Civil_Engineering_for_Multi_Hazard_Risk_Reduction_Select_Proceedings

Lecture Notes in Civil Engineering
K. S. Sreekeshava
Sreevalsa Kolathayar
N. Vinod Chandra Menon Editors
Civil Engineering
for Multi-Hazard
Risk Reduction
Select Proceedings of IACESD 2023
@seismicisolation
@seismicisolation
Volume 457
Series Editors
Marco di Prisco, Politecnico di Milano, Milano, Italy
Sheng-Hong Chen, School of Water Resources and Hydropower Engineering,
Wuhan University, Wuhan, China
Ioannis Vayas, Institute of Steel Structures, National Technical University of
Athens, Athens, Greece
Sanjay Kumar Shukla, School of Engineering, Edith Cowan University, Joondalup,
WA, Australia
Anuj Sharma, Iowa State University, Ames, IA, USA
Nagesh Kumar, Department of Civil Engineering, Indian Institute of Science
Bangalore, Bengaluru, Karnataka, India
Chien Ming Wang, School of Civil Engineering, The University of Queensland,
Brisbane, QLD, Australia
Zhen-Dong Cui, China University of Mining and Technology, Xuzhou, China
@seismicisolation
@seismicisolation
Lecture Notes in Civil Engineering (LNCE) publishes the latest developments in
Civil Engineering—quickly, informally and in top quality. Though original research
reported in proceedings and post-proceedings represents the core of LNCE, edited
volumes of exceptionally high quality and interest may also be considered for publi-
cation. Volumes published in LNCE embrace all aspects and subfields of, as well as
new challenges in, Civil Engineering. Topics in the series include:
• Construction and Structural Mechanics
• Building Materials
• Concrete, Steel and Timber Structures
• Geotechnical Engineering
• Earthquake Engineering
• Coastal Engineering
• Ocean and Offshore Engineering; Ships and Floating Structures
• Hydraulics, Hydrology and Water Resources Engineering
• Environmental Engineering and Sustainability
• Structural Health and Monitoring
• Surveying and Geographical Information Systems
• Indoor Environments
• Transportation and Traffic
• Risk Analysis
• Safety and Security
To submit a proposal or request further information, please contact the appropriate

Springer Editor:
– Pierpaolo Riva at pierpaolo.riva@springer.com (Europe and Americas);
– Swati Meherishi at swati.meherishi@springer.com (Asia—except China,
Australia, and New Zealand);
– Wayne Hu at wayne.hu@springer.com (China).
All books in the series now indexed by Scopus and EI Compendex database!
@seismicisolation
@seismicisolation
K. S. Sreekeshava · Sreevalsa Kolathayar ·
N. Vinod Chandra Menon
Editors
Civil Engineering
for Multi-Hazard Risk
Reduction
Select Proceedings of IACESD 2023
@seismicisolation
@seismicisolation
Editors
K. S. Sreekeshava Sreevalsa Kolathayar
Department of Civil Engineering Department of Civil Engineering
Jyothy Institute of Technology National Institute of Technology
Bengaluru, Karnataka, India Mangaluru, Karnataka, India

Amrita Vishwa Vidyapeetham
Kollam, Kerala, India
ISSN 2366-2557 ISSN 2366-2565 (electronic)

ISBN 978-981-99-9609-4 ISBN 978-981-99-9610-0 (eBook)
https://doi.org/10.1007/978-981-99-9610-0
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature
Singapore Pte Ltd. 2024
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
Paper in this product is recyclable.
@seismicisolation
@seismicisolation
Preface
The field of civil engineering stands as a cornerstone of human progress, shaping

our world through the creation of infrastructure and the development of resilient,
sustainable communities. In an era defined by the increasing complexity and inter-
dependence of our society, the challenges faced by civil engineers have evolved to
encompass a broader spectrum of hazards and risks than ever before. From natural
disasters such as earthquakes, floods and hurricanes to emerging threats like climate
change and pandemics, the need for a multidisciplinary approach to civil engineering
has become increasingly evident.
This volume presents the select proceedings of the G20 C20 International Confer-
ence on Interdisciplinary Approaches in Civil Engineering for Sustainable Devel-
opment (IACESD-2023) which showcase the remarkable diversity of thought and
expertise that defines the field of civil engineering today. Through a rich collection of
research papers, case studies and best practices, this volume offers a comprehensive
exploration of the strategies, technologies and methodologies that are at the forefront
of multi-hazard risk reduction. From seismic retrofitting of critical infrastructure to
the development of climate-resilient urban environments, the contributions gathered
here reflect the commitment of the global civil engineering community to address
the multifaceted challenges of our time.
We extend our heartfelt gratitude to all the authors, presenters, participants and
reviewers who contributed to the success of the G20 C20 International Conference
on Interdisciplinary Approaches in Civil Engineering.
We thank all the staff of Springer for their full support and cooperation at all
the stages of the publication. We hope that this book shall be beneficial to students,
academicians, professionals and researchers.
Bengaluru, India Dr. K. S. Sreekeshava

Surathkal, India Dr. Sreevalsa Kolathayar
Kollam, India Prof. N. Vinod Chandra Menon
v
@seismicisolation
@seismicisolation
Contents
Civil Engineering for Multi-hazard Risk

Reduction-An Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
K. S. Sreekeshava, Sreevalsa Kolathayar, N. Vinod Chandra Menon,
and Bhargavi. C
Accessibility and Convenience

Exploring Walkable Localities as Organizing Principle of City Plan . . . . 13
Sumana Jayaprakash and Vimala Swamy
Prediction of Noise Pollution of Delhi City Using Machine
Learning: A Case Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Rajashri Khanai, Rajkumar Raikar, and Mrutyunjay Uppinmath
Smart Water Management: Using Machine Learning to Analyze
Water Quality Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
B. K. Monnappa, B. M. Shiva Kumar, T. S. Pushpa, and S. Shilpa
Artificial Neural Networks Modelling for Predicting Water Quality
in the Surface Waters of Western Godavari Delta, India . . . . . . . . . . . . . . . 45
G. Sri Bala, P. A. R. K. Raju, and G. V. R. Srinivasa Rao
IOT-Based Patrolling Robot for Construction Sites . . . . . . . . . . . . . . . . . . . 57
Smita Agrawal, N. Amrutha, B. M. Punyashree, V. Raksha,
and D. R. Yuktha
Innovative Exploration Techniques: Utilizing IoT-Enabled Robots
for Safe and Efficient Underground Tunnel Investigation . . . . . . . . . . . . . . 71
N. Shravan, M. Manoj Kumar, Bharatesh Chakravarthi, and C. Bhargavi
Event-Based Sensing for Improved Traffic Detection and Tracking
in Intelligent Transport Systems Toward Sustainable Mobility . . . . . . . . . 83
Bharatesh Chakravarthi, M. Manoj Kumar, and B. N. Pavan Kumar
vii
@seismicisolation
@seismicisolation
viii Contents
Human Activity Recognition in Construction Industry Using

Machine Learning Pose Estimation Technique . . . . . . . . . . . . . . . . . . . . . . . 97
M. Manoj Kumar, Bhuvaneshwari Hegde, S. P. Veda Murthy,
M. K. Akhila, and A. S. Bhoomika
Geotechnical Engineering
Soil Moisture Detection Using Arduino Sensor and ANN Prediction . . . . 109
Rajkumar Raikar, Basavaraj Katageri, Rajashri Khanai,
Dattaprasad Torse, and Praveen Mannikatti
Reliability Analysis of Clayey Soil Slope Stability Using GMDH
and RFC Soft Computing Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Rahul Ray
Prediction of Residual Strength After Liquefaction Using Artificial
Intelligence Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
Shubhendu Vikram Singh and Sufyan Ghani
Application of Digital Image Correlation Technique in Geotechnical
Engineering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
Sukanta Das
SSI on Geodesic Dome Using RSM and Comparison with ANN . . . . . . . . 153
M. Roopa, Jayachandra, and Manjunath Vatnalmath
Risk Analysis and Approaches

Overlap of the Safety Distance of the Gas and Oil Pipeline Network
with the Urban Area of the City of Mohammedia Morocco . . . . . . . . . . . . 167
Abderrahmane Jadouane, Abdelmajid Essami, Azzeddine Chaouki,
C. Bhargavi, Ibtissam Gourich, and Mustafa Nadraoui
AHP Approach for Risk Factors Prioritisation in Tunnel
Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
Preetesh Band, Abhaysinha Shelake, and Nivedita Gogate
Identification and Prioritization of Risk Factors Impacting Cost
Overrun in Indian Road Construction Projects . . . . . . . . . . . . . . . . . . . . . . . 189
Rohan Vishal Patil and Mahesh Balwant Sonawane
Enhancing Flood Forecasting Accuracy Through Machine
Learning Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
Halappanavar Ruta Shivarudrappa, S. P. Nandhini, T. S. Pushpa,
and K. P. Shailaja
Monitoring of Smart Greenhouse Using Internet of Things (IoT) . . . . . . . 213
S. Asha Bharathi, B. Meghana, S. Meghana, M. Akshatha, and S. Hamsa
@seismicisolation
@seismicisolation
Contents ix
Machine Learning Algorithms for Classifying Land Use and Land

Cover . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
N. R. Asha Rani and M. Inayathulla
Comparative Analysis of Machine Learning Models for Earthquake
Prediction Using Large Textual Datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
K. R. Niteesh, T. S. Pooja, T. S. Pushpa, P. Lakshminarayana,
and K. Girish
Assessment and Planning of Solid Waste Management System
for the Ichamati Riverside Area of Pabna City . . . . . . . . . . . . . . . . . . . . . . . 245
Ashik Iqbal, Imtiaz Ahmed Emu, Fahreen Hossain,
Sirajum Monira Popy, Rashed Uz Zzaman, and Md. Helal Ahmmed
Flood Inundation Mapping and Flood Intimation Using IoT
for KRS Dam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
D. Ashwini, R. Revuprasad, R. G. Ranjith Kumar,
C. S. Mallikarjun Swamy, and G. P. Bharath
Monitoring Climate Hazards, Rice Production Risks
and Management Practices in Bharathapuzha River Basin (BRB),
Palakkad, Kerala . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
P. Dhanya and K. Jayarajan
Investigating Spatio-Temporal Trends and Anomalies
in Long-Term Meteorological Variables to Determine If
Maharashtra is an Emerging Warming State in India . . . . . . . . . . . . . . . . . 281
Aman Srivastava, Rajib Maity, and Venkappayya R. Desai
Structural Analysis
A Study on Transient and Modal Analysis of a GFRP Bridge
Deck Under the Action of Heaviest Main Battle Tank Currently
Available in India (T-90S Bhishma) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
Arpita Mandal and Hiranmoy Barman
Investigation of Crack Propagation of Fly Ash-Based Geopolymer
Concrete Using Digital Image Segmentation Approach . . . . . . . . . . . . . . . . 339
Aashish Lamichhane, Gaurav Basnet, Amrit Panta, Shankar Shah,
and Nishant Kumar
Building Surface Crack Detections Using Deep Convolutional
Neural Network (DCNN) Architectures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
Rajashri Khanai, Basavaraj Katageri, Dattaprasad Torse,
and Rajkumar Raikar
Using Construction Sequence Analysis to Mitigate Risk
and Prevent Failures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365
Danette M. Gonsalves and Sumitra S. Kandolkar
@seismicisolation
@seismicisolation
x Contents
Seismic Study on Step Back Buildings and Step Back Setback

Buildings by Providing Bracing in the Soft Storey . . . . . . . . . . . . . . . . . . . . 377
Sinju Jose and Binu M. Issac
Analysis of Multistory Steel Framed Structure with Different
Infills Subjected to Seismic Loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389
Anjan Kumar, A. R. Pradeep, M. Vijayanand, and H. Siddesha
Application of Internet of Things (IoT) in Seismic Performance
Evaluation of 3D Printed Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405
Bhumika Jay, G. M. Basavanagowda, Archana Kumari,
Aishwarya Chauhan, and Bhavana Prasad
Experimental and Numerical Study on Hysteretic Behavior
of Laminated Rubber Bearing Under Quasi-Static Loading and Its
Performance on Secondary System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 415
Bharat Chalise
A Study on Self-Compacting Concrete at High Elevated
Temperatures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 429
M. Maniknata, Shaik Subhan Alisha, Durga Vara Prasad Bokka,
Gottumukkala Sravya, and V. Siva Rama Raju
Optimization Studies on Bracing Systems and Its Effective
Placement to Counteract Earthquakes in Very High Damage Risk
Zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443
S. Usha, H. A. Ajay, D. T. Abhilash, and B. A. Brunda
Performance Evaluation of Fire Exposed RC Structure Using
Pushover Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455
K. Mahammad Jafar and V. Sachin
Internet of Things Enabled Structural Health Monitoring Using
Fiber Bragg Grating Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 469
K. Chethana, Somesh Nandi, A. P. Guruprasad, and S. Ashokan
Seismic Analysis of Five-Storied Building Using U-Shaped Hybrid
Isolator and Lead Rubber Bearing Isolator . . . . . . . . . . . . . . . . . . . . . . . . . . 477
Srushti Gaikwad, Priti R. Satarkar, and A. Manchalwar
Structural Health Monitoring of Bridges Using IoT . . . . . . . . . . . . . . . . . . . 487
Deepak V. Ingale, K. Chethana, Gowthami P. Jain, S. Aditya,
and V. Venkatesh
@seismicisolation
@seismicisolation
About the Editors
Dr. K. S. Sreekeshava is currently working as Associate Professor and Head of

the Department of Civil Engineering, Jyothy Institute of Technology, Bengaluru.
He obtained Ph.D. from BMS College of Engineering, Visvesvaraya Technolog-
ical University, Belagavi, in the year 2020. His research interests are in the field of
masonry structures, bio-composites and structural design. He has published more
than 30 research articles in reputed journals, Elsevier and Springer publishing
houses. He has successfully coordinated funded AICTE-ISTE and AICTE Teaching
And Learning (ATAL) Academy Sponsored Faculty Development Programme. He
has the honors of life member of ISET ISTE ICI, INSC NICEE and IAENG. He
has successfully completed the research funding grant under competitive research
funding scheme initiated under scheme of TEQIP by Visvesvaraya Technological
University, Belagavi. Dr. Sreekeshava is Organizing Chair of G20 C20 International
Conference on Interdisciplinary Approaches in Civil Engineering for Sustainable
Development (IACESD 2023).
Dr. Sreevalsa Kolathayar pursued M.Tech. from IIT Kanpur, Ph.D. from IISc and
served as International Research Staff at UPC Barcelona Tech Spain. He is Associate
Professor in the Department of Civil Engineering, National Institute of Technology
Karnataka (NITK) Surathkal India. He has authored five books, edited twelve books,
published over 100 research articles and holds two patents. He is on the Editorial
Board of several International Journals. In 2017, The New Indian Express honored
him with South India’s Most Inspiring Young Teachers Award. He received the ISET
D. K. Paul Research Award from the Indian Society of Earthquake Technology for
the best Ph.D. thesis on Earthquake Risk Reduction in India. He received the “IEI
Young Engineers Award” from The Institution of Engineers (India) in 2019. He is on
the roster of two technical committees of ASCE Geo-Institute and is Member of the
Working Groups of BIS CED 39 for three IS codes. Dr. Sreevalsa has eight funded
R&D projects and completed over 50 civil engineering consultancy projects.
xi
@seismicisolation
@seismicisolation
xii About the Editors
Prof. N. Vinod Chandra Menon has worked as Professor in the Centre for Disaster
Management at the Yeshwantrao Chavan Academy of Development Administra-
tion (YASHADA), Pune. He has worked in charge of Emergency Preparedness and
Response in UNICEF India Country Office in New Delhi. He is currently Adjunct
Professor at Amrita Vishwa Vidyapeetham, India; President of RedR India and
Regional Director of Asia of the International Emergency Management Society
(TIEMS) Oslo. Professor Menon has over 37 years of working experience, of which
more than a quarter century has been in the fields of disaster risk reduction, climate
change adaptation and public policy analysis. He was Member of the High-Powered
Committee (HPC) on Disaster Management established by the Government of India
in 1999. He is also one of the Founder Members of the National Disaster Manage-
ment Authority (NDMA), Government of India, chaired by the Prime minister of
India, and he served in this capacity from 2005 to 2010.
@seismicisolation
@seismicisolation
Civil Engineering for Multi-hazard Risk
Reduction-An Introduction
K. S. Sreekeshava , Sreevalsa Kolathayar , N. Vinod Chandra Menon ,

and Bhargavi.C
1 Introduction
The modern built environment faces an array of potential hazards, both natural and
man-made, which underscores the importance of engineering practices that prior-
itize safety and resilience. This introductory exploration encompasses key facets
including Accessibility and Convenience, Geotechnical Engineering, Risk Analysis
and Approaches and Structural Analysis. By exploring these fundamental compo-
nents, it aims to provide readers with a foundational understanding of the multidis-
ciplinary approaches employed in civil engineering to mitigate risks and develop
infrastructural systems. As the need for hazard-resilient structures becomes increas-
ingly evident, this offers a gateway into the intricate world of civil engineering’s
important role in risk reduction and safety enhancement. The discipline of Civil
Engineering for Multi-Hazard Risk Reduction addresses the concerns to design,
construct and maintain infrastructure that can withstand an array of potential threats.
These hazards encompass a broad spectrum, ranging from seismic events, extreme
weather phenomena and geological instabilities to man-made risks. In this context,
civil engineering assumes an important role in not only creating aesthetically pleasing
structures but also in ensuring the safety and sustainability of these constructions and
its counterparts.
K. S. Sreekeshava (B) · Bhargavi.C

Jyothy Institute of Technology, Bengaluru, Karnataka, India
e-mail: sreekeshava.ks@jyothyit.ac.in
S. Kolathayar
National Institute of Technology-Karnataka, Surathkal, Karnataka, India
Amrita Vishwa Vidyapeetham, Amritapuri, Kollam, India
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 1
K. S. Sreekeshava et al. (eds.), Civil Engineering for Multi-Hazard Risk Reduction,
Lecture Notes in Civil Engineering 457, https://doi.org/10.1007/978-981-99-9610-0_1
@seismicisolation
@seismicisolation
2 K. S. Sreekeshava et al.
2 Accessibility and Convenience
Ultimate capacity, energy absorption capacity, axial stiffness, durability, etc., are
prominent parameters that are considered when we discuss about the strength criteria
of a structural member. The study titled Combined Metakaolin and Ground Granu-
lated Blast-furnace Slag-induced Concrete for Marine Environment by Thomas et al.
presented an experimental investigation on combined usage of metakaolin and GGBS
by replacing cement partially in various percentages. Various tests including dura-
bility properties like water absorption, permeable void, sorptivity, and compressive
and flexural strength tests of concrete exposed to saline water were conducted. The
comparison of resistance towards sulphuric acid attack of an eco-friendly alternative
to OPC, alkali activated concrete (AAC) with OPC is discussed in the study titled
Comparative study between alkali activated mortar and conventional mortar towards
sulphuric acid by Amina et al. The study with respect to OPC blends have been
extensively studied previously [1, 2]. Further, alkali activated geopolymer concrete
(AAGPC) was studied for its mechanical properties and compared with conventional
concrete by Najma et al. in their study titled Mechanical properties of alkali activated
geopolymer concrete. LC3 which is a ternary blended cement has been researched
to be utilized as a partial alternative to the cement in the contribution titled Use of
Coconut Coir Fibre in Limestone Calcined Clay Cement (LC3) Concrete by Sachin
et al. Mechanical and thermal properties of alkali activated concrete with fine aggre-
gates that are partially replaced with quartz sand is discussed by Sai et al. in their
research titled Study on Properties of Alkali Activated Concrete by Replacement
of Fine Aggregate with Quartz Sand. Polyethylene glycol has been recently used
to study its contribution towards enhancement of concrete’s properties [3, 4]. The
study by Rinu et al., titled Analyzing the Strength and Self-Curing Properties of Recy-
cled Concrete with PEG400 Addition discusses about the utilization of polyethylene
glycol-400 in recycled concrete due to its ability to maintain optimum hydration
thus reducing the consumption of excessive water required for curing. A comparison
of binary cementitious systems of mechanical properties to determine the optimum
percentage of red mud, silica fume with manufactures sand is discussed by Sara-
vanan et al. in their contribution titled Relationships between compressive, tensile
and flexural strengths of concrete using binary blends of red mud and silica fume with
M-Sand. Sustainable alternatives as potential replacement for coarse aggregates in
concrete is discussed by Bhakti et al. in their contribution titled Mechanical and Dura-
bility Properties of Concrete using Hemp Shives as a Partial Replacement of Coarse
Aggregates. Varying partially replaced specimen are tested for their mechanical prop-
erties including compressive strength, split tensile strength and flexural strength.
Lilesh et al. in their study titled Assessment of Optimum Percentages of Chemical
Admixture in composite Self-compacting concrete have discussed the determination
of ideal admixture percentages for a composite self-compacting concrete in which
they have varied the GGBS and Fly ash contents and tested for flexural strength and
split tensile strength along with observing their behaviour at various ages.
@seismicisolation
@seismicisolation
Civil Engineering for Multi-hazard Risk Reduction-An Introduction 3
Bamboo as an alternative construction material when pre-treated with certain

resins has been researched excessively in the past decade. Due to its enhanced
mechanical qualities, specially, tensile property, it has been extensively used in
concrete construction [5]. It further has contributed towards a sustainable society
[6]. Alisha et al. in their study titled Epoxy resin treated bamboo reinforced concrete
beams for rural construction buildings discussed about performing tests on flex-
ural and durability properties of bamboo reinforced concrete beams and compared
them with steel concrete beams in low-cost construction buildings. The study titled
GFRP Strengthened Timber Filled Concrete Composites under Axial Load by Ann
et al. focuses on composite action of Glass Fibre Mesh and timber filled concrete
with varying timber cores. Various properties including failure, ultimate capacity,
displacement behaviour, absorbed energy and axial stiffness are evaluated. Valuable
contributions to the field of strengthening of beams using GFRP strips have indicated
various advantages that include the customizable ability and orientation ease [7]. The
contribution titled An experimental study on the structural performance of full-scale
RC beams strengthened for shear using NSM GFRP strips by Rohin et al. discussed
their experimental work where a full-scale RC beam models were strengthened with
near surface mounted GFRP strips and observed for load–deflection relationship,
ultimate load carrying capacity, cracking pattern and mode of failure. Carbon being
non-corrosive and lightweight is a good replacement to steel [8]. The work titled
Study on development of FRC and its application in sluice gate by Deepthi et al.
discusses about the application of FRC using carbon fibres in construction of sluice
gates in small dams. CFRP sheathing with rectangular, rounded rectangular and
elliptical openings is studied for the ultimate strength of the beam on the basis of
the load–deflection curve observations in the paper titled Experimental study on
effect of CFRP sheathing on the behaviour of RCC beams with opening by Neethu
et al. Strength of stabilized geopolymer mud block with varying compositions of
mortar utilizing masonry prisms was experimentally determined in the study titled
Experimental Investigation on the Strength of Stabilized Geopolymer Mud Blocks
with various types of Mortar by Ramya et al. Various tests including the dimension-
ality, waster absorption test, initial rate of absorption and density of the blocks were
performed.
The role of mortar in masonry unit is to bond the units and distribute the stress
[9]. The contribution titled Study on compression strength of masonry prism using
cementitious grouting material as mortar by Kavyashree et al. aims to study the effect
of increase in strength of mortar and bond by utilizing the grouting material as mortar
at 3, 7 and 28 days. Further, cementitious grouting materials were also studied for
their shear and flexural bond strength in the research work titled Studies on Flexural
and Shear bond strength of masonry using cementitious grouting materials as mortar
by Maniranganath et al. The study indicated faster rate of construction when the
thickness of mortar joints is reduced and an increased bond strength on utilization
of cementitious grouting materials as mortar.
Experimental studies that investigate the free vibration analysis of fixed base and
column supported cooling tower shell which includes the study on different modes
of vibration including Torsion mode, circumferential mode and Meridional mode
@seismicisolation
@seismicisolation
have been presented by Sachin et al. in their paper titled Study on Circumferen-
tial and Meridional Modes of Free Vibration Response for Fixed base and Column
Supported Cooling Tower Shell. Janhavi et al. in their study titled Effect of Wind
Load on RCC and Steel Buildings in Different Terrain Category have analysed a
G+40 building for wind loads by varying wind zones, utilizing both RCC and steel
in four terrain categories using ETABs v20 to evaluate bending moment, storey drift
and displacement.
3 Geotechnical Engineering
Subsurface characteristics must be concentrated upon, when we are considering the

safety and risk assessments. This plays a major role in various civil engineering
aspects such as construction, structural safety and agriculture.
For instance, moisture content regulation is an important criterion in deciding the
crop pattern, type and other agricultural aspects [3]. The study by Rajashri et al. titled
Soil Moisture Detection Using Arduino Sensor and ANN Prediction focuses on the
installation of soil moisture sensor with Arduino UNO platform. Soft computing
techniques have led the domain of geotechnical engineering with advancements of
AI. Reliability analysis of Clayey Soil Slope Stability Using GMDH and RFC Soft
Computing Techniques by Rahul examines the applications of two soft computing
techniques, which are Group Method of Data Handling (GMDH) and Random Forests
Classifier (RFC) to address the nonlinear problem such as stability of soil slope.
Different notable probabilistic approaches have been used in the previous engineering
applications [4, 5]. One of the notable studies with respect to applications of Digital
Image Correlation in geotechnical engineering was performed to study the surface
displacements and strains in soil models subjected to varying loading conditions [6].
The study by Sukanta et al., titled Application of Digital Image Correlation Technique
in Geotechnical Engineering, discusses about utilizing DIC in depicting the failure
mechanism of footing on the slope and compares it with theoretical concepts known.
Prediction of Residual Strength after Liquefaction using Artificial Intelligence
Model by Sufyan et al. aims to develop a hybrid AI model to predict residual strength
required to ensure stability against soil movement post liquefaction.
Design and analysis of steel structures and their interaction with the subsurface soil
is inclusive aspect of Soil structure Interaction. Study on the performance of various
types of structures and their effective analysis techniques has been performed to
construct distinctive moment frame buildings subjected to seismic loading [7]. The
study titled SSI on Geodesic Dome using RSM and comparison with ANN by Roopa
et al. evaluates and compares the base shear value of geodesic dome under different
soil conditions.
@seismicisolation
@seismicisolation
4 Risk Analysis and Approaches
A notable contribution to this volume is by Rajib et al. study titled Investigating

spatio-temporal trends and anomalies in long-term meteorological variables to deter-
mine if Maharashtra is an emerging warming state in India, where they discuss about
variables such as rainfall, surface temperature and relative humidity in various parts
of Maharashtra. Their study claims that the state is becoming an emerging warming
state during the term 2011–2020 and have examined their hypothesis on the same
basis. They have analysed the details using nonparametric Mann-Kendall and Sen’s
Slope Methods.
Climate smart agriculture has contributed towards sustainable developments
which was initiated with Sendai framework on Disaster Risk Reduction (SFDRR) and
is now working towards Multi-hazard early warning systems [8–10]. Dhanya et al. in
their study titled Monitoring Climate Hazards, Rice Production Risks and Manage-
ment Practices In Bharathapuzha River Basin (BRB), Palakkad, Kerala, have made
an attempt to monitor the climatic hazards due to rainfall variability and patterns in
Palakkad using the Climate Hazard Group Infrared Precipitation with Station Data
(CHIRPS) which is one of the latest satellite-based rainfall data sets having high
resolution. This research work facilitates early hazard warning to the rice farmers
and in turn facilitates risk reduction. This also supports agriculturists, hydrologists
and other resource persons working on irrigation pertaining to surface water and
groundwater.
Flood plain zones have been extensively studied in the recent times and the
study titled Flood Inundation Mapping and Flood Intimation using IoT for KRS
Dam by Ashwini et al. details about utilization of 2D Water powered model for
examining dam break of KRS Dam situated in Mandya, Karnataka, India. Inunda-
tion maps created using Arc-GIS along with HECRAS and water level sensors are
used to detect flooding areas. Availability and provision of timely data with respect
to land use and land cover can facilitate substantial amount of pattern of hazard
occurrences prediction which in turn can help us approach it in a prepared manner.
The advantages and disadvantages of identifying changes in Land Use and Land
Cover is discussed in the research titled Machine Learning Algorithms for Clas-
sifying Land Use and Land Cover by Asha et al. The study by Halappanavar et al.
titled Enhancing Flood Forecasting Accuracy through Machine Learning Approaches
proposes learning algorithms such as decision tree, random forest, K-nearest neigh-
bours and logistic regression algorithms to build predictive models to train and
evaluate flood predictions.
Solid waste dumping scenarios and possible solid waste management plan along-
side the Ichamati River are discussed in the study titled Assessment and Planning of
Solid Waste Management System for the Ichamati Riverside Area of Pabna City by
Ashik et al.
@seismicisolation
@seismicisolation
Greenhouse monitoring to maintain optimal conditions for growth of plants using

wireless sensor networks is discussed in the study titled Monitoring of Smart Green-
house using Internet of Things (IoT) by Asha et al. The parameter such as temper-
ature, humidity, soil moisture and pH is collected using various IoT devices and is
transmitted via Global System for Mobile Communication module to cloud platform
Thingspeak. They also utilize image processing techniques to detect and prevent crop
loss.
Occurrence of disasters and resilience towards them is a challenging facet of the
city. One such disaster is earthquake and the study titled Comparative Analysis of
Machine Learning Models for Earthquake Prediction using Large Textual Datasets
by Niteesh et al. discusses about various machine learning approaches to predict the
earthquakes and their real-world applicability in order to reduce the damage to life
or resources.
Rohan et al. in their study titled Identification and Prioritization of Risk Factors
Impacting Cost Overrun in Indian Road Construction Projects make an approach
towards identifying important risk factors in the early stages of road construc-
tion projects to reduce the complexity throughout the later stages using Relative
Importance Index (RII) technique.
Infrastructure development has a crucial role to play in achieving sustainable and
efficient society. Tunnelling is one of the most demanding infrastructural projects in
the industry as such. The study titled AHP approach for risk factors prioritization in
tunnel construction by Preetesh et al. discusses about the severity of work, under-
ground conditions and outcomes resulting in cost overruns and scheduled delay. The
study aims to improve these risk analysis practices and ensure an accurate framework
is formed.
Safety regulations have been laid out far before with respect to any areas prone
to risk factors. Yet, one or the other compromised situations have been observed
with respect to the safety of manpower or the materials. The study titled Overlap
of the safety distance of the gas and oil pipeline network with the urban area of the
city of Mohammedia Morocco by Jadouane et al. discusses the measures to check
whether the existing safety distance regulations for gas and oil pipelines are adequate
to ensure the safety of the population, thus mitigating he risk associated with respect
to explosions, leaks or fire outbreak.
5 Structural Analysis
Structural Integrity and reliability contribute towards a safer and resilient construction
areas [11]. Arpita et al. in their study titled A study on transient and modal analysis of
a GFRP bridge deck under the action of heaviest main battle tank currently available
in India (T-90S Bhishma) have modelled and simulated FRP bridge deck slab of a
16 m long bridge subjected to moving load of heaviest battle tank currently available
in India by using ANSYS software.
@seismicisolation
@seismicisolation
There have been excellent contributions towards study on crack propagation [12]
and their compactness [13]. The paper titled Investigation of Crack Propagation of
Fly Ash-Based Geopolymer Concrete using Digital Image Segmentation Approach
by Aashish et al. studied different parameters including Otsu threshold value, size of
area of interest of crack and fly ash content using the histograms of each crack images.
This facilitated deciding optimal quantity of fly ash in order to reduce cracks and their
propagation. Further, Rajashri et al. presented their research titled Building Surface
Crack Detections using Deep Convolutional Neural Network (DCNN) Architectures
discussed about surface crack detection using TensorFlow by fine tuning a few DCNN
models such as VGG 19, VGG16, MobileNetV2 and DenseNet.
Gonsalves et al. in their study titled Using Construction Sequence Analysis to
Mitigate Risk and Prevent Failures employed ETABs v20 to analyse multi-storey
RC buildings of varying bay widths, that are subjected to wall loads and self-weight.
Construction stage analysis is performed to check for the cumulative weight resulting
from addition of each storey and their susceptibility to sequential effects is studied.
Another notable contribution towards soft-storey research is titled Seismic Study on
Step Back Buildings and Step Back Set Back Buildings by Providing Bracing in the
Soft Storey by Sinju et al. where chevron bracings were found to be a suitable lateral
load resisting system in case of step back set back buildings.
Study on steel framed structures and their response to seismic loading has been
researched extensively. They exhibit better performance when compared to RC struc-
tures [14] and reducing the impact of seismic loading can be achieved by provision of
infill [15]. The study tiled Analysis of multistorey steel framed structure with different
infills subjected to seismic loading by Anjan et al. compares the base shear and
story displacement of steel bare framed model, steel framed model with burnt brick
masonry infill and steel framed model with concrete block masonry infill for a multi-
storied commercial complex by using linear static analysis (Equivalent static anal-
ysis) and linear dynamic analysis (Response spectrum analysis) methods in E-tabs
software. Further, Application of Internet of Things (IoT) in Seismic Performance
Evaluation of 3D Printed Structure by Bhumika et al. has studied the scaled model
of proposed structure and extrapolated the results to check the response of prototype
model in terms of displacement and acceleration. Ajay et al. in their research titled
Optimization Studies on Bracing Systems and its Effective Placement to Counteract
Earthquakes in Very High Damage Risk Zone aim to analyse an RCC building model
in Zone V subjected to peak acceleration using peak response spectrum method in
ETABs. Their response when lateral stiffness systems such as X-Bracings, Inverted
V-Bracings (concentric) and K-Bracings (eccentric) are placed at various locations in
the building models to resist lateral loads are studied. Priti et al. in their contribution
titled Seismic Analysis of five storied building using U-shaped hybrid isolator and
lead rubber bearing isolator aim to assess the effectiveness of U-shaped isolator to
reduce the base shear and the storey drift and compare it with the results obtained
when lead rubber bearing isolator is used in a five-storey building. Considerable
number of studies have been performed previously which indicate that values of
storey shear, relative drift and acceleration play a prominent role in ensuring that the
structure is not in the range of earthquake’s resonance range [16]. Also, comparison
@seismicisolation
@seismicisolation
of other type of base isolators such as Friction Damper Devices (FDD) with U-shaped
dampers has been studied [17].
A notable contribution with respect to the risk reduction is by Jafar et al.
titled Performance Evaluation of Fire Exposed RC Structure using Pushover Anal-
ysis where displacement-controlled method is adopted and Wickstrom’s empirical
formulae are used to obtain post fire properties of the structural elements subjected
to multiple fire spread scenarios in order to determine the criticality of fire and their
seismic effects.
In the study titled Experimental and Numerical Study on Hysteretic Behaviour
of Laminated Rubber Bearing under Quasi Static Loading and its Performance on
Secondary System by Bharat, hysteretic behaviour of the laminated rubber bearing
is determined by cyclic shear and vertical stiffness test. It was observed that the
laminated rubber bearing is very much useful for the lower frequency excitations.
High temperatures can be a result of various sources such as environmental and
waste management related issues. The study by Manikanta et al. titled A study on self-
compacting concrete at high elevated temperatures discusses about consequences
of partial replacement with slag-cement by-product Ash at different percentages
subjected to varying high temperatures up to 1000 °C for up to 120 min.
Structural Health monitoring requires a real time sensing devices to avoid risks
from multiple environmental conditions. Fibre Bragg Grating sensors with the advan-
tage in being immune to electromagnetic radiations make them a key enabler to
monitor the structure and its performance and the same is presented with a demon-
stration of experimental setup in the research work titled Internet of Things Enabled
Structural Health Monitoring using Fiber Bragg Grating Sensors by Somesh et al.
Bridges, being one of the prominent parts of transportation domain, are to be moni-
tored for their structural integrity for the safety of people [18]. Real-time date with
respect to the health of the bridges is an at most necessity [19]. Various methods
and technologies such as Wireless Sensor Networks (WSNs), Acoustic Emission,
Vibration-Based Monitoring, Remote Sensing Technologies, Sensors and Instru-
mentation, Non-Destructive Testing (NDT), etc., have been considered previously to
study the structural Health of a structure [20]. The study titled Structural Health Moni-
toring of Bridges using IoT by Deepak et al. introduces IoT-based data acquisition
system that can capture strain data and temperature of bridge efficiently.
6 Summary
The aforesaid researches and applications in the domain of Civil Engineering

for Multi-Hazard Risk Reduction promise to yield advancements that continually
upgrade the standards of safety and resilience in our built environment. As the world
witnesses escalating urbanization, environmental shifts and an increasing recogni-
tion of potential risks, the role of civil engineering in safeguarding lives and property
becomes ever more critical. By taking suitable measures and approaches, we can
collectively contribute to the creation of a future wherein our structures not only
@seismicisolation
@seismicisolation
withstand the trials of time but also stand resilient against the challenges that an
unpredictable world presents.
References
1. Moreno C, Allam Z, Chabaud D, Gall C, Pratlong F (2021) Introducing the “15-Minute City”:
sustainability, resilience and place identity in future post-pandemic cities. Smart Cities 4:93–
111. https://doi.org/10.3390/smartcities4010006
2. Gopalkrishnan K, Thiruvenkatasamy S, Pabhakar E, Aarathi R (2019) Nightvision patrolling
rover navigation system for women’s safety using machine learning posted. Int J Psychosoc
Rehabil 23. ISSN: 1475-7192
3. Ungar SG, Layman R, Campbell JE, Walsh J, McKim HJ (1992) Determination of soil moisture
distribution from impedance and gravimetric measurements. J Geophys Res 97(D17):18969–
18977
4. Kisi O, Shiri J, Tombul M (2013) Modeling rainfall-runoff process using soft computing
techniques. Comput Geosci 51:108–117. https://doi.org/10.1016/j.cageo.2012.07.001
5. Lal Bahadur R, Praveen K (2022) Study of soil erosion by using remote sensing and GIS
techniques in Sone command area in Bihar, India. Mater Today Proc 62:1664–1670. https://
doi.org/10.1016/J.MATPR.2022.04.739
6. Stanier SA, White DJ (2013) Improved image-based deformation measurement in the centrifuge
environment. Geotech Test J 36(6):1–14
7. Vaseghiamiri S, Mahsuli M, Ghannad MA, Zareian F (2020) Probabilistic approach to
account for soil-structure interaction in seismic design of building structures. J Struct Eng
146(9):04020184
8. Denton F, Halsnæs K, Akimoto K, Burch S, Diaz Morejon C, Farias F, Jupesta J, Shareef
A, Schweizer-Ries P, Teng F, Zusman E (2022) Accelerating the transition in the context of
sustainable development. In: IPCC, 2022: climate change 2022: mitigation of climate change.
contribution of working group III to the sixth assessment report of the intergovernmental panel
on climate change, P.R. Shukla
9. Skea J, Slade R, Al Khourdajie A, van Diemen R, McCollum D, Pathak M, Some S, Vyas P,
Fradera R, Belkacemi M, Hasija A, Lisboa G, Luz S, Malley J (eds). Cambridge University
Press, Cambridge, UK and New York, NY, USA. https://doi.org/10.1017/9781009157926.019
10. https://www.gfdrr.org/en/publication/designing-inclusive-accessible-early-warning-systems-
good-practices-and-entry-points. Accessed on 22 Apr 2023
11. ILO (International Labor Organization) (2022) The importance of early warning systems
in disaster risk reduction. https://www.ilo.org/global/topics/employment-promotion/recovery-
and-reconstruction/WCMS_858123/lang--en/index.htm?
12. Safiuddin M et al (2018) Early-age cracking in concrete: causes, consequences, remedial
measures, and recommendations. Appl Sci 8(10):17–30
13. Oderji S et al (2019) Fresh and hardened properties of one-part fly ash-based geopolymer
binders cured at room temperature: effect of slag and alkali activators. J Clean Prod 225:1–10
14. Pradeep AR, Rangaraj C, Jayaramappa N (2022) Analysis of infilled steel frames subjected
to lateral loading. In: ICAMP-international conference organized by St. Martin’s Engineering
College Dhulapally, Secunderabad, Telangana state, India on 08th–09th July 2022 published
in Materials today conference proceedings on 23 Nov 2022
15. Sanjay HS, Pradeep AR, Kiran T, Jayaramappa N (2022) Analysis of multistorey steel structure
with different infills. Int Res J Eng Technol 9
16. Torunbalci N, Ozpalanlar G (2008) Earthquake response analysis of mid-story buildings
isolated with various seismic isolation techniques. In: The 14th world conference on earthquake
engineering, Beijing, China
@seismicisolation
@seismicisolation
17. Bagheri S, Barghian M, Saieri F, Farzinfar A (2015) U-shaped metallic-yielding damper in

building structures: seismic behavior and comparison with a friction damper. In: Structures,
vol 3. Elsevier, pp 163–171
18. Anwar N, Alnuaimi A, Alhassan M, He X (2020) A review of bridge health monitoring systems
for structural safety assessment
19. Fujino Y, Abe M, Takeda T (1996) A review of structural health monitoring techniques for
civil structures
20. Reynolds TP, Veeraragavan PS, Fuhr PL (2016) A review of fiber optic sensing technologies
for structural health monitoring of civil infrastructure
@seismicisolation
@seismicisolation
Accessibility and Convenience
@seismicisolation
@seismicisolation
Exploring Walkable Localities
as Organizing Principle of City Plan
Sumana Jayaprakash and Vimala Swamy
1 Introduction
A walkable locality is characterized by the availability of most amenities required for

daily life within the proximity of walking distance, in a medium-density neighbor-
hood of mixed-income groups and mixed land uses. The environment with parks and
public spaces and a good pedestrian design has schools, health centers, markets, and
places of worship, shopping, and other recreational facilities within walking distance
from homes. Socially people here live in a close-knit community and actively partic-
ipate in decision-making pertaining to the matters of the locality [1–3]. Man is the
center of such a spatial development of urban planning and proximity is the key
indicator that binds the locality, making urban life desirable, mutually fulfilling, and
pleasant for the citizens. It reduces the carbon footprints of the cities. The compo-
nents of the city ecosystem are efficiently managed locally for good quality of life.
Historically, this pattern of urban planning was evident across all cities globally [4].
The Industrial Revolution, with the accumulation of wealth and power in the hands
of a few capitalists, had caused enough harm to the humane pattern of communi-
ties by relocating the populace with the promise of higher wages and easy working
conditions. Trade and commerce as permanent places of production with reduced
transportation costs brought people to live together as the basis of the spatial devel-
opment called cities. The deterioration of the natural environment by establishing
S. Jayaprakash (B)
School of Architecture, REVA University, Bengaluru, India
e-mail: sj@mcehassan.ac.in
Malnad College of Engineering, Hassan 573201, India
Visvesvaraya Technological University, Belagavi, Karnataka 590018, India
V. Swamy
REVA University, Bengaluru, India
Gopalan School of Architecture and Planning, Bengaluru, Karnataka, India
@seismicisolation
@seismicisolation
14 S. Jayaprakash and V. Swamy
industries along the water bodies and polluting them went unaccounted for in the
eyes of the urban planner. Inappropriate production and consumption practices had
finally reduced the thriving communities to polluted and unhealthy cities. In addition,
the advent of the automobile system put man aside, deconstructed the humane scale
of the cities, and favored the tree-lined wide avenues and the sub-urban pattern of
city design. This invasion of the community spaces in cities by cars continues even
today.
Despite the advancement in communication, the reality that home and locality
could be the center of the organizing principle of city design became a reality in the
pandemic of 2020 [5]. In this context with the search for human-centric urban living
environments, although Christopher Alexander and Jane Jacobs had long established
pedestrian sheds as a principle of the compact city, the concept of the 15 minute city
emerged with the promise of a more local, sustainable, healthy, and equitable way
of life. It has had an international impact and seen the transformation of cities like
Paris and Barcelona and is supported by international organizations such as the
Un-Habitat, UNEP, and the C40 [6]. The 15-minute city is an intelligent city with
complete streets, multimodal transportation, and connectivity to the last mile, where
the physical and the digital environments do not oppose, but complement each other,
and technology is efficiently used and managed in the articulation of the components
of the self-sustained people-centric ecosystem to improve the quality of life. Such
socially conducive, spatially well-designed, environmentally friendly, economically
sustainable, and technologically facilitative localities enrich the daily experience of
its citizens [7]. In India, The Ministry of Housing and Urban Affairs promoted this
very concept in its Smart-city initiative for 100 cities of the country, emphasized the
regeneration of walkable localities with mixed land uses and optimal open spaces
[8, 9]. The HRIDAY initiative in particular has been successful in enhancement of
the quality of life of the communities by the regeneration of the walkable localities
of historical core areas across the country [10].
2 The Study Area: Hassan
Hassan is a small city, located in the south of Karnataka State of India on the
Bengaluru-Mangaluru National Highway, in a semi-malnad region. The evolu-
tion and growth of Hassan took place around the place named after the goddess
with a smiling face, hence called “Hasana” in the Kannada dialect. The temple of
Hasanamaba is the center of the geometrical circular growth that Hassan witnessed
before colonial rule. The historic core of Hassan aligns with the concept of a 15-
minute city as can be experienced in the Ameer Mohalla and Pension Mohalla.
The tropical climate and absence of the automobile also necessitated the compact
and humane organization of the city. With the colonial rule and the advent of the
automobile, grid iron pattern of streets, defined park areas and, the bungalow as an
independent unit of building form enclosed with compound walls emerged as the
basic unit of the city plan. The K R Puram Area and the Race Course Road are
@seismicisolation
@seismicisolation
Exploring Walkable Localities as Organizing Principle of City Plan 15
expressions of this transformation in Hassan. The distance of availability of ameni-

ties increased and there was a reduction in community ties. At present, the Building
Planning and Bye-laws are in alignment with this form of development, which is
vividly expressed in the newer residential areas of Vidyanagar and Krishnanagar.
The city is plagued by inefficient transit systems and pedestrian networks [11].
3 Methodology
3.1 Primary Data Collection of Hassan City
Data collection was carried out by on site observations and researcher understanding
of the city as a resident. The city maps and the Development Plan report were collected
from HUDA. Google Maps were used for identifying locations of amenities. CAD
software was used for extracting information from the maps.
The Conceptual Understanding of the 15-minute city as sustainable communities:
The “15-minute city” may be defined as an ideal geographical parcel of land where
the most basic needs of the citizens can be accessed within a travel distance of
15 min from homes. Although there is the use of the automobile the form and scale
of the city are independent of the automobile. It employs conceptual radii drawn
on plans of cities. The first shed level as defined by the concept of 15-min city is
a 5-min walk shed, 400 m from the center to the periphery of the imaginary circle
on the city map. This area with a population of 2600 has access to daily needs, and
has a small community shared area with some mixed land uses, small shops, and
open space. The 15-min walk shed is 1200 m from the center to the periphery of
the imaginary circle on the city map. It is assumed as the maximum distance that
most people will walk. This shed provides connectivity to the bus stations for city
level and regional access. This shed provides for the daily and weekly needs of the
citizens like groceries, pharmacies, schools, large parks, and bus stations. It also
provides efficient cycling tracks, making this shed a 5-min bicycle shed as well. The
population of this shed is assumed to be 23,000. The next level of the shed is the
15-min bicycle shed which provides connectivity to the city-level higher educational
institutions, city-level parks and open spaces, office spaces, shopping centers, and
inter-city bus stations. The population of this shed is 350,000. The radius of this shed
is 3600 m. An important indicator would be the network of pedestrian movements
well connected to ensure the workability of the concept [7]. As distance is perceived
as time in the cities, the 5-min walk shed is easily understood and supported by all
alike. The distance and time required to access city-level amenities clearly defined
by appropriate transit appeals to the user of the city, the urban and transportation
planner as well the political leadership [12].
The concept is not a mere collection of physically divided entities, but all-inclusive
places where neighbors, workers, shopkeepers local store owners know each other
and there is social cohesion [13]. The areas are reflective of local practices of food,
@seismicisolation
@seismicisolation
cultural, and traditional settings. Digital and other technologies are used for connec-
tivity and building a restorative relationship to knit ties with families, community,
and nature. Such communities are resilient and able to reduce the impact of the
shocks and stresses from climate change and other global conflicts. They may not
be materialistically wealthy but will be safe and marked by fearlessness [14].
Issues of concern in the present context: Hassan’s local planning area divided into
eight districts, designated for planning purposes specified the areas designated for
residential, public-semi-public, parks and open spaces, industrial, commercial, and
transportation. They also included the agricultural areas of the local planning area.
The residential areas were not demarcated as convenient and equal-sized localities;
they were interspersed with public, commercial, and even industrial activities. As
there were no designated public zones that accommodated city-level spaces, the
new proposals inevitably made their way into the planning districts indiscriminately.
There were no urban design guidelines for the form-based design of the city. To
evaluate the neighborhood public spaces, appropriate divisions of neighborhoods did
not exist. Although the city plan was divided into areas called nagars or pura, they
were not of equal extent, and could not be considered for evaluation. These areas were
not the sub-divisions of the planning districts as some of them extended across the
planning districts. It was observed that the Rangolihalla, a residential area, extended
across planning districts 7 and 8. The original village profile had not been retained;
some of them were partially integrated into newly created nagars. Alternatively, the
present divisions of wards were limited to municipal areas, utilized for electoral
purposes, urban administration, and other matters. They also varied in size based
on the population. The areas outside the municipality but within the local planning
area were under the gram panchayats. There were no urban design guidelines for
the clear geographical demarcation of the residential localities. The building bylaws
specified the guidelines for individual plots only. Hence to evaluate the availability of
amenities in the neighborhood, the process of geographically dividing the city plan
into neighborhoods was carried out. The overlaying of the 15-min city plan on the
existing plan of Hassan was done as defined by Carlos Moreno with the assumption
that the historical medium-sized city of Hassan may have an underlying structure of
a walkable city, damaged by recent developments and an increase in population.
3.2 Study and Analysis of the Present Conditions
The existing map of Hassan was analyzed for its location, inter-city connectivity,
distribution of city-level institutions and public administrative offices, historic zones,
environmental features, and other city and regional-level functions. A zone was
demarcated and designated as the public zone, which represented the city-level transit
movement corridor with connectivity to the residential areas. Having demarcated this
15 km long and 0.5 km wide, zone as the public level zone, the 15-minute city plan
was overlaid on the remaining residential areas as shown in Fig. 1.
@seismicisolation
@seismicisolation
Fig. 1 Methodology (source author’s work)
Documentation of the various amenities available in the neighborhood: After the

designation of the public zone, the rest of the predominantly residential areas were
earmarked as neighborhoods and the 15-min city plan was superimposed on the city
map of Hassan as shown in Fig. 2. By employment of a conceptual radius of 400 m,
the pedestrian shed was defined as a locality, in the adapted model of the 15-min
city for the context of Hassan. Seven such localities were defined to constitute a
neighborhood as shown in Fig. 3. The edges of the neighborhood area were defined
as mixed-use areas and connected to the city-level public spaces. A total of seven
neighborhoods were identified. Each neighborhood had an area of 224 ha and was
subdivided into seven localities, each with an area of 32 ha. The density of population
was assumed a 100 per ha as mentioned in the development report of Hassan.
3.3 Data Analysis
Evaluation of the Neighborhood Localities on the Overlaid 15-min City Map.

