the axial fan. The speed of the fan is controlled by using a Variac, which
can vary the speed from 0 to 2000 rpm. The size of the test section is 300 × 300 ×
450 mm. A static pitot tube is placed over the entry section to measure the velocity
of the air flowing through the test section. The velocity of the air flowing through
the tunnel can be varied from 0 to 45 m/s. The 3-D rotors modeled are developed
in rapid prototyping machine by fusion depositing process in ABS material and are
shown in Fig. 7.
5.1 Experimental Setup
The rotors are mounted on a spindle and bearing arrangement and are supported from
the bottom of the test section as shown in Fig. 8. Initially, the rotors were coupled
to a 5 W DC motor generator. Since the rotors could not produce any significant
power, the DC generator is detached, and the test is carried out for the free rotation
Fig. 8 Rotor at wind tunnel testing
of the rotors at various speed. The speed of the rotors is measured by a non-contact
IR tachometer. The velocity of the air is increased at an interval of 1 m/s from 0 to
15 m/s and is ablated in the same manner in the range of experiment. The experiment
is repeated for few number of times for consistency.
5.2 Results and Discussions
From the experiment, observations were made for cut-in speed and the rotor speed
for different wind velocities. Theoretically, the ABS rotors could produce negligibly
small powers; therefore, apt comparisons were made between the speeds of the
normal baseline and the modified rotors at different wind speeds. Figure 9 illustrates
the speeds of the rotors observed for various wind velocities.
The normal baseline bladed rotor started rotating close to 4 m/s of air velocity,
which can be considered as the cut-in speed of the rotor. The speed increases from
100 rpm above 4 m/s to 3080 rpm for 15 m/s of air velocity inside the tunnel. High
Fig. 9 Speeds of the rotors at different wind velocities

898 T. Prabu et al.
flap-wise and edge vibrations observed from 10 m/s of air velocity. The humpback-
modified rotor initiated its rotation at 2.2 m/s of wind velocity, gained a steady speed
of 30 rpm at 3 m/s, and reached the maximum of 4100 rpm at 15 m/s of wind velocity.
Figure 9 illustrates that the modified rotor has produced the higher speeds for all the
air velocities compared to the baseline rotor.
The percentage increase in the speeds varies from 5% to 22% for wind velocities
lower than 10 m/s and increases over 100% at higher wind velocities. The bionic
modifications, when incorporated onto the leading edge of a rotor blade with S-
1223 airfoil, have lowered the cut-in speed and achieved the design speed, which is
required to produce the rated power at lower air velocities. This validates the positive
results of the numerical simulations of the blade sections and the rotors. From the
above experiment, it is evident that the bionic feature of the humpback whale will
enhance the performance of the wind turbine rotors in the field by generating more
power.
6 Annual Power Generation
The average power produced by the wind turbine rotors based on the cut-in speed,
rated wind speed, and the cut-out speed can be found using the Eq. 11, and the annual
energy output in kWh is calculated from the Eq. 12 (Fig. 10).

exp −(Uc /C)k − exp −(U R /C)k
Pe,ave = Pe R − exp −(U F /C) k

W (11)
(U R /C)k − (Uc /C)k
Fig. 10 Probability distribution of 10-min averaged wind speed of Peelamedu, Coimbatore [PV]
Annual Power = Pe,ave × 8760 kWh (12)
where Pe,ave is the average power produced in a year, PeR is the rated power of the
wind turbine, U c is the cut-in speed of the wind turbine, U R is the speed at which the
rated power is generated, U F is the cut-out or furling speed of the turbine, and C and
K are the Weibull scale and shape parameters. The energy generation is computed
for the wind prevailing over Peelamedu area of the Coimbatore district, India, which
lies at 11.0248° N latitude and 77.0028° E longitude, where the wind is continuously
monitored and recorded by 10-minute average scheme.
The shape factor ‘K’ and scale factor ‘C’ are computed as 1.68 and 3.22 m/s for the
above site from the recorded wind data. The annual power generation by the rotors
with normal and modified airfoil blades is computed based on the experimental data.
The cut-in speed U C for the rotors is considered as 4 and 2.2 m/s from the experiment
for the normal and the modified rotors, respectively. The rated wind speed U R is the
wind speed at which the turbine will achieve the design revolutions to produce the
rated power. The rated wind speed to produce 300 rpm, the design rotor rpm to
produce the rated power of 1 kW is considered as 5.5 m/s for both the rotors from
Fig. 9. The cut-out speed U F is considered as the maximum wind speed prevailing
in this location and is considered as 12 m/s. From Eqs. 11 and 12, the average power,
Pe,ave , and the annual energy production are computed, and the results are given in
Table 2.
From the above results, the following observations were made as follows:
1. The normal baseline rotor will operate and produce power for 1388 h in a year.
2. The annual average energy density produced will be 148.5 kWh per square
meter area of the rotor.
3. The wind energy system with the baseline rotor will produce a total energy of
1301 kWh for one meter square area of the rotor.
4. The modified rotor will operate for 3858 h and produce an annual average energy
density of 261.3 kWh per square meter area of the rotor.
5. The system with modified rotor will produce 2288 kWh of energy for one meter
square rotor area.
Table 2 Annual energy generation by the rotors of S 1223