The list of amenities were decided as per the standards in the Annexure 2 of the
URDPFI guidelines [15]. The work was done through site visits and by using Google
Maps. The tabulated results were subdivided into essential, recreational, convenience,
@seismicisolation
@seismicisolation
Fig. 2 Mapping of the “15-minute city” plan (source HUDA and author work)
Fig. 3 15-min city plan overlaid on the city map of Hassan (source HUDA and author work)
movement, and character, counted, and compared to the standards The proposed
public zone contained most of the city-level public functions and the transit corridor.
The data analysis revealed as shown in Fig. 4 that zone seven had excess availability
of social infrastructure and amenities. In the absence of clearly demarcated city-
level public spaces, this zone was performing the function of city-level public space
without being designed for it. It also suggested that neighborhood seven was more of
a city-level public space than a neighborhood. Zone 3 seemed to inaccessible place
for essential amenities and recreational facilities.
3.4 Proposal of Neighborhood Structure Plan for Hassan
The existing city map of Hassan had the transportation line, the Mangaluru–
Bengaluru national highway that ran across the city in the east-west direction. The
State highway ran along the south-north direction toward Tejur in the north. There-
fore, they were demarcated as city-level transportation lines. The commercial, public,
and semi-public activities had extended along these transportation lines. It was found
that the city-level transportation lines were located on the geographical ridge lines.
@seismicisolation
@seismicisolation
20
standards
15 Zone 1
10 Zone 2
Zone 3
5
Zone 4
0 Zone 5
Zone 6
Zone 7
Fig. 4 Availability of amenities in the proposed division of neighborhoods (source author work)
Tracking the lines along these transportation lines along the lakes and valleys, a
green zone consisting of the large lakes, the contextual natural feature of Hassan
were mapped as a continuous corridor along the transit lines parallel to the trans-
portation corridor and this emerged as the city-level public space. The valley along
the east-west axis from Haluvagilu to Bhuvanalli Lake together with the crescent
shaped valley encircling the city and connected to this valley line was earmarked as
public space. It contained in itself the Hunsinkere, the Tejur kere, and the Satyaman-
gala kere and served as a recreational and public zone of the city-level public space.
The city-level transportation lines and the city-level public spaces were thus defined
as seen in Fig. 5.
Having earmarked the city-level transit zone and the city-level public and recre-
ational zone, for the basic unit of neighborhood locality, the 400-m radius area was
considered based on the 15-minute city concept. The conceptual circular shape was
redefined as a square and the neighborhood plan was contextualized with the existing
lines of the planning districts, as shown in Fig. 5.
Within the framework of the existing land uses, the division of urban area as form-
based city-level and neighborhood-level spaces was proposed as shown in Table 1.
The historic core and the hinterland were neighborhoods which were defined as
sensitive zones to be worked upon with special urban policies.
Characteristics of the neighborhood spaces/ methodology for transformation
• Definition of the baseline benchmarking of the neighborhood for proximity and
accessibility to the essentials, convenience, recreational, transit system, etc.
• People participation in decision-making with decentralized administration.
• Low-density development and compact mix-use developments with the avail-
ability of everyday needs within the locality within walking distance within a
radius of 400 m. Location of primary social amenities within the neighborhood.
• Planning principles on three-dimensional building forms and participatory plan-
ning with the presence of parks/ open spaces for everyday recreation, convenience
shopping, and primary health amenities within the neighborhood.
@seismicisolation
@seismicisolation
Fig. 5 Proposed structure plan for neighborhoods of Hassan (source author work)
Table 1 Proposed parameters of city (HUDA and author work)

Land use analysis Area (ha) % Area divisions Area (ha) %
Land use 2930.14 Historic core 128 × 2 = 256 46.57
Residential 182.14 43.4 Neighborhoods 128 × 25 = 3200
Commercial 384.97 2.71 Neighborhood level 100 × 15 = 1500 20.21
Public spaces
Industrial 529.15 5.71 City-level public 1500 20.21
spaces
Public and semi public 576.71 7.84 City-level transport 966 13.21
lines
Park and open spaces 12.02 8.55 Total 7420
Public utility 2131.38 0.18 Hinterland 1790
Transportation 6747.51 31.5 Total 9210
Total 689.11 100
Water bodies 1790.38
Agriculture 9227
Total
@seismicisolation
@seismicisolation
• Innovative financial and regulatory mechanisms to enhance the public–private

partnerships in restoring the characteristic features of the neighborhood enhanced
by contextual features like the presence of water bodies, old temples, etc.
• Defined cycling and walking tracks within the neighborhood and an expansion of
green space at various scales of spaces available.
• Relocation of incompatible functions. Encouragement of infill development with
tools such as vacant land taxation. Adaptive reuse of buildings. The unused and
underutilized land to be used for temporary activities. Ground floor activities in
specific areas are to be incentivized by tools such as density bonus.
4 Discussions and Conclusions
The existing planning districts were divided into seven neighborhoods of 128 ha each.
27 such neighborhoods were proposed for Hassan city, with two of them designated
as Historic core areas. Each of the neighborhoods was divided into seven localities
of 16 ha each encircled by a zone of neighborhood public spaces. This was done
after the designation of areas of hinterland, city-level public spaces and transporta-
tion lines. The proposed division of the city plan into convenient neighborhoods
qualified for socio-economic equality; accessible daily needs of people with limited
use of automobiles. Due to the convenient location of services, transportation needs
were minimized; therefore reduced fuel consumption mitigated global warming.
Cycling and walking improved the health and well-being of the citizens, created
better quality of air, saved money, and improved the quality of life. The assumed
average density of housing was defined as 100 persons per hectare. The concept
evolved out of contextual environmental and social factors and was instrumental in
guiding the future growth of the city. The inclusion of surrounding villages into the
city should also be done keeping the concept of walkable localities as the organizing
principle of the city plan. This research work has contributed to the methodology of
understanding and analyzing an urban context, in light of the 15-minute city. It has
defined a methodology to overlay the 15-minute city plan on an existing city plan and
assess the adequacy and proximity of amenities. It has proposed spatial definition of
neighbor hoods as well as how a city can transform into walkable localities, in terms
of proximity to amenities, defined transit movement and experience of urban city
spaces. The proposed structure plan at the neighborhood level for the city of Hassan
assured the quality of neighborhood spaces in the city, and provided a clear strategy
for new interventions and guidelines for urban transformations which will go a long
way in the creation of sustainable communities with net zero targets.
@seismicisolation
@seismicisolation
References
1. Garau C, Pavan VM (2018) Evaluating urban quality: indicators and assessment tools for smart
sustainable cities. Sustainability 10(3). https://doi.org/10.3390/su10030575
2. Streets for people challenge (2021). [Online]. Available: https://smartnet.niua.org/indiastreetc
hallenge/
3. Correa-Parra J, Vergara-Perucich JF, Aguirre-Nuñez C (2020) Towards a walkable city: prin-
cipal component analysis for defining sub-centralities in the Santiago metropolitan area. Land
9(10):1–15. https://doi.org/10.3390/land9100362
4. Jayaprakash V, Swamy S. In: Nandagiri L, Narasimhan MC, Marathe S (eds) Recent advances
in civil engineering. Lecture notes in civil engineering, vol 256. Springer, Singapore
5. Pozoukidou G, Chatziyiannaki Z (2021) 15-minute city: decomposing the new urban planning
Eutopia. Sustainability 13(2):1–25. https://doi.org/10.3390/su13020928
6. Sharifi A, Murayama A (2013) A critical review of seven selected neighborhood sustainability
assessment tools. Environ Impact Assess Rev 38:73–87. https://doi.org/10.1016/j.eiar.2012.
06.006
7. Moreno C, Allam Z, Chabaud D, Gall C, Pratlong F (2021) Introducing the ‘15-Minute City’:
sustainability, resilience and place identity in future post-pandemic cities. Smart Cities 4(1):93–
111. https://doi.org/10.3390/smartcities4010006
8. Swamy V (2020) Smart keys for smart eco-friendly cities. BUUKS
9. Shenvi A, Slangen RH (2018) Enabling smart urban redevelopment in India through floor area
ration incentives. ADB South Asia working paper series, vol 58, no 58. [Online]. Available:
https://doi.org/10.22617/WPS189452-2
10. HRIDAY National Project Management Unit (2016) Heritage city development and augmen-
tation Yojana, p 44. [Online]. Available: http://hridayindia.in/
11. Report of Master Plan (Final) for Hassan Planning Area (2021)
12. Han Z, Long Y, Wang X, Hou J (2020) Urban redevelopment at the block level: methodology
and its application to all Chinese cities. Environ Plan B Urban Anal City Sci 47(9):1725–1744.
https://doi.org/10.1177/2399808319843928
13. Nair AK, Basu S (2016) Regeneration of a mixed use area in the historic core of a city—a case
study of ‘Chala’ in Thiruvananthapuram, Kerala. J Herit Manag 1(1):35–58. https://doi.org/
10.1177/2455929616638776
14. Gandhi MK (1968) An autobiography or the story of my experiments with truth (trans: Desai
M, Narayan S). Navjivan Publishing House, Ahmedabad, India. ISBN: 81-7229-278-3
15. URDPFI Guidelines: Ministry of Urban Development. [Online]. Available: http://mohua.gov.
in/link/urdpfi-guidelines.php
@seismicisolation
@seismicisolation
Prediction of Noise Pollution of Delhi City
Using Machine Learning: A Case Study
Rajashri Khanai, Rajkumar Raikar, and Mrutyunjay Uppinmath
1 Introduction
Noise pollution that includes excessive unwanted sound is considered as the third
most hazardous pollution after water and air. It has many harmful effects on human
health, wildlife, and environmental quality [1]. Sources of noise pollution are due
to vehicular and air traffic, neighborhood, industrialization, machineries, public
addressing systems, loudspeakers, firecrackers, and many more. Noise pollution
affects human health both physiologically, psychologically, and also on quality of
life [2]. Noise pollution also has an adverse effect on birds and animals. Hence, it
became essential to understand the trend of noise pollution from a particular cause
and analyze it for future effect. Ahirwar and Bajpai [1], Dhanwate [2], Chauhan et al.
[3], and Yadav et al. [4] have measured the noise levels during and after Deepawali
festival using instruments and compared them with the Central Pollution Control
Board (CPCB) levels. The noise pollution levels during Deepawali festival of Raipur
city measured by SLM-100 (made by Envirotech) meter were compared with the
noise pollution rules, 2000 by Ahirwar and Bajpai [1] and they reported that the
noise levels were higher than the CPCB levels and recommended for ecofriendly
celebration and ban of firecrackers [3].
Dhanwate [2] compared the noise levels of Nashik city’s Ravivar Karanja using
sound level meter YF-20 and found that they are generally higher. Similarly, noise
level in the Chittorgarh city of Rajasthan during Deepawali festival was monitored
by Chauhan et al. [3] and Gorakhpur city by Yadav et al. [4]. However, these all
studies refer to the comparison of noise levels with CPCB standards. No attempts
are made toward the application of machine learning (ML) algorithms in modeling
and prediction of noise pollution particularly pre and during Deepawali festival.
During the festival days noise level considerably increases in general, because of
R. Khanai (B) · R. Raikar · M. Uppinmath

KLE Technological University, Belagavi Campus, Belagavi, India
e-mail: rajashrikhanai@klescet.ac.in
@seismicisolation
@seismicisolation
24 R. Khanai et al.
the vehicular movement, burning of fire crackers, sound systems. In this regard,
the present study is conducted to analyze the noise pollution levels pre- and post-
Deepawali festival.
On the other hand, machine learning algorithms are a computational process that
uses input data to accomplish a desired task without being literally programmed to
achieve a specific result. In a sense, these algorithms are soft coded, with iteration
automatically changing or adapting their architecture to better perform their intended
task. The adaptation process is called training, and this training is the learning part
of machine learning [5, 6]. It incorporates ideas from various fields such as infor-
mation theory, computer science, probability and statistics, psychology, artificial
intelligence, philosophy, and control theory. The application of ML is diversified.
Some examples on the application to air pollution is after Hu et al. [7] and Guan
and Sinnott [8]. They used different techniques such as supportive vector regres-
sion (SVR), extreme gradient boosting (XGB), multilayer perceptron (MLP), linear
regression (LR), random forest regression (RFR), decision tree regression (DTR),
adaptive boosting regression (ABR) and long short-term memory neural network
(LSTM) to mention. In this study, the noise-level data were collected from Lajpat
Nagar, Mayur Vihar-II, Kamla Nagar, Pitam Pura locations in Delhi and analyzed
for the future prediction using machine learning techniques.
The growing danger of non-communicable diseases is highlighted by the most
recent Global Burden of Disease report. The current work’s non-auditory impacts
study adds to the risk factors for non-communicable diseases. Therefore, the problem
of noise pollution and the related health hazards, particularly for the vulnerable popu-
lation, must be addressed [9]. It is crucial to develop a PM2.5 concentration fore-
casting model with high precision for its monitoring and control due to its harmful
impacts on the environment and human health. The study offers a unique hybrid
model based on least squares support vector machine (LSSVM) and principal compo-
nent analysis (PCA), with cuckoo search (CS) optimization [10, 11]. The analysis
of the literature revealed that the respondents’ main source of discomfort is noise
from the road. Studies concentrating on effects on health provide the majority of the
evidence. Only 10% of articles listed noise pollution by geographic region. With the
aid of a questionnaire instrument, 44.89% of articles provided details on subjective
response data, while 14.3% of articles provided information about workplace noise in
various parts of India [12]. In order to continuously measure noise levels throughout
the year, the study discusses a pilot effort to construct the National Ambient Noise
Monitoring Network (NANMN) throughout seven important Indian cities [13]. Here
the yearly average Lday (06–22 h) and Lnight (22–06 h) values that have been
measured over the last three years. The 35 places under study—14 in a commercial
zone, 5 in an industrial zone, 7 in a residential zone, and 9 in quiet zones—show that
ambient noise levels have somewhat increased during the past three years. Methods
used to predict air pollution traditionally have drawbacks. One method that can open
up new possibilities for air pollution prediction is machine learning. However, there
are numerous machines learning techniques, making it difficult to determine which is
appropriate for a given issue. In this study, data on air pollution, specifically PM2.5,
were gathered from a number of web-based resources. After data cleansing, various
@seismicisolation
@seismicisolation
Prediction of Noise Pollution of Delhi City Using Machine Learning … 25
machine learning models, including linear regression, artificial neural networks, and
long short-term memory recurrent neural networks, were used to analyze the results
[14].
2 Study Area
In the present study, four places of Delhi city, namely Lajpat Nagar, Mayur Vihar-II,
Kamla Nagar, and Pitam Pura are considered. Delhi is the largest metropolitan city
with its latitude is 28° 70' 41'' N, longitude 77° 10' 25'' E. Delhi city resides on the
bank of river Yamuna having an area of 1483 km2 and with 200–250 m above MSL.
The average annual rainfall of city Delhi is about 653.6 mm. The coolest month is
January with an average low temperature of 8.33 °C and hottest month is May with
average high temperature is about 45 °C.
3 Methodology
The noise-level data of Deepawali and pre-Deepawali days of four places of Delhi
were initially analyzed using classifier learner and regression learner on MATLAB
platform. Due to the small amount of data, the predictions were not promising. Hence,
Python algorithms were employed to build the models. In current study, different
regression algorithms were employed. They are linear regression, Gaussian Naïve
Bayes Regression, Multinomial Naïve Bayes Regression, Bayesian Ridge Regres-
sion, Bayesian Ridge Regression, Logistic Regression, and Quantile Regression.
These algorithms were run in the Google Colaboratory. The data available were only
72 in number. By using random number technique, random noise pollution data were
generated up to 150 and 300. Using these three datasets, the prediction is performed.
4 Results and Discussions
The results of seven regression models used in the analysis are presented in this
section.
The noise pollution data and generation of random noise pollution data
The spatial noise pollution data of pre-Deepawali and Deepawali collected from the
CPCB for four locations in Delhi city (Lajpat Nagar, Mayur Vihar-II, Kamla Nagar,
and Pitam Pura) given in Table 1 were used to generate the random data. Table 2
gives the average values of noise pollution and respective highest and lowest values
data of four locations.
@seismicisolation
@seismicisolation
26
Table 1 Noise pollution data (values in dB)

Year Lajpat Nagar Mayur Vihar-II Kamla Nagar Pitam Pura
Pre-Deepawali Deepawali Pre-Deepawali Deepawali Pre-Deepawali Deepawali Pre-Deepawali Deepawali
2004 66 74 67 87 69 79 59 82
2005 62 78 71 89 62 77 74 77
2006 65 73 67 85 69 76 59 80
2007 66 71 62 87 71 78 69 –
2008 66 76 53 85 64 76 57 76
2009 61 75 66 82 62 74 58 74
2010 63 76 63 85 56 81 58 71
2011 67 81 79 91 58 81 58 75
2012 58 81 48 73 61 80 56 75
2013 56 79 60 83 63 81 61 73
2014 – – 69 83 59 80 53 71
@seismicisolation
@seismicisolation
2015 61 76 60 79 61 86 55 74
2016 55 76 59 86 60 74 43 –
2017 62 74 62 75 61 69 55 69
2018 62 69 68 74 59 74 43 75
2019 63 69 63 68 62 70 49 68
2020 60 71 60 72 62 70 55 57
The temperature drops below 10 °C in winter season. Table 1 furnishes the noise pollution data of the four places obtained from Central Pollution Control Board
(CPCB) during 2003 to 2020
R. Khanai et al.
Table 2 Average and range of noise pollution at four location (values in dB)
Locations Pre-Deepawali Deepawali
Average Lowest Highest Average Lowest Highest
Lajpat Nagar 62.29 55 67 75.76 71 89
Mayur Vihar-II 63.11 48 79 81.39 72 91
Kamla Nagar 62.61 56 71 76.89 69 86
Pitam Pura 56.59 43 74 68.56 57 82
Firstly, the average is calculated for the dataset. Then the lowest and highest
numbers were identified in the dataset and used in the generation of the random
datasets. Two sets of data were created using random numbers. The results of the
regression models are discussed in the below section.
4.1 Linear Regression
The linear regression model was initially used to analyze noise pollution data. The
coefficient of determination (R2 ) value computed was 0.45 as given in Table 3,
which is very low for the prediction model. To check whether the result was correct
or not, the other two random datasets generated were tried to analyze the regression
model. The second dataset contains 150 number of data points, and the third dataset
contains 300 number of the data point. The R2 value for both sets is 0.56 and 0.52,
respectively, as in Table 3. Further, from Fig. 1, the regression line shows wide
scattered data indication poor linear regression. Hence, it can be noted that Linear
Regression model is not suitable for the prediction of noise pollution data in this
study.
4.2 Gaussian Naïve Bayes Regression
The noise pollution data were fed in the Gaussian Naïve Bayes Regression to check
the level in the prediction of the noise pollution data. Initially the original dataset
was used. The coefficient of determination values for 72 data is 0.39 as given in
Table 3, which is too low. To compare the results, another two sets of data were
used. The second dataset has 150 number of data points, and the third dataset has
300 number of data points. The coefficient of determination (R2 ) values for the
second and third datasets are 0.34 and 0.22, respectively, as furnished in Table 3.
This indicates Gaussian Naïve Bayes Regression failed to predict accurately the noise
pollution. The Gaussian Naïve Bayes Regression depends on the Bayes theorem, and
it is dependent on the Gaussian distribution (or normal distribution). As shown in
@seismicisolation
@seismicisolation
28 R. Khanai et al.
Table 3 Coefficient of determination of different regression models

Sl. Regression Regression coefficient Model comparison Remarks
No. models Original Random data
data (72) (150) (300)
1 Linear 0.45 0.56 0.52 The data shall follow linear Not
regression regression conditions suitable
2 Gaussian 0.39 0.34 0.22 The data shall fit Gaussian Not
Naïve Bayes distribution suitable
regression
3 Multinomial 0.153 0.13 0.16 It depends on the Not
Naïve Bayes multinomial distribution suitable
regression which is popular in natural
language processing
4 Bayesian − 0.0378 − − It uses probability Not
ridge 1.019 0.201 0.035 distribution rather than point suitable
regression 1.036 1.06 distribution
5 Huber 0.2426 − − It is dependent on outliers Not
regression 0.019 0.031 suitable
6 Logistic 0.091 0.043 0.19 It depends on unconditional Not
regression distribution or continuous suitable
function
7 Quantile 0.792 0.803 0.801 It depends on conditional Suitable
regression distribution or discontinuous
function
Fig. 1 Linear regression

plot
Fig. 2, the dataset is clustered and not following Gaussian distribution. Hence, noise
pollution prediction is not possible with the Gaussian Naïve Bayes Regression.
@seismicisolation
@seismicisolation
Fig. 2 Gaussian Naïve

Bayes regression plot
4.3 Multinomial Naïve Bayes Regression
As the Gaussian Naïve Bayes Regression is not found to satisfy with the results, the
multinomial Naïve Bayes Regression is employed in the analysis. The coefficient
of determination (R2 ) value for the original dataset of 72 numbers came as 0.153.
Further, the other two datasets were tested. The R2 value obtained is 0.13 and 0.16, for
150 and 300 data, respectively. The value is too low to use as a model for the prediction
of noise pollution data, hence not considered. Figure 3 shows the graph of multino-
mial Naïve Bayes Regression. The Multinomial Naïve Bayes Regression model also
depends on the Bayes theorem and depends on the multinomial distribution. Hence
not supporting the noise prediction.
Fig. 3 Multinomial Naïve

Bayes regression plot
@seismicisolation
@seismicisolation
30 R. Khanai et al.
Fig. 4 Bayesian ridge

regression
4.4 Bayesian Ridge Regression
In Bayesian Ridge Regression, there are two limits, upper limit and lower limit. The
upper limit is 97.5% and the lower limit is 2.5%, where the results will be of two
values because of the limit. Initially, the first set of 72 numbers of data (Table 1)
is used in the analysis. The coefficient of determination (R2 ) value appeared as −
0.0378, 1.019. To check the accuracy of the model, other two datasets were used. The
second dataset and the third dataset gave value of the coefficient of determination
(R2 ) as − 0.201, 1.036 and − 0.035, 1.06, respectively. These values are relatively
low. Figure 4 presents the graph of the Bayesian Ridge Regression. The Bayesian
Ridge Regression uses probability distribution rather than point distribution. Hence,
the Bayesian Ridge Regression is not suitable in the prediction of noise pollution
data.
4.5 Huber Regression
The first dataset of 72 numbers of data points from Table 1 is tested in the Huber
Regression algorithm. The coefficient of determination (R2 ) value came after running
the program as 0.243. To check model accuracy, the other two more datasets 150 data
points and 300 number of data points were used in the analysis. The respective value
of the coefficient of determination (R2 ) for the second and third sets is − 0.019 and −
0.031, respectively. These values were not acceptable. Hence, this method is not used
for the prediction of noise pollution. Figure 5 shows the Huber Regression graph.
As the Huber Regression depends on the outliers, and due to outlier dependency, the
model failed to give the expected outcome.
@seismicisolation
@seismicisolation
Fig. 5 Huber regression plot
4.6 Logistic Regression
Initially, the original dataset of 72 numbers given Table 1 is employed in the

Logistic Regression algorithm. The coefficient of determination value comes as
0.091. Further, with the other two sets of data such as second dataset of 150 data
and the third set of 300 data points, the coefficients of determination (R2 ) values
are 0.043 and 0.19, respectively. This method also fails to predict the noise pollu-
tion data because the Logistic Regression depends on the unconditional distribution
or continuous function, hence, found not suitable. Logistic Regression variation is
presented in Fig. 6.
Fig. 6 Logistic regression

plot
@seismicisolation
@seismicisolation
32 R. Khanai et al.
Fig. 7 Quantile regression

plot
4.7 Quantile Regression
As the linear regression conditions were not met, Quantile Regression is tried. Quan-
tile Regression is an extension of linear regression. The first dataset of 72 data points
given in Table 1 is fed in the Quantile Regression algorithm. The coefficient of deter-
mination (R2 ) value came as 0.792. To check whether the value came is correct or
not, two more sets were used in the analysis. The second dataset consists of random
150 data points and the third set with 300 data points were used in the analysis.
The corresponding value of the coefficient of determination (R2 ) is 0.803 and 0.801,
respectively, as given in Table 3. From this result, it is found that the Quantile Regres-
sion is best suited in the prediction of noise levels. As Quantile Regression model
depends on the conditional distribution or discontinuous function, it fitted very well
to the data of present study. Figure 7 shows the graph of the Quantile Regression.
5 Conclusions
In the present study, machine learning is used for the prediction of noise pollution of
the four locations of Delhi city, namely Lajpat Nagar, Mayur Vihar-II, Kamla Nagar,
and Pitam Pura. Regression algorithms were employed to build the prediction model
for noise pollution. The regression algorithms used in the study are linear regression,
Gaussian Naïve Bayes Regression, Multinomial Naïve Bayes Regression, Bayesian
Ridge Regression, Huber Regression, Logistic Regression, and Quantile Regression.
Initially, the original dataset of 72 points collected from Central Pollution Control
Board (CPCB) during 2003–2020 is used in the model. Further, two sets of random
data are generated with 150 dataset and 300 dataset, which are used for analysis. All
the datasets were used in regression algorithms to get the best regression models for
the prediction.
@seismicisolation
@seismicisolation
It is found that, out of all regression models, the Quantile Regression model
fits very well with the present data and is suitable for prediction of noise level.
The R2 -Value for original data (72 numbers) and two random data (150 and 300
numbers) is found to be 0.792, 0.803, and 0.801, respectively. Hence, it is suggested
to use Quantile Regression for future prediction of noise levels at the study locations.
However, at other location, the suitability of a particular regression model can be
determined and recommended.
References
1. Ahirwar AV, Bajpai S (2015) Assessment of noise pollution during Deepawali festival in
Raipur city of Chhattisgarh, India. In: International conference on chemical, environmental
and biological sciences (CEBS-2015), 18–19 Mar, Dubai
2. Dhanwate SV (2017) Study of noise level at Ravivar Karanja in Central Part of Nashik City
during normal days and Diwali festival days. Int J Res Anal Rev 4(1):774–780
3. Chauhan GS, Wani S, Sharma V (2020) Comparative analysis of noise pollution during
Deepawali festival in Chittorgarh City of Rajasthan. Compliance Eng J 11(3):294–300
4. Yadav M, Patel RK, Yadav A, Sharma G, Pandey G (2021) Assessment of noise pollution at
various locations of Gorakhpur. Int J Eng Sci Technol 13(1):131–137
5. Naqa I, Murphy MJ (2015) Machine learning in radiation oncology: theory and application.
Springer International Publishing, Switzerland, pp 3–11
6. Rebala G, Ravi A, Churiwala S (2019) An introduction to machine learning. Springer Nature,
Switzerland, p 263
7. Hu K, Rahaman A, Bhrugubanda H, Sivaraman V (2017) HazeEst: machine learning based
metropolitan air pollution estimation from fixed and mobile sensors. IEEE Sens J 17(11):3517–
3525
8. Guan Z, Sinnott RO (2018) Prediction of air pollution through machine learning approaches
on the cloud. In: IEEE/ACM 5th international conference on big data computing applications
and technologies (BDCAT), pp 51–60
9. Khaiwal R et al (2016) Assessment of noise pollution in and around a sensitive zone in North
India and its non-auditory impacts. Sci Total Environ 566:981–987
10. Garg N, Maji S (2016) A retrospective view of noise pollution control policy in India: status,
proposed revisions and control measures. Curr Sci 29–38
11. Sun W, Sun J (2017) Daily PM2.5 concentration prediction based on principal component
analysis and LSSVM optimized by cuckoo search algorithm. J Environ Manag 188:144–152
12. Goswami S, Swain BK (2017) Environmental noise in India: a review. Curr Pollut Rep 3:220–
229
13. Garg N et al (2016) A pilot study on the establishment of national ambient noise monitoring
network across the major cities of India. Appl Acoust 103:20–29
14. Sinnott RO, Guan Z (2018) Prediction of air pollution through machine learning approaches on
the cloud. In: 2018 IEEE/ACM 5th international conference on big data computing applications
and technologies (BDCAT). IEEE
@seismicisolation
@seismicisolation
Smart Water Management: Using
Machine Learning to Analyze Water
Quality Index
B. K. Monnappa, B. M. Shiva Kumar, T. S. Pushpa , and S. Shilpa
1 Introduction
Water quality is the overall state or status of water, which encompasses its chem-
ical, physical and biological attributes or properties. It describes the composition and
condition of water based on various parameters that affect its suitability for different
uses [1]. Traditional water quality monitoring methods involve manual testing of
water samples, which can be time-consuming, costly and not always accurate. Moni-
toring the quality of water poses a challenging task due to the intricate analysis of
a multitude of variables and their high variability caused by human activities and
natural factors [2]. Several studies in India have highlighted the significance of water
quality indices (WQI) as a measure to evaluate water quality, which was originally
proposed by Horton in 1965 [2].
WQI can be evaluated on the basis of various physical, chemical and bacteriolog-
ical parameters [3]. Physical analysis methods are used to determine transparency,
color, temperature, turbidity and odor of water. Chemical analysis methods involve
analyzing various parameters like EC, pH, BOD, COD, fluorides, TDS, TSS, hard-
ness, nitrates, phosphates, sulfates, chlorides, DO, free CO2 , potassium and sodium
[4]. Biological monitoring methods focus on the qualitative analysis of planktons
[5].
B. K. Monnappa (B) · B. M. Shiva Kumar · T. S. Pushpa · S. Shilpa

Department of Computer Applications, B.M.S. College of Engineering, Bengaluru, Karnataka,
India
e-mail: monnappabk.mca21@bmsce.ac.in
B. M. Shiva Kumar
e-mail: shivakumarbm.mca21@bmsce.ac.in
T. S. Pushpa
e-mail: pushpa.mca@bmsce.ac.in
S. Shilpa
e-mail: shilpa.mca@bmsce.ac.in
@seismicisolation
@seismicisolation
36 B. K. Monnappa et al.
Machine learning has emerged as a promising approach for water quality predic-
tion, leveraging historical water quality data to predict future quality levels of water
[6]. WQI is a mathematical technique that can effectively summarize a vast amount
of water quality data into a single number, providing a comprehensive assessment
of the quality of water [3]. Water quality prediction is a topic of extensive research.
The values for each variable must then be determined to be acceptable and unac-
ceptable [7]. Water is deemed suitable for a given purpose when it complies with the
established requirements. In this paper, machine learning-based approach for water
quality prediction to segregate its purpose of uses such as drinking, agriculture and
industrial purpose and evaluate the performance of several popular machine learning
algorithms.
2 Literature Survey
The various models that have been developed to predict the quality of water. Some of
the important models include decision trees (DT), support vector machines (SVM),
artificial neural networks (ANNs) and multiple linear regression (MLR) [8].
Water quality prediction techniques: This area covers the various techniques that
have been used to predict the quality of water. Some of the key techniques include
machine learning, statistical analysis, data mining and fuzzy logic [9].
Data sources: This area covers the different sources of the data that are used to
predict the quality of water. Some of the key data sources include satellite data,
re-mote sensing data, weather data and historical water quality data.
Applications: This area covers the different applications of water quality predic-
tion systems. Some of the key applications include water treatment plants, water
resource management and water quality monitoring.
Table 1 summarizes the literature survey done on few papers that are most relevant
to Smart Water Management paper and implementation methods.
3 Related Works
Studies have explored the need of machine learning techniques to predict quality of
the water. For example, Kang et al. [11] proposed a method on random forest for
predicting water quality, achieving high accuracy [11]. Similarly, support vector
regression is used to predict water quality in a river system, achieving good
performance [12].
However, these studies focuses on individual machine learning algorithms and
do not explore the potential benefits of ensemble methods for water quality predic-
tion. “Water quality prediction using machine learning techniques: A review” by
Sharma et al. [9] provides comprehensive review on predicting the water quality
using machine learning techniques. The study highlights the need for appropriate
@seismicisolation
@seismicisolation
Smart Water Management: Using Machine Learning to Analyze Water … 37
Table 1 Literature survey

References Author Title of the Area of Algorithm Outcome
name paper examination
[5] Ashwini Intelligent IoT and ML-based K-nearest Minimizing human
et al. Model for system neighbor intervention and
Predicting ensuring
Water Quality affordability
[6] Prasad Smart Water IoT and wireless None Effectively
et al. Quality sensors are used to triggering alerts
Monitoring upload water based on parameter
System (2015) quality data to the thresholds to
internet prompt immediate
action
[10] Koditala Water Quality IoT, ML and cloud None Low-cost sensors
et al. Monitoring computing are used for water
System Using quality assessment
IoT and ML
feature selection, model selection, and hyper parameter tuning for accurate predic-
tions. “Water quality prediction using artificial neural network: A case study of the
Ganga River, India” (2022) by Bhardwaj, R., and Singh, R. K, proposes an artifi-
cial neural network-based approach for water quality prediction and achieving high
accuracy. The study highlights the importance of feature selection and the need for
an adequate amount of data for accurate predictions [13].
4 Methodology
Figure 1 demonstrates the process flow of the system in determining the water quality
index [9].
1. Data collection and preprocessing: Collecting accurate and representative data
is the first step. The dataset is taken from Kaggle. The data includes various water
properties parameters, such as dissolved oxygen, pH, conductivity, biochemical
oxygen demand, nitrate level, temperature and coliform measured at different
time intervals and different locations [14]. The data is preprocessed by removing
missing values and outliers, and normalized the data to ensure consistent scaling
across all features [12]. This is done to increase the accuracy of the ML model.
2. Feature selection and engineering: Feature selection is done to identify the most
relevant features for predicting the water quality. Used several methods, including
correlation analysis, mutual information and feature importance ranking, to select
the top features. After selecting the features, select the appropriate machine
learning algorithm for the prediction task. Some ML algorithms used for
predicting the water quality include decision trees, linear regression, random
forests and XG Boost regression [15].
@seismicisolation
@seismicisolation
Fig. 1 Process flow diagram of the system
3. Model training and validation: Several popular machine learning algorithms,

including linear regression, decision tree is trained on the preprocessed and engi-
neered data. The ML model is then trained using a portion of the data, and its
performance is evaluated using the remaining data. This is done to confirm that
the model is not overfitting and results in accurate predictions on new data.
4. Model evaluation: Evaluate the performance and accuracy of each ML model is
evaluated using different metrics such as Root Mean Squared Error (RMSE) and
Mean Squared Error (MSE). This model can be used to predict quality of water
for new data.
• The data is split into two parts: training (80%) and testing (20%).
• In order to classify, repeated cross-validation is carried out for training dataset
with the fixed number of iterations.
• The model’s parameter configuration is selected, resulting in the maximum
accuracy. The model was scrutinized.
• Accuracy: Accuracy gives the ratio of correctly classified instances to the total
instances.
@seismicisolation
@seismicisolation
Accuracy = (TP + TN)/(TP + TN + FP + FN) (1)
• Mean Squared Error (MSE) is a metric used for evaluating the performance
of regression models.
1
n
MSE = ∗ (y_ true − y_ pred)2 (2)
n i=1
• Root Mean Squared Error (RMSE) is an evaluation metric for regression

models.
√
RMSE = MSE (3)
5 Classification
5.1 Linear Regression
A linear regression model is trained on the preprocessed data, using the input param-
eters as features and the quality of water as the target variable. The model is then
evaluated and tested using various performance metrics such as R-squared, mean
squared error and correlation coefficient [16]. Based on the evaluation results, the
linear regression model can be classified as a good or poor predictor of water quality.
A good predictor model would have high accuracy and low error rates, and can be
used for making predictions on new data.
5.2 Decision Tree
The use of decision tree for predicting the water quality involves creating a tree-like
model that represents a decision-making process based on different water parameters.
The decision tree algorithm works by recursively partitioning the data into smaller
subsets based on the selected parameters. At each level of the tree, the algorithm
selects the parameter that best separates the data into different classes or categories
of water quality. The resulting decision tree model can be used for water quality
prediction by inputting new data into the model and following the path of decisions
until a leaf node is reached.
@seismicisolation
@seismicisolation
5.3 Random Forest Regression
The algorithm works by building multiple decision trees, each trained on a subset of
the available data. The trees are then combined to create a final prediction based on the
majority vote of all the trees. Random forest has several advantages for water quality
prediction. It can handle large datasets with many features and can handle missing
data. It is also robust to outliers and can handle nonlinear relationships between the
input variables and the output [17].
5.4 XG Boost Regression
The water parameters are used as input features for the XG Boost regression model,
which is trained on historical water quality data to predict the quality of water in
future. XG Boost regression constructs an ensemble model of tree-based models,
where each tree is trained on a subset of data. The predictions are combined, giving
more weight to important features. XG Boost is effective for handling nonlinear
relationships and can be valuable for real-time water quality prediction in monitoring
and managing water resources.
6 Model Evaluation
In any machine learning algorithms, data is split into training model and testing
model.
Training the model: In this step, the algorithm is fed with a dataset to learn patterns
and relationships between the input data and the target variables. The algorithm
adjusts its internal parameters to reduce the error between the predicted and actual
values.
Testing/Evaluation: After the model is trained, it is tested on a new dataset to
evaluate its performance. The performance and accuracy of model is measured
using metrics such as accuracy, precision, recall, F 1 score, and others. Based on
the evaluation results, the model is either accepted or fine-tuned to improve its
performance.
7 Modeling and Analysis
The modeling and analysis of this system involve various techniques and algorithms
to forecast the quality of water, based on different factors. Decision tree, random
forest, boost, and linear regression are some commonly used algorithms for this
@seismicisolation
@seismicisolation
Fig. 2 Dataset taken from Kaggle [15]
purpose. For these applications, it is essential to know, various parameters of the

water attribute toward the quality of the water. This information is collected from
a domain expert or historical data collections [11]. Using water quality parameters,
overall quality of the water is evaluated, in terms of portability for each dataset.
These variables are pH, nitrate, conductivity, BOD, dissolved oxygen, temperature,
coliform are the terms that can be used for predicting the water quality [18]. The
choice of parameters with well-defined water quality standards is very important
to determine water quality. To develop effective water quality prediction systems,
it is crucial to analyze the different models and select the most appropriate ones
based on the specific requirements and data availability [19]. This involves eval-
uating the performance of different models using various statistical metrics, such
as Mean Regression (MR), Root Mean Squared Error (RMSE) and Mean Squared
Error (MSE). The predictive modeling as described in this paper, on the other hand,
is flexible enough to function with any number of parameters [13]. The reliability and
accuracy of predicting the water quality systems depend on the quality and quantity
of data used. Therefore, data analysis techniques, such as machine learning, statis-
tical analysis, fuzzy logic and data mining, are used to preprocess and transform raw
data into meaningful information for the models. To analyze overall quality of water
in terms of the water quality index, water quality factors are used for each data set.
One such dataset of Goa state indicated in Fig. 2 is taken from Kaggle which is used
to determine water quality index and studied. These parameter’s value depends on
the various factors of the environment [15].
The results of this experiment demonstrate that the random forest method achieves
the best performance compared to other algorithms linear regression, decision tree
and XG Boost regression used, with a mean squared error of 0.048 and Root Mean
Squared Error of 0.220.
Figure 3 scatter plot shows individual data points of water quality index (WQI)
against year. Each blue point represents a specific year’s WQI value. The plot provides
a visual representation of how WQI values vary over different years.
@seismicisolation
@seismicisolation
Fig. 3 Scatter plot

representation of WQI using
random forest
Figure 4 scatter plot shows mean WQI for each year. It aggregates the WQI values
by year, calculating the average WQI for each year using a group by operation. The
blue dots on the plot represent the average WQI values for respective years. This plot
helps in visualizing the overall water quality, highlighting any potential patterns or
changes over time in the quality of water. This gives us proper understanding of the
WQI over the years.
Fig. 4 Scatter plot

representation of mean WQI
using random forest
@seismicisolation
@seismicisolation
Table 2 Comparison of algorithms

Sl. No. Algorithm type Accuracy score MSE RMSE MR
1 Linear regression 40.45 101.63 10.08 57.15
2 Decision tree 99.15 6.1 7.8 57.55
3 Random forest regression 99.97 0.048 0.220 57.56
4 XG boost regression 98.94 1.80 1.34 57.51
8.1 Performance Metrics Results
The performance metrics of random forest model is compared with decision tree,
linear regression and XG Boost regression and experimented results are tabulated
below:
Table 2 compares the performance of different algorithms. Decision tree and
random forest regression exhibit notably high accuracy scores of 99.15% and 99.97%,
respectively, with low RMSE values, showcasing their robustness in predictive tasks.
Linear regression and XG boost regression also perform reasonably well but with
slightly lower accuracy scores.
9 Conclusion
In this paper, based on the Water Quality Indices (WQI), the water is classified into
drinking purpose, agricultural purpose and industrial purpose. If the WQI value is
greater than 70 it is fit for drinking, WQI greater than 50 is fit for agricultural purpose
and below can made use for Industrial purpose [15]. This system provides accurate
WQI to classify the usage of water [18]. The algorithms used provide high accurate
results that match the predicted output. The proposed approach has demonstrated
the potential of machine learning-based methods for predicting the quality of water
accurately. Using this developed model, water quality can be predicted for different
types of water usage [20]. The approach can be further improved by incorporating
additional data sources such as weather data, geological data and land use data and
developing more sophisticated algorithms.
References
1. Najah A, El-Shafie A, Karim OA, El-Shafie AH (2013) Application of artificial neural networks
for water quality prediction. Neural Comput Appl 22:187–201
2. Kangabam RD, Bhoominathan SD, Kanagaraj S, Govindaraju M (2017) Development of a
water quality index (WQI) for the Loktak Lake in India. Appl Water Sci 7:2907–2918
@seismicisolation
@seismicisolation
3. Singh P, Singh G, Kansal A (2017) Water quality prediction using artificial neural network:
a case study of River Yamuna, India. Environ Monit Assess 189(11):570. https://doi.org/10.
1007/s10661-017-6294-4
4. Srivastava G, Kumar P (2013) Water quality index with missing parameters. Int J Res Eng
Technol 02(04):609–614. https://doi.org/10.15623/ijret.2013.0204035
5. Ashwini K, Janice Vedha J, Diviya D, Deva Priya M (2019) Intelligent model for predicting
water quality. Int J Adv Res Ideas Innov Technol 5(2). www.IJARIIT.com
6. Prasad AN, Mamun KA, Islam FR, Haqva H (2015) Smart water quality monitoring system.
In: 2015 2nd Asia-Pacific world congress on computer science and engineering
7. Thukral AK, Bhardwaj R, Kaur R (2005) Water quality indices. Ministry of Statistics and
Programme Implementation, New Delhi 1, p 99
8. Haghiabi AH, Nasrolahi AH, Parsaie A (2018) Water quality prediction using machine learning
methods. Water Qual Res J 53(1):3–13
9. Sharma N, Kumar D, Sharma S (2020) Water quality prediction using machine learning
techniques: a review. J Environ Manag 268:110667. https://doi.org/10.1016/j.jenvman.2020.
110667
10. Koditala NK, Pandey PS (2018, Aug) Water quality monitoring system using IoT and machine
learning. In: 2018 International conference on research in intelligent and computing in
engineering (RICE). IEEE, pp 1–5
11. Kang G, Gao JZ, Xie G (2017) Data-driven water quality analysis and prediction: a survey.
In: 2017 IEEE Third international conference on big data computing service and applications
(BigDataService), pp 224–232. https://doi.org/10.1109/BigDataService.2017
12. Nair JP, Vijaya MS (2021) Predictive models for river water quality using machine learning
and big data techniques—a survey. Paper presented at the 2021 International conference on
artificial intelligence and smart systems (ICAIS), Coimbatore, India, pp 1747–1753. https://
doi.org/10.1109/ICAIS50930.2021.9395832
13. Bhardwaj R, Singh RK (2022) Water quality modeling of the river Ganga in the northern region
of India using the artificial neural network technique. J Water Manag Model. https://doi.org/
10.14796/JWMM.C486
14. Ragi NM, Holla R, Manju G (2019, May) Predicting water quality parameters using machine
learning. In: 2019 4th International conference on recent trends on electronics, information,
communication & technology (RTEICT). IEEE, pp 1109–1112
15. Indian water quality analysis and prediction. Kaggle (2018). https://kaggle.com/code/anbari
van/indian-water-quality-analysis-and-prediction
16. Veeramsetty V, Shadamaki N, Pinninti R, Guduri N, Ashish G (2022) Water quality index
estimation using linear regression model. AIP Conf Proc 2418(1):040033. https://doi.org/10.
1063/5.0081794
17. Devi SVSG (2019) Random forest advice for water quality prediction in the regions of Kadapa
district. Int J Innov Technol Explor Eng 8:1–3
18. Environmental Protection Agency (2021, Apr 22) Water quality criteria. EPA. https://www.
epa.gov/wqc/water-quality-criteria
19. Ahmed U, Mumtaz R, Anwar H, Shah AA, Irfan R, García-Nieto J (2019) Efficient water
quality prediction using supervised machine learning. Water 11(11):2210. https://doi.org/10.
3390/w11112210
20. Rani JPA, Nivasini R, Rubavathi CY, Jona P (2023, Feb) Machine learning based real time water
quality monitoring system. In: 2023 Third international conference on artificial intelligence
and smart energy (ICAIS). IEEE, pp 1366–1370
@seismicisolation
@seismicisolation
Artificial Neural Networks Modelling
for Predicting Water Quality
in the Surface Waters of Western
Godavari Delta, India
G. Sri Bala, P. A. R. K. Raju, and G. V. R. Srinivasa Rao
1 Introduction
Water plays a crucial role in maintaining both environmental and social well-being.
As the most abundant natural resource on our planet, water resources hold a privi-
leged status. However, the rapid progress of urbanization and industrial processes has
led to the production of hazardous toxins that degrade the quality of surface waters
[1]. Various types of waste, including organic pollutants, heavy metals, and microbi-
ological pathogens, find their way into water bodies, posing a threat to local wildlife
and human health [2]. It is imperative to monitor water quality levels to address
this pressing issue. Moreover, the world has been experiencing a global shortage of
freshwater and a decline in water quality over the past few decades, primarily due to
population growth and climate change [3]. To assess water quality, the Water Quality
Index (WQI) offers a straightforward scale format, presenting data on a range from 1
to 100 [4]. Higher index values correspond to better water quality, with scores above
80 indicating “clean rivers”, scores between 40 and 80 indicating “slightly polluted
rivers”, and scores below 40 representing “polluted rivers” that fail to meet the water
quality requirements [5].
The development and use of the WQI, which combines multiple criteria into a
single comprehensive index, has seen exponential growth [6]. The challenge lies in
reducing costs whilst developing more sophisticated computer-aided methods for
evaluating water quality levels [7]. The WQI serves as an effective tool to explain
and categorize different water quality levels, allowing for an assessment of suitability
for specific applications [8]. Several types of WQI have been designed for efficient
evaluation of overall water quality in specific regions, with India and other nations
G. Sri Bala (B) · P. A. R. K. Raju

Department of Civil Engineering, SRKR Engineering College, Bhimavaram, India
e-mail: sribala.3883@gmail.com
G. V. R. Srinivasa Rao
Department of Civil Engineering, Andhra University, Visakhapatnam, India
@seismicisolation
@seismicisolation
46 G. Sri Bala et al.
relying on the WQI scale [9]. In India, for instance, most of coastal areas exhibit
poor water quality despite extensive efforts and expenditures by the government to
regulate and treat polluted water bodies [10]. This high level of contamination poses
significant economic and social consequences, underscoring the need for accurate
water quality prediction as a vital component of managing and protecting the water
environment [11]. In recent years, Government of India has launched similar schemes
like the National River Conservation Plan (NRCP) and the National Mission for
Clean Ganga (NMCG) to aid in the clean-up efforts of other significant national
streams [12]. These programs strongly emphasize installing solid waste management
infrastructure, sewage treatment plants, and effluent treatment facilities and reducing
industrial pollution along rivers [13].
Numerous researchers have contributed to the advancement of water quality index
evaluation through modelling [14, 15]. There are two main categories of research
methods for predicting water quality: mechanistic and non-mechanistic approaches.
Mechanistic models employ theoretical mathematical and physical models to predict
the evolution of water quality mechanisms [16]. By examining the chemical, physi-
ological, and biological changes throughout the water resource cycle, these models
reflect the relationships between the components and solve corresponding equations.
Whilst mechanistic models offer a thorough understanding of water quality changes
by considering various processes, they tend to be complex and require extensive
background knowledge and data [17].
On the other hand, non-mechanistic prediction methods rely on statistical and
mathematical models constructed based on historical data. These models provide a
means to forecast water quality by analysing specific water quality systems using
mathematical techniques such as regression models, mathematical probability simu-
lations, grey prediction models, and time series models [18]. However, achieving a
consistent distribution of water quality over time and location remains a challenge,
and these models have their limitations, especially when characterizing complex
aquatic systems.
In recent years, many researchers have turned to neural networks and other
machine learning algorithms to forecast water quality, yielding promising results.
Few researchers have adopted hybrid models, such as ANN with graphical user inter-
face [19], ANN with wavelet transforms [20], and ANN with principal component
analysis [21] for successful water quality prediction.
In this study, ANNs are utilized to forecast the water quality index in India’s
extensive inland farming zone. By leveraging these advanced techniques, it becomes
possible to make reliable predictions and mitigate abrupt changes in water quality.
Accurate forecasting of water quality changes has a positive impact on guiding
irrigation systems, preserving biodiversity, and ensuring the safety of drinking water
supplies.
@seismicisolation
@seismicisolation
Artificial Neural Networks Modelling for Predicting Water Quality … 47
2 Site Description
The study area is in a coastal region of India, bounded by the Bay of Bengal, the
Upputeru River, and the Godavari River with latitudes (north) of 16° 19' 06'' and
16° 56' 10'' and longitudes (east) of 81° 18' 25'' and 81° 52' 45'' (Fig. 1). The tidal
patterns of the Bay of Bengal significantly impact this area. The climate is charac-
terized by summer from March to June, accompanied by the southwest monsoon.
The region experiences an average annual rainfall of approximately 875 mm, whilst
temperatures range between 28 and 37 °C. As a result of the direct influence of tidal
amplitudes from the Bay of Bengal, the studied area is predominantly affected by
saline water [22].
Over the past few decades, the water quality in this area has been significantly
degraded due to the discharge of untreated wastewater from intensive inland aquacul-
ture practices. The delta regions, particularly, have focused on aquaculture activities,
leading to the deterioration of surface water quality. The Godavari River, originating
from its upstream flow, and the Upputeru River, flowing at its base, are intercon-
nected by canals primarily supplying freshwater for agricultural practices in the
western Godavari delta region.
The area is complex, and intensive aquaculture systems, characterized by a high
density of prawn ponds, contribute to the increased use of fertilizers and supplemental
feeds [23–25]. However, it has been observed that pollutants originating from anthro-
pogenic activities in the upstream areas and along the canals have minimal influence
Fig. 1 Study area
@seismicisolation
@seismicisolation
on the research area. Instead, the aquaculture practices along the river reach have
emerged as the primary contributors to the changes observed in water quality within
the research area.
3 Methodology
In the current study, the pre-monsoon and post-monsoon data of Venkayya Vayyeru
stream over a period of 9 years from 2014 to 2022 are used for model training and
validation. In the stream, a total of 25 random locations were selected and each year,
25 samples were tested for various water quality parameters such as pH, EC, TDS,
total alkalinity (TAlk), hardness, Ca2+ , Mg2+ , chlorides, nitrates, dissolved oxygen
(DO), biochemical oxygen demand (BOD), and WQI (as obtained by Weighted
Arithmetic Method). In this model generation, WQI is calculated using the input
parameters. The input and target attributes for the model are as shown in Table 1.
This study employed an ANN as the machine learning model to forecast water
quality characteristics. ANN models have gained recognition in engineering and
research fields for their ability to solve complex problems effectively [26, 27]. In
addition, they are known for their excellent performance and computational effi-
ciency, enabling them to make accurate predictions without the requirement of
explicit pattern formalization or prior knowledge. Through the interconnected input,
hidden, and output layers, an ANN establishes a memory that can associate a vast
number of input patterns with corresponding output results. The network forms a
multilayer perceptron structure by incorporating a hidden layer and nodes within the
input and output layers. Water quality forecasting models frequently adopt a three-
layered feed-forward neural network. Neurons within each hidden layer receive input
data from the input layer through a linear summation operation, and their outputs
Table 1 Input and target data

Modelling data
variables used for modelling
Input variables pH
EC (µmhos/cm)
TDS (mg/L)
TAlk (mg/L)
Hardness (mg/L)
Ca2+ (mg/L)
Mg2+ (mg/L)
Chlorides (mg/L)
Nitrates (mg/L)
DO (mg/L)
BOD (mg/L)
Output variable WQI-WA
@seismicisolation
@seismicisolation
are determined by passing the linear summation through an activation function. This
process is replicated from the hidden layer to the output layer. Whilst several machine
learning models are available for forecasting water quality metrics, the choice of ANN
in this study was driven by its compatibility with integration into the application with
minimal complexity. In addition, it was selected over other models due to its ease
of implementation and absence of intricate subroutines. Therefore, the ANN model
was chosen as the preferred approach in this study, leveraging its capabilities in accu-
rately forecasting water quality characteristics and its suitability for integration into
the application.
Before developing any model, it is important to know the relations between the
variables. It is necessary to find a significant correlation whether it is stronger or
weaker. If the correlation coefficient is greater than 0.8, the relation is said to be
stronger, otherwise it is said to be weaker. It is clear from the heat map that EC and
TDS are significant variables affecting WQI (Fig. 2).
Fig. 2 Heat map of input and output variables
@seismicisolation
@seismicisolation
4 ANN Prediction Results
The capacity of ANN models to detect intricate nonlinear correlations and patterns in
the data makes them ideal for WQI prediction. These models are modelled after the
way the human brain works, which is made up of interconnected artificial neurons
arranged in layers. Each neuron processes the information it receives from the layer
before it transmits it to the layer above, ultimately producing a prediction for the
output. Several stages are routinely taken to create an ANN model for forecasting
the WQI. First, a large dataset is gathered, which contains information on the conduc-
tivity, pH, dissolved oxygen, and other aspects of water quality. To ensure the gener-
alizability of the model, this dataset should include a wide variety of water sources,
such as rivers, lakes, and groundwater. The training set and the testing set are then
created from the dataset. By modifying the weights and biases of the neurons based
on the input–output patterns, the training set is utilized to train the ANN model. On
the other hand, the trained model’s performance and predicted accuracy are evaluated
using the testing set.
By contrasting the projected WQI values with the actual values from the training
set, the ANN model iteratively adjusts its parameters during the training process. This
adjustment is made by utilizing optimization strategies like backpropagation, which
analyses the discrepancy between the anticipated and actual values and modifies
the weights accordingly. Finally, the training is continued when a predetermined
stopping criterion is satisfied, or the model reaches an acceptable degree of accuracy.
The quality and representativeness of the input data, as well as the model’s archi-
tecture and training procedure, all have a role in how accurate and reliable the ANN
model’s predictions are. Proper data preparation, feature selection, and model opti-
mization procedures are essential to improve the model’s performance. The present
study includes 9 years post-monsoon data of Venkayya Vayyeru stream and Naras-
apur stream. Deep learning in Artificial Intelligence after several simulations gener-
ates model predicted values for the target variable. The comparison is as shown in
Tables 2 and 3. The model values are predicted with the following Python algorithm
using different training and testing accuracies like 70–30, 80–20, 85–15, 90–10, and
95–15 (see Figs. 3, 4, 5, 6 and 7). After several iterations, each time, the set of models
predicted values are generated in the form of output. Those output values are now
compared to the average of calculated WQI’s for each year successively.
It can be seen from Table 2, % deviation between the model WQI and calculated
WQI varies from 4.157 to 13.574 with 70–30 ratio. Similarly, % deviation varies
from 0.090 to 3.389 with 80–20 ratio. The average % deviation varies from 0.09 to
13.574 with 5 ratios. Figures 3, 4, 5, 6 and 7 show the predicted and calculated WQI
with varying proportions of datasets.
@seismicisolation
@seismicisolation
Table 2 Comparison of ANN model and WQI computed values

S. No. Calculated WQI Predicted WQI with various ratios of datasets
(training–testing)
70–30 80–20 85–15 90–10 95–05
1 96.54 103.62 99.69 99.41 98.97 98.75
2 102.22 107.54 103.18 105.49 107.28 103.38
3 96.69 109.70 95.22 99.94 100.53 90.47
4 99.60 103.74 99.69 102.76 108.25 91.17
5 103.35 109.70 103.18 104.49 107.28 104.72
6 104.04 110.92 100.63 105.49 108.25 107.45
7 107.37 114.79 109.70 110.05 108.25 113.22
8 111.61 126.76 114.85 113.85 111.25 118.34
9 111.31 125.05 114.85 111.85 113.65 120.71
Table 3 Model and calculated values % deviations

S. No. % deviations with various ratios of datasets (training–testing)
70–30 80–20 85–15 90–10 95–05
1 7.334 3.263 2.973 2.517 2.289
2 5.204 0.939 3.199 4.950 1.135
3 13.455 1.544 3.361 3.971 6.875
4 4.157 0.090 3.173 8.685 9.246
5 6.144 0.165 1.103 3.803 1.326
6 6.613 3.389 1.394 4.047 3.278
7 6.911 2.170 2.496 0.820 5.448
8 13.574 2.903 2.007 0.324 6.030
9 12.344 3.180 0.485 2.102 8.445
160
Calc WQI/Model WQI
Calc WQI Model WQI
100
40
1 2 3 4 5 6 7 8 9
Number of years
Fig. 3 Model and calculated WQI’s comparison (70–30 ratio)
@seismicisolation
@seismicisolation
160
Calc WQI 80-20 ratio

Calc WQI
100
40
1 2 3 4 5 6 7 8 9
Number of years
160
Calc WQI
Calc WQI
85-15 ratio
100
40
1 2 3 4 5 6 7 8 9
Number of years
160
Calc WQI
100
40
1 2 3 4 5 6 7 8 9
Number of years
5 Conclusion
The management of water quality in the canals of the western Godavari delta region
poses significant challenges, mainly due to the presence of intensive aquaculture
ponds. This issue occurs frequently and with greater severity in such areas. To address
@seismicisolation
@seismicisolation
160
Calc WQI
100
40
1 2 3 4 5 6 7 8 9
Number of years
this concern, statistical interpolation mapping has become increasingly utilized in

water management to analyse and monitor water quality. This study explores the
potential of an artificial intelligence tool called ANN for water management in inland
aquaculture zones, presenting a comprehensive approach.
The study reveals that the ANN model produces consistent results; however, the
outcomes vary depending on the training dataset used. Nevertheless, the ANN model
is effective for long-term applications, as it can accommodate multiple input datasets
and generate highly accurate forecasts by learning from the entire dataset.
Researchers continue to seek solutions for the prevalent challenge of poor perfor-
mance when insufficient input data is available in data-driven methodologies like
ANN. Therefore, it is inferred that augmenting the network’s training with additional
experimental data would likely enhance the accuracy of the task.
In conclusion, this study highlights the difficulty in managing water quality in the
canals of the western Godavari delta region, primarily due to intensive aquaculture
ponds. It proposes using an ANN model as a valuable tool for water management
in these areas. However, it acknowledges the need for further exploration to address
issues related to limited input data and performance optimization in data-driven
methodologies.
References
1. Ukaogo PO, Ewuzie U, Onwuka CV (2020) Environmental pollution: causes, effects, and the
remedies. In: Microorganisms for sustainable environment and health, pp 419–429
2. Madhav S, Ahamad A, Singh AK, Kushawaha J, Chauhan JS, Sharma S, Singh P (2020) Water
pollutants: sources and impact on the environment and human health. In: Sensors in water
pollutants monitoring: role of material, pp 43–62
3. Schaefer M (2008) Water technologies and the environment: ramping up by scaling down.
Technol Soc 30(3–4):415–422
4. Sutadian AD, Muttil N, Yilmaz AG, Perera BJC (2016) Development of river water quality
indices—a review. Environ Monit Assess 188:1–29
5. Muyibi SA, Ambali AR, Eissa GS (2008) The impact of economic development on water
pollution: trends and policy actions in Malaysia. Water Resour Manag 22:485–508
@seismicisolation
@seismicisolation
6. Abba SI, Pham QB, Saini G, Linh NTT, Ahmed AN, Mohajane M et al (2020) Implementation
of data intelligence models coupled with ensemble machine learning for prediction of water
quality index. Environ Sci Pollut Res 27:41524–41539
7. Gao W, Zhang Y, Ramanujan D, Ramani K, Chen Y, Williams CB et al (2015) The status,
challenges, and future of additive manufacturing in engineering. Comput Aided Des 69:65–89
8. Fulazzaky MA (2010) Water quality evaluation system to assess the status and the suitability
of the Citarum river water to different uses. Environ Monit Assess 168:669–684
9. Uddin MG, Nash S, Olbert AI (2021) A review of water quality index models and their use for
assessing surface water quality. Ecol Ind 122:107218
10. Rajaram T, Das A (2008) Water pollution by industrial effluents in India: discharge scenarios
and case for participatory ecosystem specific local regulation. Futures 40(1):56–69
11. Davies JM, Mazumder A (2003) Health and environmental policy issues in Canada: the role of
watershed management in sustaining clean drinking water quality at surface sources. J Environ
Manag 68(3):273–286
12. Dhingra UG, Bedarkar M, Rao P (2023) Sustainable restoration strategies for rivers Mula-
Mutha, Pune, India-adaptive learnings from global river restoration experiences and thematic
analysis. Int J Bus Globalisation 33(1–2):171–203
13. Denny P (1997) Implementation of constructed wetlands in developing countries. Water Sci
Technol 35(5):27–34
14. Egbueri JC (2022) Incorporation of information entropy theory, artificial neural network, and
soft computing models in the development of integrated industrial water quality index. Environ
Monit Assess 194(10):693
15. Pesce SF, Wunderlin DA (2000) Use of water quality indices to verify the impact of Córdoba
City (Argentina) on Suquı́a River. Water Res 34(11):2915–2926
16. Kiparissides C (1996) Polymerization reactor modeling: a review of recent developments and
future directions. Chem Eng Sci 51(10):1637–1659
17. Jager T, Vandenbrouck T, Baas J, De Coen WM, Kooijman SA (2010) A biology-based approach
for mixture toxicity of multiple endpoints over the life cycle. Ecotoxicology 19:351–361
18. DeAngelis DL, Franco D, Hastings A, Hilker FM, Lenhart S, Lutscher F et al (2021) Towards
building a sustainable future: positioning ecological modelling for impact in ecosystems
management. Bull Math Biol 83:1–28
19. Rizal NNM, Hayder G, Yusof KA (2022) Water quality predictive analytics using an artificial
neural network with a graphical user interface. Water 14(8):1221
20. Wu J, Wang Z (2022) A hybrid model for water quality prediction based on an artificial neural
network, wavelet transform, and long short-term memory. Water 14(4):610
21. Ibrahim A, Ismail A, Juahir H, Iliyasu AB, Wailare BT, Mukhtar M, Aminu H (2023) Water
quality modelling using principal component analysis and artificial neural network. Mar Pollut
Bull 187:114493
22. Nagaraju TV, Malegole SB, Chaudhary B, Ravindran G (2022) Assessment of environmental
impact of aquaculture ponds in the western delta region of Andhra Pradesh. Sustainability
14(20):13035
23. Nagaraju TV, Sunil BM, Chaudhary B (2022) Understanding the role of biological oxygen
demand in aquaculture waters in the western delta region of Andhra Pradesh. In: Recent
advances in sustainable environment: select proceedings of RAiSE 2022, pp 13–20
24. Nagaraju TV, Sunil BM, Chaudhary B (2022) Impact of aquaculture solid waste on environment
in the delta region of Andhra Pradesh: a case study. In: Transportation and environmental
geotechnics: proceedings of the Indian geotechnical conference, vol 4, pp 369–374
25. Nagaraju TV, Chaudhary B, Prasad CD, Gobinath R (2023) Prediction of ammonia contami-
nants in the aquaculture ponds using soft computing coupled with wavelet analysis. Environ
Pollut 121924
@seismicisolation
@seismicisolation
26. Nagaraju TV, Gobinath R, Awoyera P, Abdy Sayyed MAH (2021) Prediction of California
bearing ratio of subgrade soils using artificial neural network principles. In: Communication
and intelligent systems: proceedings of ICCIS 2020, pp 133–146
27. Alisha SS, Nagaraju TV, Onyelowe KC, Dumpa V, Sireesha M (2022) Prediction of strength
and stiffness behavior of glass powder stabilized expansive clay using ANN principles. In:
International conference on trends and recent advances in civil engineering, pp 211–219
@seismicisolation
@seismicisolation
IOT-Based Patrolling Robot
for Construction Sites
Smita Agrawal, N. Amrutha, B. M. Punyashree, V. Raksha, and D. R. Yuktha
1 Introduction
The construction industry is one of the largest industries in the global economy.
However, it is one among the slowest sectors to start the path of automation and
digitization, so the integration of robots in construction has been late compared to
other sectors. This is due to many factors, such as labor costs or a lack of planning
in the process. Today, safety of construction sites is a major concern in many parts
of the world. The lack of good security measures has led to an increase in construc-
tion thefts, particularly equipment theft. Construction sites should be monitored and
inspected periodically before, during, and after construction. The longer the site is
under surveillance, the harder it is for intruders to commit the crime. A number of
thefts are required to examine the workplace. In this context, “Robotics” has a great
potential to change the construction industry and provide it with many advantages.
Construction is a labor-intensive industry, and robotic technology can be effective in
reducing labor costs by improving the safety, productivity, and quality of construc-
tion with the advent of autonomous mobile robots designed to tackle patrol and guard
duties. Additionally, it becomes more difficult for an intruder to commit crimes on the
property, if it is under surveillance. The jobsite needs to be reviewed through multiple
assessments. Most venues require a security personnel to do foot patrol rounds inside
of guarded areas in order to receive reliable protection. Illegal trespassers are warned
just by the security guard’s simple presence in the restricted area rather than just
while they are “on duty” in a guard hut. Foot patrol is frequently a pricey service that
depends on the professionalism and dedication of the person engaged; as a result, it
requires a more expensive long-term investment of Money and Time. Despite its enor-
mous economic significance, the building industry is plagued by inefficiency. In this
S. Agrawal (B) · N. Amrutha · B. M. Punyashree · V. Raksha · D. R. Yuktha

Department of Electronics and Communication Engineering, Jyothy Institute of Technology,
Bangalore, India
e-mail: smitha.agarwal@jyothyit.ac.in
@seismicisolation
@seismicisolation
58 S. Agrawal et al.
circumstance, “robotics” has the potential to revolutionize the construction sector and
bring about a number of benefits. Numerous industrial and non-industrial applica-
tions exist where mobile robots can really make a difference. To analyze, control, and
safeguard the areas against unwelcome threats, this project intends to create a multi-
purpose surveillance robot that can execute surveillance tasks at construction sites
[1]. This study offers an overview of contributions that reflect current research trends
to address the challenges faced by robotic patrolling. Robotic patrolling involves
sending a group of autonomous mobile robots into a physical place with the goal of
securing it [2]. This article reviews mobile robots and the way it can maneuver in the
real world without interacting with people to achieve their goals. It also discusses
some of the tendencies such as AI, automatic driving, network connectivity, and
other new trends are pushing these changes. Additionally, such news trends are
utilized in numerous industries, including manufacturing, distribution of commodi-
ties, service robotics, sports, ergonomics, and medicine. In the upcoming years, these
trends will continue to develop [3]. This literature review explores the problems that
mobile robots are now facing. A thorough examination of tools/sensors and well-
established sensor fusion methods created to address problems with localization,
estimation, and navigation in mobile robots is also offered, and they are arranged
according to relevance, strengths, and shortcomings. The work provides important
insights for future research into creating strategies to address the inconsistencies
encountered by autonomous mobile robots [4]. A common trend in the wide range of
heterogeneous approaches and techniques is the introduction of limitations, dynamic
settings, limited attacker capabilities, and richer strategy representations in order to
create more realistic models. The findings offer up a number of intriguing open
problems and give complementarities and synergies toward more effective robotic
patrolling systems [5]. This work covers the prior research in the area of multi-robot
patrolling from the view-points of regular and adversarial patrolling, in accordance
with various objectives. A number of deterministic tactics are suggested for routine
patrolling, which calls for robots to visit crucial sites as regularly as feasible, while
an adversarial approach concentrates on unpredictable robot movement to increase
the livelihood and it detects. A thorough analysis is conducted for each category,
including problem formulation and modeling, patrolling goals and standards of eval-
uation and typical patrolling tactics [6]. Excellent examples of mobile robots include
humanoid robots, unmanned rovers, entertainment dogs, drones, and so forth. They
differ from other robots in that they can move independently and have the intelligence
to respond and decide depending on perceptions of their surroundings. The world
of mobile robots, including the newest trends, is examined in the article. These
emerging themes are being driven by artificial intelligence, autonomous driving,
network communication, cooperative work, nano robotics, amiable human–robot
interfaces, secure human–robot interaction, and emotion expression and perception
[7]. The issue of many robots patrolling an area is one that affects both military
operations and other security-related situations. The robot wireless connections may
be jammed in a hostile environment and their sensor ranges may be constrained
to visibility. This makes the issue more challenging, yet regardless of the setting,
answers will be broadly applicable. Deterministic robot tracks can be monitored and
@seismicisolation
@seismicisolation
IOT-Based Patrolling Robot for Construction Sites 59
predicted by an adversary, allowing for the exploitation of known coverage gaps.

Additionally, centralized and Markov chain-based patrolling strategies are discov-
ered that reduce the average projected journey time for the subset of locations while
allowing each robot to cover a specific area of the environment and stay in touch
with a centralized base station. The chance of catching an opponent and the longest
period of time a place will be unsecured are the last two metrics are used to assess
how vulnerable our patrolling policies are [8]. In this work, a Raspbian Operating
System with an Automated Obstacle Detection Robot with a controlled algorithm
that is equipped with a night-vision camera and IoT is used for motion control as well
as useful for machine-based image or video capture and transmission to the user via
the Internet using the Thing Speak Application [9]. Supporting autonomous or tele-
operated robots and multi-robot systems can be very beneficial for search and rescue
(SAR) operations. These can help with situational analysis and mapping, moni-
toring, and surveillance, setting up communication networks, and victim recovery.
In this study, multi-robot systems that support SAR operations are reviewed, with
system-level considerations and an emphasis on algorithmic approaches for multi-
robot perception and coordination [10]. This work is home security and surveillance,
and the major goal of this research is to use an intelligent robot to monitor home
safety. A smart, user-friendly surveillance robot is developed that can address the
shortcomings of the traditional CCTV system. The robot’s design also incorporates
the transfer of monitored data to the Thing Speak IoT Cloud Server, from which the
resident receives immediate email alerts in the event [11]. Security on construction
sites is a significant problem, especially for the commercial construction industry.
Theft and vandalism can have a direct impact on a project’s success and cause delays,
operator downtime, increased insurance costs, a possible cancellation of an insurance
policy, and a loss in the project’s profitability. The theft of tools and equipment cost
the US construction industry close to $1 billion in 2001, according to the National
Insurance Crime Bureau. In 2001, over 1000 thefts of construction equipment were
reported in Japan, while annual insurance claims for theft losses in Canada average
over $46 million. The creation of sensor-assisted intelligent mobile robots that are
capable of performing complex navigational tasks for indoor construction site secu-
rity and safety is the subject of ongoing research at the Peter Kiewit Institute (PKI) in
Omaha, Nebraska, as described in this study. The ultimate goal of the research is to
fully integrate sensor-equipped robots into the WSN system to visualize sensed data
information in a 3D graphical control interface for advanced planning and execution
against any anomalies that have been detected in warehouses, offices, manufacturing
facilities, and various construction sites—anywhere there is a need for advanced
frontline security.
@seismicisolation
@seismicisolation
2 Methodology
Here, we present an autonomous security robot that continuously runs and patrols
vast areas by itself to safeguard the structure. A robotic gadget equipped with sensors
and a camera is employed to safeguard the area. A buzzer sound warns the owner if
an intruder is spotted when the robot searches a specified area for them. It is equipped
with cameras and other sensors, and it moves in a predefined or random direction.
When patrolling, it uses its own predefined or randomly created path to follow the
path. It always ends, especially if a suspect path is discovered, then carries on to
the next location. This device keeps an eye on any area using an IR-based patrolling
system. It maintains track of the area. The device starts to move in that direction along
a predetermined course when there is no activity in the calm region. Then, using the
cameras attached on the device, the area is checked for any suspicious activities.
Then, it uploads the photos it took of the incident to the IOT website. Here, we
use the IOT Local Area Network (LAN) to receive sent photos, display them to the
user, and play warning noises. The Patrolling Robot allows for skillful moving video
monitoring at medium distances. The Patrolling is designed to keep an ongoing eye
on things, which gives it effectively keeping control over any size of space. The best
method for keeping an eye on areas is to use the Patrol Robot. Different Blocks of
construction area can be observed with many separate constructions, walls, or other
elements stumbling blocks that prevent conducting video observation from remote
sites. So, by placing a few robots in one general area that is being watched, sufficient
patrolling can be achieved. If smartly observing, it is difficult for a gate-crasher to
get access to the premises made possible by a few robots using a few unique travel
routes.
2.1 Flowchart for Obstacle Avoidance and Path Following
Figure 1 shows the flowchart.

• The robotic vehicle moves at predefined path and is equipped with camera and
sound sensors.
• It stops at particular points and moves to next point, if sound is detected else
continues in the same path. If any sound is detected, the images and videos are
sent to the user.
Line Follower: A computerized guided vehicle called a “Line Follower Robot,”

as the name suggests, follows a visible line that has been implanted in the floor or
roof. The visual line, which is typically a dark line on a white surface, indicates the
direction that the line supporter robot travels. Undoubtedly, the infrared beam (IR)
sensors installed under the robot should be able to detect the queue. From that point
on, explicit change transporters are used to transmit the information to the processor.
From this point forward, the system will select the appropriate recognizes, send them
@seismicisolation
@seismicisolation
Fig. 1 Flowchart
to the driver, and as a result, the line adherent robot will follow the route. It is typically
a self-operating, multi-functional machine. Two IR sensors will be used to determine
whether the robot is on the correct track, and two engines will be used to correct
it if it veers off the path. We use an engine driver module like the L298N for these
engines, because they need high flow and should be bi-directional. Additionally, we
will need a computing device like the Raspberry Pi to instruct the engines based on
the characteristics of the IR sensor.
Obstacle Detection: Perhaps, the most important aspect of portable robotics
is deterrent aversion. In self-governing driver assistance frameworks that fire off
antecedents to self-ruling automobiles, obstacle discovery is a requirement. A new
area of robotics is obstacle avoidance innovation. Three main modules, including
camera and raspberry pi drivers, make up the raspberry pi-based obstruction recog-
nizing framework. The camera module receives the information picture that is gath-
ered during ongoing action. The Raspberry Pi is a platform with all the basic equip-
ment modules gathered on it. Photos are obtained through the camera module. If
there is an obstruction, it will inform the engine driver further and accordingly. The
engines receive the signal from the raspberry pi, and in the case that an obstruction
appears in its path, they function similarly to a flag, moving left or right with the
help of the left and right engine to avoid the obstructions. The robot is capable of
avoiding the obstacle in its path by using a deterrent discovery calculation. The rasp-
berry pi will receive information from the pi camera module about the obstruction
@seismicisolation
@seismicisolation
on an ongoing basis. Similar to assembly technology, mechanical snag identification

frameworks use ultrasonic sensors as well.
Intrusion Detection: Our model is capable of accurately identifying sound with the
use of sound sensors. When one of these events occurs, the camera is activated, and
the circumstance involving the notable interruption is captured and sent over SMTP
technology to the customer’s email. A device that checks a company or set of frame-
works for retaliatory action or strategy infringement is known as an interruption iden-
tification framework. Any interruption action or infringement is typically announced
interruption location has generally been managed using modern programming and
quantifiable examination, though occasionally it must be completed by directors,
either by gradually identifying the gate-crashers or by re-examining network logs,
making this a time-consuming and laborious task. This has been made easier by the
completion of an interruption identification test using visual, audible, or material
palpable data in PC interfaces. We offer a framework for attack moderation that has
been developed and tested.
2.2 Block Diagram
Figure 2 shows the block diagram and Fig. 3 shows the implementation of proposed
project. Different components of project are: an IR sensor is used to direct the robot to
travel in a specific direction, a sound sensor is used to determine the sound in a certain
area, a USB camera is integrated with Raspberry Pi Power supply for Raspberry Pi.
Fig. 2 Block diagram
@seismicisolation
@seismicisolation
Fig. 3 Patrolling robot
2.3 Hardware Components
Raspberry Pi: Fig. 4 represents a model of Raspberry Pi. A single-board computer

of the size of a credit card is called Raspberry Pi. It enables users of all educational
backgrounds and computing skills to experiment with, and learn about computation.
It is an improved motherboard that the Raspberry Pi Foundation created in the UK and
is now generally acknowledged as a component of developing computer technology.
Other pieces of peripheral gear, including a keyboard, mouse, and monitor, can be
connected to the minicomputer. Robots can have the best core available thanks to
the Raspberry Pi. It is simply ideal due to its small size and lightweight, as well as
its affordable pricing. It is perfect for robotics due to its compact size and powerful
processor.
Sound Sensor: Fig. 5 represents a model of Sound Sensor. A module that is used
to hear sound is the sound sensor. Typically, this module is used to measure sound
intensity. Switch, security, and monitoring are the key applications of this module.
This sensor’s accuracy can be adjusted for convenience of use. The microphone
used by this sensor serves as the input source for the peak detector, amplifier, and
buffer. This sensor detects sound, interprets an o/p voltage signal, and sends it to a
microcontroller. The necessary processing is then carried out.
Night Vision Camera: Fig. 6 represents a model of Night vision Camera. To
increase a video surveillance system’s efficiency in low light, night vision cameras
are used. Presence of IR lighting is the key component that unifies all of these cameras;
without IR, it is difficult to even consider night-time video monitoring. You can keep
an eye on your property day and night using night vision cameras. Many burglars
will wait until it is dark outside to break into a home or any work site to avoid being
seen. They will be captured on video, thanks to night vision technology. They have
@seismicisolation
@seismicisolation
Fig. 4 Raspberry Pi
Fig. 5 Sound sensor
motion detection. Numerous night vision cameras feature motion detectors that may
detect movement.
DC Motor: Fig. 7 represents a model of DC motor. A direct current motor is a
device that converts electrical power from a DC source into mechanical power. DC
motor operation depends on the conductor’s ability to conduct current while in an
attractive field where the conductor is subject to mechanical force. This motor is very
well-liked for several reasons, and it is employed in robots to “actuate” things, such
Fig. 6 Night vision camera
@seismicisolation
@seismicisolation
Fig. 7 DC motor
Fig. 8 IR sensor
as tracks, arms, or cameras. Due to their great torque and efficiency, DC motors are
especially practical for robotics applications.
IR Sensor: Fig. 8 represents a model of IR Sensor. The IR sensor is an electronic
module that contains the transmitter and the receiver, the transmitter emits infrared
radiation which hits the obstacles and is reflected by that obstacle which is detected
by the receiver of the module and it changes the output to high. Infrared Sensors or IR
Sensors are one of the frequently used sensor modules by electronics hobbyists and
makers. They are often used as Obstacle Detecting Sensors or Proximity Sensors.
2.4 Software Components
Python: Python is an interpreted, object-oriented, high-level, dynamically semantic

programming language. It is particularly desirable for Rapid Application Develop-
ment as well as for usage as a scripting or glue language to tie existing components
together due to its high-level built-in data structures, dynamic typing, and dynamic
binding. Python’s straightforward syntax emphasizes readability and makes it simple
to learn, which lowers the cost of program maintenance. Python’s support for modules
and packages promotes the modularity and reuse of code in programs. For all popular
platforms, the Python interpreter and its comprehensive standard library are freely
distributable in source or binary form.
Raspbian OS: Raspberry Pi OS is a free, open-source operating system built
on Debian Linux and designed for use with Pi boards. Raspberry Pi OS is also
@seismicisolation
@seismicisolation
installed on a number of single-board computers with an ARM processor. The first

iteration, formerly known as Raspbian, was released in 2013, and starting in 2015,
the Raspberry Pi Foundation made it available as a Pi distribution that had received
official approval. Raspbian was developed by developers Peter Green and Mike
Thompson and was initially an unaffiliated project.
3 Other Applications of Patrolling Robot
3.1 Preventing Break-Ins and Burglary
Retail burglary, which involves breaking into a store with the intention of stealing
goods, frequently causes expensive structure damage and the loss of priceless goods.
Due to the opportunity to enter the facility undetected, these crimes tend to happen
at night or after regular work hours. Many businesses rely on night patrol services in
addition to putting in place external security measures like reinforced windows and
doors to deter burglars. Criminals are less likely to try a break in and burglary when
there are qualified and experienced security officers on present.
3.2 Eliminating Vandalism
A firm might lose an average of $3370 to vandalism, according to the U.S. Small firm
Administration. Businesses are at danger of incurring expensive property damage
due to the spike in vandalism. In addition to the cost of the repairs and replacements
needed after a criminal incident, vandalism can also force a firm to temporarily close,
which results in lost sales. Vandalism can be stopped and property damage can be
avoided by having a professional security presence. Professional police officers are
prepared to deal with all criminal offences and are aware of what to do to deter
vandalism.
3.3 Immediate Response
It is crucial to have someone on site to manage the situation as promptly as possible

when a security event occurs inside or outside of a business. With a dedicated patrol
officer on the property, thieves, vandals, and other criminals can be apprehended
quickly, minimizing or even preventing any losses or damage. The speedy response
to commercial alarm systems is another potential aspect of night security services.
An on-site security officer will validate and reset the alarm as soon as a break-in
@seismicisolation
@seismicisolation
attempt alarm sounds, and if a crime actually takes place, they will also call the local
police.
3.4 Contacting Police
The initial responders to security incidents inside or outside of a business are

frequently patrol policemen. Businesses may rely on the security guards they’ve
employed to call the police as soon as an incident occurs, allowing law enforcement
to proceed with the arrest and prosecuting of any suspects. Many times, it takes hours
for businesses to become aware of a crime; by then, the offender has already left the
scene. With the help of night patrol services, law enforcement can be alerted shortly
after a crime to improve the likelihood that a business will receive justice.
3.5 Protecting Neighborhoods
Today’s security personnel offer a wide range of services, including neighborhood

protection. Depending on the size of the region, a security officer may choose to patrol
a neighborhood on foot, by bicycle or by car. Security personnel who have received
training are aware of the situations that raise suspicion and know when to take action.
Crimes are typically committed when people are asleep because perpetrators like to
hide in the shadows. With night patrol services, communities can relax knowing that
a skilled security guard is deterring trespassers.
3.6 Get Started with Professional Security Services
Employers cannot constantly be on the premises, which is why business owners

frequently worry that their establishments will be broken into at night while nobody
is around. Many business owners hire night patrol services to keep the structure and
its contents secure. Employing security experts can assist firms in a variety of ways.
They can execute business checks, perform lock-up services, remove trespassers,
discourage criminal behavior, assure prompt responses to security threats, and keep a
close eye on a facility. Numerous other industrial and non-industrial applications exist
as well, such as those for surveillance, planetary exploration, patrolling, emergency
rescue operations, reconnaissance, petrochemical applications, industrial automa-
tion, construction, entertainment, museum guides, personal services, intervention in
harsh environments, transportation, and healthcare.
@seismicisolation
@seismicisolation
4 Conclusion
In this proposed framework, the entire construction area will be covered by the robot
equipped with sound sensors and a night vision camera. As the size of the robot is
less, it can enter into smaller areas also. The patrolling robot patrols in a predefined
path autonomously. If any obstacle comes in the patrolling path, it can change the
route on its own and moves to the next predefined path. Once the construction work is
stopped for the day in the evening, it patrols the entire area and moves to that specific
area where any sound is detected. In this case of sound detection, it stops there and
takes pictures of the area from different angles and sends it to the administrator of the
area. In case of no sound detection, it keeps on monitoring the area. This patrolling
robot is highly beneficial when used in combination with a human security system,
as this robot is very useful in carrying out routine tasks. Using this robot for securing
construction sites can also compensate in case of scarcity of human resources. It also
helps in saving money and reduces the dangers to the life of workers as the number of
human personal involvement would be less by using this low-cost robot. Therefore, it
can be said that usage of patrolling robots will be a revolutionary step in the security
of construction sites.
References
1. Salman H, Acheampong S., Xu H (2018) Web-based wireless controlled robot for night vision
surveillance using shell script with Raspberry Pi. In: Advances in intelligent systems and
computing, pp 550–560, June 2018. https://doi.org/10.1007/978-3-319-93659-8_49
2. Kanagaraj R, Amsaveni MM, Binsha S, Chella Keerthana S (2022) Raspberry Pi-based spy
robot with facial recognition. In: Part of the Advances in intelligent systems and computing
book series (AISC), vol 1415. https://doi.org/10.17577/IJERTV8IS120298
3. Gopalkrishnan K, Thiruvenkatasamy S, Pabhakar E, Aarathi R (2019) Nightvision patrolling
rover navigation system for Women’s safety using machine learning posted. Int J Psychosoc
Rehabil 23. ISSN: 1475-7192. https://doi.org/10.37200/IJPR/V23I4/PR190440
4. Singh D, Nandgaonkar A (2018) IOT-based Wi-Fi surveillance robot with real-time audio and
video streaming. In: Part of the Advances in intelligent systems and computing book series
(AISC), vol 810. https://doi.org/10.1007/978-981-13-1513-8_65
5. Khalid F, Albab IH, Roy D, Asif AP, Shikder K (2021) Night patrolling robot. In: 2nd Interna-
tional conference on robotics, electrical and signal processing techniques (ICREST). https://
doi.org/10.1109/ICREST51555.2021.9331198
6. Devakunchari R, Bhowmick S, Bhutada SP, Shishodia Y (2019) Analysis of crimes against
women in India, using regression. Int J Eng Adv Technol (IJEAT) 8(2S2). ISSN: 2249-8958
7. Divya BN, Hegde B, Chaithanya BR, Moulya Raju MT, Shambhavi S (2021) Smart motion
detection surveillance rover with night patrolling for women’s safety and monitoring. Int J Res
Eng Sci Manag 4(7). ISSN (Online): 2581-792
8. Kitchagiri P, Vaddi S, Rajanala SR, Manthena B, Sanepalli S, Pindoo IA (2021) Night vision
patrolling robot for security patrolling using Raspberry Pi. Int J Eng Res Appl 11(5), Series-VI.
ISSN: 2248-9622
9. Ratnaparkhe VR, Bokade AU (2016) Video surveillance robot control using smartphone and
Raspberry Pi. In: 2016 International conference on communication and signal processing
(ICCSP). IEEE Conference Publication. https://doi.org/10.1109/ICCSP.2016.7754547
@seismicisolation
@seismicisolation
10. Yoon C, Park M, Eum H, Lee J (2013) Human action recognition for nightvision using temporal
templates with infrared thermal camera. In: 2013 10th International conference on ubiquitous
robots and ambient intelligence (URAI). IEEE Conference Publication. https://doi.org/10.1109/
URAI.2013.6677407
11. Tao W, Liu J, Zhang Y (2013) Sound source localization system based on microphone array. J
Comput Appl 32(5):1457–1459
@seismicisolation
@seismicisolation
Innovative Exploration Techniques:
Utilizing IoT-Enabled Robots for Safe
and Efficient Underground Tunnel
Investigation
N. Shravan , M. Manoj Kumar , Bharatesh Chakravarthi ,

and C. Bhargavi
1 Introduction
In recent years, the use of IoT-enabled robots has become increasingly popular in the
investigation of tunneling industry. These robots are equipped with various sensors
and cameras that allow them to navigate through underground tunnels, collecting
data and performing tasks that would be difficult or dangerous for humans to do.
The use of IoT-enabled robots has several advantages, including increased safety for
workers, improved efficiency, and reduced costs. By using robots to perform tasks
such as mapping, surveying, and inspecting underground structures, companies can
reduce the risk of accidents and injuries, while also improving the accuracy and speed
of data collection. In this article, we will explore the benefits of utilizing IoT-enabled
robots in analysis and maintenance of tunnels, as well as some of the challenges that
must be overcome to successfully implement this technology.
LIDAR is a technology that is used in several autonomous tasks and functions by
calculating the distance by measuring the time infrared light source takes to reflect
back to the photo-detectors. Additionally, Camera can also be used to get a real-world
view of how the environment in front of the robot is. This research is mainly based
on understanding of how robots can be easily operated with ROS and 3D mapping
and navigation of the robot.
N. Shravan (B) · M. Manoj Kumar