Rotor Normal HB modified
Operating time in hours 1952 3853
Wind speed of the rotors to produce rated power in m/s 5.5 5.5
Operating hours with rated output in hours 1388 1654
Average energy density in kWh/m2 148.5 261.3
Cumulative annual energy developed by the rotors in kWh/m2 1301 2288
Percentage increase in the energy production by the modified turbines – 75%
900 T. Prabu et al.
On comparison, the modified rotor will produce 75% of increased energy in a

year than the normal baseline bladed rotor. The experiment has concluded that the
performance of the wind turbines can be enhanced by incorporating the bionic feature
on the blades, which will facilitate to produce more energy from low wind speeds.
7 Conclusion
Numerical and experimental investigations were carried out on S-1223 airfoil by

incorporating bionic bumps of humpback whale. The 2-D numerical analysis showed
11% increased lift coefficient, and the 3-D rotor simulations produced 20% increased
torque coefficient for the bionic-modified blades over the normal baseline bladed
rotor. The results of the experiments conducted in a simulated environment had also
proved that the modifications made on the leading edge reduce the cut-in speed and
increase the rpm corresponding to wind velocity.
From the experimental data, it is computed that the bionic-modified rotor will
produce 75% more energy than normal baseline rotors annually. In overall, the inves-
tigations had showed that the bionic feature of humpback whales’ bumps, when
incorporated onto the leading edge of the rotor with S-1223 airfoil, would enhance
the performance of wind turbine at low wind regimes. The future scope of work
involves fabricating the blades with and without bumps, tests the performance of the
blades at the field, and evaluates the actual performance of modifications.
References
1. Statistics I (2017) Key world energy statistics 2017. Int Energy Agency
2. Fuglsang P, Aagaard Madsen H (1999) Optimization method for wind turbine rotors. J Wind
Eng Ind Aerodyn 80:191–206
3. Lin JC (2002) Review of research on low-profile vortex generators to control boundary-layer
separation. Prog Aerosp Sci 38:389–420
4. Quackenbush T, McKillip R, Whitehouse G (2010) Development and testing of deployable
vortex generators using SMA actuation. In: 28th AIAA applied aerodynamics conference
5. Von Stillfried F (2009) Computational studies of passive vortex generators for flow control.
Diss. KTH, 108
6. Fish FE, Weber PW, Murray MM, Howle LE (2011) The tubercles on humpback whales’
flippers: application of bio-inspired technology. Interact Comp Biol 51:203–213
7. Carreira Pedro H, Kobayashi M (2008) Numerical study of stall delay on humpback whale
flippers. In: 46th AIAA aerospace sciences meeting and exhibit
8. Čarija Z et al (2014) Numerical analysis of aerodynamic characteristics of a bumped leading
edge turbine blade. Eng Rev 34:93–101
9. Malipeddi AK, Mahmoudnejad N, Hoffmann KA (2012) Numerical analysis of effects of
leading-edge protuberances on aircraft wing performance. J Aircr 49:1336–1344
10. Gawad AFA (2012) Numerical simulation of the effect of leading-edge tubercles on the
flow characteristics around an airfoil. In: International mechanical engineering congress and
exposition
11. Karthikeyan N, Sudhakar S, Suriyanarayanan P (2014) Experimental studies on the effect of

leading edge tubercles on laminar separation bubble. AIAA J 1279:1–16
12. Corsini A, Delibra G, Sheard AG (2013) On the role of leading-edge bumps in the control of
stall onset in axial fan blades. J Fluids Eng 135:
13. Viswanathan P, Prabu T, Sivasubramaniam S, Rudramoorthy R (2016) Aerodynamic perfor-
mance evaluation of humpback whale inspired wind turbine blade using CFD. Adv Appl Fluid
Mech 19:837
14. Selig MS, Guglielmo JJ (1997) High-lift low Reynolds number airfoil design. J Aircr 34:72–79
15. Custodio D, Henoch CW, Johari H (2015) Aerodynamic characteristics of finite span wings
with leading-edge protuberances. AIAA J 53(7):1878–1893
16. Mellen CP, Frohlich J, Rodi W (2003) Lessons from LESFOIL project on large-eddy simulation
of flow around an airfoil. AIAA J 41: 573–581
17. Beri H, Yao Y (2011) Effect of camber airfoil on self starting of vertical axis wind turbine. J
Environ Sci Technol 4:302–312
18. Eleni DC, Athanasios TI, Dionissios MP (2012) Evaluation of the turbulence models for the
simulation of the flow over a National Advisory Committee for Aeronautics (NACA) 0012
airfoil. J Mech Eng Res 4(3):100–111
19. Gharali K, Johnson DA (2013) Dynamic stall simulation of a pitching airfoil under unsteady
freestream velocity. J Fluids Struct 42:228–244
20. Wilcox DC (2008) Formulation of the kw turbulence model revisited. AIAA J 46(11):2823–
2838
21. Abbott IH, von Doenhoff AE (2015) Theory of wing sections: including a summary of airfoil
data dover publications. Dover Publications Inc., New York
22. Sadraey MH (2012) Aircraft design: a systems engineering approach. John Wiley & Sons
23. Tangler JL, Somers DM (1995) NREL airfoil families for HAWTs. National Renewable Energy
Laboratory, pp 117–123
24. Timmer WA, Van Rooij RPJOM (2003) Summary of the Delft University wind turbine dedicated
airfoils. J SolEnergy Eng 125:488–496