Department of Computer Science and Engineering, Jyothy Institute of Technology, Bangalore,
India
e-mail: shravan.nanjunda@gmail.com
B. Chakravarthi
School of Computing and Augmented Intelligence, Arizona State University, S Mill Avenue,
Tempe, AZ, USA
C. Bhargavi
Department of Civil Engineering, Jyothy Institute of Technology, Bangalore, India
@seismicisolation
@seismicisolation
72 N. Shravan et al.
Autonomous Navigation means ability of the robot to plan its path and execute its
plan with minimal human intervention. Navigation in unstructured environments is
a problem with many variants. The basic idea depends on the environment like land,
air, and water. The type of locomotion also depends on the structure of the robot-like
legs, wheels, propellers, magnetic field effect, etc. In this research, we will limit our
work to land environment, which has flat surface and we will use wheeled robot for
locomotion.
Autonomous systems play a vital role in assisting humans in various domains. For
example, driver-less cars, humanoid robots, military systems, etc. Assistive robotics
is a vast area of autonomous systems that help humans who require medical, mobility,
domestic, physical, and mental assistance. Data is one of the major components in
autonomous systems, examples of such data could be robots state and location coor-
dinates in the environment, obstacles, speed, acceleration etc. This data is updated
periodically and hence obtained directly from the system through a secure channel
and protocol. We review the past and present research using various sensors and
software. The main objective of the robot is to navigate in unknown environments
without damage and collision. Later after analyzing the map, the travel path can be
optimized.
The autonomous navigation system consists of the following concepts and Fig. 1
depicts the general outline of the robot.
• Perception: Interpretation by robot using sensors to sense the obstacles and locate
in its environment. By this, the robot can easily calculate the x and y coordinate
and prevent collision.
• Map Building: Generating a map will make the robot know the environments
boundary and obstacles present. It is required for easy navigation and collision
free path planning.
• Location: Calculating the x and y coordinates with respect to the generated map.
Using this, the robot can calculate its next position and travel there. LIDAR data
Fig. 1 A general outline of the robot
@seismicisolation
@seismicisolation
Innovative Exploration Techniques: Utilizing IoT-Enabled Robots … 73
has to be processed and any noise present must be addressed, so that mapping and
navigation is performed accurately.
• Planning: Deciding the next possible movements, the robot can perform without
collision and calculating the shortest path is necessary.
• Control: The robot should be capable of calculating alternative path in case the
previously calculated path is blocked for some reason. Also, the robot should not
collide with the obstacles in the environment.
• Obstacle Avoidance: The robot should be capable of sensing obstacles dynami-
cally and should not collide with them.
There are various applications where this concept of autonomous mapping and
navigation is used such as reconnaissance, military actions, hazardous material
handling, cleaning, inspection, and surveillance. The importance of these applica-
tions has led to the innovation in the field of autonomous locomotion and hence, new
technology and applications are under research and development.
Previous research conducted by scholars has explored the application of drones or
Unmanned Aerial Vehicles (UAVs) in various fields. These studies have highlighted
the advantages of using UAVs for aerial surveillance, mapping, and data collection
in inaccessible or hazardous areas. UAVs offer the ability to capture high-resolution
images, create detailed maps, and collect real-time data, providing valuable insights
for decision-making. However, in the context of underground tunnel, the use of
Unmanned Ground Vehicles (UGVs) presents distinct advantages. Unlike UAVs,
UGVs can operate within confined underground spaces, navigate complex terrains,
and directly interact with the environment. UGVs equipped with sensors can collect
data in real time, assess underground conditions, and ensure accurate resource explo-
ration. Their presence in underground environments enhances safety, efficiency, and
resource extraction optimization, making them well-suited for our proposed work.
2 Requirements in Tunnel Inspection and Preference to IoT

Applications
Civil Infrastructure requires inspecting and assessing along with efficient mainte-
nance and safe operation. This projects to be an interesting challenge when IoT is
included in the scenario. Tunnels, being one of the main infrastructural develop-
ments, have the ability to transport materials and men to far off places within shorter
time span and least distance that can possibly be covered. Tunnels are subjected
to various deteriorations like damages due to environmental factors, due to human
factors, loading factors, and ageing along with many other reasons. Examples of dete-
riorations include Ceiling collapse of Big Dig in 2003, Boston and Sasago Tunnel
Collapse, Tokyo in 2012 [3]. This indicates that there is an urgent need for tech-
nology to perform exhaustive inspection along with the same being automated and
cost effective.
@seismicisolation
@seismicisolation
Visual testing being one of the non-destructive tests plays a pivotal role in finding
discrepancies or damages in the tunnel. These discrepancies can be workmanship
errors, deterioration of materials among others. A well-trained eye can easily analyze
the indications of distress. Currently, tunnel inspections are performed periodically
by inspectors’ visual observations and a conclusion is made upon based on structural
defects and extent of these defects.
Tunnel inspection process has strict safety protocols and guidelines that must be
adhered to and in most cases, these safety procedures are not followed and thus lead to
increased risk of the personal involved. Multiple tools or equipment must be carried
to evaluate the various aspects of tunnels which can again cause inconvenience due to
limited working space inside the tunnels. Adequate lighting is required to ensure thor-
ough inspection and identify potential issues in the tunnels, else, might lead to missed
or misinterpreted findings. Absence of adequate walkways, platforms, or guiderails
might cause harm to the personal involved. The inspection process should incorporate
standardized methods for capturing and recording inspection data, including visual
evidence, sensor readings, and measurements [4]. This data serves as a reference
for assessing the tunnel’s condition, identifying trends, and planning maintenance
or repair activities. The activity would require large number of skilled labors to be
involved, imposing risk to each one of them. Tunnel inspections must be conducted
efficiently, providing accurate assessments of tunnel conditions, identifying potential
risks or issues, and enabling appropriate maintenance and repair actions which are
few shortcomings when human errors are involved.
Hence, there is an advent requirement for involving IoT-based technology so as
to minimize the risk and errors due to human intervention.
3 Challenges for IoT in Tunnel Inspection
There are few challenges that any IoT-enabled technologies might face during tunnel
inspections. They include:
• Limited connectivity, which makes the transmission of data difficult.
• Environmental conditions, which affect performance and durability of the devices.
• Locating and placing of the sensors to ensure that the right data is collected.
The proposed work tends to overcome these short falls in order to provide a
system, which is less cumbersome and more user friendly with efficient working
conditions.
@seismicisolation
@seismicisolation
4 Materials and Method
To investigate the utilization of IoT-enabled robots for safe and efficient underground
tunnel operations, a comprehensive approach combining hardware, software, and
experimental methodologies was employed [5]. The following materials and methods
were utilized in the research:
1. Robotic Hardware: A range of IoT-enabled robotic platforms and devices are
utilized in the study. These included autonomous robots equipped with sensors
such as LIDAR, cameras, gas detectors, temperature sensors, and proximity
sensors. The robot is designed to navigate challenging underground terrains,
perform tasks such as mapping and data collection, and interact with the
environment.
2. Sensor Technologies: Advanced sensor technologies are integrated into the
robotic platforms to gather real-time data about the underground tunnel envi-
ronments [6]. These sensors can provide information on geological conditions,
gas concentrations, temperature variations, structural stability, and other relevant
parameters. The collected data forms the basis for decision-making algorithms
and operational planning.
3. Communication and Connectivity: IoT protocols and communication technolo-
gies are employed to establish connectivity between the robotic platforms and a
centralized command center. Wireless networks, such as Wi-Fi or cellular connec-
tions, enable real-time data transmission, remote control, and monitoring of
the robots’ activities [7]. This communication infrastructure facilitates seamless
coordination and supervision of the operations.
4. Software Development: Customized software applications and algorithms are
developed to support the functionalities of the IoT-enabled robots. These included
mapping and navigation algorithms, data processing and analysis tools, risk
assessment models, and autonomous decision-making algorithms. The software
facilitates efficient data interpretation, optimization of exploration strategies, and
adaptive operations in dynamic underground environments.
5. Experimental Setup: Experimental setups are created to simulate underground
tunnel and mining environments. These setups consist of mock tunnels, varying
geological structures, and controlled conditions to replicate real-world scenarios.
The robots are deployed in these setups to evaluate their performance in terms
of safety, efficiency, and resource exploration capabilities.
6. Data Collection and Analysis: Data collected by the robotic platforms and sensors
are systematically recorded and analyzed. Various metrics such as mapping
accuracy, resource identification, gas concentrations, temperature variations, and
robot performance parameters are quantified. Statistical analysis, data visualiza-
tion techniques, and comparative evaluations are employed to derive meaningful
insights from the collected data [8].
7. Evaluation and Validation: The performance of the IoT-enabled robots and the
effectiveness of the implemented techniques are evaluated and validated through
field tests and real-world deployments. The robots’ capabilities in terms of safety
@seismicisolation
@seismicisolation
enhancements, efficiency improvements, and collaboration with human workers

are assessed. Comparative studies with traditional methods are conducted to
highlight the advantages of the IoT-enabled approach.
5 Proposed Work
The proposed robot system consists of three subsystems, they are sensors, motors,
and compute node. For the navigation system, we considered three options which are
GPS, LIDAR, and Camera. After considering the advantages and disadvantages, we
considered that LIDAR will be the best option. LIDAR is easy to use, cost effective,
and does not depend on the lighting conditions and weather conditions.
The robot consists of LIDAR of 8 m range which makes up the perception part.
The Raspberry pi 4B+ 8 GB SBC (Single Board Computer) makes up the computing
part. Motors are connected to the L298N 2A H-bridge motor driver which takes up
the locomotion part. The power to the whole system is supplied by a li-ion battery
pack of 11.1 V 2500 mAh. The raspberry pi requires 5v regulated constant power
supply, hence, we need to use a buck converter which regulates the 12–5 V. The inter-
connecting between all the components is shown in Fig. 2. The robot and computer
communicate with each other via Wi-Fi.
Later, we developed a program using ROS using which the Robot (Fig. 3) was able
to communicate to the Computer via Wi-Fi by steaming the pre-processed LIDAR
data. The navigation commands can be sent to the robot using the arrow keys on the
keyboard. This way the Map of the environment can be generated and saved. After
that the robot can move to the specified location in the environment autonomously
choosing the shortest path. If any obstacle is detected in its path and is not able to
proceed to the destination in the calculated path, then it can find the next feasible
alternative path to reach the destination.
Fig. 2 Interconnection of components
@seismicisolation
@seismicisolation
Fig. 3 Robot prototype
6 Integrated Civil Engineering Applications
The integration of IoT-enabled robots in civil engineering presents numerous appli-

cations that can enhance the efficiency, safety, and quality of construction projects.
Some key civil engineering applications of IoT-enabled robots include:
Construction Site Monitoring: IoT-enabled robots equipped with sensors and
cameras can be deployed to monitor construction sites in real time. These robots can
capture high-resolution images, record videos, and collect data on various parameters
such as site conditions, progress monitoring, and worker safety [9]. The collected
data enables project managers to make informed decisions, identify potential issues,
and ensure adherence to project schedules.
Automated Inspection and Maintenance: Robots equipped with advanced sensing
capabilities can conduct inspections of critical infrastructure elements such as
bridges, tunnels, and dams. These robots can identify structural defects, measure
dimensions, and detect signs of deterioration. Through IoT connectivity, inspection
data can be transmitted to a centralized system for analysis, allowing engineers to
proactively plan maintenance and repairs, reducing the risk of failures and improving
asset lifespan.
Smart Infrastructure Monitoring: IoT-enabled robots can be deployed to monitor
the structural health of infrastructure such as buildings, bridges, and roads. Equipped
with sensors, these robots can collect data on factors like vibrations, temperature vari-
ations, and load distribution. The real-time data enables early detection of anomalies
@seismicisolation
@seismicisolation
and potential risks, facilitating timely maintenance interventions and ensuring the
safety of the infrastructure.
Autonomous Construction and Excavation: Robots can be employed for
autonomous construction tasks, such as bricklaying, concrete pouring, and exca-
vation. These robots can be programmed to perform repetitive and physically
demanding tasks with precision and consistency [10]. By automating these tasks,
construction projects can benefit from increased productivity, reduced labor costs,
and improved quality control.
Hazardous Environment Exploration: In civil engineering projects involving
hazardous environments such as contaminated sites or areas with high radiation
levels, IoT-enabled robots can play a crucial role. These robots can be equipped with
specialized sensors to detect and analyze hazardous substances, allowing for safe
and efficient site investigations. The robots can navigate through challenging terrain
and provide valuable data for decision-making during remediation efforts.
Geotechnical Monitoring: Robotic systems can be utilized for geotechnical moni-
toring in civil engineering projects. These robots can be deployed to collect data on
soil conditions, groundwater levels, and stability of slopes [11]. The data obtained
helps engineers in assessing the feasibility and safety of construction projects,
optimizing foundation designs, and implementing appropriate mitigation measures.
Disaster Response and Recovery: During natural disasters or emergencies, IoT-
enabled robots can be deployed for search and rescue operations, structural assess-
ments, and damage evaluation. These robots can access hazardous areas, collect vital
information, and assist in the recovery process. They enable rapid response, improve
situational awareness, and help ensure the safety of emergency responders.
IoT-enabled robots in civil engineering applications, construction projects can
benefit from improved efficiency, reduced costs, enhanced safety measures, and
better-quality control. The use of these technologies offers opportunities to opti-
mize construction processes, mitigate risks, and enhance the overall sustainability of
infrastructure development.
7 Results
A visualization of surrounding objects of the robot, shown in Fig. 4, was gener-

ated using python and matplotlib library. The blue color dot in the center indicates
the origin and represents the location of the robot. The surrounding obstacles are
measured by LIDAR sensor and by calculating the angle and distance of each reading,
a graph can be plotted which represents the top view.
We conducted some experiments at different environments. Figure 5 represents
the point cloud inside a cave and Fig. 6. represents point cloud inside a tunnel. Green
points represent the boundary and red points indicate that specific part needs to be
processed in order to get more clarity.
@seismicisolation
@seismicisolation
Fig. 4 LIDAR data plotted on a graph
Fig. 5 Point cloud inside a cave
@seismicisolation
@seismicisolation
Fig. 6 Point cloud inside a tunnel
8 Conclusion
The utilization of IoT-enabled robots in underground tunnel inspection has the

potential to revolutionize the industry, providing safe and efficient exploration
techniques.
• The integration of advanced sensor technologies, autonomous decision-making
algorithms, and collaborative frameworks between humans and robots offers
significant advantages such as enhanced safety measures, real-time data gathering,
remote monitoring and control, adaptability, and improved productivity.
• Previous research in this field has paved the way for further advancements, demon-
strating the feasibility and benefits of employing IoT-enabled robots in under-
ground environments. The proposed work aims to build upon these advancements
by developing advanced sensing technologies, autonomous decision-making
algorithms, and human–robot collaboration frameworks.
@seismicisolation
@seismicisolation
• By embracing these innovations, the industry can achieve safer working condi-
tions, increased operational efficiency, optimized resource extraction, and sustain-
able practices. The integration of IoT-enabled robots in civil engineering appli-
cations further expands the benefits to tunneling projects, enabling real-time
monitoring, automated inspections, and improved management.
• Overall, the adoption of IoT-enabled robots in underground tunneling holds
immense potential for the industry, creating safer, more efficient, and sustainable
means to facilitate maintenance of tunnels built many years ago.
References
1. Yan H, Zhang W, Wang L (2019) IoT-enabled tunnel monitoring system based on wireless
sensor networks. Int J Distrib Sens Netw 15(5):1550147719846301
2. Valente A, Jimenez LM, Calderon A, Skarmeta AF (2020) IoT-based environmental monitoring
for tunnel construction projects. In: 2020 Global IoT summit (GIoTS). IEEE, pp 1–6
3. Montero R, Victores JG, Martínez S, Jardón A, Balaguer C (2015) Past, present and future of
robotic tunnel inspection. Autom Constr 59:99–112
4. Wu Y, Liao C, Shih C (2020) A review of the applications of IoT sensing technology in tunnel
engineering. Adv Civ Eng
5. Huang Y, Zhou S, Ma X, Hu X (2020) A framework for IoT-based tunnel safety monitoring
and early warning systems. Sensors 20(17):4976
6. Wang D, Zhang Y, Li J, Zhang Y, Li Y (2020) IoT-based system for tunnel health monitoring
and safety early warning. Appl Sci 10(15):5286
7. Zhao C, Wang Y, Hu C (2020) An IoT-enabled real-time monitoring system for road
infrastructure safety. IEEE Access 8:135002–135012
8. Liu J, Wu J, Han R, Han J (2021) An IoT-enabled sensing system for landslide monitoring in
mountainous areas. Sensors 21(4):1364
9. Tsuchiya Y, Zhang T, Higashino T, Kato N (2021) An IoT-enabled framework for continuous
structural health monitoring of civil infrastructure. Sensors 21(7):2500
10. Kurnaz E, Sert M, Kurban M, Capar H (2021) IoT-enabled structural health monitoring system
for bridges. In: Proceedings of the 14th international conference on electronics, computer and
computation (ICECCO), pp 199–202
11. Lee DH, Kim DS (2021) IoT-based real-time monitoring system for slope instability in
geotechnical engineering. Sustainability 13(7):3847
@seismicisolation
@seismicisolation
Event-Based Sensing for Improved
Traffic Detection and Tracking
in Intelligent Transport Systems Toward
Sustainable Mobility
Bharatesh Chakravarthi , M. Manoj Kumar , and B. N. Pavan Kumar
1 Introduction
Roadside traffic monitoring plays a vital role in intelligent transportation systems,

contributing to efficient traffic management, enhanced safety, and improved overall
mobility. Traditional traffic monitoring systems primarily rely on cameras and
sensors to capture visual information regarding traffic flow, speed, and volume [1].
However, these systems encounter various limitations, such as high data processing
requirements, reduced accuracy in challenging lighting conditions, and difficulties
in detecting fast-moving objects [2]. To overcome these limitations and advance the
field of traffic monitoring, event-based vision systems have emerged as a promising
alternative. Event-based vision systems capture changes in the visual scene as they
occur, providing a stream of asynchronous and sparse events rather than traditional
frame-based data. This unique approach reduces the amount of data that needs to be
processed and transmitted, enabling real-time detection and tracking of fast-moving
objects with high accuracy, even in challenging lighting conditions.
While event-based vision holds great potential for traffic monitoring applications
[3], its adoption is still in its early stages. In this study, we aim to understand the
current state-of-the-art in event-based vision for real-time traffic monitoring systems.
Our focus is on the challenging task of detecting and tracking traffic participants in
B. Chakravarthi (B)
School of Computing and Augmented Intelligence, Arizona State University, S Mill Avenue,
Tempe, AZ 85281, USA
e-mail: chakravarthi.bsb@asu.edu
M. Manoj Kumar
Computer Science and Engineering, Jyothy Institute of Technology, Off Kanakapura Road,
Bengaluru, Karnataka 560082, India
B. N. Pavan Kumar
Computer Science and Engineering, Indian Institute of Information Technology, Sri City, Gnan
Marg, Sri City, Andhra Pradesh 517646, India
@seismicisolation
@seismicisolation
84 B. Chakravarthi et al.
complex traffic scenarios characterized by dynamic movement and difficult lighting

conditions. To address this problem, we have developed a pipeline that enables the
visualization of both real-time and synthetic event-based data streams in 2D and 3D
space, providing a better visual understanding of the data. This pipeline incorporates
preprocessing features, such as adaptive event rate control and optical flow estimation
to improve the quality of the raw event data. Furthermore, we have demonstrated a
pipeline for event-based traffic participant detection and tracking using automotive
datasets and sample recordings offered by Prophesee Metavision Intelligence Suite.
This study involves detecting and tracking different types of traffic participants,
including vehicles on highways, pedestrians crossing the street, and wheelchair users,
in both stationary and moving vehicle scenarios. To gain detailed insights into the
distribution of events over space and time, we utilized techniques such as histogram
difference and exponential-decay time surface analysis.
Event-based vision systems offer advantages over traditional fixed-rate cameras,
addressing limitations in data processing, accuracy in challenging lighting condi-
tions, and capturing fast-moving objects [1, 2]. These systems generate asynchronous
and sparse events, reducing data volume and enabling more efficient processing [4].
Event-based cameras excel in accurately detecting moving vehicles and pedestrians,
utilizing techniques such as event-based optical flow and neural networks [3, 5–7].
Challenges remain in the limited availability of event-based datasets, necessitating
efforts for diverse dataset creation [8–11]. Researchers have developed pipelines
incorporating preprocessing, visualization, adaptive rate control, and optical flow
estimation to enhance event-based traffic monitoring capabilities [12, 13]. Further
research is needed to fully exploit the potential of event-based systems for sustainable
traffic management and improved road safety. To summarize, this pilot study aims to
advance the adoption of event-based vision in real-time traffic monitoring systems.
By exploring available datasets, developing tools for synthetic event generation, visu-
alization, and preprocessing, and demonstrating the pipeline for traffic participant
detection and tracking, we aim to provide a comprehensive understanding of the
current state-of-the-art and potential applications of event-based vision for traffic
monitoring.
2 Methods to Enhance Event Stream Analysis
Event cameras, also known as neuromorphic cameras or dynamic vision sensors,

capture visual information in a fundamentally different way compared to traditional
cameras. Instead of capturing frames of pixels at a fixed rate, event cameras produce
a continuous stream of asynchronous events, each indicating a change in luminance
at a specific pixel location.
@seismicisolation
@seismicisolation
Event-Based Sensing for Improved Traffic Detection and Tracking … 85
Fig. 1 Event data generation: (a) raw event data of a fidget spinner captured using an event sensor,
(b) frame-based to event-based representation
2.1 Event Generation
The generation of events in an event camera is based on the logarithmic change in

light intensity at each pixel. The sensor outputs a stream of events that encode the
change in logarithmic intensity over time. This is mathematically modeled by the
equation:

It+1
Δ log(I ) = log(It+1 ) − log(It ) = −α log (1)
It
where Δlog(I) represents the change in logarithmic intensity, I t+1 is the new intensity
level, I t is the previous intensity level, and α is a constant determining the sensor’s
sensitivity. Event frames can be generated periodically over time or at user-defined
intervals, allowing precise control over the event generation rate. Figure 1a shows
an example of raw events captured by a Prophesee Evaluation Kit 3.1 (EVK3.1)
VGA sensor [14]. Figure 1b demonstrates an event-based frame converted from a
conventional camera-based video stream using a frame-to-event simulator.
2.2 Event Format
An event stream consists of ‘n’ events, where each event contains the following
information encoded in (x, y, p, t) format:
@seismicisolation
@seismicisolation
Fig. 2 Event polarity filter: (a) positive polarity events, (b) positive and negative polarity events,
and (c) negative polarity events
• The pixel location (x, y) where the event occurred.

• The polarity ‘p’ of the event, with 1 indicating an increase in intensity (ON event)
and 0 indicating a decrease in intensity (OFF event).
• The timestamp ‘t’ indicates the time when the event was generated.
Figure 1a illustrates a raw event represented in (x, y, t) format in different views
(front and top). A polarity filter can be applied to the event sensor output to extract
events with specific polarities, filter out noise, and identify key features. Figure 2
demonstrates the polarity filter representation of an event frame showing a pedestrian.
2.3 Event Visualization
Event visualization is an important aspect of event sensing as it helps in understanding

the output of the sensor. One common way to visualize the event stream is using an
event display to provide a compact way of visualizing the output of an event sensor,
and they can be used to detect motion, track objects, and perform other computer
vision tasks. An event display can be a 3D plot (as shown in Fig. 1a) or a 2D plot
where each pixel location is represented as a dot, and the polarity of the event is
indicated by the color of the dot (as in Fig. 2). The timestamps of the events can be
represented by the brightness of the dot. Figure 3 depicts a 2D event visualization of
different traffic scenarios.
@seismicisolation
@seismicisolation
Fig. 3 Event visualization: event data visualization in 2D space with three different color themes
(dark, light, and gray in top-to-bottom order): (a) shows an event frame of cars moving on a
highway, (b) shows pedestrians crossing on a street, and (c) shows a person in a wheelchair and
other pedestrians on a sidewalk of a road
2.4 Adaptive Rate Event Filtering and Optical Flow

Estimation
The adaptive rate event filter is a technique to extract a sequence of sharp event frames
from the asynchronous event stream, using a variable time window that depends on
the content of the event stream itself. The filter adaptively adjusts the time duration
and number of events per frame, so that the frames contain enough information to
be useful, but not so much that the frames become blurred or redundant. The time
window is of a fixed duration and slides along the timeline of the event stream. For
each time window, the filter computes the event rate within the window and compares
it to a threshold value. If the event rate is higher than the threshold, the events within
the window are passed through the filter, and if the rate is lower, the events are
discarded. Also, the filter uses a variance threshold to determine when to generate
a new frame, and the down sampling factor ‘M’ can also be adjusted to reduce the
resolution of the frames.
The threshold value is adaptive and is updated after each time window based on
the event rate. If the event rate is higher than the threshold, the threshold is increased,
and if the event rate is lower than the threshold, the threshold is decreased. The rate of
increase or decrease is determined by two parameters, ‘α’ and ‘β’. The ‘α’ controls
@seismicisolation
@seismicisolation
the speed at which the threshold is increased when the event rate is high, and ‘β’
controls the speed at which the threshold is decreased when the event rate is low.
Algorithm 1 Adaptive Rate Event Filter
1: Let E i be the event rate within the ith time window

2: Let T i be the adaptive threshold for the ith time window
3: if E i > T i then
4: Pass events within the ith time window through the filter
5: else if E i ≤ T i then
6: Discard events within the ith time window
7: end if
8: if E i > T i then
9: T i + 1 ← T i + α(E i − T i )
10: else if E i ≤ T i then
11: T i+1 ← T i + β(E i − T i )
12: end if
Algorithm 1 depicts the mathematical model of the adaptive rate event filter. It
takes in the event rate ‘E i ’ for the ‘i’th time window and uses an adaptive threshold
‘T i ’ to determine whether to pass events through the filter or discard them. The
threshold is updated after each time window based on the event rate, with the
rate of increase or decrease determined by the constants ‘α’ and ‘β’. The overall
resulting frames can then be used in a range of downstream applications, such as
object detection, tracking, or optical flow estimation.
Figure 4a shows the adaptive rate event filter applied to raw event streams of
different scenarios. Processing raw event data generated by neuromorphic sensors
or simulators and extracting sparse optical flow information is essential for further
analysis. To handle large amounts of event data in real time, it is more efficient to
generate clusters of events in a sparse way instead of processing each event individu-
ally. Each cluster of events is represented as a pixel, and the optical flow is computed
for each pixel that represents a cluster of events. The mathematical model to compute
the optical flow is represented by the following equation.

Ix2 Ix I y Ix I y I y2 (uv)( I x It I y It ) (2)
where ‘I x ’ and ‘I y ’ are the spatial gradients of the image patch, it is the temporal
gradient of the image patch, and u and v are the x and y components of the optical
flow vector. The optical flow vector ‘[u; v]’ is then computed using the following
equation.

[u : v] = inverse Ix2 Ix I y I y2 I x It (3)
Equations (2) and (3) are applied to each window of events around each pixel, and
the resulting system of linear equations is solved to estimate the optical flow vector
@seismicisolation
@seismicisolation
Fig. 4 Adaptive rate event filtering and optical flow estimation: (a) event filter (T i = 0.1, ‘M’ =
2) applied to three scenarios (left to right)—vehicles moving on a highway, pedestrians crossing,
and driving on a street road. (b) Sparse optical flow for pedestrian crossing scenarios
for that pixel. The optical flow vector minimizes the sum of squared errors between
the image patch in the current frame and the corresponding patch in the reference
frame.
The resulting optical flow map represents the estimated motion of the events
within the corresponding cluster. The optical flow map can then be visualized as a
vector (arrows) field as shown in Fig. 4b, where the direction and length of each
vector represent the direction and magnitude of the estimated motion. The computed
optical flow can be used for intention prediction in the form of trajectory as the
optical flow provides information about the movement of objects in an event frame,
which can be used to track the trajectory of objects. For example, if a pedestrian
is walking toward a particular location, the optical flow vectors in the event frame
sequence will point in the direction of their movement. By analyzing the direction
and magnitude of these optical flow vectors, the pedestrian’s movement intention
toward a destination can be predicted.
@seismicisolation
@seismicisolation
3 Experimental Study
In this section, we conducted an experimental study to evaluate the current level of

performance achieved by event-based traffic participant detection and tracking. To
accomplish this, we utilized the metavision intelligence package [13] and leveraged
a pre-trained model [15] based on feed-forward event-based architectures. Due to the
shortage of labeled data available for event-based vehicle and pedestrian detection,
we restricted our experiments to datasets that only encompassed certain scenarios.
Our preliminary findings provide insights into the potential benefits of using an event-
based vision system to enhance traffic monitoring and observations of some failure
cases.
3.1 Datasets
In our experiments, we used an open-source detection and tracking inference pipeline

implemented in Python. We used datasets provided by the Metavision Intelligence
software framework for event-based vision, as well as the Prophesee 1 Megapixel
automotive detection dataset [13]. The datasets we used included recordings of cars
on a highway (30 s) captured by a static event camera (Prophesee GEN 3.0) [12],
pedestrians’ data (28 s) recorded using a static event camera (Prophesee GEN 4.1)
[16], and driving on a street road data (60 s) captured using 1 Megapixel event camera
mounted behind the windshield of a car [17]. Prior to conducting the experiments,
we preprocessed the datasets to match the pre-trained model format. Each dataset is
composed of 14 bits for the ‘x’ position, 14 bits for the ‘y’ position, and 1 bit for the
polarity.
3.2 Experiments
We implemented an inference pipeline for vehicle and pedestrian detection and

tracking using a pre-trained TorchScript model. The model was trained using 2D
bounding box annotations in numpy format, with the boxes containing the fields
such as ‘x’ the horizontal coordinate of the top-left corner of the box in pixels, ‘y’
the vertical coordinate of the top-left corner of the box in pixels, ‘w’ the width of the
box in pixels, ‘h’ the height of the box in pixels, ‘t s ’ the timestamp of the box in the
sequence in microseconds, indicating when the object was detected, and the ‘class id’
a numeric identifier for the object class, ranging from 0 to 6, with classes including
pedestrians, two wheelers, cars, trucks, buses, traffic signs, and traffic lights.
We included several functionalities to the pre-trained model, including input defi-
nition (specifying the input source, sampling period, start and end timestamps),
output generation (producing inference video and exporting detected and tracked
@seismicisolation
@seismicisolation
bounding boxes), object detection (defining the pre-trained detection model and its
calibrated hyperparameters and setting inference thresholds for detection confidence
and NMS-IoU level), geometric preprocessing (providing geometric preprocessing
of the event stream, such as input transposition and filtering events outside of a
Region of Interest), noise filtering (using Trail and STC filters), and data association
(defining matching thresholds for tracking confidence and NMS-IoU level, along
with other association parameters).
3.3 Preliminary Findings
We conducted the experiment using a pre-trained TorchScript model for object

detection and tracking in scenarios involving vehicles and pedestrians. The model
uses a histogram-based preprocessing technique to process the event data before
performing object detection. This technique involves dividing the event data into
multiple time bins and generating histograms of the events that occur within each
bin. These histograms are then used as input to the detection model, which allows
it to effectively capture both the spatial and temporal characteristics of the event
data, resulting in improved object detection performance. We conducted preliminary
testing on this model and found that it has faster inference times and lower memory
usage compared to other preprocessing methods. We input the preprocessed event
data into the model’s forward method, which produced a set of bounding boxes and
class probabilities for each detected object. The input and output data, containing the
detected and tracked bounding boxes, were displayed in two windows side by side as
shown in Fig. 5. The input event stream was shown in the left pane, and the output in
the right window, with colors indicating the object’s class membership and tracking
drawn as a trajectory with colors indicating the track ID and confidence level.
In order to analyze and compare the event characteristics of the datasets used
in this study, we employed a technique called histogram difference. This technique
involves computing the difference between the positive and negative event data at
each pixel position ‘(y, x)’ and time ‘t’, using the formula,
D(t, p, y, x) = H (t, 1, y, x) − H (t, 0, y, x); x, y, t ∈ [0, width]

× 0, length × [0, A] (4)
By doing this, we were able to visualize the resulting difference as histograms,

which revealed important insights that were not immediately obvious from the raw
data itself as shown in Fig. 6. It should be noted that if there is a significant amount of
activity in one area of the field of view, the same histogram cell may receive several
positive and negative events. Therefore, computing the difference between positive
and negative events that occur in the same histogram cell is useful for understanding
the spatial and temporal distribution of the differences between the positive and
negative events.
@seismicisolation
@seismicisolation
Fig. 5 Event-based detection and tracking results: the left side of the window shows the input raw
event stream, while the right side shows the resulting output of the model, including detection and
tracking indicated with 2D bounding boxes and trajectories
Fig. 6 Event histograms: shows the spatial and temporal distribution of the differences between
the positive and negative events. A more gradual change of the number of events over a time bin
could be observed with respect to different scenarios (left to right)
To evaluate the effectiveness of event-based detection in low-light situations, we

utilize the exponential-decay time surface method to analyze the temporal distri-
bution of events over time. This allows us to visualize recent events in each pixel,
giving more weight to newer events than those in the distant past as shown in Fig. 7.
By examining the time surface, we can identify vehicle and pedestrian structures
that may not be immediately discernible from the raw event data. This approach
provides a detailed insight into the performance of event-based detection under chal-
lenging lighting conditions. The equation for the Exponential-Decay Time Surface
for a single event with polarity ‘p’, position ‘(y, x)’, and timestamp ‘t i ’ is,
@seismicisolation
@seismicisolation
Fig. 7 Exponential-decay time surface: shows a particular event frame in different scenarios: (left
to right) driving on a road, pedestrians crossing, and vehicles on a highway. Each scenario is
represented by two-time surfaces, one for each polarity (0 and 1), and the decay parameter is set to
τ = 0.05 s

−(ti − t)
TSt ( p, y, x) = exp (5)
t≤ti
τ
where ‘τ ’ is the decay time constant. Equation 5 sums up the exponential decay of past
events at each pixel location ‘(y, x)’ and polarity ‘p’, with the summation starting from
the beginning of time up to the timestamp ‘t i ’ of the current event. The resulting value
represents the temporal activity of the pixel and reflects the contribution of all past
events that have occurred at or before ‘t i ’, with more recent events contributing more
to the value due to the exponential decay. During our experiments, we adjusted the
parameters, such as the minimum and maximum size of the objects to track in pixels,
while using the datasets used in this study for event-based vehicle and pedestrian
detection. We noted instances where the performance of the pre-trained models was
suboptimal in detecting and tracking vehicles and pedestrians, as depicted in Fig. 8.
Fig. 8 Failures in detection and tracking: illustrative examples of failures in detection and tracking:
showing incorrect detection and tracking scenarios from the three datasets used in this study
@seismicisolation
@seismicisolation
4 Conclusion
This paper presents a pilot study on event sensing systems for traffic monitoring
in the context of Intelligent Transport Systems (ITS) for Sustainable Mobility. The
study showcases the advantages of event-based cameras over traditional cameras in
detecting and tracking traffic participants, particularly in difficult lighting conditions.
A comprehensive pipeline is introduced, synthesizing event frames and providing
intuitive visualizations of real-time and synthetic event data in both 2D and 3D
space, enabling a comprehensive understanding of traffic dynamics. The pipeline
incorporates an adaptive event rate filter, optical flow estimation, and a pre-trained
TorchScript model for object detection and tracking. The study highlights the need for
more event-based datasets for traffic monitoring and suggests future work involving
advanced preprocessing techniques and the integration of event-based and frame-
based approaches. Overall, this research contributes to the development of event-
based sensing for sustainable mobility, offering potential improvements in traffic
monitoring accuracy and reliability in challenging environments.
References
1. Datondji SR, Dupuis Y, Subirats P, Vasseur P (2016) A survey of vision-based traffic monitoring
of road intersections. IEEE Trans Intell Transport Syst 17(10):2681–2698
2. Chen J, Wang Q, Cheng HH, Peng W, Xu W (2022) A review of vision-based traffic semantic
understanding in ITSs. IEEE Trans Intell Transport Syst
3. Gallego G, Delbrück T, Orchard G, Bartolozzi C, Taba B, Censi A, Leutenegger S et al (2020)
Event-based vision: a survey. IEEE Trans Pattern Anal Mach Intell 44(1):154–180
4. Holešovský O, Škoviera R, Hlaváč V, Vítek R (2021) Experimental comparison between event
and global shutter cameras. Sensors 21(4):1137
5. Mitrokhin A, Fermüller C, Parameshwara C, Aloimonos Y (2018) Event-based moving object
detection and tracking. In: 2018 IEEE/RSJ international conference on intelligent robots and
systems (IROS). IEEE, pp 1–9
6. Mondal A, Giraldo JH, Bouwmans T, Chowdhury AS (2021) Moving object detection for event-
based vision using graph spectral clustering. In: Proceedings of the IEEE/CVF international
conference on computer vision, pp 876–884
7. Wan J, Xia M, Huang Z, Tian L, Zheng X, Chang V, Zhu Y, Wang H (2021) Event-based
pedestrian detection using dynamic vision sensors. Electronics 10(8):888
8. Hu Y, Liu S-C, Delbruck T (2021) v2e: from video frames to realistic DVS events. In:
Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp
1312–1321
9. Binas J, Neil D, Liu S-C, Delbruck T (2017) DDD17: end-to-end DAVIS driving dataset. arXiv
preprint arXiv:1711.01458
10. Muglikar M, Gehrig M, Gehrig D, Scaramuzza D (2021) How to calibrate your event camera.
In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp
1403–1409
11. Rebecq H, Gehrig D, Scaramuzza D (2018) ESIM: an open event camera simulator. In:
Conference on robot learning. PMLR, pp 969–982
12. Perot E, De Tournemire P, Nitti D, Masci J, Sironi A (2020) Learning to detect objects with a
1 megapixel event camera. Adv Neural Inf Process Syst 33:16639–16652
@seismicisolation
@seismicisolation
13. Machines PM (2023) Metavision Intelligence Suite—v3.1.2. https://docs.prophesee.ai/.

Accessed 20 Mar 2023
14. Metavision P (2023) Metavision® EVK3—VGA/HD. https://www.prophesee.ai/event-based-
evk-3/. Accessed 20 Mar 2023
15. Metavision P (2023) Pre-trained model zoo—object detection TorchScript models. https://docs.
prophesee.ai/stable/metavisionsdk/modules/ml/pre-trainedmodels.html. Accessed 20 Mar
2023
16. Metavision P (2023) Recordings and datasets. https://docs.prophesee.ai/stable/datasets.html.
Accessed 20 Mar 2023
17. Metavision P (2023) Prophesee automotive dataset toolbox. https://github.com/prophesee-ai/
prophesee-automotive-dataset-toolbox. Accessed 20 Mar 2023
@seismicisolation
@seismicisolation
Human Activity Recognition
in Construction Industry Using Machine
Learning Pose Estimation Technique
M. Manoj Kumar , Bhuvaneshwari Hegde, S. P. Veda Murthy,

M. K. Akhila, and A. S. Bhoomika
1 Introduction
Human activity recognition is a rapidly growing area of research that aims to auto-
matically identify and track human physical activities. This technology has become
increasingly popular in recent years due to the widespread adoption of wearable
devices, such as smartwatches and fitness trackers, which provide an easy way to
collect biometric data and monitor the activities [1]. The recognition of human activ-
ities has become a prominent area of interest within the research field due to its
complex nature as a time series classification. The process of predicting human activ-
ities involves recording sensor data related to specific subjects’ actions, followed by
training a machine learning model to generalize patterns for unseen data. The applica-
tions of activity recognition are diverse, encompassing behavior analysis, monitoring
workouts, analyzing gait, recognizing gestures, and enhancing video surveillance [2].
Among all the applications of Human activity recognition, this work is focusing
on activities done in construction industries. Construction activity recognition is
the process of automatically identifying and classifying construction-related activ-
ities using sensor data and machine learning algorithms. It involves analyzing data
from various sources such as video cameras, wearable sensors, and environmental
sensors to extract meaningful information about construction activities [2]. Construc-
tion activity recognition has become increasingly important in recent years due to
its potential to improve safety, productivity, and efficiency on construction sites.
By recognizing and understanding construction activities, it is possible to identify
potential safety hazards and take corrective actions to prevent accidents. It also allows
project managers to monitor progress, identify areas for improvement, and optimize
workflows.
M. Manoj Kumar (B) · B. Hegde · S. P. Veda Murthy · M. K. Akhila · A. S. Bhoomika

Department of Computer Science and Engineering, Jyothy Institute of Technology, Bengaluru,
India
e-mail: dbamanoj@gmail.com
@seismicisolation
@seismicisolation
98 M. Manoj Kumar et al.
Activity recognition model across various problems, sensors, and activities have
significant challenge in this field. The activity signals can exhibit substantial varia-
tions among individuals, and even the same person may perform an activity differ-
ently at different times. Moreover, distinguishing between different activities can be
difficult as they may share similar signal patterns, leading to potential confusion in
the learning process [2]. Additionally, challenges include managing the computa-
tional cost for implementation on embedded and portable devices, ensuring accurate
data annotation for training purposes, addressing the wide range of complex daily
activities, and safeguarding the privacy of the subjects involved [3].
Numerous innovative ideas are emerging, and ongoing experiments are continu-
ously pushing the boundaries of model architecture in order to achieve more accu-
rate recognition of human activity. These efforts encompass domains including deep
learning, recurrent neural networks, convolutional neural networks, attention mech-
anisms, and ensemble methods, among others. In many of the recent studies [4–
7], the author proposed framework for detecting and tracking 33-body joint key
points, including body, torso, arms, leg, and face, is impressive and ideal for fitness
applications and mobility activities. The use of Percentage of correct key points
(PCK) evaluation metrics during training and testing operations to learn weight
distribution is a sound approach that has yielded state-of-the-art performance and
better results. Computer vision algorithms are employed to identify human pose,
body movements, and actions using various techniques. While previous works have
focused on improving accuracy or efficiency in terms of speed, many techniques
struggle with intensive computational demands and low-latency requirements. In
this study, they have presented a novel approach for single-person pose estimation
and action recognition specifically designed for fitness applications and mobility
activities. This proposed framework utilizes a base network for initial pose esti-
mation, followed by refinement through the Intensive Feature Consistency (IFC)
network. The IFC network applies high-level constraints for global body intensity
correction and local body part adjustments, mitigating the impact of body joint move-
ment diversity. Experimental results of this paper demonstrate the effectiveness of
our framework, showcasing improved pose estimation accuracy on two benchmark
datasets. Notably, the IFC network achieves a state-of-the-art performance with a
real-time processing speed of 31 FPS on the CPU platform, surpassing existing
methods with 99.3% PCK body accuracy and 95.1% PCK torso accuracy [8].
While these hybrid methods occasionally yield superior outcomes, the majority
of these approaches tend to be difficult and suffer from computational inefficiency
[6, 9–11]. In this work, our system utilizes two camera perspectives to accurately
capture body landmarks that we use to calculate a set of heuristic values as shown in
Fig. 1. These heuristics were specifically selected to capture the fundamental aspects
of painting form. Our framework yielded promising preliminary results and opened
various possibilities for future enhancements.
In our proposed approach for wall painting in construction industry, computer
vision-based motion tracking and analysis is utilized through the MediaPipe frame-
work. This involves extracting the key points of the human skeleton from the input
video or image data, which are then used to apply specific motion constraints based
@seismicisolation
@seismicisolation
Human Activity Recognition in Construction Industry Using Machine … 99
Fig. 1 A fine-grained multi-level hierarchical model for accurate human pose estimation and joint
localization in dynamic environments
on the instructions provided by an expert. The proposed system tackles the draw-
backs of current human pose estimation models by integrating domain expertise
from painting professionals. This fusion of knowledge bridges the gap between pose
estimation and optimal postures, enabling the device to provide users with more
precise and meaningful feedback. This innovative integration of technology holds
great promise in elevating the effectiveness of wall painting in construction industry
and enhancing overall performance level.
In this work, our contributions are (a) Enhanced Activity Recognition in Construc-
tion industry by accurately tracking human body posture from video, the technology
can enable activity recognition in civil work. (b) Real-Time Performance on Various
devices, which include modern mobile phones, desktops/laptops, Python, and even
web-based platforms. (c) Machine Learning Solution for High-Fidelity Body Posture
Tracking by utilizing the Machine Learning Pose Detection Application program-
ming interface, predefined 3D landmarks, and background segmentation masks for
the entire body from RGB video frames. To summarize, our study aims to recognize
the human activity in wall painting and track the human body motion continuously
to provide real-time feedback.
The rest of the paper is organized as follows: In Sect. 2, Materials and methods
for human activity recognition are presented. In Sect. 3, we presented our proposed
work. In Sect. 4, results and analysis are shown. In Sect. 5, conclusions are drawn.
@seismicisolation
@seismicisolation
2 Materials and Method
To develop a fitness application with the main components mentioned, you would
need the following materials and methods:
2.1 Materials
Model Selection
To get accurate results, it’s important to pick the right human posture model and key
point detection methods. Human posture estimate frequently employs three typical
categories of human models:
Skeleton-Based Model: The Kinematic Model is used to estimate 2D and 3D
poses. To reflect the structure of the human body, this model comprises a variety of
joint positions and limb orientations that capture the relationships between various
body parts.
Contour-Based Planar Model: This model is used to estimate 2D posture. The
appearance and shape of the human body are depicted using planar models. Multiple
rectangles with approximate human body contours are used to represent the body
parts. Principal component analysis is used to capture the deformations of the
silhouette as well as the entire human body graph.
Volume-Based Model: This model is used to estimate 3D poses. It consists of
many common 3D human body models and postures that are represented by human
geometric meshes and forms, typically captured for machine learning-based 3D
human pose estimation. In this work, we have used the processing and assembly
of various media data, including audio, video, and image data, is made possible via
the open-source framework known as MediaPipe, which provides pre-built examples
and pre-trained models for a number of applications, such as Face Detection, Face
Mesh, and Pose. MediaPipe BlazePose is used by Pose, a machine learning system
that infers 33 3D landmarks and a backdrop segmentation mask for the full body to
achieve high-fidelity body pose tracking.
Camera Selection
In this work, RGB and 170-degree fisheye cameras were chosen. The fisheye camera
offers a broader field of view, allowing us to position the camera as near to the appa-
ratus as 3 feet away while still getting the whole image of the exerciser. Depending
on the type of exercise, the fisheye camera is utilized to collect video data from the
side and front perspectives, while the RGB camera is used for the top view, in order
to get the most data for corrective feedback. But before analysis, the output from
the fisheye cameras needs to be corrected, because it is distorted. Wide angle or
fisheye lenses frequently produce fisheye camera distortion, which makes straight
lines appear bent or deformed in the image that is taken. This distortion is undesirable
in many applications, including ours.
@seismicisolation
@seismicisolation
A common camera calibration method that has been used needs an understanding
of the camera matrix and distortion coefficients to fix the fisheye camera distortion.
When compared to the distortion, the camera matrix comprises inherent properties
like focus length and main point. The radial and tangential distortion brought on by
the lens is described by coefficients. We will be able to undistort the photos and get
a more accurate depiction of the scene for analysis with the use of this information.
2.2 Methods
Install the necessary libraries which are as follows:

OpenCV is a widely used open-source computer vision library. It provides a
comprehensive set of tools and functions for image and video processing tasks,
including extracting frames from video data. OpenCV allows to manipulate images
and videos, apply filters and transformations, detect objects, perform feature
extraction, and much more.
MediaPipe is another open-source framework specifically designed for building
multimodal perceptual applications. It provides a set of pre-built solutions for
computer vision tasks, such as pose estimation, hand tracking, object detection, and
facial recognition. MediaPipe offers a flexible and efficient pipeline for processing
and analyzing video streams, making it easier to develop applications that require
real-time computer vision capabilities.
TensorFlow is an open-source machine learning library developed by Google. It
provides a flexible and comprehensive ecosystem for building and deploying machine
learning models.
NumPy is a fundamental library for numerical computing in Python. It provides
support for large, multi-dimensional arrays and matrices, along with a collection of
mathematical functions to operate on these arrays.
Scikit-learn is a machine learning library in Python that provides a wide range
of algorithms and tools for classification, regression, clustering, and dimensionality
reduction.
BlazePose model loading: Open the MediaPipe library and load the pre-trained
BlazePose model. The following steps below can be used to include the Blazepose
model into a project:
Using pip, set up the MediaPipe and OpenCV-python libraries are done. The
Blazepose model can be loaded and used with the help of these libraries.
Create an mp_pose object. Using the mediapipe.solutions.pose module, create an
mp_pose object. The interface to load and use the Blazepose model is provided by this
object. Load the Blazepose model: The mp_pose.Pose method can be used to load
the Blazepose model. The arguments for this function are static_image_mode, model
complexity, and enable segmentation. With the use of these parameters, modify the
model’s behavior to suit unique requirements. Process images or video frames: Once
the Blazepose model is loaded, use it to process images or video frames. To process
an image, call the blazepose.process(image) method, where image is a NumPy array
@seismicisolation
@seismicisolation
representing the image. To process a video stream, loop over the frames in the video
and call the blazepose.process(frame) method for each frame.
Extract Pose Features: The output of the Blazepose model is a set of key points
representing the estimated pose of the person in the image or video frame. Extract
these key points using the results.pose_landmarks attribute of the blazepose.process()
method. These key points can be used as input to a machine learning algorithm for
further analysis and classification.
3 Proposed Work
Human Activity Recognition (HAR) is a crucial component in the construction

industry to ensure the safety of workers. In this work, the Blazepose method has been
used for pose estimation to recognize human activities in the construction industry.
Blazepose is a real-time pose estimation algorithm that uses a lightweight neural
network architecture to estimate human poses from video frames. To address the
fisheye camera distortion, we employed a widely used camera calibration method
that involves understanding and utilizing the camera matrix and distortion coeffi-
cients. This approach enabled us to effectively correct the distortions present in the
fisheye camera imagery.
The camera matrix and distortion coefficients are computed as follows:
Step 1: A set of calibration images of nine different various camera angles of a
checkerboard pattern of a size 10 × 7 was collected as shown in Fig. 2.
Step 2: Utilized the standard camera calibration function program, as shown
in Fig. 3, provided by OpenCV to determine the camera matrix and distortion
coefficients. This program identifies the pattern in the calibration images and
calculates the necessary parameters.
Fig. 2 Distorted image from

fisheye camera
@seismicisolation
@seismicisolation
Fig. 3 Caliberating the

camera
Step 3: The obtained camera matrix and distortion coefficients are used to undistort
the images. This can be done using the undistort program provided by OpenCV.
Algorithm
In the proposed approach for painting in civil work, computer vision-based motion
tracking and analysis is utilized through the MediaPipe framework. This involves
extracting the key points of the human skeleton from the input video or image data,
which are then used to apply specific motion constraints based on the instructions
provided by an expert. These motion constraints differ for each movement and are
governed by a specific algorithm. The algorithm defines a set of rules to ensure that
each exercise is executed correctly. To provide real-time feedback to the user, the
pixel coordinates of the key point skeleton points are monitored continuously. If any
constraints are violated, the user is immediately notified, allowing them to adjust
their movements and perform the activity correctly.
The algorithm utilized in this study comprises three main parts—Measurement,
Evaluation, and Visualization.
Measurement: It involves the extraction of relevant 2D key points from video
data using the MediaPipe framework. These key points are used to calculate body
measurements using pixel coordinates, creating a baseline for evaluation. The
measurements are normalized, and they vary depending on the type of exercise being
monitored, established using fixed points on the equipment as a reference.
Evaluation: The algorithm applies motion constraints to the user’s movements
based on the output from the Measurement block. The pixel coordinates and body
measurements calculated from the video data are used to create constraints that dictate
the user’s movements in accordance with the expert’s guidelines. The constraints vary
considerably for each hand movement, and the algorithm is designed to account for
these differences. The Evaluation component is based on three metrics—Completion,
Posture, and Intensity.
@seismicisolation
@seismicisolation
Fig. 4 Flowchart of body posture detection using computer vision
Completion accounts for 60% of the evaluation, and the criteria for completing
one repetition are 80% of “perfect” completion. The perfect completion is defined
based on the average human completion measurement. Posture accounts for 30%,
and points are given based on the number of correct repetitions performed. The
definition of correct posture varies based on the exercise being performed. Intensity
accounts for 10%, and the score is based on the number of repetitions per minute
performed, with a scoring matrix defined based on average human performance.
Visualization: It provides real-time feedback to the user. The system’s display
visualizes key point detection and provides visual cues highlighting incorrect execu-
tion of hand movements, enhancing the user’s understanding of their performance.
The visual cues are presented in different colors according to the user’s performance.
The screen also monitors repetition count in real time, allowing users to track their
progress throughout the activity. At the end of each movement, the system generates
a detailed report on the user’s performance.
The flowchart for body posture detection using computer vision outlines the steps
involved in the process as shown in Fig. 4. It begins with input frames, proceeds with
pre-processing and human detection, extracts key body key points, classifies postures,
analyzes them, and generates output. This flowchart enables accurate detection and
analysis of body postures, facilitating applications like fitness tracking, ergonomic
assessment, and human–computer interaction.
4 Results and Analysis
To evaluate the performance of painting activity recognition for construction industry

using BlazePose method to identify the skeleton system, experiments can be
conducted on a dataset of video recordings of painters performing various painting
@seismicisolation
@seismicisolation
Fig. 5 Accurate key point detection for human pose estimation during painting activities
activities. The dataset should be divided into training and testing sets, with a portion
of the data reserved for testing [9].
In summary, the experiment and results for painting activity recognition for civil
engineering using BlazePose method to identify the skeleton system involve pre-
processing the data, extracting features, training a machine learning model, eval-
uating the model’s performance, and fine-tuning its parameters. The performance
metrics can be presented in a confusion matrix and reported as accuracy, precision,
recall, and F1-score. Figure 5 shows the key point detection of a man painting.
5 Conclusion
In summary, leveraging the BlazePose method for painting activity recognition in

civil engineering can greatly enhance safety and efficiency. By accurately identifying
the skeleton system through key point detection, this technology enables real-time
feedback and alerts for unsafe painting practices. The combination of computer vision
techniques, feature extraction, and machine learning algorithms ensures precise and
@seismicisolation
@seismicisolation
reliable recognition. Experimental results demonstrate the model’s potential for opti-
mization and improved accuracy. The BlazePose model serves as a valuable asset
for construction professionals, enabling effective monitoring and management of
painting activities at construction sites.
References
1. Chakravarthi B, Patil AK, Ryu JY, Balasubramanyam A, Chai YH (2022) Scenario-based

sensed human motion editing and validation through the motion-sphere. IEEE Access
10:28295–28307
2. Chakravarthi B, Prasad BP, Cethana B, Kumar BP (2022) Real-time human motion tracking and
reconstruction using IMU sensors. In: 2022 ınternational conference on electrical, computer
and energy technologies (ICECET). IEEE, pp 1–5
3. Maurya R, Teo TH, Chua SH, Chow H-C, Wey I-C (2022) Complex human activities recog-
nition based on high performance 1D CNN model. In: IEEE 15th international symposium on
embedded multicore/many-core systems-on-chip (MCSoC)
4. Chakravarthi B et al (2019) Quaternions based intuitive visualization for tracking weightlifting
exercises. In: Korean society of information science and technology academic conference, pp
1052–1054
5. Sheu M-H, Salahuddhin Morsalin SM, Hsu C-C, Lai S-C, Wang S-H, Chang C-Y (2023)
Improvement of human pose estimation and processing with the intensive feature consistency
network. IEEE Access. 110-2221-E-224-052-MY2
6. Sajjad F, Ahmed AF, Ahmed MA (2017) A study on the learning based human pose recognition.
In: 9th IEEE-GCC conference and exhibition (GCCCE)
7. Shuvo MMH, Ahmed N, Nouduri K, Palaniappan K (2020) A hybrid approach for human
activity recognition with support vector machine and 1D convolutional neural network. IEEE
8. Saravanan MS, Charan S (2022) Prediction of insufficient accuracy for human activity recog-
nition using convolutional neural network in compared with support vector machine. In: 5th
international conference on contemporary computing and informatics (iC3I) 2022
9. Bhuvaneswari S, Ashlin Jinushia R, Madhavan K, Thandapani T (2020) Automated exterior
wall painting robot using Raspberry Pi. In: IEEE, 2020 6th international conference on advanced
computing & communication systems (ICACCS)
10. Manoj Kumar M, Nandakumar AN (2020) Conceptual multilateral security framework for
authentication of cloud computing services. Int J Grid Distrib Comput 13(1):1622–1627. ISSN:
2005-4262
11. Manoj Kumar M, Nandakumar AN (2017) Exploring multilateral cloud computing secu-
rity architectural design debt in terms of technical debt. In: Smart innovation, systems, and
technologies (SCI), vol 2. Springer
@seismicisolation
@seismicisolation
Geotechnical Engineering
@seismicisolation
@seismicisolation
Soil Moisture Detection Using Arduino
Sensor and ANN Prediction
Rajkumar Raikar, Basavaraj Katageri, Rajashri Khanai,

Dattaprasad Torse, and Praveen Mannikatti
1 Introduction
Appropriate soil moisture content is a pre-requisite for optimum plant growth. Irri-
gation process is a dominant consumer of water, which necessitates the regulation of
water supply. Fields should neither be over-irrigated nor under-irrigated for the best
yield of crops. Hence, the smart system of detecting the existing moisture content
of soil along with irrigation system controllers is the need of an hour. Soil moisture
detection is a major concern in many fields of engineering and agriculture. It can
be determined using laboratory as well as field techniques [1–4]. The traditional
laboratory techniques and other methods [11] require considerable time. However,
in the agricultural system, it is required to determine soil moisture content instantly.
As a consequence of the development of IoT (internet of things) techniques, the
quick soil moisture detection sensors using Arduino are gaining importance in the
field. The two conducting plates that make up the soil moisture sensor serve as the
probe, and the difference in resistance between them measures the moisture content
of the soil. The resistance between the two conducting plates varies inversely with
the amount of moisture present in the soil. When the moisture content of soil is zero
percent, the sensor acts ideally as an open circuit giving infinite resistance. Further,
as the moisture content increases, the resistance value becomes finite and decreases
correspondingly. These sensors can be calibrated and used for soil moisture detec-
tion in the field [5]. On the other hand, Arduino is an open-source hardware platform
that can be used to design and build digital devices that can interact with the real
world. The microcontrollers present on the Arduino board facilitate the interaction
with the sensors that can sense and control both physically and digitally [6]. Many
applications of soil moisture sensors are reported in the literature [5]. Raghuveer
et al. [7] used Node MCU to measure soil moisture in the field and transfer the data
R. Raikar · B. Katageri · R. Khanai (B) · D. Torse · P. Mannikatti

KLE Technological University, Belagavi Campus, India
@seismicisolation
@seismicisolation
110 R. Raikar et al.
to HYDRUS-2D software through which moisture levels are predicted up to 30 cm

depth of soil and water supply is automatically controlled for irrigation. Archana and
Priya [8] employed the humidity and soil moisture sensors in the root zone of the
plant. Based on the sensor values of the microcontroller, the supply of water to the
field is controlled. The various soil parameters such as pH, humidity, moisture, and
temperature are measured for getting high yield from soil by Gainwar and Rojatkar
[9]. Subalakshmi et al. [10] proposed an automation system using microcontroller
and GSM to make irrigation system simpler. Prahlad and Vasudha [11] used FC28
Hygrometer and DHT11 sensors integrated with Arduino UNO to measure soil mois-
ture and temperature and humidity, respectively. Satish Kumar et al. [12] developed
a low-cost soil moisture sensor to measure the moisture content of soil at different
locations in horizontal plane as well as in the vertical direction. They obtained the
soil moisture distribution in the vertical direction. Athani et al. [13] monitored soil
moisture in farmland using IoT-enabled Arduino sensors along with neural network
to predict future rainfall for planning future harvesting. All these studies emphasize
the application of integrated soil moisture sensor with Arduino to measure the soil
moisture content. However, the measurement of soil moisture in different soil grain
size and the graded soil is not attempted so far. Hence, the study focuses on the on-
board installation of soil moisture sensor with Arduino UNO platform to measure
moisture content of soil samples with different grain sizes as well as well graded
soil sample, which is for geotechnical implementation. Further, the sensor reading,
soil moisture content measured, and sediment characteristics are used to develop the
prediction models. Numbers of prediction models are used to fit the experimental
results.
2 Materials and Methodology
2.1 Experimental Components
In the present study, the Arduino UNO kit and soil moisture sensor FC-28 were
used. Arduino is an open-source platform with a facility to integrate microcontrollers
having peripheral embedding system and the Arduino software (IDE) for developing
Arduino programming language. The schematic for the Arduino UNO kit, which
uses a microcontroller board based on the ATMEGA 328P, is shown in Fig. 1. For
storing code, the ATMEGA 328P contains 32 kB of flash memory. The board features
a USB port, an ICSP circuit, 14 digital input and output pins, 6 analogue inputs, a
quartz crystal operating at 16 MHz, and a reset button.
The soil moisture sensor used in this study is FC-28 (see Fig. 2) having both
analog and digital modem connections. The input and output voltage ranges were,
respectively, 3.3–5 V and 0–4.2 V. The input current limit is 35 mA. The main sensor
and the control board made up the Soil Moisture Sensor. Sensor portion of the Soil
Moisture Sensor consists of pair of conductive probes that can be used to measure
@seismicisolation
@seismicisolation
Soil Moisture Detection Using Arduino Sensor and ANN Prediction 111
Fig. 1 Schematic diagram of Arduino-UNO kit
Fig. 2 Soil moisture sensor
the volumetric content of water in soil. The control board, on the other hand, is a
voltage comparator and is composed of LM393 IC.
2.2 Interfacing Soil Moisture Sensor with Arduino
Figure 3 illustrates the interfacing circuit diagram between the sensor, Arduino UNO
kit, and the LCD display unit. In the process to relate the sensor value of soil moisture
with the moisture content of the soil, the analog signal received from the soil mois-
ture module is fed into the Analog input of Arduino, which yielded the percentage
of moisture in the soil precisely [2]. When taking the analog output from the soil
moisture sensor FC-28, the sensor gives the value of 0–1023. The percentage soil
@seismicisolation
@seismicisolation
Fig. 3 Circuit diagram of soil moisture sensor interfacing with Arduino kit
Table 1 Program for

int sensor_pin = A0;
calibration of soil moisture
int output_value;
sensor
void setup() {
Serial.begin(9600);
Serial.println(“Reading from the sensor”);
delay(2000);
}
void loop() {
output_value = analogRead(sensor_pin);
output_value = map(output_value, 1023,540,0,100);
Serial.print(“Moisture:”);
Serial.print(output_value);
Serial.println(“%”);
delay(1000);
}
moisture values ranging 0–100 are mapped with sensor readings and then these
values are displayed on the serial monitor. The code is given in Table 1. Further,
different ranges of moisture values can be set, and the water pump can be turned on/
off accordingly.
2.3 Experimentation
In the present study, 5 uniform soils (poorly graded) and one non-uniform soil (well
graded) samples were tested. The poorly graded soil samples considered were having
mean sediment size, d50 = 850, 600µ, 425, 300, and 150 µm (C u ranging from 1.1
@seismicisolation
@seismicisolation
to 1.35 and C c = 0.95 to 1.24). The well graded soil sample, considered in the present
study, contains equal percentages of above 5 soil sizes having d50 = 325 µm (C u =
4.55 and C c = 1.6).
The dry soil samples were added with moisture having varying degree of saturation
from 10 to 100% computed from:

Sr γw
w = e = Sr −1 (1)
G γd
where: w = moisture content (%), e = voids ratio, S r = degree of saturation (%), G

= specific gravity of soil sample, γ w = unit weight of water (kN/m3 ), and γ d = unit
weight of soil (kN/m3 ). The value of γ d = 1.86 kN/m3 and G = 2.68.
The soil moisture computed using Eq. (1) by varying degree of saturation from
10 to 100% at equal increments of 10% was added to dry soil sample. The soil
moisture sensor FC-28 probe was inserted into the soil sample and sensor reading
was recorded. Then, the soil moisture added was determined using oven drying
method. The procedure is adapted to all soil samples. The variation between the
sensor reading with degree of saturation (SR) and corresponding moisture content
of soil samples were obtained.
The experimental readings obtained between the soil moisture sensor and the mois-
ture content obtained after oven-drying furnished in Table 2 are used to plot the
variation 6 between the degree of saturation, corresponding moisture content and the
respective sensor readings obtained from the soil moisture sensor FC-28 are shown in
Fig. 4a–e for uniform soils and Fig. 4f for non-uniform soil sample. The figures indi-
cate that with an increase in degree of saturation from 10 to 100%, the soil moisture
sensor reading decreases (variation shown in red). This is in accordance with Garg
et al. [5], where it is indicated that the resistance between the two conducting plates
is inversely proportional to the amount of moisture content present in the soil. The
variation between the soil moisture sensor reading and degree of saturation is found
to be polynomial with 4th degree fitting the data trend satisfactorily. It can also be
witnessed that for all soil sample sizes of uniform (Fig. 4a–e) as well as non-uniform
soil (Fig. 4f), the variation between the soil moisture sensor reading and degree of
saturation follows the same 4th order polynomial satisfactorily as seen by R2 value
given in Fig. 4a–f. Further, Fig. 4 can be employed to find the moisture content of
the soil using the soil moisture sensor reading.
For examples: in case of soil sample d50 = 850 µm, if the soil moisture sensor
reading (SR) is 800 (see Fig. 4a), the corresponding degree of saturation (S r ) can be
obtained as 26% from the curve between degree of saturation (S r ) and soil moisture
sensor reading (SR), the corresponding soil moisture content will be 14%, which can
be observed from soil moisture (w) and degree of saturation (S r ) curve. Similarly,
@seismicisolation
@seismicisolation
114
Table 2 Soil moisture and sensor readings based on degree of saturation

Degree of Uniform soil (poorly graded) Non-uniform soil
saturation 850 µm 600 µm 425 µm 300 µm 150 µm (well graded)
(%)
Sensor Moisture Sensor Moisture Sensor Moisture Sensor Moisture Sensor Moisture Sensor Moisture
reading content reading content reading content reading content reading content reading content
10 1006.5 6.4 986.25 12.2 980 8.78 1004.25 11.64 979 11.4 1002.8 11.4
20 953.67 9.76 924.5 20 809.67 26.28 985.5 14.28 939 17.64 961.17 17.64
30 725.5 17.34 652.5 29.23 650.17 32.42 890.5 19.64 712.84 25.49 749.34 25.49
40 435.5 22.51 360.5 38.46 483 35.7 595.5 25 486.67 33.33 537.5 33.33
50 245.67 27.11 273.75 41.23 253 47.37 471.5 33.3 341.01 37.52 492.25 38.52
60 185 32.19 207 44 203.84 52.45 247.5 41.6 195.34 45.71 247 47.71
@seismicisolation
@seismicisolation
70 158.34 35.1 126.09 53.25 159.67 60.11 147.25 51.55 152.92 50.21 199.5 52.21
80 136.84 36.94 110.17 62.5 149 67.17 147 61.5 110.5 62.7 152 64.7
90 116.17 39.21 106.59 72.16 130.17 77 131.5 68.25 105.59 72.35 146.84 72.35
100 114.34 41.68 104 79.81 110 89 116 75 106.67 80 142.67 80
R. Raikar et al.
Fig. 4 Variation between the degree of saturation, moisture content, and sensor readings
for soil sample having d 50 = 600 µm, for the sensor reading = 200 (see Fig. 4b), the
corresponding degree of saturation is 56% and the respective soil moisture content
will be 46%. This indicates that the soil moisture content is inversely related to the
sensor reading. Further, all the figures in Fig. 4a–f between the degree of satura-
tion (Sr) versus sensor readings (SR) are combined by normalizing the values and
presented in Fig. 5. In order to compare and combine the data of all soil uniform
and non-uniform soil as in Fig. 5. Further, Fig. 4a–f presents the variation between
the degree of saturation (S r ) and moisture content (shown in blue), which follows
2nd order polynomials with satisfactory acceptance seen by R2 value as shown in
Fig. 4a–f.
Figure 6a–f illustrates the variation between moisture content (w) of the soil in
percentage and the sensor reading (SR) for each case of uniform and non-uniform
soil samples.
The moisture content follows power law variation with sensor reading (y = axb,
where y = moisture content w, x = sensor reading SR, and a and b are constants
obtained by the fitting). The variation of the normalized moisture content with
@seismicisolation
@seismicisolation
Fig. 5 Variation between the degree of saturation and normalized sensor readings
normalized sensor reading is illustrated in Fig. 7, which indicates a satisfactory

similarity of all data collapsible on a single curve indicated by y = 1.4074(x) −
0.714 with coefficient of determination of R2 = 0.9635. This corroborates strong
similarity of sensor readings with moisture content readings.
Furthermore, the experimental data was used to develop a good prediction model
using different models such as ANN (Artificial Neural Network), KNN (K-Nearest
Neighbors) and SVM (Support Vector Machines) in JMP Statistical Software. The
ANN architecture with 3-58-1 (3—input neurons, 58 neurons in single hidden layer,
and 1 output neuron) gave a good prediction model with R2 = 0.981 for training
and 0.985 for validation. The total data points were 60 with 50 uniform soil and 10
non-uniform soil. This data is divided into 75–25 proportions with 45 for training
and 15 for validation.
The activation function employed was sigmoidal with learning rate 0.1. Figure 8a
shows the ANN structure, Fig. 8b and c presents the actual and predicted values of
moisture contents for training and validation, respectively.
Similarly, the KNN model gave good prediction with K = 6 having R2 = 0.932 for
validation. Figure 9 shows the results of K-nearest neighbor with K = 2 for training
(R2 = 0.927) and K = 6 for validation (R2 = 0.932). Even though the KNN is used
for unsupervised clustering, the clustered data is further employed for prediction. In
addition, SVM having 31 support vectors, radial basis function as kernel function
and gamma = 0.333 was employed as a prediction model after classification through
JMP software. The prediction model yielded R2 = 0.978 and 0.98 for training and
@seismicisolation
@seismicisolation
Fig. 6 Variation between moisture content and sensor reading for different sediment sizes
validation. Figure 10 presents the actual and predicted values of moisture contents
for training and validation using SVM. On comparison, the ANN model resulted
better prediction than other two.
@seismicisolation
@seismicisolation
Fig. 7 Variation between normalized moisture content and normalized sensor reading
(a)
(b) (c)
Fig. 8 a ANN architecture b, c variation between actual and predicted values of moisture contents
for training and validation
@seismicisolation
@seismicisolation
Fig. 9 KNN predictions with K = 2 for validation and K = 6 for validation
Training Validation
Fig. 10 Variation between actual and predicted values of moisture contents for training and
validation using SVM
4 Conclusion
The moisture sensor that is included within the Arduino platform is used in the
experimental results of this study to assess the moisture content of soil samples of
both uniform and non-uniform types. The following conclusions are drawn from the
study:
The moisture sensor readings are used to estimate the existing moisture content of
the soil sample. The variation between the moisture sensor reading and soil moisture
follows a power equation with an R2 value of 0.96. The sensor readings are connected
to the soil moisture through degree of saturation of the soil. With normalized values
of sensor reading, a promising similarity is found between degree of saturation (S r )
versus normalized sensor readings (SR) for all the cases of uniform and non-uniform
@seismicisolation
@seismicisolation
soil. This indicates that the sensor readings can be used to predict the soil moisture
content irrespective of soil size and type.
Three machine learning models namely ANN, KNN, and SVM are used to predict
the moisture content based on sensor reading, sediment size, and soil uniformity
type. It is observed that out of three predictors, ANN resulted in a better solution as
compared to KNN and SVM with R2 = 0.981 for training and 0.985 for validation.
Generally, KNN and SVM are used for clustering and classification, respectively;
however, they can be employed for prediction. In the present case, the results obtained
by both KNN and SVM were not promising as compared to ANN. Hence, ANN can
be used as a good predictive model.
References
1. Punmia BC, Jain AK, Jain AK (2005) Soil mechanics and foundation engineering, 16th edn.
Laxmi Publications, New Delhi
2. Johnson AI (1962) Methods of measuring soil moisture in the field. Geological Survey Water
Supply Paper, 1619-U, U. S. Geological Survey, Federal Center, Denver
3. Ledieu J, De Ridder P, De Clerck P, Dautrebande S (1986) A method of measuring soil moisture
by time-domain reflectometry. J Hydrol 88(3–4):319–328
4. Ungar SG, Layman R, Campbell JE, Walsh J, McKim HJ (1992) Determination of soil moisture
distribution from impedance and gravimetric measurements. J Geophys Res 97(D17):18969–
18977
5. Garg A, Munoth P, Goyal R (2016) Application of soil moisture sensors in agriculture: a
review. In: Proceedings of international conference on hydraulics, water resources and coastal
engineering (HYDRO2016), CWPRS Pune, India
6. Badamasi YA (2014) The working principle of Arduino. In: Proceedings of 11th international
conference on electronics, computer and computation (ICECCO)
7. Raghuveera E, Pavan Kumar EN, Sai Yeswanth A, Satya Mani Pavan L (2019) Soil moisture
monitoring system using IoT. Int J Innov Technol Expl Eng 8(7)
8. Archana P, Priya R (2016) Design and implementation of automatic plant watering system. Int
J Adv Eng Glob Technol 4(01):1567–1570
9. Gainwar SD, Rojatkar DV (2015) Soil parameters monitoring with automatic irrigation system.
Int J Sci Eng Technol Res 4(11):3817–3820
10. Subalakshmi R, Anu Amal A, Arthireena S (2016) GSM based automated irrigation using
sensors. Int J Trend Res Dev Special Issue 4–6
11. Bhadani P, Vashist V (2019) Soil moisture, temperature and humidity measurement using
Arduino. In: Proceedings of 9th international conference on cloud computing, data science &
engineering (confluence), pp 567–571
12. Kumar MS, Ritesh Chandra T, Pradeep Kumar D, Sabarimalai Manikandan M (2016) Moni-
toring moisture of soil using low cost homemade soil moisture sensor and Arduino UNO.
In: Proceedings of 3rd international conference on advanced computing and communication
systems
13. Athani S, Tejeshwar CH, Patil MM, Patil P, Kulkarni R (2017) Soil moisture monitoring
using IoT enabled Arduino sensors with neural networks for improving soil management for
farmers and predict seasonal rainfall for planning future harvest in North Karnataka—India.
In: Proceedings of international conference on I-SMAC (IoT in social, mobile, analytics and
cloud), pp 43–48
@seismicisolation
@seismicisolation
Reliability Analysis of Clayey Soil Slope
Stability Using GMDH and RFC Soft
Computing Techniques
Rahul Ray
1 Introduction
For soil-based building projects like dams, retaining wall, and embankments, slope
stability of the soil is a crucial consideration. Because of the randomness in its produc-
tion, soil is heterogeneous by nature, which makes its prediction difficult. Therefore,
it is necessary to take into account all of the variability in the soil characteristics whilst
taking soil as construction. Limit Equilibrium methodology and Strength Reduction
methodology are two commonly used approaches for slope stability issues. However,
because they take root of deterministic approach, they may result in overly conserva-
tive slope designs. In addition, errors made during the testing process and differences
in soil depositions result in discrepancies dataset shown by Phoon and Kulhawy [1].
To get a realistic design, with the inclusion of all the soil characteristics in consider-
ation, probabilistic approach is considered useful. Reliability analysis is employed
for the probabilistic approach, taking into account the soil factors that determine
the stability of the soil slope as input variables for the models created using soft
computing techniques, and then analysing the models’ output responses.
In past, researchers [2–6] used different probabilistic approach methods to show
its applicability in different sections of engineering, Christian et al. [7] using field
data for performing probabilistic analysis and also to give proof that uncertainties of
soil characteristics can lead to unrealistic design. Researcher also used various soft
computing technique to find many failure modes by the help of both deterministic
and probabilistic models [8–11]. Author coded a method, random forest machine
to predict the stability of shallow landslides [12]. A typical reinforced soil slope
was analysed using various soft computing technique including RFC and GMDH for
counting their applicability [13]. Settlement of shallow foundation was also analysed
using soft computing techniques and the models show high applicability [14]. After
R. Ray (B)
GLA University Mathura, Mathura, Uttar Pradesh, India
e-mail: rahul.ray@gla.ac.in
@seismicisolation
@seismicisolation
122 R. Ray
many research, it was found that the probabilistic analysis covers all the variability
in soil attributes. The main purpose of this work is to perform reliability analysis of
stability of cohesive soil slope stability using two soft computing models, GMDH
and RFC. In addition, both the models were also analysed based on assessment
parameters and statistical tests.
2 Soft Computing Models’ Theoretical Background
2.1 GMDH Model
This model uses iteration-based technique with a topology resembling multilayered,

self-learning neural network systems. GrMDH was developed by Ivakhnenko et al.
[15] to address the eq.’s complicated nature and linear dependability was observed
in conventional regression. The GMDH is a modelling method that consists of a
number of neuron layers connected to one another by quadratic equations to produce
new neurons. For the multivariate to optimize models, Eqs. (1) and (2) are used.
g = arg min .X (g) (1)
( )
X (g) = f R, W, e2 , S, U (2)
Here,
g set for GMDH,
X criteria for GMDH,
R no. of attributes,
W complex part of GMDH,
e2 noise,
S number of transformation inside dataset,
U equations for reference.
The GMDH arrangement is shown in Fig. 1, with solid objects standing in for the
chosen model of matching layers [16]. Current paper has 3 inputs (c, γ , and φ) and
1 output (F) for the GMDH model.
2.2 Random Forests Classifier
RFC is an effective way to form a most accurate decision tree for detecting multi-
parameter and extensive patterns [12]. This is achieved when k different decision
tree is trained, each of them is connected to a certain region of inputs. The actual
training dataset is then sampled with replacement n < N dataset, and one tree k is
@seismicisolation
@seismicisolation
Reliability Analysis of Clayey Soil Slope Stability Using GMDH … 123
3rd LAYER
INPUT
LAYER 2nd LAYER
1st LAYER
Fig. 1 GMDH model structure
then produced as shown in Fig. 2. RFC model decision tree uses Gini impurity index
(G) and Product (P) to get best node using following equations.
P = G1G2 (3)
∑
n kc
Gk = 1 − qki2 (4)
i=1
where
nkc no. of class and

qki phase class i ratio in current dataset.
Input Data Soil Stability

Parameter
Decision Tree 1 Decision Tree 1 Decision Tree K
w1 w2 wTN
Majority Predicted
Voting Soil Stability
Factor of Safety
Fig. 2 Random forest classifier (RFC) structure
@seismicisolation
@seismicisolation
124 R. Ray
3 Model Development
In this paper, embankment of 10 m height with two thick stiffer stratum described
by Griffiths et al. [17] was analysed to find the stability. Other descriptions about
embankment are mentioned in Fig. 3. For reliability analysis of the embankment
stability, the variability in soil attributes was taken under consideration for unit weight
(γ ) and cohesion (c) by taking coefficient of variation for γ = 0.03 and for c as 0.1,
0.1, and 0.15 for 3 layers top to bottom, respectively.
For slope stability analysis, Morgenstern-Price method was taken under consid-
eration with the help of GeoStudio 2016. By taking c, γ , and φ as input and factor of
safety (F) as output for embankment slope. The permissible range of γ , c, and φ was
taken as base to generate 100 dataset and based on that F values from GeoStudio
2016 were obtained. After that the soft computing models were trained and tested
using MATLAB, for that normalization of data is done by using Eq. 5.
X − X min .
X nor = (5)
X max . − X min .
Here,
X nor normalized form

X actual form
X min . min. value
X max . max. value
After training and testing of the models, predicted output of the two soft computing
model is analysed using assessment parameters to get a comparative analysis to find
the most reliable model for prediction.
Fig. 3 Multilayered embankment [17]
@seismicisolation
@seismicisolation
4 Assessment Parameters
Assessment of the all the soft computing models is done to know which model fits
best to the considered problem. To access the best model, following are used: Rela-
tive Percentage Difference (RPD) [18], Nash–Sutcliffe (NS) [19], Root Mean Square
Error (RMSE) [20], R2 (Coefficient of determination) [21], Adj. R2 (adjusted Coef-
ficient of determination), Expanded uncertainty (U95 ), Performance Index (PI) [22],
Bias Factor [23], RSR [24], Variance Account Factor (VAF), MAPE (Mean Abso-
lute Percentage Error) [25], Global Performance Indicator (GPI) [26], Willmott’s
Index for agreement (WI), Mean Absolute Error (MAE) [27], Legate and McCabe’s
Index (LMI), Normalized Mean Bias Error (NMBE) [28, 29], t-statistic [30], and
Reliability Index (β) [31].
∑n
(di − yi )2
NS = 1 − ∑n i=1 (6)
i=1 (di − dmean )
2
[
|
|1 ∑ n
RMSE = | (di − yi )2 (7)
N i=1
( )
var(di − yi )
VAF = 1 − × 100 (8)
var(di )
∑n ∑n
i=1 (di − dmean ) − i=1 (di − yi )
2 2
R =
2
∑ n (9)
i=1 (di − dmean )
2
(n − 1) ( )
AdjR 2 = 1 − 1 − R2 (10)
(n − p − 1)
PI = adj.R 2 + 0.01VAF − RMSE (11)
1 ∑ yi
n
Bias Factor = (12)
N i=1 di
RMSE
RSR = / ∑n (13)
i=1 (di − dmean )2
1
N
∑n
i=1 (yi − di )
1
NMBE(%) = N
1 ∑n × 100. (14)
N i=1 di
n | |
1 ∑|| di − yi ||
MAPE = . (15)
N i=1 | di |
SD
RPD = (16)
RMSE
@seismicisolation
@seismicisolation
126 R. Ray
[ ∑N ]
i=1 (di − yi )2
WI = 1 − ∑ N . (17)
i=1 (|yi − dmean |+|di − dmean |)2
1 ∑
n
MBE = (yi − di ) (18)
N i=1
1 ∑
n
MAE = |(yi − di )| (19)
N i=1
[ ∑ ]
N
|di − yi |
LMI = 1 − ∑ N i=1 (20)
i=1 |di − dmean |
U95 = 1.96 ∗ (SD2 + RMSE2 )1/2 (21)

/
(N − 1)MBE2
t-stat = (22)
RMSE2 − MBE2
( )
GPI = MBE × RMSE × U95 × tstat × 1 − R 2 (23)
μF − 1
β= (24)
σF
In this context, d i represents the observed value of the ith point, whilst yi signifies
the predicted value of the same point. Furthermore, d mean denotes the mean value of
the observed data, whereas SD denotes the standard deviation. In addition, the mean
value of the factor of safety of soil slope is taken into consideration as μ F alongside
the standard deviations of the factor of safety of soil slope as σ F .
5 Results and Discussion
The input values for the models GMDH and RFC, during their training and testing,
were obtained from the normalized values of all soil characteristics. The projected
values, or model output, were subsequently evaluated for the purposes of a reliability
study on the soil slope stability of the section under investigation.
Figures 4 and 5 depict graphical comparisons of the factual and predicted values
of the soil slope factor of safety for both the training and test sets of data. These
figures have been employed to explicitly demonstrate the efficacy of the models. The
GMDH model plot, in particular, indicates that the training and testing data are more
closely aligned with the actual equals to predicted line amongst the two models.
@seismicisolation
@seismicisolation
Notably, the plots reveal that the models’ predictions are satisfactory, i.e., the values
are in close proximity to the actual equals to predicted line.
Various parameters are used to evaluate each model’s performance, as shown in
Table 1. Table 1’s NS value for both the models reveals that GMDH’s NS value
is closer to 1 than RFC model, indicating that GMDH has the highest predictive
1.8
1.6
PREDICTED VALUES OF FOS
1.4
1.2
1
Actual = Predicted
0.8 Training Data
Testing Data
0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
ACTUAL VALUES OF FOS
Fig. 4 GMDH model performance
1.8
1.6
PREDICTED VALUES OF FOS
1.4
1.2
1
Actual = Predicted
0.8 Training Data
Testing Data
0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
ACTUAL VALUES OF FOS
Fig. 5 RFC model performance
@seismicisolation
@seismicisolation
128 R. Ray
capability. When comparing the models using RMSE and VAF, GMDH has a lower
prediction error, indicating that it performed better than RFC model. The R2 and Adj.
R2 values for the GMDH model are closest to 1 and also closer to each other RFC,
indicating that GMDH has captured the majority of the variability in soil parameter.
GMDH model is less biased from its actual and mean value and good in predicting
the FOS, as determined by the Bias factor, RSR, PI, MAPE, and NMBE factors.
In comparison to the RFC model, the GMDH soft computing technique performed
well having lower error and highly accurate, as indicated by the values for RPD, WI,
MAE, MBE, and LMI. Both the models are doing well based on the values for U95
and t-stat, and the GPI value indicates that the GMDH model is more accurate than
RFC. Both GMDH and RFC have reliability indices between 3 and 4 [25], indicating
that the models perform in the good category.
A Taylor diagram [32] shows how models perform. Figure 6’s representation
of the diagram’s Standard deviation (SD), root mean square error (RMSE), and
correlation coefficient all works together to provide the most precise model possible.
The findings demonstrate that GMDH performs better than expected, which suggests
that there is good agreement between real and predicted values for this model.
In addition, Mann–Whitney U (M–W) test [33] is performed for checking the
homogeneity of the models to get the acceptable acceptancy. As per the Table 2,
P-value for GMDH and RFC models is greater than 0.05, which shows that models
Table 1 Assessment
Parameters GMDH RFC Ideal values
parameters of GMDH and
RFC NS 0.8619 0.5855 1
RMSE 0.0430 0.0771 0
VAF 86.1998 61.9034 100%
R2 0.8619 0.5855 1
Adj. R2 0.8556 0.5377 1
PI 1.6746 1.0796 > 1.0
Bias factor 1.0002 0.9855 1
RSR 0.3716 0.6438 0
NMBE (%) − 0.0877 − 1.6525 0
MAPE 0.0250 0.0417 0
RPD 2.6909 1.5533 > 2.5
WI 0.9612 0.8845 0.0–1.0
MAE 0.0331 0.0559 0
MBE − 0.0012 − 0.0219 0
LMI 0.6531 0.4443 1
U95 0.2421 0.2793 0
t-stat 0.2255 1.5971 Smaller value
GPI 0.0000 − 0.0003 Higher value
β 3.7221 3.0186 >3
@seismicisolation
@seismicisolation
Standard deviation
Standard deviation
(a) Training (b) Testing
Fig. 6 Taylor diagram for GMDH and RFC model
Table 2 Mann–Whitney statistics of GMDH and RFC

M–W test statistics Training Testing
GMDH RFC GMDH RFC
M–W U 2854 2536 458 452
P 0.9562 0.8562 0.9853 0.8652
follow null-hypothesis, indicates that both the models follow the trend of normally
distribution. GMDH follows the normal distribution trend in most likely amongst all
the soft computing techniques.
6 Conclusion
In the present investigation, two soft computing methodologies, namely Group

Method of Data Handling (GMDH) and Random Forest Classifier (RFC), were
employed to conduct a reliability analysis of soil slope stability. The efficacy of
the factor of safety prediction of soil slope for slope stability by each of the models is
demonstrated by their high reliability indices. Nevertheless, it has been observed that
GMDH outperforms all the other models when evaluated using various performance
measures, which highlights a novel approach towards the probabilistic analysis of soil
slope stability. Hence, GMDH can be considered as a dependable tool for reliability
analysis to estimate the factor of safety of a soil slope for slope stability studies.
@seismicisolation
@seismicisolation
130 R. Ray
References
1. Phoon KK (2002) Potential application of reliability-based design to geotechnical engineering.