Fluid Mechanics and Fluid Power: T. Prabu P. Viswanathan Amit Agrawal Jyotirmay Banerjee Editors

Uploaded by

Copyright:

Available Formats

Fluid Mechanics and Fluid Power: T. Prabu P. Viswanathan Amit Agrawal Jyotirmay Banerjee Editors

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Fluid Mechanics and Fluid Power: T. Prabu P. Viswanathan Amit Agrawal Jyotirmay Banerjee Editors

Uploaded by

Copyright:

Available Formats

Lecture Notes in Mechanical Engineering

To submit a proposal or request further information, please contact the Springer

More information about this series at http://www.springer.com/series/11236

Fluid Mechanics and Fluid

Amit Agrawal Jyotirmay Banerjee

ISSN 2195-4356 ISSN 2195-4364 (electronic)

Transient Thermal Stratification in Sodium Pool of a Fast Reactor . . . . . 1

Sharp Interface Level Set Method on a Co-located Grid for High

Lattice Boltzmann Simulations of a Lid-Driven Cavity at Different

Mixed Convection MHD Flow of a Nanofluid in a Rotating System

Double Diffusive Mixed Convection in a Lid-Driven Enclosure

Numerical Study of Effect of Asymmetry on Performance

Investigation of Second Throat Supersonic Exhaust Diffusers

Numerical Analysis of Coupled Heat and Mass Transfer in Solid

Effect of Reynolds Number on Rotational Behavior of Low Speed

Dr. T. Prabu is Professor in the Department of Mechanical Engineering, PSG

Dr. P. Viswanathan is Associate Professor in the Department of Mechanical Engi-

Dr. Amit Agrawal is currently Institute Chair Professor in the Department of

Dr. Jyotirmay Banerjee is Professor in the Department of Mechanical Engineering

K. Natesan, K. Velusamy, and N. Kasinathan

K. Natesan (B) · K. Velusamy · N. Kasinathan

Fig. 1 Schematic of primary sodium circuit

Thermal stratification in sodium pool can be prevented if inertial force is made

3 Results and Discussion

Fig. 2 Schematic model of hot pool

3.1 Reactor SCRAM

3.2 Power Failure

D. V. S. Bhagavanulu and R. Balamurugan

Fig. 1 Definition sketch of a wall jet

Fig. 2 Curvature parameter

Fig. 3 Experimental arrangement

3 Results and Discussion

3.1 Decay of Maximum Velocity

Fig. 5 Decay of maximum 1

3.2 Mean Velocity Profiles in the Longitudinal Direction

Fig. 6 Mean velocity

Fig. 7 Variation of V in the

Fig. 8 Variation of W in the

Fig. 9 Mean velocity

3.3 Mean Velocity Profiles in the Spanwise Direction

3.4 Variation of Length Scales

Fig. 10 Variation of V in the

Fig. 12 Variation of length

However, there is a marginal shift of the position of the maximum velocity

1. Glauert MB (1956) The wall jet. J Fluid Mech. 1:625–643

Sujyesh Aanandh Manjunathan and Rajneesh Bhardwaj

2.1 Numerical Details

The governing equations are Navier–Stokes equations for viscous, incompressible,

S. A. Manjunathan · R. Bhardwaj (B)

where i, j = 1, 2, 3 and ν, p, τ and Re denote dimensionless velocities, pressure,

2.2 Simulation Setup

2.3 Performance Parameters

3 Results and Discussion

3.1 Mechanism of Thrust Generation

The mechanism of thrust generation of rigid and elastic/flexible plate is explained

Rigid Plate Pitching at k = 6/π .

position, a negative vortex starts developing and intensifying, finally creating an S

3.2 Comparison Between Propulsive Performance of Rigid

3.3 Impact of Reynolds Number