In: Proceedings of 4th Colombian geotechnical seminar, Medellin, pp 1–22
2. Yadav P, Shah K (2021) Quinolines, a perpetual, multipurpose scaffold in medicinal chemistry.
Bioorg Chem 109
3. Sharma H, Jalal AS (2021) Visual question answering model based on graph neural network
and contextual attention. Image Vis Comput
4. Verma P, Agrawal P, Amorim I, Prodan R (2021) WELFake: word embedding over linguistic
features for fake news detection. IEEE Trans Comput Soc Syst 8:881–893
5. Ghosh S, Singh D, Kumar R, Maharaj S (2021) Phase transition of AdS black holes in 4D EGB
gravity coupled to nonlinear electrodynamics. Ann Phys 424
6. Praveen K, Roy LB (2022) Assessment of groundwater quality using water quality indices: a
case study of Paliganj Distributary, Bihar, India. ETASR 12:8199–8203
7. Christian JT, Ladd CC, Baecher GB (1994) Reliability applied to slope stability analysis. J
Geotech Eng 120:2180–2207. https://doi.org/10.1061/(ASCE)0733-9410(1994)120:12(2180)
8. Khandelwal M, Singh TN (2013) Prediction of blast-induced ground vibration using artificial
neural network. Int J Rock Mech Min Sci 46:1214–1222. https://doi.org/10.1016/j.ijrmms.
2009.03.004
9. Ray R, Roy LB (2021) Reliability analysis of soil slope stability using Ann, Anfis, PSO-Ann
soft computing techniques. NVEO-Nat Volatiles Essent Oils 8:3478–3491
10. Ray R, Choudhary SS, Roy LB (2021) Reliability analysis of layered soil slope stability using
ANFIS and MARS soft computing techniques. Int J Performability Eng 17:647. https://doi.
org/10.23940/ijpe.21.07.p9.647656
11. Ray R, Choudhary SS, Roy LB (2021) Reliability analysis of soil slope stability using MARS,
GPR and FN soft computing techniques. Model Earth Syst Environ 2021:1–11. https://doi.org/
10.1007/S40808-021-01238-W
12. Dang V, Hoang N, Nguyen L, Bui D, Forests PS (2020) A novel GIS-based random forest
machine algorithm for the spatial prediction of shallow landslide susceptibility. For MDPI
11:1–20. https://doi.org/10.3390/f11010118
13. Ray R, Choudhary SS, Roy LB, Kaloop MR, Samui P, Kurup PU, Ahn J, Hu JW (2023)
Reliability analysis of reinforced soil slope stability using GA-ANFIS, RFC, and GMDH soft
computing techniques. Case Stud Constr Mater 18:e01898. https://doi.org/10.1016/J.CSCM.
2023.E01898
14. Ray R, Samui P, Roy LB (2023) Reliability analysis of a shallow foundation on clayey soil
based on settlement criteria. J Curr Sci Technol 13
15. Ivakhnenko A, Krotov G, Visotsky V (1979) Identification of the mathematical model of a
complex system by the self-organization method. Theor Syst Ecol Adv Case Stud
16. Nguyen TN, Lee S, Nguyen-Xuan H, Lee J (2019) A novel analysis-prediction approach for
geometrically nonlinear problems using group method of data handling. Comput Methods Appl
Mech Eng 354:506–526. https://doi.org/10.1016/J.CMA.2019.05.052
17. Griffiths DV, Huang J, Fenton GA (2012) Risk assessment in geotechnical engineering: stability
analysis of highly variable soils. In: GeoCongress 2012. https://doi.org/10.1061/978078441
2138.0004
18. Ray R, Kumar D, Samui P, Roy LB, Goh ATC, Zhang W (2020) Application of soft computing
techniques for shallow foundation reliability in geotechnical engineering. Geosci Front 12:375–
383. https://doi.org/10.1016/j.gsf.2020.05.003
19. Jain SK, Sudheer KP (2008) Fitting of hydrologic models: a close look at the Nash-Sutcliffe
index. J Hydrol Eng 13:981–986. https://doi.org/10.1061/(ASCE)1084-0699(2008)13:10(981)
20. Kisi O, Shiri J, Tombul M (2013) Modeling rainfall-runoff process using soft computing
techniques. Comput Geosci 51:108–117. https://doi.org/10.1016/j.cageo.2012.07.001
21. Babu GLS, Srivastava A (2007) Reliability analysis of allowable pressure on shallow foundation
using response surface method. Comput Geotech 34:187–194. https://doi.org/10.1016/j.com
pgeo.2006.11.002
@seismicisolation
@seismicisolation
22. Kung GT, Juang CH, Hsiao EC, Hashash YM (2007) Simplified model for wall deflection and
ground-surface settlement caused by braced excavation in clays. J Geotech Geoenviron Eng
133:731–747. https://doi.org/10.1061/(ASCE)1090-0241(2007)133:6(731)
23. Prasomphan S, Machine SM (2013) Generating prediction map for geostatistical data based
on an adaptive neural network using only nearest neighbors. Int J Mach Learn Comput 3
24. Moriasi DN, Arnold JG, Van Liew MW, Bingner RL, Harmel RD, Veith TL (2007) Model
evaluation guidelines for systematic quantification of accuracy in watershed simulations. Trans
ASABE 50:885–900. https://doi.org/10.13031/2013.23153
25. Armstrong JS, Collopy F (1992) Error measures for generalizing about forecasting methods:
empirical comparisons. Int J Forecast 8:69–80. https://doi.org/10.1016/0169-2070(92)900
08-W
26. Viscarra Rossel RA, McGlynn RN, McBratney AB (2006) Determining the composition of
mineral-organic mixes using UV–vis–NIR diffuse reflectance spectroscopy. Geoderma 137:70–
82. https://doi.org/10.1016/j.geoderma.2006.07.004
27. Raventos-Duran T, Camredon M, Valorso R, Mouchel-Vallon C, Aumont B (2010) Structure-
activity relationships to estimate the effective Henry’s law constants of organics of atmospheric
interest. Atmos Chem Phys 10:7643–7654. https://doi.org/10.5194/acp-10-7643-2010
28. Matérn B (1960) Spatial variation, volume 36 of lecture notes in statistics, 2nd edn. Springer-
Verlag, New York (First edn. Publ. by Meddelandan fran Statens …, 1960)
29. Lal Bahadur R, Praveen P (2022) Study of soil erosion by using remote sensing and GIS
techniques in Sone command area in Bihar, India. Mater Today Proc 62:1664–1670. https://
doi.org/10.1016/J.MATPR.2022.04.739
30. Stone RJ (1993) Improved statistical procedure for the evaluation of solar radiation estimation
models. Sol Energy 51:289–291. https://doi.org/10.1016/0038-092X(93)90124-7
31. USACE (1997) Risk-based analysis in geotechnical engineering for support of planning studies,
engineering and design. Dept. Army, USACE Washington, DC
32. Taylor KE (2001) Summarizing multiple aspects of model performance in a single diagram. J
Geophys Res Atmos 106:7183–7192. https://doi.org/10.1029/2000JD900719
33. Mann HB, Whitney DR (1947) On a test of whether one of two random variables is stochastically
larger than the other. Ann Math Stat 18:50–60. https://doi.org/10.1214/aoms/1177730491
@seismicisolation
@seismicisolation
Prediction of Residual Strength After
Liquefaction Using Artificial Intelligence
Model
Shubhendu Vikram Singh and Sufyan Ghani
1 Introduction
Predicting residual strength is a crucial task in the field of geotechnical engineering.

Residual strength refers to the ability of soil to bear loads after being subjected to
damage or deformation. The residual strength of soil can be affected by various
conditions such as earthquakes, sinkholes, landslides, and many other geotechnical-
related failures. Accurately predicting the residual strength of soil is essential for
determining its remaining useful life and ensuring its safe operation on the ground
[1].
The impact of soil’s drained residual shear strength on the stability analysis of
reactivated landslides was initially examined by Skempton [2]. He proposed that
modifications to clay particle orientation during unidirectional shearing may have
contributed to the reduction in shear strength. Later, reports on the impact of soil
mineralogy on residual strength came [3]. Mesri and Shahien [4] have discovered
that residual strength also manifests in initial slope failures after studying 99 instances
of slope failure.
Due to the limited analytical capacity of the human intellect, it is difficult to
analyze complex geoscience data, so AI tools have great potential to analyze vast data
with better precision in the very limited time frame. By using AI models, researchers
can develop accurate predictions of residual strength, which can help in improving the
design, maintenance, and safety of soil [5]. Researchers have concluded that AI and
ML-based algorithms have provided better outcomes in comparison to conventional
methods (limit equilibrium methods) for predicting residual strength [6, 7].
S. V. Singh · S. Ghani (B)

Department of Civil Engineering, Sharda University, Greater Noida, India
e-mail: Sufyan.ghani@sharda.ac.in
S. V. Singh
e-mail: 2022453567.shubhendu@dr.sharda.ac.in
@seismicisolation
@seismicisolation
134 S. V. Singh and S. Ghani
Some of the popular machine learning algorithms used for predicting residual
strength include artificial neural networks, support vector machines, decision trees,
and random forests. These algorithms have been shown to provide accurate predic-
tions of residual strength for various soils [8]. In the past decade, researchers have
used the backpropagation method over empirical methods for analyzing the residual
shear strength of the soil. In this article, the backpropagation method has been applied
to predict the residual shear strength of Noida Metro (Sector 51 to Greater Noida
Sector 2) [9–16].
Nevertheless, charts are used to develop the majority of the relationships. Tiwari
and Marui [17] correlation is based on the clay’s mineralogical composition. Finding
out the precise surface and mineralogical content is not always useful for geotech-
nical engineers in practice. Therefore, creating a correlation or approach that allows
residual shear strength to be defined in terms of all pertinent variables is much desired.
With a great degree of accuracy, Das and Basudhar [18] predicted the residual friction
angle of clay using artificial neural network models. However, the study mentioned
above only looks at tropical soil from a certain area. Consequently, there is a pressing
need to create a strategy that is appropriate and effective for different origins of soil
[19].
In this article, authors have developed an ANN model to facilitate the prediction of
residual shear strength from the available soil report of the Noida metro from Sector
51 to Greater Noida Sector 2. Ten geotechnical parameters (N1)60, percentage of
fine sand, silt clay, moisture content, angle of friction, LL, PL, PI, and dry density are
used in predicting residual shear strength. To obtain and develop a multiple regression
model that can accurately establish a relationship between these factors. The accuracy
of the predictions made by the ANN model is assessed using statistical performance
metrics, such as over fitting ratio, maximum absolute error, average absolute error,
and root mean square error. The above study is used to create a model equation
that illustrates correlation, which may be calculated using spreadsheet inputs. ANN
models are used to determine the relevance of the input parameters.
2 Research Methodology
A database was created for analysis using soil reports from the Noida Metro Rail as
shown in Figs. 1 and 2. To ensure quality assurance, a total of 197 soil sample data
were taken out of available 300 data from 10 boreholes for which soil report is being
prepared.
2.1 Artificial Neural Network
The use of artificial neural networks in civil engineering has been started long
ago. Since then, these networks have been applied in various subfields, including
@seismicisolation
@seismicisolation
Prediction of Residual Strength After Liquefaction Using Artificial … 135
Fig. 1 Boreholes from Noida Sector 51 to Greater Noida Sector 2
Fig. 2 Map for Noida Sector 51 to Greater Noida Sector 2 by NMRC
environmental and water resources engineering, geotechnical engineering, traffic

engineering, highway engineering, and construction engineering and management.
Researchers have published numerous articles on neural networks in archival research
journals. In this study, a back propagation neural network algorithm was employed
to establish a link between the input parameters and the residual shear strength value.
As indicated in Fig. 3, the model was trained with 132 data points, and its accuracy
was evaluated using the remaining 65 data points. The effectiveness of the artificial
intelligence model was gauged using the R-squared value and mean squared error
(MSE) metrics.
@seismicisolation
@seismicisolation
Fig. 3 Diagram illustrating the stages in the creation of ANN models
Table 1 provides statistical information of training data on various properties of a

type of soil, including its SPT N corrected value, percentage of fine sand, silt, and
clay, moisture content, angle of friction, liquid limit, plastic limit, plasticity index,
dry density, and residual shear strength. The table includes maximum and minimum
values, mean and median values, as well as the mode, standard deviation, and variance
for each property. This information can be useful for understanding the properties
and behavior of the soil, as well as for making decisions about its use in construction
or other applications.
Table 2 presents statistical data regarding testing data and the characteristics of a
specific soil type, such as SPT N corrected value, percentages of fine sand, silt, and
clay, moisture content, angle of friction, liquid limit, plastic limit, plasticity index,
dry density, and residual shear strength. It provides various measures, including the
maximum and minimum values, mean and median values, mode, standard deviation,
and variance, for each of these properties. This information can be significant in
comprehending the soil’s traits and behavior, as well as making informed decisions
about its application in construction or other fields.
In Fig. 4, the plot between Actual Residual Shear Strength and Predicted Residual
shear strength shows a coefficient with a correlation value of 0.96, which indicates a
strong positive linear relationship between the variables. This suggests that the ANN
model is highly predicted. The plot in Fig. 5 depicting the relationship between Actual
Residual Shear Strength and Predicted Residual Shear Strength exhibits a correlation
coefficient of 0.93, indicating a robust positive linear association between the two
variables. This implies that the ANN model is highly accurate in its predictions.
@seismicisolation
@seismicisolation
Table 1 Values of descriptive statistics for different training dataset factors
SPT N % fine sand % of silt % of clay % moisture Angle of Liquid limit Plastic limit Plasticity Dry density Residual
corrected content friction (LL) (PL) index (PI) (gm/cc) shear
strength
(kPa)
Max 19.18 96.00 86.85 55.49 29.20 31.80 65.26 28.74 47.17 1.61 37.33
Min 11.04 1.36 3.93 0.08 5.60 6.70 3.52 0.00 0.00 1.50 20.00
Mean 16.40 32.93 52.99 16.71 16.20 23.59 25.65 13.15 9.09 1.56 28.56
Mode 16.50 12.00 60.00 0.00 8.06 30.60 0.00 0.00 0.00 1.52 0.00
@seismicisolation
@seismicisolation
Median 16.59 26.88 60.00 13.89 17.90 28.00 22.21 15.13 6.66 1.56 28.69
SD 2.34 20.17 22.36 11.31 6.02 10.27 13.74 9.72 10.37 0.04 4.15
Prediction of Residual Strength After Liquefaction Using Artificial …
Variance 5.48 406.82 500.16 127.89 36.20 105.55 188.82 94.47 107.53 0.00 17.22
137
138
Table 2 Values of descriptive statistics for different testing dataset factors

SPT N % fine sand % of silt % of clay % moisture Angle of Liquid limit Plastic limit Plasticity Dry density Residual
corrected content friction (LL) (PL) index (PI) (gm/cc) shear
strength
(kPa)
Max 18.67 90.00 86.42 38.51 29.20 31.80 64.57 30.01 46.07 1.61 37.40
Min 9.92 10.20 9.57 0.05 5.60 6.70 4.29 0.00 0.00 1.50 20.00
Mean 15.97 33.82 54.70 14.11 16.34 22.42 25.87 13.60 8.64 1.56 28.71
Mode 9.92 0.00 60.00 0.00 7.66 30.60 0.00 0.00 0.00 1.52 20.00
@seismicisolation
@seismicisolation
Median 16.07 29.88 60.00 13.75 17.90 28.00 21.85 16.02 5.12 1.56 28.56
SD 2.69 19.68 21.37 9.43 5.97 11.01 13.29 9.39 10.50 0.04 4.77
Variance 7.25 387.45 456.57 88.98 35.65 121.12 176.54 88.17 110.24 0.00 22.76
S. V. Singh and S. Ghani
36.00
Predictded Residual Shear Strength
31.00
26.00
21.00
16.00
16.00 21.00 26.00 31.00 36.00
Actual Residual Shear Strength
Fig. 4 Actual versus predicted residual shear strength values for training data
36.00
Predicted Residual Shear Strength
31.00
26.00
21.00
16.00
16.00 21.00 26.00 31.00 36.00
Actual Residual Shear Strength
Fig. 5 Actual versus predicted residual shear strength values for testing data
4 Conclusion
Based on the analysis and predictions made using the Artificial Intelligence model
for residual shear strength of soil, the following conclusions can be drawn:
@seismicisolation
@seismicisolation
• The results of the AI model show a robust correlation value of 0.93 in testing data
between actual vs. predicted shear strength. This indicates that AI models can be
used to predict residual shear strength of soil with high accuracy for the specified
area.
• The input parameters such as angle of internal friction, cohesion, and density were
found to have a significant impact on the prediction of residual shear strength of
soil. Therefore, it is essential to consider these parameters when developing an
AI model for predicting residual shear strength.
• The results of the AI model indicate that more data is required to improve the
accuracy of the model. More data for soils with different characteristics, such
as grain size distribution, plasticity, and mineralogy, could help to enhance the
accuracy of the model.
• The results of the AI model can be used for practical applications, such as
designing slopes, embankments, and foundations. By providing accurate predic-
tions of the residual shear strength of soil, the AI model can assist engineers and
designers in making informed decisions and reducing the risk of failure.
References
1. Schulz WH, Wang G (2014) Residual shear strength variability as a primary control on the
movement of landslides reactivated by earthquake-induced ground motion: Implications for
coastal Oregon, U.S. J Geophys Res Earth Surface 119(7):1617–1635. Available at https://doi.
org/10.1002/2014JF003088
2. Skempton AW (1964) The long-term stability of clay slopes. Geotechnique 14:77–101
3. Kenney TC (1967) Slide behavior and shear resistance of a quick clay determined from a study
of the landslide at Selnes, Norway. In: Proceedings of the geotechnical conference, Oslo, vol
1, pp 57–64
4. Mesri G, Shahien M (2003) Residual shear strength mobilized in first-time slope failures. J
Geotechn Geoenviron Eng 129(1)
5. Baghbani A et al (2022) Application of artificial intelligence in geotechnical engineering:
a state-of-the-art review. Earth-Sci Rev. Available at https://doi.org/10.1016/j.earscirev.2022.
103991
6. Goh AT (2002) Probabilistic neural network for evaluating seismic liquefaction potential. Can
Geotechn J 39(1):219–232
7. Kerh T, Chu D (2002) Neural networks approach and microtremor measurements in estimating
peak ground acceleration due to strong motion. Adv Eng Softw 33:733–742
8. Khan SZ et al (2016) Prediction of the residual strength of clay using functional networks.
Geosci Front 7(1):67–74. Available at https://doi.org/10.1016/j.gsf.2014.12.008
9. Ghani S, Kumari S (2021) Liquefaction study of fine-grained soil using a computational model.
Innov Infrastruct Solut. https://doi.org/10.1007/s41062-020-00426-4
10. Ghani S, Kumari S (2021) Sustainable development of prediction model for seismic hazard
analysis. In: Sustainable development through engineering innovations. Springer, pp 701–716
11. Kaya Z (2016) Predicting liquefaction-induced lateral spreading by using neural network and
neuro-fuzzy techniques. Int J Geomech 16(4):1–14. https://doi.org/10.1061/(ASCE)GM.1943-
5622.0000607
12. Kumar M, Bardhan A, Samui P, Hu JW, Kaloop MR (2021) Reliability analysis of pile foun-
dation using soft computing techniques: a comparative study process. https://doi.org/10.3390/
pr9030486
@seismicisolation
@seismicisolation
13. Kutanaei SS, Choobbasti AJ (2019) Prediction of liquefaction potential of sandy soil around a
submarine pipeline under earthquake loading. J Pipeline Syst Eng Pract 10(2):4019002
14. Mughieda OS, Bani-Hani K, Abu Safeh BF (2009) Liquefaction assessment by artificial neural
networks based on CPT. Int J Geotech Eng 3(2):289–302. https://doi.org/10.3328/IJGE.2009.
03.02.289-302
15. Sabbar AS, Chegenizadeh A, Nikraz H (2019) Prediction of liquefaction susceptibility of clean
sandy soils using artificial intelligence techniques. Indian Geotech J 49(1):58–69. https://doi.
org/10.1007/s40098-017-0288-9
16. Samui P, Sitharam TG (2011) Machine learning modeling for predicting soil liquefaction
susceptibility. Nat Hazards Earth Syst Sci 11(1):1–9. https://doi.org/10.5194/nhess-11-1-2011
17. Tiwari B, Marui H (2005) A new method for the correlation of residual shear strength of the
soil with mineralogical composition. J Geotech Geoenviron Eng 131(9):1139–1150. https://
doi.org/10.1061/(asce)1090-0241(2005)131:9(1139)
18. Das SK, Basudhar PK (2008) Prediction of residual friction angle of clays using artificial neural
network. Eng Geol 100:142–145
19. Das SK et al (2011) Machine learning techniques applied to the prediction of residual strength
of clay. Central Euro J Geosci 3(4):449–461. Available at https://doi.org/10.2478/s13533-011-
0043-1
@seismicisolation
@seismicisolation
Application of Digital Image Correlation
Technique in Geotechnical Engineering
Sukanta Das
1 Introduction
Digital Image Correlation (DIC) has several important applications in civil engi-
neering. In structural deformation analysis, DIC can be used to measure the defor-
mation and displacement of structures under various loading conditions. It helps in
understanding the structural behavior, identifying potential failure points, and vali-
dating structural design models. In material characterization, DIC allows for the
measurement of strain and deformation in materials such as concrete, metals, and
composites. It helps in understanding the material’s mechanical properties, including
elasticity, plasticity, and failure behavior. Similarly, DIC can be employed to monitor
the behavior of soil and rock masses, including slope stability analysis, tunneling,
and excavation processes. It helps in assessing ground movements, deformations,
and potential failure mechanisms. DIC can assist in the structural health monitoring
of bridges, buildings, and other infrastructure. It enables the detection of cracks,
displacements, and deformations, allowing for timely maintenance and repair inter-
ventions. This technique can also be used to analyze the response of structures during
earthquakes. It helps in evaluating the performance of structural systems, identifying
areas of excessive deformation or stress concentration, and improving seismic design
approaches.
DIC technique can complement traditional NDT methods by providing surface
strain and displacement measurements, allowing for a more comprehensive evalua-
tion of structural integrity and detecting hidden defects or damages. The advantage
of DIC in civil engineering lies in its ability to provide full-field measurements of
displacement and strain, offering detailed information about localized behavior and
capturing the response of complex structures or materials. It is a non-contact, non-
destructive technique that can be applied in laboratory testing as well as real-world
S. Das (B)
Department of Earthquake Engineering, Indian Institute of Technology Roorkee, Roorkee, India
e-mail: sada@eq.iitr.ac.in
@seismicisolation
@seismicisolation
144 S. Das
field conditions, making it a valuable tool for research, design, and monitoring in
civil engineering applications.
One of the early pioneering works in the application of DIC to geotechnical
engineering was published by Dr. Charles W. Swan from the University of Cali-
fornia, Davis, in the late 1990s. Swan and his team used DIC to measure surface
displacements and strains in soil models subjected to various loading conditions.
Their work demonstrated the feasibility and potential of DIC for geotechnical anal-
ysis. Since then, numerous researchers and institutions have adopted DIC in geotech-
nical engineering. They have used it to investigate soil-structure interaction, monitor
the behavior of slopes, analyze tunneling-induced ground movements, study the
performance of retaining walls, and assess the stability of foundations, among other
applications. It is important to note that DIC is a continually evolving technology,
and its application in geotechnical engineering is expanding as researchers and prac-
titioners explore its capabilities and refine its methodologies for specific geotechnical
problems.
Rechenmacher et al. [1] focused on the next generation of sand behavioral models
to understanding the relationships between these mesoscale deformational entities
and corresponding macroscale response. This understanding may also help with
efforts to understand jamming-unjamming transitions in dense granular flows more
generally. They have also DIC technique. Stanier and White [2] carried out measure-
ments in DIC during element and model testing. In contrast to fluid mechanics, which
is where PIV originated, geotechnical engineering frequently deals with extremely
low strain levels. This prompted the creation of methods to enhance the accu-
racy of PIV-based deformation measurements through modifications to the experi-
mental processes as well as the imaging equipment and software. Stanier et al. [3]
demonstrate how recent developments in PIV-DIC algorithms, when combined with
photogrammetric correction routines, provide more accurate deformation measure-
ments for geotechnical applications. For the benchmarking, a number of synthetic
pictures that resembled soil were subjected to set displacements and deformations.
Eichhorn et al. [4] reported the use of ChAruUco binary fiducial marker in DIC
technique. The costing of the setup be reduced with this marker. Chavda et al. [5]
used image based deformation analysis for open caisson foundation in GeoPIV-RG
on Matlab. They also validated the displacement observed from GeoPIV-RG and dial
gauge. The image based deformation analysis successfully captured soil dissipation
surrounding the open caisson. Janabi et al. [6] discussed DIC technique for failure
mechanism of circular footing on layered soil. The similar methodology can be used
for dynamic analysis of slope [7]. However, the load-settlement curve is obtained
from conventional instruments. Finally, they proposed the influence factor for the
determination of the elastic settlement of footing on layered soil. The distribution of
load beneath the footing is observed from DIC clearly in static condition.
@seismicisolation
@seismicisolation
Application of Digital Image Correlation Technique in Geotechnical … 145
2 Methodology
The working principle of Digital Image Correlation (DIC) involves comparing pixel
intensities between consecutive images to determine the displacement and deforma-
tion of an object or surface. The methodology for Digital Image Correlation (DIC)
in geotechnical engineering typically involves the following steps, as given in Fig. 1.
Sample Preparation: The sample should be prepared in such a way that the texture
of the area of interest has been looks like a grid. The colored materials can be mixed
with the natural soil; speckle can be made on the rock surface. Other materials steel
ball and other granular material having different textures than soil can also be used
for the formation of good texture.
Image Acquisition: Capture a series of digital images of the specimen using a high-
resolution camera. It is important to maintain consistent lighting conditions and
camera settings throughout the image acquisition process.
Image Pre-processing: Prior to DIC analysis, perform image pre-processing
steps, which may include image cropping, resizing, noise reduction, and image
enhancement techniques. These steps help optimize the image quality for accurate
displacement measurement.
Image Correlation: Apply DIC algorithms to the image sequence to track the
displacement and strain patterns on the specimen’s surface. DIC algorithms compare
Fig. 1 Flowchart for the

performing DIC in
geotechnical problems
@seismicisolation
@seismicisolation
146 S. Das
pixel intensities between consecutive images to determine the displacement vectors

for each point or subset on the specimen’s surface. Different DIC software packages
offer various algorithms, such as subset-based or point-based DIC, which have their
own advantages and limitations.
Post-processing and Analysis: Analyze the displacement and strain data obtained
from DIC. This may involve filtering the displacement data, calculating strain fields,
and extracting relevant parameters such as deformation profiles, displacement gradi-
ents, and strain localization zones. Comparisons with numerical models or theoretical
predictions can also be performed to validate experimental results.
Interpretation and Reporting: Interpret the results obtained from DIC analysis and
draw conclusions regarding the behavior of the geotechnical system under investiga-
tion. Generate visualizations such as displacement maps, strain contours, or anima-
tions to facilitate a better understanding of the deformation mechanisms. Document
the findings in technical reports or research papers.
It is worth noting that specific applications within geotechnical engineering may
require additional considerations and adaptations to the DIC methodology. These
can include challenges related to surface preparation, dealing with large deforma-
tions, handling soil/rock discontinuities, and accounting for material heterogeneity.
Therefore, it is recommended to refer to relevant research literature, guidelines, and
experienced practitioners for specific methodologies tailored to the desired geotech-
nical analysis. The key idea behind DIC is to track the movement of small, discrete
points or patterns (subsets) on the object’s surface by comparing their pixel intensities
over time. By analyzing the displacements and deformations, DIC provides valuable
insights into the mechanical behavior of the object or surface under investigation. It
is important to note that the actual implementation of DIC may involve more sophis-
ticated algorithms and techniques, and various software packages are available to
perform DIC analysis with greater accuracy and efficiency. In the present study, the
post-process has been done using VIC 2D. The version and detailed step by step
procedure are available in their document section. The specification required for the
DIC measurement is also explained in Correlated Solutions [8].
3 Present Study
In the present study, a state of art Digital Image Correlation (DIC) technique is used
to capture photographs and analyze the strip footing failure mechanism. In Fig. 2, the
pictorial view of the DIC setup is shown. An 8.8 Megapixel camera with 16 mm lens
is used for capturing the images during the test. It is linked to VIC-Snap software
[9] in a computer. The camera is set at a fixed distance from the transparent sheet of
the model. Two LED white lights were installed beside the camera to improve the
visibility of the surface texture of the sand. In two-dimensional DIC, measurements
are taken of the surface texture of the sample. To track the particle movement, DIC
used many patterns like dots, grids, and pixel boxes (subset). These subsets should
@seismicisolation
@seismicisolation
Fig. 2 A DIC test setup for strip footing on horizontal ground under centric vertical loads
be unique, with a range of grayscale contrast ratio varying from 0 to 255 for white
to black.
3.1 Material Used
The Solani River sand was used in the present study. The dyed color is mixed with
the sand at low and high intensity for better texture or grid formation. The overall
30% of colored sand mixed with the 70% natural Solani sand. In Fig. 3, the texture of
the individual sand has been presented. Author has also checked the index properties
of the natural sand remained unchanged due to adding of dyed colored sand. The
dyed colored sand was used by Stanier and White [2] and Chavda et al. [5] for appro-
priate range contrast. The material properties of Solani sand have been considered
as reported by Das and Maheshwari [10]. All tests have been conducted at a relative
density of 80%, by weight and tamping method.
The detailed test plane and loading mechanism have been reported by Das and
Maheshwari [7]. In the present study, an application of the DIC to mimic the failure
wedge developed beneath the footing has been examined.
@seismicisolation
@seismicisolation
148 S. Das
Fig. 3 Sample preparation for DIC, sand mixed with dyed colored at different percentage
3.2 Results and Discussions
The centric vertical load applied to the strip footing of width (B) 50, 70, and 100 mm.
The DIC setup has been mounted at the transparent site of the tank, as shown in
Fig. 2. The images have been captured with the increment of the loads. The vertical
displacement developed beneath the footing has been shown Fig. 4. In Fig. 4a, the
vertical displacement of footing of width of 50 mm has been presented. Similarly,
Fig. 4b, c for footing width 70 mm and 100 mm, respectively.
It has been observed that the contours have been extending as the footing width
increases at a relative density of 80%. Further, the zones beneath the footing devel-
oped similar to conventional theory, as given in Fig. 5. Active zones beneath the
center of footing and radial at the sides.
4 Future Scope
The future scope of Digital Image Correlation (DIC) in geotechnical engineering has
been found promising and can bring advancements in various areas. Here are a few
potential areas of development:
1. Advanced Deformation Analysis: DIC can be further enhanced to handle
large deformations and complex failure mechanisms in geotechnical materials.
This includes developing algorithms and methodologies to accurately track
and analyze the deformation of soils, rocks, and geosynthetic materials under
challenging conditions.
@seismicisolation
@seismicisolation
(a)
(b)
(c)
Fig. 4 Vertical displacement of strip footing on horizontal ground surface subjected centric vertical
load for relative density 80% a B = 50 mm, b B = 70 mm, c B = 100 mm
@seismicisolation
@seismicisolation
150 S. Das
Fig. 5 Conventional wedge failure of footing on horizontal ground surface under centric loads
2. Real-Time Monitoring: Integrating DIC with real-time monitoring systems can

enable continuous and automated measurement of ground displacements and
deformations. This would provide valuable data for geotechnical infrastructure,
such as monitoring embankments, slopes, retaining walls, and tunnels in real-
time, enhancing early warning systems for potential failures.
3. Multiphysics Analysis: DIC can be combined with other measurement techniques
and numerical modeling tools to enable multiphysics analysis in geotechnical
engineering. For instance, integrating DIC with sensors for pore pressure, temper-
ature, or strain gauges can provide a comprehensive understanding of coupled
processes such as soil–water interaction, thermal effects, and deformation
behavior.
4. Field-Scale Applications: DIC has predominantly been used in laboratory
settings, but there is potential for its application in field-scale geotechnical
projects. Expanding DIC techniques to large-scale field experiments and full-
scale geotechnical structures can provide valuable insights into real-world
behavior and help validate numerical models and design assumptions.
5. Data-Driven Approaches: DIC generates a significant amount of data, and with
advancements in machine learning and data analytics, there is potential for devel-
oping data-driven approaches for geotechnical engineering. This includes auto-
mated feature extraction, pattern recognition, and predictive modeling based on
DIC data, which can aid in better understanding and predicting geotechnical
behavior.
6. Integration with Remote Sensing: Combining DIC with remote sensing tech-
niques, such as satellite imagery and aerial photogrammetry, can provide a
broader spatial coverage for monitoring large-scale geotechnical phenomena,
including subsidence, landslides, and ground stability over regional areas.
These are just a few potential areas where DIC can contribute to the future
development of geotechnical engineering. As technology continues to advance and
researchers explore new applications, the scope and capabilities of DIC in geotech-
nical engineering are likely to expand, leading to improved understanding, design,
and monitoring of geotechnical systems.
@seismicisolation
@seismicisolation
5 Summary
In the present study, the working principle of Digital Image Correlation (DIC) has
been summarized. The steps involved in the performing the DIC have also been
explained. Further, the future scope of DIC has been presented. Based on the above
discussions for use of DIC in geotechnical engineering, the following conclusions
can be drawn.
• The application of the DIC technique is wide, particularly measurement and
monitoring of the Geostructures.
• Many limitations of NDT (Non Distractive Test) can be overcome using DIC
technique. Though the calibration and tracking of soil in three-dimensional space
is still a challenging task in geotechnical engineering.
• In two-dimensional DIC technique, the failure mechanism of the strip foundation
on horizontal ground can be successfully captured. DIC techniques also indicate
the influence of footing width and failure wedges developed beneath the footing.
• The future application of the DIC has been found wide, especially health moni-
toring of Geostructures. Some of the potential areas such as advanced deforma-
tion analysis, real-time monitoring, multiphysics analysis, field-scale applications,
data-driven approaches, and integration with remote sensing where DIC can be
used.
References
1. Rechenmacher AL, Finno RJ (2004) Digital image correlation to evaluate shear banding in
dilative sands. Geotech Test J 27(1):13–22
2. Stanier SA, White DJ (2013) Improved image-based deformation measurement in the centrifuge
environment. Geotech Test J 36(6):1–14
3. Stanier SA, Blaber J, Take WA, White DJ (2015) Improved image-based deformation
measurement for geotechnical applications. Can Geotech J 53:727–739
4. Eichhorn GN, Bowman A, Haigh SK, Stanier SA (2020) Low-cost digital image correlation
and strain measurementfor geotechnical applications. Special issue, Strain
5. Chavda JT, Mishra S, Dodagoudar GR (2020) Experimental evaluation of ultimate bearing
capacity of the cutting edge of an open caisson. Int J Phys Model Geotech 20(5):281–294
6. Janabi HF, Sakleshpur VA, Prezzi M, Salgado R (2022) Strain influence diagrams for settlement
estimation of square footings on layered sand. J Geotech Geoenviron Eng 148(5):04022025
7. Maheshwari BK, Das S (2023) Experimental and numerical study for seismic response of strip
footing on slopes. In: Soil dynamics and earthquake engineering. Elsevier (In press)
8. Correlated solutions: the Vic-2D system: technology overview. Available at http://www.correl
atedsolutions.com/vic-2d/ (2015)
9. Cintrón R, Saouma V (2008) Strain measurements with digital image correlation system Vic-
2D, CU-NEES-08-06 Center for Fast Hybrid Testing. Department of Civil Environmental and
Architectural Engineering, University of Colorado
10. Das S, Maheshwari BK (2023) Failure mechanism of strip footing on slopes under vertical and
eccentric loads: experimental and numerical study. Int J Phys Model Geotechn (Under review)
@seismicisolation
@seismicisolation
SSI on Geodesic Dome Using RSM
and Comparison with ANN
M. Roopa, Jayachandra, and Manjunath Vatnalmath
1 Introduction
Engineers and architects have long been fascinated by constructions that can span a
vast distance without the need for intermediary columns. In this instance, it seems
that a spherical dome with a grid-like structure is suitable [1]. The seismic response
spectrum for the design basis earthquake level (DBE) was employed to analyze a
single-layer reticulated dome, revealing the base shear and lateral displacement [2].
The interaction between the soil and structure (SSI) was investigated with shallow
foundation structures using the response spectrum method (RSM), and it was discov-
ered that RSM is an effective technique for analyzing SSI [3]. Vaseghiamir et al.
[4] proposed a methodology that accounts for uncertainties in soil-structure system
parameters and seismic loading. This approach is utilized for constructing distinctive
steel moment frame buildings with surface foundations. The research suggests that
structures built on moderately soft to solid soils with shear wave velocities exceeding
150 m/s do not require adjustments in design base shear. The impact of SSI was
examined using seismic data from 58 locations in California. An expert system,
specifically a neural network, is employed for assessing SSI and land use. Back-
propagation is used for models 1 and 2, while a general regression neural network
technique is applied to models 3, 4, and 5. To avoid overfitting in backpropagation,
a three-stage strategy involving training, testing, and cross-validation is utilized [5].
Different height-to-width ratios are considered for various soil conditions, number
of stories, and local site depths. The direct method is used to solve SSI problems
M. Roopa (B)
Department of Civil Engineering, Siddaganga Institute of Technology, Tumakuru, India
e-mail: roopam@sit.ac.in
Jayachandra
School of Civil Engineering, REVA University, Bengaluru, India
M. Vatnalmath
Department of Mechanical Engineering, Siddaganga Institute of Technology, Tumakuru, India
@seismicisolation
@seismicisolation
154 M. Roopa et al.
under seismic loads. The response parameters evaluated to extract training and testing
sets include period, top story displacement, and top story accelerations. The numer-
ical model response is significantly influenced by the hysteretic damping ratio, soil
Poisson’s ratio, and footing size [6]. Machine learning techniques, particularly ANN
and support vector machines, are robust strategies that consistently yield reliable
results. Machine learning methodologies like ANN and support vector machines are
employed to assess the impact of soil-structure interactions on the structural seismic
response under various earthquake conditions [7]. Furthermore, the backpropagation
mechanism is utilized in the feedforward approach to compare the outcomes of the
suggested ANN with the results of the finite element method (FEM) [8]. Sbartai and
Goudjil [9] developed a model that predicts the response of the soil-structure inter-
action system using a statistical approach based on an ANN, eliminating the need
for advanced numerical techniques.
The seismic analysis of geodesic domes has become a critical part of the design
and construction of such structures. Many studies have been carried out FEA on
geodesic domes for SSI, however, the studies using RSM method for analyzing SSI
of geodesic dome are scanty. Hence, the present study aims to analyze the SSI on
a geodesic dome of 30 m diameter by considering the in-situ soil properties using
SAP 2000 and ANN is employed to correlate the effect of SSI.
1.1 Soil-Structure Interaction
Soil-Structure Interaction (SSI) impacts the response of structures during seismic

events. It involves the mutual influence between the motion of soil particles in
response to ground shaking and the resulting motion of the structure, as shown in
Fig. 1. The diagram illustrates how seismic waves propagate through the soil medium,
affecting both the free-field motions along the base and sides of the model and the
response of the interacting system, which is predicted using equations of motion in
a finite element model. In this case, the dynamic interaction is not considered, and
the motion of the foundation and superstructure is determined based on the free-field
motion.
The inertia force acting on the structure causes vibrations in the superstructure,
foundation, and soil at the interface. The motion equation is given by,
m ẍ + c ẋ + kx = −M I ü (1)
m Mass matrix of the total system includes structure, foundation, and soil.
ẍ, ẋ, x acceleration, velocity, and displacement with respect to the base or
foundation.
c Damping matrix of the structure and soil.
ü Free field ground acceleration.
k Stiffness matrix of the structure and soil.
@seismicisolation
@seismicisolation
SSI on Geodesic Dome Using RSM and Comparison with ANN 155
Fig. 1 Schematic
representation of
soil-structure interaction and
transmission of seismic
waves through soil media
M Mass matrix of the structure having non zero masses.

I Mass matrix having non zero masses for the structural DOF.
1.2 Artificial Neural Network Model
Artificial Neural Networks (ANNs) offer a viable alternative to mathematical

modeling by incorporating statistical nonparametric and nonlinear models, effec-
tively addressing a range of problems including identification, decision support,
diagnostics, prediction, and more. ANNs are inspired by biological systems and
emulate the functioning of biological neurons through mathematical models. Typi-
cally, an ANN consists of three layers: input, output, and hidden. The neurons within
the input layer receive external inputs to the system. The Back-Propagation (BP)
algorithm is a type of Artificial Neural Network (ANN) model that utilizes super-
vised learning. In this model, the network generates predicted output values based
on its input values. The sigmoid function is commonly employed as the activation
function, which produces output values ranging from 0 to 1. The BP-ANN algorithm
propagates the errors from the output layer toward the inner layers, aiming to mini-
mize the difference between the network’s outputs and the expected outputs. This
process is depicted in Eq. 2.
e = d − y(k) (2)
where
e Error value,
d Expected output value,
y Output value of k.
@seismicisolation
@seismicisolation
156 M. Roopa et al.
1.3 Performance Criteria
The coefficient of determination R2 is the most commonly used statistical measure

for assessing the goodness-of-fit of a model’s results. It ranges between 0 and 1,
representing the extent to which the observed outcomes are accurately reproduced
by the model. A higher R2 value (with a maximum of 1) indicates a better-fitting
model. The relationship is illustrated in Eq. 3.
2
yi − y
R =1−
2
(3)
(yi − y)2
where
y Predicted value of y,
y Mean value of y, i = 1 to n.
2 Methodology
2.1 Soil Foundation Modeling (MAT)
The values for soil springs are computed using the properties specified in the Richart
and Lysmer (1970) formula, as presented in Table 1. These spring values calculated
as shown in Table 2 are then incorporated into the geodesic dome model, along with
a mat foundation, as depicted in Fig. 2.
The geodesic dome was modeled using SAP2000, V22 with suitable elements
for the frame and glass. The steel tubular pipes, spherical node connectors, and ring
beams were modeled using frame elements and glass members were modeled using
the shell elements with suitable dimensions. The soil properties considered for the
SSI analysis are shown in Table 3.
Table 1 Soil spring values

Direction Spring values Equivalent radius
according to Richart and /
Lysmer (1970) Vertical Kz = 4Gr z
r z = LπB
(1−ϑ)
/
32(1−ϑ)Gr x
Horizontal Kx = Ky = (7−8ϑ) r x = LπB
3 /
8Gr∅x B3
Rocking K ∅x= 3(1−ϑ) r Øx = 4 L3π
3 /
8Gr∅y B3
K ∅y= 3(1−ϑ) r Øy = 4 L3π
3 /
16Gr∅z 3 +B L 3
Twisting K ∅z= 3 r Øz = 4 L B 6π
@seismicisolation
@seismicisolation
Table 2 Soil spring values of

Direction Notation Spring values (kN/m)
geodesic dome (30 m, MAT)
Vertical Kz 1.54 E + 07
Horizontal Kx = Ky 1.31 E + 07
Rocking K ∅x=K ∅y 3.05 E + 07
Twisting K ∅z 4.27 E + 07
Fig. 2 Geodesic dome of

30 m diameter with mat
foundation
Table 3 Soil properties

SPT No. N 50 50
Shear wave velocity Vs 80 × N 1/3 294.722 m/s
Mass density ρV ρ 1834.86 kg/m3
Shear modulus G ρV s 2 159.3 MPa
Poisson’s ratio υ 0.25–0.3 0.3
SSI Using ANN

The ANN is modeled using MATLAB 2021Ra, which allows for the creation and
execution of the ANN model. The values of soil springs and corresponding base
shear outputs from all zones are employed to train the ANN. A large dataset of
soil properties is obtained by varying the properties namely SPT number, shear
wave velocity, density, and Poisson’s ratio. The soil spring values were computed
according to Table 1 and then incorporated to foundation of geodesic dome to obtain
FEA results for base shear. The ANN is then trained with four inputs representing
soil spring values (Kz, Kx = Ky, Kx = Ky, and Kz), and four outputs representing
base shear values for zones II, III, IV, and V (Fig. 3).
The Levenberg–Marquardt model is utilized for training, validation, and testing
of the dataset, as depicted in Fig. 4
@seismicisolation
@seismicisolation
158 M. Roopa et al.
Fig. 3 Structure of neural network model
Fig. 4 Training network details for different combinations of soil properties
Figure 5 displays the validation performance and mean squared error of the network.
It consists of three separate plots representing the training, validation, and testing
processes. Additionally, a fourth plot demonstrates the excellent correlation between
the Finite Element Analysis (FEA) and ANNs findings for diverse soil conditions.
Furthermore, Fig. 6 presents an error histogram illustrating that the extensive training,
validation, and testing samples exhibit minimal error, with the majority of values
being extremely close to zero.
The training process iterates on the training vectors until the model reaches a
point where the error of the network on both the training and validation vectors is
minimized. This helps to prevent overfitting of the data sets. Figure 7 illustrates this
process, and it indicates that the best validation performance is achieved at epoch
258.
The base shear results of FEA and obtained results of the ANNs with the regression
curve for each zone are shown from Figs. 8, 9, 10 and 11. It is evident from these
figures that both the FEA and ANNs model produced base shear values with variation
of less than 1% error.
@seismicisolation
@seismicisolation
Fig. 5 Regression curves for training, validation, and test data for different soil conditions
@seismicisolation
@seismicisolation
160 M. Roopa et al.
Fig. 6 Error histogram with 20 bins for the training, validation, and test steps
Fig. 7 Best validation performance in artificial neural network
@seismicisolation
@seismicisolation
Zone II (FEA/ANNs)
700
Base Shear (kN) ANNs
600
500
400
300
200
100
0
0 100 200 300 400 500 600 700
Base Shear (kN)
Fig. 8 Normal plot for base shear results of ANNs and FEA for zone II
Zone III (FEA/ANNs)

1000
900
Base Sher (kN) ANNs
800
700
600
500
400
300
200
100
0
0 200 400 600 800 1000
Base Shear (kN)
Fig. 9 Normal plot for base shear results of ANNs and FEA for zone III
@seismicisolation
@seismicisolation
162 M. Roopa et al.
Zone IV (FEA/ANNs)
1600
1400
Base Shear (kN) ANNs
1200
1000
800
600
400
200
0
0 200 400 600 800 1000 1200 1400 1600
Base Shear (kN)
Fig. 10 Normal plot for base shear results of ANNs and FEA for zone IV
Zone V (FEA/ANNs)
2500
Base Sher (kN) ANNs
2000
1500
1000
500
0
0 500 1000 1500 2000 2500
Base Shear (kN)
Fig. 11 Normal plot for base shear results of ANNs and FEA for zone V
@seismicisolation
@seismicisolation
4 Conclusion
In this study, MATLAB is utilized to develop ANN for analyzing the base shear values
of a 30 m geodesic dome under different soil conditions. The ANN model demon-
strates a high level of accuracy in predicting base shear results when compared to
findings obtained through FEA. ANN proves to be a reliable technique for analyzing
large datasets in this context. The study focuses on using ANN to analyze the effects of
various soil characteristics on base shear values. The ANN model accurately predicts
the outcomes of base shear with an error margin of less than 1%. The reliability of the
ANN model is demonstrated through regression curves, error histograms, validation
curves, and normal plots for both in-situ and diverse soil parameters. Overall, the
investigation confirms that the proposed ANN model closely aligns with the FEA
model for the existing geodesic dome’s soil-structure interaction.
References
1. Barbieri N, Machado RD, Barbieri L (2017) Dynamic analysis of a geodesic dome. In: Proceed-
ings of the 24th ABCM international congress of mechanical engineering. ABCM 2017 24–28.
https://doi.org/10.26678/ABCM.COBEM2017.COB17-5951
2. Anwar N, Norachan P, Aung TH (2015) Case study: challenges of a single-layer reticulated
dome. In: IABSE conference—structural engineering: providing solutions to global challenges.
J Constructional Steel Res
3. Gao Z, Zhao M, Du X, Zhao X (2020) Seismic soil–structure interaction analysis of structure
with shallow foundation using response spectrum method. Bull Earthq Eng 18(8):3517–3543.
https://doi.org/10.1007/s10518-020-00827-x
4. Vaseghiamiri S, Mahsuli M, Ghannad MA, Zareian F (2020) Probabilistic approach to
account for soil-structure interaction in seismic design of building structures. J Struct Eng
146(9):04020184. https://doi.org/10.1061/(ASCE)ST.1943-541X.0002741
5. Sariyar O, Ural DN (2010) Expert system approach for soil structure interaction and land
use. J Urban Plan Dev 136(2):135–138. https://doi.org/10.1061/(ASCE)0733-9488(2010)136:
2(135)
6. Dao ND, Ryan KL (2022) Soil–structure interaction and vertical-horizontal coupling effects
in buildings isolated by friction bearings. J Earthquake Eng 26(4):2124–2147. https://doi.org/
10.1080/13632469.2020.1754969
7. Ali T, Eldin MN, Haider W (2023) The effect of soil-structure interaction on the seismic
response of structures using machine learning, finite element modeling and ASCE 7–16
methods. Sensors 23(4):2047. https://doi.org/10.3390/s23042047
8. Pala M, Caglar N, Elmas M, Cevik A, Saribiyik M (2008) Dynamic soil–structure interaction
analysis of buildings by neural networks. Constr Build Mater 22:330–342. https://doi.org/10.
1016/j.conbuildmat.2006.08.015
9. Richart FE, Hall JR, Woods RD (1970). Vibrations of soils and foundations.
10. Sbartai B, Goudjil K (2012) Response of soil-structure interaction system using artificial neural
network (ANN). In: Uncertainty modeling in knowledge engineering and decision making,
1017–1022. https://doi.org/10.1142/9789814417747_0163
@seismicisolation
@seismicisolation
Risk Analysis and Approaches
@seismicisolation
@seismicisolation
Overlap of the Safety Distance of the Gas
and Oil Pipeline Network with the Urban
Area of the City of Mohammedia
Morocco
Abderrahmane Jadouane, Abdelmajid Essami, Azzeddine Chaouki,

C. Bhargavi, Ibtissam Gourich, and Mustafa Nadraoui
1 Introduction
Gas and oil pipeline networks play a critical role in transporting energy resources
from their production source to their consumption destination. These networks are
comprised of a complex system of pipelines, pumps, valves, and storage facilities
that work together to transport oil and gas over long distances. The industry heavily
depends on these networks to transport petroleum products from extraction sites
to processing plants, refineries, and storage facilities. There are several methods to
transport these materials, but pipelines are considered the safest, most efficient, and
cost-effective way to transport large quantities of oil and gas over long distances,
compared to other modes of transportation such as trucks, trains, or ships. Despite
their importance in the global energy supply chain, gas and oil pipeline networks
also pose significant risks to public safety and the environment, especially with
increasing population growth and urbanization which leads to an increase in demand
for petroleum materials and an increase in the construction and use of pipelines
[1], which can be located in urban areas, near water sources, or in environmentally
sensitive areas, leading to accidents and spills that can have catastrophic conse-
quences. To address these risks, pipeline operators must implement safety measures
and emergency response plans, while governments enforce regulations and standards
to ensure pipeline safety. Despite the challenges and risks associated with gas and oil
pipeline networks, they remain a vital component of the global energy supply chain.
Understanding the safety, regulatory, and environmental aspects of these networks
A. Jadouane (B) · A. Essami · A. Chaouki · I. Gourich · M. Nadraoui

LADES, Department of Geography, Faculty of Arts and Humanities-Mohammedia, Hassan II
University of Casablanca, Casablanca, Morocco
e-mail: jadouane.a@gmail.com
C. Bhargavi
Department of Civil Engineering, Jyothy Institute of Technology, Bangalore, India
@seismicisolation
@seismicisolation
168 A. Jadouane et al.
is critical to ensuring their safe operation and meeting the growing global demand
for energy resources [2].
The overlap of the safety distance of gas and oil pipeline networks with urban areas
presents significant risks and potential hazards to public safety and the environment.
This is particularly relevant in the city of Mohammedia, Morocco, where there is a
considerable concentration of industrial activities. The risk of accidents related to
gas and oil pipelines is high due to the proximity of urban areas to the pipelines,
as well as its proximity to the oil port, which is the largest importer of petroleum
materials in Morocco. Furthermore, the city has already experienced explosions of
oil refining areas due to the flood of river Oued ElMalleh, which underscores the
potential danger of such overlap [3].
The central issue is to determine whether the current safety distance regulations
for gas and oil pipelines are sufficient to ensure the safety of the population and the
environment in the urban areas of Mohammedia. As a result, the main problem is
how to evaluate the adequacy of the existing safety distance regulations for gas and
oil pipelines in the city’s urban areas and to establish effective measures to mitigate
the risks associated with the overlap of pipeline safety distance with the urban areas
of the city. This article seeks to address this problem by evaluating the safety distance
in the event of an explosion, leak, or fire outbreak and determining whether it will
reach the civilian field.
2 Study Area
The city of Mohammedia is located on the coast of the Atlantic Ocean, 25 km to

the north-east of Casablanca and 65 km south-west of Rabat in Morocco. The urban
perimeter of the city covers an area of 34 km2 , between the mouths of river Oued
Elmalleh in the west and Oued El Nfifikh river to the east, and the motorway axis
to the south. The city is bounded to the north by the Atlantic Ocean, to the east
and south by the province of Benslimane, and to the west by the commune of Aïn
Harrouda (Fig. 1). It is located between 7° 22' and 7° 25' west longitude and 33° 41'
and 33° 43' north latitude [4]. The city of Mohammedia is one of the most important
industrial cities in Morocco, and this is represented in its availability of an integrated
industrial structure, consisting of more than 600 industrial units, on an area of 720
ha, representing 29% of the city’s area (Fig. 2), the most important of which are
petrochemical industries, in addition to the presence of the first petroleum port in
Morocco, from which a network of pipelines branches transportation of oil and gas
materials from sea tankers toward reservoirs, industrial zones and refineries that are
located in the city, such as the company Samir and the reservoirs of companies
located on the banks of river Oued ElMalleh, the length of this network of pipes is
estimated at about 80 km.
@seismicisolation
@seismicisolation
Overlap of the Safety Distance of the Gas and Oil Pipeline Network … 169
Fig. 1 Location of the city of Mohammedia–Morocco
Fig. 2 Land use in the city of Mohammedia 2022
@seismicisolation
@seismicisolation
3 Materials and Methods
In order to conduct this study, a series of steps and stages must be taken to under-
stand the study area Mohammedia city, the causes of the problems in oil and gas
transportation pipelines, and the technical standards for the safety distance that must
be taken to prevent damage from these pipelines in case of danger. The methodology
adopted in this study is divided into two parts and is illustrated in Fig. 3.
The first part aims to build a database about the study area in general and the
locations of oil and gas transportation pipelines from the port to distribution areas
in particular. Various references such as maps, aerial photographs, administrative
documents, and fieldwork were used to collect the necessary data. Then, the data
was digitized and converted into a digital database and ready-to-use maps, on which
safety distance standards for pipeline damage were applied.
The second part was to evaluate and determine the extent of damage caused by
these pipeline networks by Geographic information systems using ArcGIS software
were used to determine the spatial radiation range of the safety distance by projecting
international safety distance standards for pipelines transporting oil and gas using
Buffer analysis [5]. This was done to prevent any damage that may be caused by
the occurrence of danger from these pipelines and to produce a final map of the
intersection of safety distances with urban areas that are likely to be affected. These
areas were identified in six spatial ranges arranged according to the hazardous effects
resulting from the diameter and pressure of the pipeline.
Preliminary Identify Data Collect Spatial

Star
Study Involved Data
Apply
Creating Risk Measurements Calculate
Database Assessment To The Overlap
Database
Identification
Mapping Risk Of Overlap End
Areas
Fig. 3 Flowchart of research design
@seismicisolation
@seismicisolation
4 Results
The transportation of petroleum products through the pipeline network in the city
of Mohammedia is a crucial operation that requires the provision of pipelines from
the source point (the seaport) to the delivery point, whether it is refineries or storage
tanks located in the industrial areas of the city. In addition, some external points
in the city, such as the Sidi Larbi area which is located 15 km away from the city,
receive oil materials through a direct pipeline from the port.
In the city of Mohammedia, there is a network of pipelines estimated to be 12 km
long, as shown in the following Fig. 4. These pipelines extend from the petroleum
port to various points in the city, including the river Oued ElMalleh, warehouses,
and the coastal industrial area, particularly the Samir Company. The main network
consists of 7 pipelines that vary in thickness and the material they transport. Most
of these pipelines are buried underground at a depth of one to one and a half meters,
except for some areas where pump pressure intersections occur or due to the presence
of the river Oued ElMalleh, where they are passed through special bridges.
These pipelines are made of fiberglass (polyester) material, which is resistant
to heat, pressure, moisture, corrosion, and fire, additionally it is lightweight [6].
However, some pipelines are made of aluminum and iron, and some are welded to
transport liquid materials and coal. An example of this is the pipeline linking the
power station to its supply point. Another essential element in this process is the
pressure pumps, which transport these materials inside the pipelines. The inside of
33°43'31.72"N - 7°23'23.38"W
33°42'16.25"N - 7°24'37.83"W
Moroccan Society Storage (SOMAS) + Field Research + Google Earth 2022 33°40'37.71"N - 7°24'30.25"W
Fig. 4 Petrochemical transport pipeline network in Mohammedia city
@seismicisolation
@seismicisolation
33°42'24.7"N - 7°24'12.5"W 33°42'16.42"N - 7°24'37.99"W
Fig. 5 Pictures of some petrochemical transport pipelines in the city of Mohammedia
the pipeline has a high pressure ranging from 3 to 140 atmospheric pressure units,
which moves the transported material. Gas is the fastest transport material due to
its low density. In addition to the equipment that complements the pipeline trans-
port process, there are electronic, electrical, and mechanical control and monitoring
equipment whose goal is to ensure flow, pressure, and temperature indicators in the
pipelines. Additionally, safety equipment and safety valves are installed to function
in emergencies.
Despite all the equipment and features available in pipelines, they remain vulner-
able to the risks of failure due to various factors, such as intentional human actions,
age-related wear and tear, or technical and operational errors. Therefore, the areas
that host these pipelines in the city of Mohammedia, as represented in (Fig. 5), are
at risk of accidents related to pipelines.
Apart from the direct internal causes that can cause damage to the pipes and thus
the occurrence of danger as shown in Fig. 5, there are also other external factors such
as natural forces and disasters that occur in the area, especially as it is prone to floods
and tsunamis, in addition to geological hazards [7].
Not to mention the impact of the domino phenomenon, which may occur sequen-
tially for petroleum tanks that can affect the pipes in the event of an explosion or
catching fire, because there is an overlap of the effects of thermal forces with other
infrastructure such as industrial structures, and the pipeline network that supplies oil
from the port, or those that connect the industrial area in between, the convergence
in the distance between the tanks and the overlap of thermal power ranges raises the
possibility of occurrence of the Domino phenomenon [8].
5 Discussion
One of the problems that can cause danger in pipelines is the release of the transported
material from the inside to the outside due to various reasons, as illustrated in Fig. 6
[9].
@seismicisolation
@seismicisolation
Fig. 6 Forms of hazard sources in transport pipes
1. Gas leakage.
2. Occurrence of pressure and deformation in the outer cover of the pipeline.
3. Fire ignition due to the leakage of the transported material.
4. Leakage of the transported liquid material.
5. Occurrence of corrosion and rust in the pipeline membrane.
6. Technical malfunction in the pump resulting in excessive pressure.
The required safety distance to avoid potential risks due to pipelines in case of
explosion, leakage or fire depends on the pressure and diameter of the pipeline. The
higher the pressure, the longer the safety distance needs to be, which can reach up
to 200 m. High pressure can raise the spread distance of materials in case of danger
and therefore cause damage to the environment, whether it be another pipeline,
residential areas, or properties. It is generally recommended that a distance of 10 m
is the starting point for the safety distance of oil and gas pipelines that face a pressure
unit of 0.4 MPa [10].
According to the general accepted standards for safety distances of pipelines
transporting petrochemical materials, the following Fig. 7 was developed according
to the following rules [10]:
• 10 m for gas pipelines with a pressure of less than 0.4 megapascals.
• 20 m for gas pipelines with a pressure ranging from 0.4 megapascals to 4
megapascals and a diameter of up to 350 mm.
• 50 m for gas pipelines with a pressure ranging from 0.4 megapascals to 4
megapascals and a diameter greater than 350 mm.
• 50 m for gas pipelines with a pressure above 4 megapascals and a diameter of up
to 150 mm.
• 100 m for gas pipelines with a pressure above 4 megapascals and a diameter of
up to 300 mm.
@seismicisolation
@seismicisolation
Fig. 7 Overlap of urban areas with safety distance in case of danger according to pressure and pipe
diameter
• 150 m for gas pipelines with a pressure above 4 megapascals and a diameter of
up to 500 mm.
• 200 m for gas pipelines with a pressure above 4 megapascals and a diameter above
500 mm.
• 50 m for control stations and pumps.
The map illustrates that the territorial space potentially affected by pipeline
hazards, based on previous incidents, covers a large portion of the western part of the
city, including some areas of intersection such as the port and the coastal industrial
zone. Such overlap can be considered normal as these areas primarily rely on pipeline
transport for supply, distribution, and exploitation. However, what is concerning is
the overlap of the safety distance with certain recreational facilities such as Royal
Golf Club, and residential areas including Douar (village): FABRIKA, LASHHAB,
WARDA), raising a critical question about the possibility of civilian impact in these
locations in case of pipeline accidents.
These Dawawir are small villages with random and unequipped buildings. The
fragility that lies in them can help the danger spread faster and greater, because they
are not kept up, are not flexible, have no ability to withstand them, and they do not
@seismicisolation
@seismicisolation
have the tools of warning and vigilance in order to combat gas or oil leaks in case of
occurrence.
Therefore, additional measures must be taken to protect these areas from danger
and ensure the safety of their residents, by providing the necessary instructions,
training, and resources to equip these areas with tools to prevent accidents or to
transfer them from these places to safe places to live.
6 Conclusion
This article raises the problem of overlapping the safety distance of oil and gas
transport pipes in the city of Mohammedia-Morocco, with the urban areas of the
city, with the help of the available materials and information, given the difficulty of
the matter, especially since this field is a sensitive field and has great confidentiality
by the officials assigned to it. Also, this article focused on the spatial analysis of the
extent of the danger that could result from the pipe in the event of a problem occurring
to it, and it was done with the help of geographic information systems (GIS) in order
to obtain a buffer analysis that tried to apply some international standards for the
safety distance of pipes based on the diameter of the pipe and its density.
In general, this article emphasizes the importance of respecting the distances and
not overlapping between the urban and civil space with the safety distances, in order
to avoid the occurrence of disaster and damage, especially in the four areas that were
found to be threatened in the event of the spread of danger from the pipes,
We also stress the need for the authorities to be fully prepared in order to protect
public safety and the environment and to review assessments of safety systems in
order to reduce or mitigate the event of danger.
As a general conclusion, the following areas should be considered: Royal Golf
Club, and Dawawir (village names: FABRIKA, LASHHAB, WARDA), because they
appeared as areas of interference with the safety distance of the pipes, which would
cause damage in case of a problem.
References
1. Zhao M, Liu X (2016) Regional risk assessment for urban major hazards based on GIS
geoprocessing to improve public safety. Saf Sci 87:18–24. https://doi.org/10.1016/j.ssci.2016.
03.016
2. Chen C, Li C, Reniers G, Yang F (2020) Safety and security of oil and gas pipeline trans-
portation: a systematic analysis of research trends and future needs using WoS. J Clean Prod:
123583. https://doi.org/10.1016/j.jclepro.2020.123583
3. Jadouane A, Chaouki A (2022) Simulation of the flood of El Maleh River by GIS in the City
of Mohammedia-Morocco. In: Kolathayar S, Mondal A, Chian SC (eds) Climate change and
water security. Lecture notes in Civil Engineering, vol 178. Springer, Singapore. https://doi.
org/10.1007/978-981-16-5501-2_8
4. Commune of Mohammedia (2017) Monographic Mohammedia city-Morocco
@seismicisolation
@seismicisolation
5. Wang Z, Xi M, Xiao J (2014) Risk assessment system of spatial city gas pipeline. Electron J
Geotech Eng 19:17915–17927
6. Alabtah FG, Mahdi E, Eliyan FF (2021) The use of fiber reinforced polymeric composites
in pipelines: a review. Compos Struct 276:114595. https://doi.org/10.1016/j.compstruct.2021.
114595
7. Hao MJ, You QJ, Yue Z (2017) Risk analysis of urban gas pipeline network based on improved
bow-tie model. IOP Conf Ser Earth Environ Sci 93:012059. https://doi.org/10.1088/1755-1315/
93/1/012059
8. Chen F, Zhang M, Song J, Zheng F (2018) Risk analysis on domino effect caused by pool fire
in petroliferous tank farm. Procedia Eng 211:46–54. https://doi.org/10.1016/j.proeng.2017.
12.136
9. Omira R, Baptista MA, Leone F, Matias L, Mellas S, Zourarah B, Miranda JM, Carrilho F,
Cherel J-P (2013) Performance of coastal sea-defense infrastructure at El Jadida (Morocco),
against tsunami threat: lessons learned from the Japanese 11 March 2011 tsunami. Nat Hazards
Earth Syst Sci 13:1779–1794. https://doi.org/10.5194/nhess-13-1779-2013
10. Natália K (2019) Security of pipelines in natural gas distribution network. Faculty of Security
Engineering, Univerzitná 8215/1, Žilina 010 26, Slovak Republic. https://doi.org/10.1016/j.
trpro.2019.07.197
@seismicisolation
@seismicisolation
AHP Approach for Risk Factors
Prioritisation in Tunnel Construction
Preetesh Band , Abhaysinha Shelake , and Nivedita Gogate
1 Introduction
1.1 Background
The construction business is developing faster than other industries throughout the
world. Delays in construction projects are amongst the most frequent concerns, which
have a plethora of bad implications for the project and its participants. The railway
and road transport sectors account for 70% of building projects, according to the
Ministry of Statistics and Programme Implementation’s Government of India 2020–
21 report. These projects were delayed by up to 324 weeks. While cost overruns
for railway and road projects ranged from 28 to 108%, the total cost overrun for all
projects is 65.44%, or 28 lakh crores of estimated cost. This is a severe matter that
requires quick action.
The Indian government has established the Smart India programme, which aims
to construct smart infrastructure throughout the country. It includes the develop-
ment of motorways, Metros, and new rail routes. According to the Union Economic
Budget Report for 2021–22, there are 702 km of traditional metro and 1016 km of
metro and RRTS. (Regional Rapid Transit System) under development in 27 cities.
Investment in rail infrastructure projects has also grown significantly. Tunnels for
many portions of such projects prove to be quite helpful due to the decrease in travel
time. Many Indian metropolitan cities are highly populated and lack adequate public
transit networks. Underground metro systems are an efficient mode of transportation
in such areas (Indian Infrastructure 2022).
P. Band (B) · A. Shelake · N. Gogate

Dr. Vishwanath Karad MIT World Peace University, Pune, Maharashtra 411038, India
e-mail: bandpreetesh@gmail.com
@seismicisolation
@seismicisolation
178 P. Band et al.
1.2 Motivation
India has 200+ metro tunnel projects covering 500+ km with 350 subterranean
stations. Since constructing tunnels is a complicated process, risk analysis and
management are essential, which often leads to delays and increased costs. However,
a procedural framework is lacking in this area [1].
1.3 Risk Analysis in Construction Projects
Traditionally, risk analysis for project duration uncertainty was done using proba-
bility theory. Risk management gained importance in the 1980s, resulting in new ideas
addressing risk in building projects. Modern methodologies such as AHP, decision
trees, stochastic programming, cognitive argument, and DSS are currently employed
in construction risk analysis, replacing classic methods such as probability theory
and PERT. Taroun [2] emphasises the importance of field experience and stake-
holder perceptions in measuring construction risk in his review study. Furthermore,
it emphasises the fact that accessible tools and DSS are not widely used.
Multiple studies have examined risk assessment using a three-step method [3, 4]
Different approaches to analysing construction delays include the as-planned, as-
built, and adjusted schedule methods [5], AHP and Monte Carlo simulations [3], and
machine learning models that make use of naïve Bayesian classification methods
and decision trees [6]. Prior methods were inefficient, required human scheduling
modifications, and lacked field validation. Most were restricted to certain building
stages or project types, with limited application. Managing numerous project phases
was difficult and time-consuming. There have been few studies in the literature
that address risk-based scheduling in construction projects, and a comprehensive
scheduling framework that includes risk is uncommon.
Delays Project monitoring must consider delays and their effect on the schedule
[7]. However, PERT and CPM analysis often overlooks risk variables and their
impact, leading to scheduling issues and cost overruns [8]. Including risk factors in
project planning minimises project delays and costs. Risk analysis reduces the likeli-
hood of project failure and identifies cost-overrun activities. Mapping risk variables
and their scheduling influence increases the likelihood of project success.
1.4 Need for Selection of Most Significant Risk Factors
Several academic researchers have analysed and assessed various risk variables in
building projects. These risk factors are instrumental in preventing delays and cost
overruns.
@seismicisolation
@seismicisolation
AHP Approach for Risk Factors Prioritisation in Tunnel Construction 179
To address delays and cost overruns, it is crucial to identify the associated risk vari-
ables [9]. The author conducted a study that prioritises the identification of causal risk
variables impacting profitability in building projects. Additionally, [10] extensively
examined 39 research findings on the hazards of modular integrated construction to
identify significant risk indicators classified by their occurrence frequency. Collins
et al. [11] integrated scaffolding safety hazards into the Building Information Model.
Chileshe et al. [12] assessed risk variables in construction projects in Ghana. Farid
et al. [13] investigated risk variables in Pakistan’s construction sector. Zou et al.
[14] ranked significant dangers in China’s construction industry, considering their
impact on project goals and the life cycle. Many studies from many nations have
identified and ranked significant risk variables, including Thailand [15], the United
Kingdom [16], the United Arab Emirates [17], Chile [18], and India [19]. Further-
more, literature on risk factor evaluation is reviewed in some studies pertaining
to public–private partnership projects [20]; construction project risk modelling and
assessment [2]; modular integrated construction application [10]; and many more.
These studies emphasise the crucial role of detecting and rating risk factors in risk
analysis and project planning.
1.5 Methods Used for Factor Prioritisation
In resource-constrained situations, identifying, assessing, and highlighting the major

risk variables to be controlled becomes even more vital since they pose the biggest
danger to project success [10]. To obtain usable information for risk evaluation
and analysis, risk variables must be quantitatively assessed and prioritised [10]. This
entails determining the relative relevance of risk variables associated with the chosen
project type. Estimation of risk factors can be done using a variety of methods ranging
from simple classical methods such as Monte Carlo Simulation [21, 22], Failure
mode, effects, and criticality analysis [23], cost–benefit analysis [24], decision trees.
The approach is chosen based on its usefulness, dependability, convenience of use,
data requirements, ease of result interpretation, and application to the problem [1].
Each approach has distinct properties that determine its applicability to the specified
choice issue. Probabilistic analysis is done on subjective and linguistic data using the
Monte Carlo method, which may not be suited for drawing objective and scientific
findings [25]. Furthermore, the Monte Carlo approach and certain newer methods,
such as Artificial Neural Networks [26], are data demanding. Multi-criteria decision
analysis methods simplify risk assessment by breaking down complex topics into
smaller sub-problems. They follow a hierarchical approach and are valuable for
handling multiple factors in decision-making.
According to a study of the literature, MCDM approaches have been widely used
for risk assessments in a range of building projects. Analytic Hierarchy Process
(AHP) method is used for contractor prequalification that considers qualitative and
supplementary standards that are then assessed through the use of quantifiable indica-
tors. The decision-making process is rationalised, and a solid multi-criteria technique
@seismicisolation
@seismicisolation
180 P. Band et al.
for contractor prequalification is used with this AHP-based model [27]. A suggested
equipment selection model that uses a modified AHP technique to analyse both qual-
itative and quantitative aspects, with essential characteristics such as offsetting rank
reversal, efficient evaluation of alternatives, and combining cost and benefit scores
[28].
An AHP MCDM model is being developed to rate CSFs for Lithuanian building
projects [29]. The findings highlight critical areas for successful construction project
implementation in Lithuania [29] for construction projects using AHP and deci-
sion tree [30] for infrastructure construction projects’ success factors using AHP-
based MCDM method [31]. For three different types of projects, AHP and the rela-
tive significance index were used to find the most relevant change order elements.
Increased re-planning, project management efforts, rework/demolition works, job
interruption loss of efficiency, and delayed payment are the main variables that can
assist professionals in handling changes more successfully [32] as well as for antici-
pating the main risks and quality standards of Taiwanese construction projects using a
hybrid AHP-ANN model [26]. According to a comprehensive literature evaluation,
the AHP approach is the most often used method of MCDM for risk assessment.
Because of the inherent simplicity and stability of the findings provided by these
procedures, they are the most often used techniques.
This study evaluates risk variables in tunnel building projects, focusing on the
Indian context. Limited research exists on tunnel project hazards in India, including
risk factor prioritisation. Given the growing number of projects, risk assessments for
Indian tunnelling projects are necessary.
2 Methodology
In previous work [33], the list of tunnel construction risk factors was extracted
initially. It was prioritised based on Relative importance index. The final prioritised
risk factors are given in Table 1.
The RII method, relying on subjective expert opinion, has limitations in deter-
mining relative importance. A modified approach like AHP is crucial for prioritising
factors using MCDM. The study employed literature review and stakeholder feed-
back to identify risk indicators for tunnel projects in India. The resulting critical risk
factor (CRF) list accounted for 60% of the total weightage. A three-level hierarchy
categorised 75 risk variables into 19 subgroups and 10 major clusters as illustrated
in Fig. 1 [33].
Given the complexities of tunnel construction, risk assessment becomes an essen-
tial component for project success. The step-by-step methodological process utilised
in this work for determining the most significant risk variables in tunnel building
projects is portrayed in Fig. 2 and, a brief description of the same is given below.
@seismicisolation
@seismicisolation
Table 1 Prioritised factors list [33]

S. No. List of prioritised factors
1 Soft rock present across alignment of tunnel
2 Unstable ground or unanticipated ground behaviour
3 Insufficient number of boreholes to predict general nature of rock mass
4 Excessive groundwater inflow
5 Extra dewatering work
6 Poor quality of construction
7 Fault in maintaining tunnel alignment
8 Lack of experience in executing tunnelling work
9 Poor judgement in design
10 Poor communication of instructions
11 Site condition not as predicted in contract
12 Changes in work scope stated in contract
13 TBM spare parts not available on time owing to manufacturer’s lack of supply
14 Damage to other vicinity structures
15 Unexpected major calamities like earthquake, landslide, volcanic eruption, etc.
Fig. 1 Hierarchical diagram for risk factors extractions [33]
Step 1: The extraction of 98 risk indicators via a rigorous literature research

concentrating on risk analysis in tunnel projects and insights from real-life case
studies via stakeholder inputs.
Step 2: Create clusters and categorise components inside them.
Step 3: Screening these risk variables for repeating and overlapping possibilities,
and narrowing the list to 75 risk factors.
Step 4: Ranking these criteria based on relative weightages acquired by expert
judgement utilising two distinct methodologies.
Step 4A: Ranking using the Relative Importance Index (RII) approach.
Expert comments are collected via a Likert scale questionnaire survey.
@seismicisolation
@seismicisolation
182 P. Band et al.
Fig. 2 Research methodology
@seismicisolation
@seismicisolation
The RII approach is used to give weightages to factors [33].

Step 4B: Analytical Hierarchical Process (AHP) Ranking.
In this strategy, a three-level hierarchy is created at first, with each level being
independent of the others. The top level is made up of a total of ten clusters. These are
further subdivided into 19 middle-level sub-clusters and 59 bottom-level risk factors.
The most significant risk indicators are identified by collecting expert comments on
Satty’s scale.
3 Result and Discussion
Literature research on tunnel construction projects in India yielded 98 risk indicators

initially. After eliminating duplicates and overlaps, there were 75 distinct risk factors
left. These components were arranged in a hierarchical framework, with 59 at the
bottom, separated into 19 subgroups, and consolidated into 10 primary clusters (refer
Fig. 3).
The final weightages for each risk factor are calculated by multiplying coeffi-
cients hierarchically. The highest-weighted risk factors from each cluster are selected,
resulting in a condensed list of 39 factors.
As the cumulative percentage weight is calculated it is found that amongst the
pool of 39 risk factors, 15 risk factors contribute to about 60% of the total risks
for a tunnelling project. Hence these 15 risk factors are critical in nature. Finally,
these critical risk factors are ranked on basis of their final weightage (refer Fig. 4).
The risk factor associated to each serial number is given in Table 2. The study
identified significant risk factors for tunnel construction. The highest-ranked risk is
unexpected calamities like earthquakes and landslides, with a weightage of 15.65%.
Other risks include insufficient boreholes (ranked 2, 4.82%), unsuitable TBM selec-
tion (ranked 3, 3.96%), lack of experience (ranked 4, 3.80%), commitment failure
(ranked 5, 3.67%), misinterpretation of conditions (ranked 6, 3.20%), TBM spare
parts unavailability (ranked 7, 3.09%), soil erosion (ranked 8, 3.05%), different soil
and groundwater behaviour (ranked 9, 2.92%), funds availability (ranked 10, 2.79%),
poor design judgement (ranked 11, 2.75%), inadequate safety policies (ranked 12,
2.71%), material price variations (ranked 13, 2.63%), water near construction site
(ranked 14, 2.58%), and resource unavailability (ranked 15, 2.56%). The final risk
factors are grouped into two groups based on their capacity to mitigate risks, as
illustrated, controllable risk factors and uncontrolled risk factors. The distribution
of final weightages is given in Fig. 4 corresponding to risk factor numbers given in
Table 2.
Hence, the author suggests that looking after controllable risk factors potentially
provides remedial measures for delay in tunnel construction project.
@seismicisolation
@seismicisolation
184 P. Band et al.
Fig. 3 Sample AHP cluster
2.56
2.63 2.58
15.65
2.71
2.75
2.79
2.92 4.82
3.05
3.96
3.09
3.20 3.67 3.80
10 1 2 3 4 5 6 11 12 13 7 8 14 15 9
Fig. 4 Final weightages of risk factors
@seismicisolation
@seismicisolation
Table 2 Controllable and uncontrollable risk factors

S. No. Controllable risk factors S. No. Uncontrollable risk factors
1 Insufficient number of boreholes to 10 Unexpected major calamities like
predict general nature of rock mass earthquake, landslide, volcanic
eruption, etc.
2 Selection of unsuitable TBM for strata 11 Soil erosion within the tunnel as a
given result of increased groundwater flow
3 Lack of experience in executing 12 Difference in behaviour of soil and
tunnelling work groundwater from what is often
anticipated or in comparison to the
conditions specified in the contract
4 Failure in commitment by parties in 13 Problem with funds availability
contract that are owner and contractor
5 Misinterpretation of soil and 14 Variation in construction material
groundwater conditions leading to the prices
selection of improper construction
techniques or equipment
6 TBM spare parts not available on time 15 Presence of water source near
owing to manufacturer’s lack of supply tunnel construction site
7 Poor Judgement in design
8 Inadequate safety policies
9 Unavailability of required resources
like man, equipment, and materials
4 Conclusion
Large infrastructure projects often face uncertainties and obstacles that increase the
risk of delays and cost overruns. This study examines the impact of risk variables
on project scheduling, proposing a practical approach to analysing their impact
on building operations. By understanding the occurrence of risks and their effect
on deadlines and budgets, construction businesses can take proactive measures to
enhance project performance.
The study used AHP and MCDM to rank risk variables. Initially, 98 risk vari-
ables were identified from literature analysis, which were narrowed down to 75
after eliminating overlaps. These variables were then classified into clusters. Using
Saty’s scale led AHP framework, a final list of critical risk factors (CRFs) was
established. This study improves risk identification and management for project
leaders and stakeholders. The research generated a list of 15 significant risk vari-
ables categorised into two groups based on their nature and manageability. Some
risk factors can be controlled during planning, such as TBM spare parts availability,
soil and groundwater conditions, experience in tunnelling work, borehole quantity,
and resource availability. Others require special handling beyond standard proce-
dures, including the presence of water sources, soft rock along the tunnel alignment,
@seismicisolation
@seismicisolation
186 P. Band et al.
and soil erosion inside the tunnel. Understanding these risk variables during planning
improves preparedness for minimising or managing hazards during construction.
As we can see out of the 15 most imp factors through RII only 6 factors are
identified again in AHP while the remaining are changed. Hence this shows that the
AHP gives additional and better results as compared to RII and hence the AHP work
proves.
References
1. Dziadosz A, Rejment M (2015) Risk analysis in construction project—chosen methods.

Procedia Eng 122:258–265. https://doi.org/10.1016/j.proeng.2015.10.034
2. Taroun A (2014) Towards a better modelling and assessment of construction risk: insights from
a literature review. Int J Project Manage 32(1):101–115. https://doi.org/10.1016/j.ijproman.
2013.03.004
3. Albogamy A, Dawood N, Scott D (2014) A risk management approach to address construction
delays from the client aspect. In: Computing in civil and building engineering. American
Society of Civil Engineers, Reston, VA, June 2014, pp 1497–1505. https://doi.org/10.1061/
9780784413616.186
4. KarimiAzari A, Mousavi N, Mousavi SF, Hosseini S (2011) Risk assessment model selection
in construction industry. Expert Syst Appl 38(8):9105–9111. https://doi.org/10.1016/j.eswa.
2010.12.110
5. Kraiem ZM, Diekmann JE (1987) Concurrent delays in construction projects. J Constr Eng
Manag 113(4):591–602. https://doi.org/10.1061/(ASCE)0733-9364(1987)113:4(591)
6. Gondia A, Siam A, El-Dakhakhni W, Nassar AH (2020) Machine learning algorithms for
construction projects delay risk prediction. J Constr Eng Manag 146(1). https://doi.org/10.
1061/(ASCE)CO.1943-7862.0001736
7. Ahuja V, Thiruvengadam V (2004) Project scheduling and monitoring: current research status.
Constr Innov 4(1):19–31. https://doi.org/10.1108/14714170410814980
8. Pehlivan S, Öztemir AE (2018) Integrated risk of progress-based costs and schedule delays in
construction projects. Eng Manag J 30(2):108–116. https://doi.org/10.1080/10429247.2018.
1439636
9. Jahan S, Khan KIA, Thaheem MJ, Ullah F, Alqurashi M, Alsulami BT (2022) Modeling
profitability-influencing risk factors for construction projects: a system dynamics approach.
Buildings 12(6):701. https://doi.org/10.3390/buildings12060701
10. Wuni IY, Shen GQP, Mahmud AT (2022) Critical risk factors in the application of modular
integrated construction: a systematic review. Int J Constr Manag 22(2):133–147. https://doi.
org/10.1080/15623599.2019.1613212
11. Collins R, Zhang S, Kim K, Teizer J (2014) Integration of safety risk factors in BIM for
scaffolding construction. In: Computing in civil and building engineering. American Society
of Civil Engineers, Reston, VA, June 2014, pp 307–314. https://doi.org/10.1061/978078441
3616.039
12. Chileshe N, Boadua Yirenkyi-Fianko A (2012) An evaluation of risk factors impacting construc-
tion projects in Ghana. J Eng Des Technol 10(3):306–329. https://doi.org/10.1108/172605312
11274693
13. Farid W, Kureshi NI, Babar S, Mahmood S (2020) Critical risk factors of construction industry
of Pakistan for improving project outcome. Mehran Univ Res J Eng Technol 39(1):71–80.
https://doi.org/10.22581/muet1982.2001.08
14. Zou PXW, Zhang G, Wang J (2007) Understanding the key risks in construction projects in
China. Int J Project Manage 25(6):601–614. https://doi.org/10.1016/j.ijproman.2007.03.001
@seismicisolation
@seismicisolation
15. Ghosh S, Jintanapakanont J (2004) Identifying and assessing the critical risk factors in an under-
ground rail project in Thailand: a factor analysis approach. Int J Project Manage 22(8):633–643.
https://doi.org/10.1016/j.ijproman.2004.05.004
16. Bing L, Akintoye A, Edwards PJ, Hardcastle C (2005) The allocation of risk in PPP/PFI
construction projects in the UK. Int J Project Manage 23(1):25–35. https://doi.org/10.1016/j.
ijproman.2004.04.006
17. El-Sayegh SM (2008) Risk assessment and allocation in the UAE construction industry. Int J
Project Manage 26(4):431–438. https://doi.org/10.1016/j.ijproman.2007.07.004
18. Serpella AF, Ferrada X, Howard R, Rubio L (2014) Risk management in construction projects:
a knowledge-based approach. Procedia Soc Behav Sci 119:653–662. https://doi.org/10.1016/
j.sbspro.2014.03.073
19. Subramanyan H, Sawant PH, Bhatt V (2012) Construction project risk assessment: development
of model based on investigation of opinion of construction project experts from India. J Constr
Eng Manag 138(3):409–421. https://doi.org/10.1061/(ASCE)CO.1943-7862.0000435
20. Yu Y, Chan APC, Chen C, Darko A (2018) Critical risk factors of transnational public–private
partnership projects: literature review. J Infrastr Syst 24(1). https://doi.org/10.1061/(ASCE)IS.
1943-555X.0000405
21. Arnold U, Yildiz Ö (2015) Economic risk analysis of decentralized renewable energy infras-
tructures—a Monte Carlo simulation approach. Renew Energy 77:227–239. https://doi.org/10.
1016/j.renene.2014.11.059
22. Qazi A, Simsekler MCE (2021) Risk assessment of construction projects using Monte Carlo
simulation. Int J Manag Proj Bus 14(5):1202–1218. https://doi.org/10.1108/IJMPB-03-2020-
0097
23. Suwanasri C, Saribut S, Suwanasri T, Phadungthin R (2021) Risk analysis using failure modes,
effects, and criticality analysis for transmission network assets. Energies 14(4):977. https://doi.
org/10.3390/en14040977
24. Merisalu J, Sundell J, Rosén L (2021) A framework for risk-based cost-benefit analysis for
decision support on hydrogeological risks in underground construction. Geosciences 11(2):82.
https://doi.org/10.3390/geosciences11020082
25. Sadeghi N, Fayek AR, Pedrycz W (2010) Fuzzy Monte Carlo simulation and risk assessment in
construction. Comput-Aided Civ Infrastr Eng 25(4):238–252. https://doi.org/10.1111/j.1467-
8667.2009.00632.x
26. Lin C-L, Fan C-L, Chen B-K (2022) Hybrid analytic hierarchy process-artificial neural network
model for predicting the major risks and quality of Taiwanese construction projects. Appl Sci
12(15):7790. https://doi.org/10.3390/app12157790
27. Anagnostopoulos KP, Vavatsikos AP (2006) An AHP model for construction contractor
prequalification. Oper Res Int J 6(3):333–346. https://doi.org/10.1007/BF02941261
28. Shapira A, Goldenberg M (2005) AHP-based equipment selection model for construction
projects. J Constr Eng Manag 131(12):1263–1273. https://doi.org/10.1061/(ASCE)0733-936
4(2005)131:12(1263)
29. Gudienė N, Banaitis A, Podvezko V, Banaitienė N (2014) Identification and evaluation of the
critical success factors for construction projects in Lithuania: AHP approach. J Civ Eng Manag
20(3):350–359. https://doi.org/10.3846/13923730.2014.914082
30. Maceika A, Bugajev A, Šostak OR, Vilutienė T (2021) Decision tree and AHP methods applica-
tion for projects assessment: a case study. Sustainability 13(10):5502. https://doi.org/10.3390/
su13105502
31. Belay S, Goedert J, Woldesenbet A, Rokooei S (2022) AHP based multi criteria decision
analysis of success factors to enhance decision making in infrastructure construction projects.
Cogent Eng 9(1). https://doi.org/10.1080/23311916.2022.2043996
32. Gunduz M, Mohammad KO (2019) Assessment of change order impact factors on construction
project performance using analytic hierarchy process (AHP). Technol Econ Dev Econ 26(1):71–
85. https://doi.org/10.3846/tede.2019.11262
33. Shelake AG, Gogate NG, Rajhans NR (2022) An integrated approach for identification and
prioritization of risk factors in tunnel construction. Mater Today Proc 65:1805–1812. https://
doi.org/10.1016/j.matpr.2022.04.855
@seismicisolation
@seismicisolation
Identification and Prioritization of Risk
Factors Impacting Cost Overrun
in Indian Road Construction Projects
Rohan Vishal Patil and Mahesh Balwant Sonawane
1 Introduction
In the construction industry, transportation is a crucial component, which contributes

significantly to the GDP of any country [1]. Several types of road networks exist,
including expressways, highways, district roads, bridges, and tunnels. It is vital to
development in various fields, including economics, geography, development of the
community, and urban growth [2]. Road construction projects have similar risk factors
for cost overruns as any other construction projects do. But, their effects on road
construction differ. Since road network construction projects require high investment,
situations with a lengthy duration and longitudinal site are more susceptible to cost
overruns than traditional construction projects.
The term “cost overrun” refers to the difference between actual and estimated
costs. The price is expressed in local money, at constant prices, and compared to a
reliable baseline [3]. The cost overrun in construction refers to the variation between
budget and actual expenses incurred throughout the project. There are different causes
of cost overruns, such as taxes, restrictions, and escalation, as well as internal factors
like project scale, time span, difficulty, the location, and estimation type [4].
India has 3.34 million kilometers of roadways, after China’s, the second-highest
in the entire world. According to the 13th five-year plan (2017–2023), this network
will be extended by another 18,637 km, of which more than 1000 km will be express-
ways. The Government of India aimed to construct 15,000 km of roads in the union
budgetary 2019–20. But just 8200 km were constructed [5]. Clearly, the govern-
ment is dedicated to implementing significant changes in the field of transportation
infrastructure. Despite this, the industry has not presented a positive image at the
implementation stage. Government authorities, contractors, and subcontractors must
recognize risks in order to accomplish the government’s goal and ensure that each
R. V. Patil (B) · M. B. Sonawane

Vishwanath Karad MIT World Peace University, Kothrud, Pune, Maharashtra 411038, India
e-mail: rohanvpatil533@gmail.com
@seismicisolation
@seismicisolation
190 R. V. Patil and M. B. Sonawane
project is risk-safe and that the risks are managed during construction. The Indian
researchers did not pay attention to risk and uncertainty of roadways construction
projects. Therefore, this research paper’s objectives are:
1. To determine the vital factors that contribute to the overruns in project costs to
construct roads in India.
2. To evaluate and rank the important risk factors which are affecting cost-
effectiveness of Indian road development projects.
3. To restrict these risks or avoid them in future projects before the implementation
of the work starts.
2 Literature Review
The completion costs of transportation network projects are high for approximately
90% of the projects, and this resulted in an increase in costs compared to their
original budgets [6]. There have been a number of well-known transportation projects
worldwide that have overrun their budgets [7]. For example, over 51% of government
projects are experiencing delays as of December 1, 2022, as per the governmental
data of India. According to the investigation, there has been a total cost increase of
Rs 4.5 lakh crore, roughly 22% more than the project’s initial estimate [8]. Activities
associated with international development (ID) in Asian countries suffered from cost
overruns. There were many factors that contributed to this cost escalation, including
the domestic currency’s decline, the insignificant to anticipated tender price, less
than expected utilization of the reserve funds, the scope of the project shortened,
local taxes, changes in interest rates, and international competitive tenders [9, 10].
Studies have shown that Cost overruns frequently occur in infrastructure projects,
for example, roadways construction in Norway [11], projects for constructing roads
in the United States, and infrastructural developments in Australia [12]. In [13]
examined cost overruns from different perspectives. They included more than 250
transportation-related projects from various nations in their study on cost overruns.
Numerous projects’ cost performance has already been analyzed, frequently in devel-
oped nations. However, relatively few research has been carried out in developing
nations like India. Since India has different policies about construction and topog-
raphy, the findings of cost overruns in the United States, Australia, and Europe cannot
be applied.
The primary factors affecting Australian highway projects’ cost overruns were
identified by [14]. Various factors, including modifications to the design, pricing
changes, and unanticipated events, can result in cost overruns. As per [15], the
main causes of cost overruns in Canadian road projects were design modifica-
tions, subsurface conditions, permissions, and regulations. In relation to the Chinese
highway building sector, according to, [16] cost overruns can result from a number of
issues, including low workforce productivity, increased material prices, and expen-
sive equipment expenses. In Afghanistan [17], according to research, construction
overruns in expenses are primarily caused by corruption, delayed payments by
@seismicisolation
@seismicisolation
Identification and Prioritization of Risk Factors Impacting Cost Overrun … 191
owners, financial difficulties by contractors, security, change orders, and inflationary

market conditions in road construction projects. In [18] found that the main reasons
for price increases in Jordanian infrastructure projects are weather and geography.
The majority of prior research on the construction of roads has been on the risk
components influencing cost overruns in developed nations like the United Arab
Emirates, the United States, Australia, China, etc. The most important elements
contributing to price overruns in the development of India’s roadways are currently
the subject of a very small number of studies. In India, most road construction projects
lag behind schedule due to various risk factors, so it directly impacts the project’s
cost overrun. In order to prevent cost overruns, the client, consultant, and contractors
should determine the main contributing elements.
3 Methodology
The main purpose of this research is to determine the elements that lead to cost over-
runs in projects that construct roadways in India. The procedure starts by listing all of
the risk factors related to road development. Afterward, the collection of qualitative
data was done through a questionnaire survey about the likelihood, and impact of
those variables that could cause cost overruns in projects for constructing roadways
in India. Hence, the results of these studies will help decision-makers to deal with
the problems about the budget in future projects. The steps of the methodology are
shown in Fig. 1.
Identification of Risk Factores Impacting Cost Overrun
Preprstion of Questionnaire Survey
Analysis of Data Collected from Questionnaire Survey
Preparation of Risk Matrix to Find out Critical Risk Factores
Fig. 1 Research methodology steps
@seismicisolation
@seismicisolation
3.1 Risk Identification
Before starting the identification of risk factors, the scope of work, schedule, budget,
and other relevant factors are reviewed thoroughly about the road construction
project. The next step is to conduct a risk assessment of the project by the project team,
which may involve brainstorming sessions, risk checklists, literature review analysis,
opinions of industry professionals, and other resources for identifying risk issues in
Indian roadway development projects. Risk factors can consist of a variety of factors,
such as environmental factors, political factors, permissions, design problems, man,
material, money problems, and other unforeseen events. The risk considerations
related to road development projects in India are shown in Table 1.
3.2 Preparation of Questionnaire Survey
For any construction project, the preparation of a questionnaire survey is a very

important step. Using the collected data about risk identification, the questionnaire
is designed. There were 52 cost overrun factors identified in total and they were
grouped into 2 categories which are External Factors and Internal Factors. These
internal variables are further subdivided into variables referring to both construction
activities and the overall road construction project.
The data was collected from respondents using a Five-Point Likert Scale (1. Very
Low; 2. Low; 3. Moderate; 4. High; 5. Very High).
Respondents were questioned about the likelihood and impact of each factor on
cost overruns based on their prior experience in projects comparable to this one.
Most critical risk factors are represented with the help of the Probability and Impact
matrix (Risk Matrix).
The ranking of the risk factors from the collected data can be done with different
analytical techniques. In this, relative significance factors for analysis were calculated
using descriptive statistics known as the relative importance index (RII). Data from
experts was collected through a variety of methods like face-to-face meetings, google
Forms, email, LinkedIn, etc.
3.3 Data Analysis
Data collected through questionnaire surveys conducted by experts at road construc-

tion projects is analyzed using a qualitative approach, which uses the relative impor-
tance index (RII) to determine risk levels. In this analysis, the risk level is computed
using the following equation:
Risk Level = Probability(P) × Impact(I ) (1)
@seismicisolation
@seismicisolation
Table 1 Risk factors affecting cost overrun

S. No. Risk factors impacting cost overrun Researcher
External
Inflation of price [19]
Changes in governmental regulations [20]
Changes in taxes [21]
Legal processes [17]
Interference by political parties [22]
Strikes by labors, workers [18]
Weather conditions [23]
Availability of funds [19]
Governmental ministries conflicts [20]
Disputes about land clearance [24]
Objections from local bodies/councils [25]
Force majeure [17]
Internal
a. Risk factors related to the projects
Scope of work [18]
Inaccurate cost estimation [24]
Location of the project [25]
Duration of the project [19]
Length of the project [21]
Tender deadlines [23]
Changes in design [18]
Changes in quantity [23]
Unexpected soil conditions [26]
Schedule pressures [25]
Lack of funds [27]
Variations in site conditions [17]
Order variation [20]
Delay in approvals [19]
Changes in specifications [24]
Design errors [17]
Problems accessing the site [21]
Improper supervisions [25]
Violations of supervisory authority [26]
Lack of communication between clint and contractor [19]
Improperly verified contract document [23]
Delay in payments [22]
(continued)
@seismicisolation
@seismicisolation
Table 1 (continued)
S. No. Risk factors impacting cost overrun Researcher
Delay in dispute resolutions [27]
Managerial incompetence [18]
b. Risk factors in construction-related activities
Inappropriate construction planning [21]
Complexity of the project [18]
Quality of work [17]
Lack of experience [24]
Poor material quality [21]
Equipment availability [19]
Shortage of skilled labors [23]
Inadequate quantity surveying [26]
Delay in procurement [26]
Low productivity [18]
Resource availability [21]
Dispute with labors [22]
Delay by the vendors [18]
Wrong material selection [27]
Pollution due to construction [17]
Delay due to third-party [24]
An analysis of multivariate data utilizing RII aims to extract key factors from the
complexity. As a result, RII is recommended for sorting the factors. As per equation,
all risk factors under each group’s section have had their RII computed and rated
accordingly.
RII = ∑W ÷ (A × N ) (2)
where
W Sum of weight given by respondents to the risk which is from 1 to 5.

A Maximum weight (in this case, A = 5).
N No. of respondents.
3.4 Risk Matrix Preparation
Once the ranking of all the risks is done after risk levels have been calculated, the
table shows the risk levels for each of the risks. After that, the Probability and Impact
@seismicisolation
@seismicisolation
matrix (also known as the Risk matrix) can be created. This matrix shows how high,
medium, and low risks are classified.
Red zone: For critical (C) risks of the highest priority that call for extra attention
to reduce or eliminate the adverse impacts.
Yellow zone: For risks in the moderate (M) category that need to be mitigated but
are given less importance than those in the red zone.
Green zone: Low (L) risks that can be neglected.
To categories each of the respondent’s data that has been gathered into two groups,
the following categories were created: Internal and External Factors. By calculating
Eq. 1, the severity of each element was established. To identify each factor’s impact
level as determined by the participants and then determining the risk degree for each
factor. Using Eq. 2, the RII of each factor was calculated in order to determine the
relative importance of each risk factor.
4.1 External Factors
In Table 2, an external group consisting of 12 risk factors. According to the results,

two risk factors have been identified as “critical” (C), RII of 3.778 and 3.760 is
recorded in each case. Among the risk factors, none are categorized as “moderate—
M,” and ten are categorized as “low—L.” The external group’s most critical scores
are caused by political interference and inflation and with RII of 3.778 and 3.760,
respectively, they are ranked fifth and sixth.
The project’s success may be significantly impacted by political influence in
the selection of project priorities and the timetable for implementation. All project
partners agreed, almost unanimously, according to questionnaire survey data, that
political influence significantly changed the project’s financial and schedule profiles.
In Brazil, there is a direct link between political interference and the overspending
of road construction projects [24].
In India, the economic environment serves as an external element that affects infla-
tion. Especially due to escalation of prices in construction industry due to many condi-
tions like foreign currency involvement, war conditions, etc. This directly affects cost
overruns in infrastructure projects, especially those involving road projects. Several
respondents also indicated that keeping the price of the project the same regardless
of changes in scope, location, and schedule may involve a critical risk. The devel-
opment of a method or system to identify anticipated inflation prior to finalizing the
estimation would simplify budget formulation during the appraisal process. Many
developing nations experience unanticipated price changes as a result of the unstable
global economy and volatility in the major currency exchange rates [28].
@seismicisolation
@seismicisolation
Table 2 Risk ranking

Risk factors Score (∑ W ) Degree (∑ W )/N RII Symbol Rank
impacting cost
overrun
External
Interference by 2608 18.89 3.778 C 5
political parties
Inflation of price 2595 18.80 3.760 C 6
Objections from 1455 10.54 2.108 L 27
local bodies/councils
Weather conditions 1005 7.28 1.456 L 33
Availability of funds 993 7.19 1.438 L 34
Governmental 885 6.41 1.282 L 37
ministries conflicts
Changes in 530 3.84 0.768 L 44
governmental
regulations
Strikes by labors, 400 2.89 0.578 L 46
workers
Changes in taxes 323 2.34 0.468 L 47
Legal processes 300 2.17 0.434 L 48
Disputes about land 295 2.13 0.426 L 49
clearance
Force majeure 208 1.50 0.300 L 52
Internal
a. Risk factors related to the projects
Inaccurate cost 3100 22.46 4.492 C 1
estimation
Changes in design 2973 21.54 4.308 C 2
Changes in quantity 2815 20.39 4.078 C 3
Order variation 2705 19.60 3.920 C 4
Changes in 2543 18.42 3.684 C 7
specifications
Scope of work 2485 18.00 3.600 C 8
Location of the 2395 17.35 3.470 M 9
project
Delay in payments 2373 17.19 3.438 M 10
Delay in dispute 2300 16.66 3.332 M 11
resolutions
Duration of the 2250 16.30 3.260 M 12
project
Delay in approvals 2223 16.10 3.220 M 13
(continued)
@seismicisolation
@seismicisolation
Table 2 (continued)
impacting cost
overrun
Design errors 1750 12.68 2.536 M 21
Managerial 1703 12.34 2.468 M 22
incompetence
Schedule pressures 1523 11.03 2.206 L 25
Length of the project 1505 10.90 2.180 L 26
Lack of funds 1310 9.49 1.898 L 28
Tender deadlines 1250 9.05 1.810 L 30
Variations in site 1133 8.21 1.642 L 31
conditions
Improper 950 6.88 1.376 L 35
supervisions
Unexpected soil 925 6.70 1.340 L 36
conditions
Lack of 870 6.30 1.260 L 38
communication
between clint and
contractor
Improperly verified 753 5.45 1.090 L 39
contract document
Violations of 643 4.65 0.930 L 40
supervisory authority
Problems accessing 580 4.20 0.840 L 42
the site
b. Risk factors in construction-related activities
Inappropriate 2178 15.78 3.156 M 14
construction
planning
Equipment 2118 15.34 3.068 M 15
availability
Poor material quality 2033 14.73 2.946 M 16
Complexity of the 2015 14.60 2.920 M 17
project
Delay in 1998 14.47 2.894 M 18
procurement
Inadequate quantity 1975 14.31 2.862 M 19
surveying
Resource availability 1875 13.58 2.716 M 20
Quality of work 1628 11.79 2.358 M 23
Shortage of skilled 1580 11.44 2.288 M 24
labors
(continued)
@seismicisolation
@seismicisolation
Table 2 (continued)
impacting cost
overrun
Lack of experience 1303 9.44 1.888 L 29
Low productivity 1025 7.42 1.484 L 32
Delay due to 628 4.55 0.910 L 41
third-party
Dispute with labors 550 3.98 0.796 L 43
Wrong material 508 3.68 0.736 L 45
selection
Delay by the vendors 250 1.81 0.362 L 50
Pollution due to 243 1.76 0.352 L 51
construction
“Critical” category indicating crititcal risk factors are indicated in bold C
4.2 Internal Factors
There are two risk factor sections within the internal group:
• Section 1 consists 24 of risk factors related to the project;
• Section 2 consists of 16 risk factors in construction-related activities.
Section 1. Risk Factors Related to the Project.
From Table [5], In Section A, 6 risk factors are listed as “critical,” 7 as “moderate,”
and 11 as “low.”
With an RII of 4.492, inaccurate cost estimation was the internal group’s primary
problem. Several respondents said that it was often the owner’s instructions that
determined project budget should fit within fixed bill of quantities, in spite of the
project’s location, size, and duration. Therefore, if the quantities surpass what was
anticipated, the budget may rise during the implementation phase. As a result, expe-
rienced estimators take time during the appraisal phase to fully grasp the scope and
activities of the project, with the help of final drawings and specifications of the
project. Inaccurate cost estimates may expose a project to danger, cause financial
loss, harm its reputation, and undermine its trustworthiness [29].
Furthermore, according to the respondents, modifications to design are the
project’s second-most important cause driving cost overrun in India’s road devel-
opment. Creedy et al. [30] determined that changes in project design and scope are
the main reasons for transportation project overruns in Australia. Through improved
communication and coordination between stakeholders, design modification can be
reduced.
One more critical internal factor is changes in quantities which is having RII of
4.078. Ground and terrain conditions are unpredictable, that may result in changes in
quantity. In order to avoid these problems, consultants must conduct reconnaissance
@seismicisolation
@seismicisolation
and preliminary surveys very carefully. Otherwise, project cost and duration may
increase extensively.
One of the most important risk factors in projects that construct roads is variation
in orders. It is possible to reduce variation orders by engaging design consultants with
experience in similar projects. Variation orders can be avoided by employing method
statements and a resource-rich schedule that clearly specifies roles and responsibil-
ities, equipment, and progress curves. In [18] claimed that variation orders had a
significant impact on the cost overruns of a Jordanian road construction project.
In terms of RII, changes in specifications are ranked seventh with 3.684. Consul-
tant instructions are always responsible for specification changes. To avoid poor
quality of work, consultants or owners make changes in specifications of project. But
as a result of these specification revisions, significant project flaws, cost overruns,
delays, or even project failure can frequently be identified.
Changes in the scope of work are also a critical risk factor with RII value of 3.600.
The owners are solely responsible for changes in scope, according to all respondents.
In this situation, the owner and designer are unable to provide clarity on the scope of
the project. When precise specifications, thorough modeling methods, and precise
figures are supplied prior to the scope of work being finalized, the scope may be
reduced.
Cost overruns for Australian highway projects are frequently caused by changes
to the project’s scope [30].
Section 2. Risk Factors in Construction-Related Activities.
As shown in Fig. 2, 9 risk factors are considered to be moderate, 7 are considered to be
“low,” but there is no critical risk factor. The majority of road network development
projects are affected by elements that are moderate or low because creating roads
carries less risk than developing factories, airports, buildings, and other types of
infrastructure.
Probability
Impact
Fig. 2 Risk matrix
@seismicisolation
@seismicisolation
5 Conclusion
The main purpose of this study is to determine the elements that influence cost
overruns in road network development projects in India. Reduce the effects of cost
overruns on upcoming construction activities in order to benefit the owner agencies
and other parties involved. In order to obtain the information necessary, the study
used both quantitative and qualitative research methods. 52 elements that lead to
roadway construction cost overruns in India were identified through expert inter-
views and literature reviews. Then, 138 engineers with expertise in managing and
constructing roads filled out a structured questionnaire survey. The most frequent
causes of cost overruns in Indian roadway construction projects were examined and
identified in the study. The RII was used to rank every factor that was identified.
The survey’s findings led to the identification of 8 critical factors that cause cost
overruns. Political party interference and price inflation were the first two factors
that were tied to external factors. They come in at fifth and sixth, respectively, in the
rankings. The remaining six components are internal factors. These are inaccurate
cost estimation, modifications in design, changes in quantity, variations in orders,
changes in specifications, and changes in scope of work. They came in first, second,
third, fourth, seventh, and eighth, respectively, according to their rankings.
As a result of statistical analysis of the data, future project owners are suggested
to follow the following recommendations:
• Select the right consultants and designers to provide the best specifications
possible. Within the project’s budget, the design should accommodate for the
socioeconomic demands of the area.
• There should be sufficient time given for tender preparation and making sure that
the quantities are accurately calculated in order to prevent rising in quantity during
developing.
• Prior to finalizing the budget estimate, it is necessary to design a method for
forming the budget during the appraisal process. The inflation rate should also be
taken into account.
• In order to limit project costs during construction, it is crucial to provide consul-
tants and designers enough time to make any necessary modifications to the design
or specifications during the appraisal process. This should be in line with the
project budget. Cost overruns may be caused by changes in quantity, variation in
orders, specifications, or change in scope of work, so any technical causes should
be avoided.
• It is better to employ a value estimator for cost estimation in order to generate an
appropriate budget.
• Ensure that your timetable makes sense and is pertinent.
• Ensure there is a contingency fund available throughout the implementation phase
so that any cost overruns may be met.
@seismicisolation
@seismicisolation
References
1. Sui Pheng L, Shing Hou L (2019) Construction quality and the economy. https://doi.org/10.
1007/978-981-13-5847-0
2. Shi G, Shan J, Ding L, Ye P, Li Y, Jiang N (2019) Urban road network expansion and its driving
variables: a case study of Nanjing city. Int J Environ Res Public Health 16(13). https://doi.org/
10.3390/IJERPH16132318
3. Melaku Belay S, Tilahun S, Yehualaw M, Matos J, Sousa H, Workneh ET (2021) Analysis
of cost overrun and schedule delays of infrastructure projects in low income economies: case
studies in Ethiopia. Adv Civ Eng 2021. https://doi.org/10.1155/2021/4991204
4. Kamaruddeen AM, Sung CF, Wahi W (2020) A study on factors causing cost overrun of
construction projects in Sarawak, Malaysia. Civ Eng Archit 8(3):191–199. https://doi.org/10.
13189/CEA.2020.080301
5. National Highways Authority of India, Ministry of Road Transport & Highways, Government
of India. https://nhai.gov.in/#/. Accessed 14 May 2023
6. Love PED, Sing C-P, Carey B, Kim JT (2015) Estimating construction contingency: accom-
modating the potential for cost overruns in road construction projects. J Infrastr Syst 21(2).
https://doi.org/10.1061/(ASCE)IS.1943-555X.0000221
7. Paraskevopoulou C, Boutsis G (2020) Cost overruns in tunnelling projects: investigating the
impact of geological and geotechnical uncertainty using case studies. Infrastructures 5(9).
https://doi.org/10.3390/INFRASTRUCTURES5090073
8. Ministry of Statistics and Program Implementation, Government of India. https://www.mospi.
gov.in/. Accessed 14 May 2023
9. Guidelines for the Evaluation of Public Sector Operations, Apr 2016. 10.22/JS/
JQUERY.DATATABLES.MIN.JS
10. Ahsan K, Gunawan I (2010) Analysis of cost and schedule performance of international devel-
opment projects. Int J Project Manage 28(1):68–78. https://doi.org/10.1016/J.IJPROMAN.
2009.03.005
11. Odeck J (2019) Variation in cost overruns of transportation projects: an econometric meta-
regression analysis of studies reported in the literature. Transportation 46(4):1345–1368.
https://doi.org/10.1007/S11116-017-9836-5
12. Terrill M, Coates B, Danks L (2016) Cost overruns in Australian transport infrastructure
projects. Australasian transport research forum. Accessed 14 May 2023 [online]. Available
http://www.atrf.info
13. Cantarelli CC, Flyvbjerg B, Van Wee B, Molin EJE (2008) Cost overruns in large-scale trans-
portation infrastructure projects: which explanations can be given? In: 2008 1st international
conference on infrastructure systems and services: building networks for a brighter future,
INFRA 2008. https://doi.org/10.1109/INFRA.2008.5439650
14. Love PED, Ahiaga-Dagbui DD, Irani Z (2016) Cost overruns in transportation infrastructure
projects: sowing the seeds for a probabilistic theory of causation. Transp Res Part A Policy
Pract 92:184–194. https://doi.org/10.1016/J.TRA.2016.08.007
15. Ammar T, Abdel-Monem M, El-Dash K (2022) Risk factors causing cost overruns in road
networks. Ain Shams Eng J 13(5). https://doi.org/10.1016/J.ASEJ.2022.101720
16. Mansur SA, Zin RM, Linbo L (2019) Components of cost overrun in China construc-
tion projects. IOP Conf Ser Mater Sci Eng 620(1). https://doi.org/10.1088/1757-899X/620/
1/012081
17. Niazi GA, Painting N (2017) Significant factors causing cost overruns in the construction
industry in Afghanistan. Procedia Eng 182:510–517. https://doi.org/10.1016/J.PROENG.2017.
03.145
18. Al-Hazim N, Abusalem Z (2015) Delay and cost overrun in road construction projects in Jordan.
Int J Eng Technol 4(2):288. https://doi.org/10.14419/IJET.V4I2.4409
19. Kazaz A, Ulubeyli S, Tuncbilekli NA (2012) Causes of delays in construction projects in
Turkey. J Civ Eng Manag 18(3):426–435. https://doi.org/10.3846/13923730.2012.698913
@seismicisolation
@seismicisolation
20. Chang T, Deng X, Zuo J, Yuan J (2018) Political risks in central Asian countries: factors and
strategies. J Manage Eng 34(2). https://doi.org/10.1061/(ASCE)ME.1943-5479.0000588
21. Bv B, Ma N, Pp AK (2020) A methodology for identifying critical factors influencing land
value in urban areas: a case study of Kerala, India. Prop Manag 38(5):665–681. https://doi.org/
10.1108/PM-01-2020-0004
22. Eskander RFA (2018) Risk assessment influencing factors for Arabian construction projects
using analytic hierarchy process. Alex Eng J 57(4):4207–4218. https://doi.org/10.1016/J.AEJ.
2018.10.018
23. El-Sayegh SM, Mansour MH (2015) Risk assessment and allocation in highway construction
projects in the UAE. J Manage Eng 31(6). https://doi.org/10.1061/(ASCE)ME.1943-5479.000
0365
24. França A, Haddad A (2018) Causes of construction projects cost overrun in Brazil. Int J Sustain
Constr Eng Technol 9(1):69–83. https://doi.org/10.30880/IJSCET.2018.09.01.006
25. Andrić JM, Lu D-G (2016) Risk assessment of bridges under multiple hazards in operation
period. Saf Sci 83:80–92. https://doi.org/10.1016/j.ssci.2015.11.001
26. Alsuliman JA (2019) Causes of delay in Saudi public construction projects. Alex Eng J
58(2):801–808. https://doi.org/10.1016/J.AEJ.2019.07.002
27. Issa UH, Marouf KG, Faheem H (2021) Analysis of risk factors affecting the main execution
activities of roadways construction projects. J King Saud Univ Eng Sci. https://doi.org/10.
1016/j.jksues.2021.05.004
28. Sohu S et al (2019) Controlling measures for cost overrun causes in highway projects of Sindh
province. Eng Technol Appl Sci Res 9(3):4276–4280. https://doi.org/10.48084/ETASR.2749
29. Awosina A, Ndihokubwayo R, Fapohunda J (2018) Effects of inaccurate cost estimate on
construction project stakeholders. J Constr Project Manage Innov 8(2):1886–1904. https://doi.
org/10.36615/JCPMI.V8I2.164
30. Creedy GD, Skitmore M, Wong JKW (2010) Evaluation of risk factors leading to cost overrun
in delivery of highway construction projects. J Constr Eng Manage 136(5):528–537. https://
doi.org/10.1061/(ASCE)CO.1943-7862.0000160
@seismicisolation
@seismicisolation
Enhancing Flood Forecasting Accuracy
Through Machine Learning Approaches
Halappanavar Ruta Shivarudrappa, S. P. Nandhini, T. S. Pushpa ,

and K. P. Shailaja
1 Introduction
Floods pose a serious hazard to both infrastructure and human lives. On physical
models and statistical techniques, traditional flood prediction models are built. In
order to reduce the dangers and effects of flooding, which is a crucial component
of disaster management, precise forecasting techniques are needed. The long-term
viability of buildings and the general resilience of communities in flood-prone loca-
tions depend critically on flood prediction. When determining flood hazards and
creating mitigation plans, civil engineers are crucial players. To gauge the prospec-
tive flood hazard’s potential size, the results of this flood forecast are used. Due
to their complexity, poor accuracy, and large processing costs, these methods do
have some drawbacks. As a result, river overflows and flash floods pose a threat to
the infrastructure and human populations [1]. To reduce the effects of floods and
create efficient prevention and response plans, flood prediction models are crucial
[2]. In recent years, machine learning techniques have become a potent tool for
creating accurate flood prediction models. Based on statistical methods, machine
learning algorithms are capable of making accurate predictions without being explic-
itly programmed. Because of this, they are especially beneficial for complicated
and nonlinear systems, like those used in flood prediction. Data gathering, analysis,
H. R. Shivarudrappa (B) · S. P. Nandhini · T. S. Pushpa · K. P. Shailaja

Department of Computer Applications, B.M.S. College of Engineering, Bengaluru, Karnataka,
India
e-mail: halappanavar.mca21@bmsce.ac.in
S. P. Nandhini
e-mail: nandhinisp.mca21@bmsce.ac.in
T. S. Pushpa
K. P. Shailaja
e-mail: shailaja.mca@bmsce.ac.in
@seismicisolation
@seismicisolation
204 H. R. Shivarudrappa et al.
model development, infrastructure design, emergency planning, risk assessment, and

infrastructure maintenance of civil engineering projects are all areas where machine
learning approaches can add knowledge. Combining machine learning methods with
expertise in civil engineering.
2 Literature Survey
Traditional flood prediction models are built on statistical and physical models, which
have limitations such as complexity, accuracy, and high processing cost. With the
rise of machine learning algorithms as a powerful tool for model creation over the
last several years, the field of flood prediction has seen a noticeable breakthrough.
To produce precise predictions and decisions, machine learning uses the concepts of
statistical analysis, pattern recognition and nonlinear systems, such as those involved
in flood prediction [3]. As a result, complex and nonlinear systems, such as those
involved in flood prediction, can benefit greatly from using them [3]. Numerous
studies have been undertaken to examine the application of machine learning algo-
rithms in the field of flood prediction. In order to forecast floods in a river basin
in China, [4] created an artificial neural network (ANN) model [5]. This research
showed that machine learning algorithms performed better than traditional flood
prediction models by using input variables including rainfall, river flow, and temper-
ature data. Similar to this, Roushangar et al. [6] created an SVM model to predict
floods in an Iranian river basin [6]. They outperformed conventional flood predic-
tion models by using data on rainfall, river flow, and temperature as input factors
[7]. Various machine learning methods have been combined in ensemble models in
other research to increase accuracy and decrease uncertainty in flood prediction. In
a similar vein, Sun et al. [4] created an SVM model to forecast floods in a Chinese
river basin. They outperformed more established flood prediction models by using
data on rainfall, river flow, and temperature as input factors [4].
3 Objective
Based on the rainfall data, a flood prediction model employing machine learning is
created. Predictions are made using various Machine Learning models. To determine
which model is the best, these models are contrasted with one another. The major
goal of this study is to create a real-time flood prediction model [8] that will be
helpful in areas prone to flash floods. This system analyzes rainfall data from Kerala
using a variety of machine learning algorithms. Based on the accuracy of different
models, the best model is chosen in order to assist people in advance, save lives, and
conserve a significant amount of meteorological work.
@seismicisolation
@seismicisolation
Enhancing Flood Forecasting Accuracy Through Machine Learning … 205
4 Methods and Procedures
The process is as indicated in Figs. 1 and 2 and the description is stated below:
i. Data collection: Collecting accurate and representative data is the first step.
Dataset is obtained from Kaggle. The dataset is about the parameter like rainfall.
Fig. 1 Process flow diagram

[8]
Fig. 2 Machine learning process [9]
@seismicisolation
@seismicisolation
ii. Data pre-processing: After the dataset is gathered, pre-processing becomes

necessary to remove outliers, missing values, and irrelevant features. This is
done to improve the accuracy of the ML model.
iii. Feature selection: In this step, the most relevant features are selected, that can
be used to predict floods.
iv. ML model selection: Once the features are selected, the next step is to choose
the appropriate ML algorithm for the prediction task. Some ML algorithms used
for flood prediction include logistic regression, decision tree, random forests,
and KNN classifier [9].
v. Model training and validation: The ML model is then trained using a segment
of the data, and its performance is evaluated using the remaining data. This
process is done to guarantee that the model is not overfitting and can make
accurate predictions on new data.
vi. Model evaluation: Following optimization, the performance of the refined
machine learning model is assessed through diverse metrics, including accu-
racy, recall, and ROC scores. Consequently, the model becomes capable of
forecasting floods for novel data instances.
5 Implementation and Evaluation
Data Collection: Kaggle provided the data for this study, which focused on Kerala,
a picturesque state in South India known for its beautiful beaches and backwaters.
It is frequently referred to as “God’s own country” and has been named one of
National Geographic’s top ten paradises in the world. The dataset includes Kerala’s
monthly rainfall index from 1900 to 2018, which indicates if a flood occurred in each
corresponding month. This dataset’s analysis of the rainfall patterns enables flood
forecasting. Due to the effects of climate change, Kerala has seen an increase in
recent years in both heavy rainfall events and floods. Access to this useful informa-
tion, which includes monthly and annual rainfall observations in millimeters (mm)
from 1901 to 2018, was made possible through the Indian government’s open data
initiative. This study uses the enormous Kerala flood dataset, which includes data
going back over a century, to anticipate flood risks in Kerala using machine learning
techniques [9]. Here, we employ four machine learning (ML) algorithms—KNN
Classification, Logistic Regression, Random Forest, and Decision Tree—to create
the best model for forecasting flooding conditions using rainfall data for Kerala that
spans the last 100 years of rainfall in the region. The dataset was stored in the Data
Frame with the identifier “df.” and loaded into pandas using the read csv() function.
When working with tabular data in Pandas, the dataset was automatically turned
into a Data frame because it was in tabular form. A Data Frame is a changeable,
two-dimensional data structure in Python. It has columns and rows, just like an excel
sheet.
Data Preprocessing: Before data is used by machine learning algorithms, pre-
processing is essential for getting it ready. Its goal is to turn the raw data into a
@seismicisolation
@seismicisolation
more refined dataset so that the algorithm may use it to provide better results. The
lack of standardization and varied sources of the data make it difficult to examine.
Therefore, the data needs to be suitably formatted in order to improve the perfor-
mance of the applied machine learning model. Data formats must meet specified
requirements for some machine learning models. For instance, null values are not
supported by the Random Forest method. Therefore, it is necessary to manage the
null values existing in the original raw dataset appropriately in order to apply the
Random Forest algorithm effectively.
5.1 Algorithms Implemented
Decision Tree
The decision tree classification method is an effective machine learning algorithm
employed for flood prediction [10]. It falls under the category of supervised learning
algorithms and operates by constructing a decision tree using training data to make
predictions regarding the class labels of new data points. This algorithm identifies
the feature that provides the optimal data division into two groups, determined by
the highest information gain or Gini index. Subsequently, the data is recursively
partitioned based on the selected feature until a predefined stopping criterion is met,
such as reaching a maximum depth or a minimum number of instances in a leaf node.
Each leaf node within the decision tree represents a specific class label, enabling the
classification of new data points based on the path they traverse within the tree
structure.
Logistic Regression
Logistic regression serves as another valuable machine learning algorithm for flood
prediction. As a supervised learning method, it can estimate the probability of an
event taking place based on input variables. The algorithm establishes a model that
relates the input variables to the output variable, utilizing a logistic function [3].
In the context of flood prediction, the output variable is a binary variable, with the
value 1 denoting the occurrence of a flood and 0 indicating the absence of a flood.
By employing the logistic function, the input variables are mapped to a probability
value ranging from 0 to 1, representing the likelihood of a flood transpiring based on
the given input variables.
Random Tree Forest
Random forest classification is a powerful machine learning algorithm suitable for
flood prediction purposes. Operating as an ensemble learning algorithm, it amalga-
mates multiple decision trees to generate a final prediction [11]. In this approach, the
algorithm randomly selects subsets of the training data and features for each tree.
Multiple decision trees are then constructed, utilizing these distinct subsets of data
and features. Each decision tree produces a prediction based on its specific subset
@seismicisolation
@seismicisolation
of data and features. Ultimately, the final prediction is determined by aggregating

the predictions of all the decision trees, typically through voting or averaging mech-
anisms. By leveraging the collective knowledge of multiple trees, random forest
classification enhances the accuracy and robustness of flood prediction models.
KNN Classifier
Another useful machine learning approach that can be used to anticipate floods is the
KNN (k-Nearest Neighbor) classifier [11]. It belongs to the class of instance-based
learning algorithms and bases its prediction on the similarity between new data and
the data in the training set. This approach defines the parameter k as the number of
nearest neighbors to take into account when classifying. The program then calculates
the distances between each data point in the training set and the new data point [12].
Then, it chooses the k data points that are closest to the new data point. The new data
point is then labeled according to the class labels of its k closest neighbors. The KNN
classifier provides a reliable method for predicting floods by utilizing the closeness
of surrounding data points.
Ensemble Labeling Technique
By merging various separate models, ensemble learning is a useful machine learning
technique that improves a system’s overall performance. The underlying idea behind
ensemble learning is to take use of the power of several models, which, while they
may be weak individually, result in reliable predictions when combined. This is
accomplished by combining their forecasts using different aggregation techniques.
Ensemble learning, a powerful strategy for handling complicated machine learning
issues, improves accuracy, robustness, and generalization potential by utilizing the
various perspectives and insights offered by each model.
6 Comparison
The LR method got the highest score of the four algorithms utilized, 0.916, according
to the results as indicated in Fig. 3. This shows that the LR algorithm is very accurate
in forecasting floods. The study made clear how crucial it is to choose the right
features and have a high-quality dataset for effective flood prediction. A dataset of
flood records from Kerala was used to develop and analyze four machine learning
algorithms: K-Nearest Neighbor (KNN), Logistic Regression (LR), Decision Tree
(DT), and Random Forest (RF) with Ensemble Learning (EL). It also showed that the
model’s capacity to predict outcomes accurately depends on the quantity and quality
of data available. To increase the precision of the prediction model, efforts must be
made to improve the collection, storage, and management of flood-related data [13].
@seismicisolation
@seismicisolation
Fig. 3 Accuracy versus algorithms [2]
Table 1 Comparison of
ML models Recall ROC Accuracy
accuracy results [11]
Logistic regression 73.33 86.66 91.67
KNN classifier 93.33 96.66 83.83
Decision tree 100.00 83.33 66.67
Random tree forest 80.00 78.88 79.17
The table below given below presents a comparison of the accuracy results obtained
from the Logistic Regression, KNN Classifier, Random Forest Classifier, and Deci-
sion Tree Classifier algorithms. Analyzing the calculated values reveals that the
KNN Classifier and Random Forest Classifier demonstrate relatively superior results
as described in Table 1. Further examination indicates that the Logistic Regression
algorithm yields the highest accuracy percentage value [14]. Based on these findings,
it can be concluded that Logistic Regression exhibits the highest level of accuracy
among the compared algorithms.
8 Conclusion
In conclusion, our work has shown that machine learning algorithms are good in
predicting floods, which is a crucial part of efficient disaster management. A robust
and accurate flood prediction model has been created using historical rainfall data
and cutting-edge methods like Decision Tree, K-nearest neighbors (KNN), Random
@seismicisolation
@seismicisolation
Forest, and Logistic Regression. With a focus on Kerala, India, an area prone to
frequent flood events, the main goal of this research is to improve the capacity to
anticipate the incidence and severity of floods in a particular region. The experimental
findings demonstrate how machine learning models may considerably increase the
accuracy of flood prediction. The Decision Tree method has offered interpretability
and a better knowledge of the underlying patterns, delivering useful insights into the
key factors influencing flood occurrences. The KNN algorithm shows promise in
improving prediction performance for particular circumstances due to its capacity to
capture local patterns and take nearby influences into account. The Random Forest
algorithm’s ensemble learning strategy has successfully used the combined wisdom
of several decision trees, leading to a considerable improvement in prediction accu-
racy. In the meantime, Logistic Regression has demonstrated its prowess in binary
classification tasks by excelling at estimating the likelihood of flood events.
The comparison of various machine learning algorithms has highlighted the signif-
icance of selecting the most suitable algorithm based on specific requirements and
dataset characteristics.
References
1. Mosavi A, Ozturk P, Chau K (2018) Flood prediction using machine learning models: literature
review. Water 10(11):1536. https://doi.org/10.3390/w10111536
2. Kashyap K, Karthik A, ShivaKumar GS (2021) Flood prediction using machine learning. Int
Res J Eng Technol (IRJET) 8(7):4. https://www.irjet.net/archives/V8/i7/IRJET-V8I7583.pdf
3. Ha H, Bui QD, Khuc TD, Tran DT, Pham BT, Mai SH, Nguyen LP, Luu C (2022) A machine
learning approach in spatial predicting of landslides and flash flood susceptible zones for a
road network. Modeling Earth Syst Environ 8(4):4341–4357. https://doi.org/10.1007/s40808-
022-01384-9
4. Chen S, Ren M, Sun W (2021) Combining two-stage decomposition based machine learning
methods for annual runoff forecasting. J Hydrol 603:126945. https://doi.org/10.1016/j.jhydrol.
2021.126945
5. Zhang D, Shi X, Xu H, Jing Q, Pan X, Liu T, Wang H, Hou H (2020) A GIS-based spatial
multi-index model for flood risk assessment in the Yangtze River Basin, China. Environ Impact
Assessment Rev 83:106397. https://doi.org/10.1016/j.eiar.2020.106397
6. Roushangar K, Ghasempour R (2023) Chapter 24—supporting vector machines. In: Eslamian
S, Eslamian F (eds) Handbook of hydroinformatics. Elsevier, pp 411–422. https://doi.org/10.
1016/B978-0-12-821285-1.00009-9
7. Cui C, Cui L (2019) An innovative flood prediction system using improved machine learning
approach. Can Sci Fair J 2(2):5. https://static1.squarespace.com/static/5a63b41dd74cff19f4
0ee749/t/5e6bace8099e5e661b0ec845/1584114921879/An+Innovative+Flood+Prediction+
System.pdf
8. Ambore AK, Charan TSS, Reddy UR, Reddy TSS, Tarun G (2023) Flood prediction using
machine learning. Int J Res Appl Sci Eng Technol 11(5):363–367. https://doi.org/10.22214/ijr
aset.2023.51528
9. Kunverji K, Shah K, Shah N (2021) A flood prediction system developed using various machine
learning algorithms. SSRN Electron J. https://doi.org/10.2139/ssrn.3866524
10. Ruslan FA, Samad AM, Adnan R (2017) Modelling of flood prediction system using hybrid
NNARX and Extended Kalman Filter. In: 2017 IEEE 13th international colloquium on signal
processing & its applications (CSPA), 149–152. https://doi.org/10.1109/CSPA.2017.8064941
@seismicisolation
@seismicisolation
11. Vignesh Baalaji S (2020) Flood prediction system using multilayer perceptron classifier and
neural networks. Int Res J Eng Technol (IRJET) 7(5):10. https://www.irjet.net/archives/V7/i5/
IRJET-V7I51189.pdf
12. Ali MHM, Asmai SA, Abidin ZZ, Abas ZA, Emran NA (2022) Flood prediction using deep
learning models. Int J Adv Comput Sci Appl 13(9). https://doi.org/10.14569/IJACSA.2022.
01309112
13. Dazzi S, Vacondio R, Mignosa P (2021) Flood stage forecasting using machine-learning
methods: a case study on the Parma River (Italy). Water 13(12):1612. https://doi.org/10.3390/
w13121612
14. Puttinaovarat S, Horkaew P (2020) Flood forecasting system based on integrated big and
crowdsource data by using machine learning techniques. IEEE Access 8:5885–5905. https://
doi.org/10.1109/ACCESS.2019.2963819
@seismicisolation
@seismicisolation
Monitoring of Smart Greenhouse Using
Internet of Things (IoT)
S. Asha Bharathi , B. Meghana, S. Meghana, M. Akshatha,

and S. Hamsa
1 Introduction
India’s agriculture industry continues to operate conventionally and lags in adopting

modern advancements. Around 55% of Indians are employed in agriculture and
related activities, but these activities only account for 15% of the country’s GDP
[1]. As a result, farmers must move away from conventional farming methods and
modernize their operations using technology. While many individuals continue to
labor in the agricultural sector, the economic contribution of agriculture to India’s
GDP is steadily falling due to the nation’s broad-based economic growth. As a result,
there is an urgent need to upgrade the system so that it can enhance productivity and
create nutritious organic food [2].
New technologies that have been developed in the fields of information science
and communication have a wide range of uses. In particular, IoT applications in
agriculture are rapidly growing because they provide farmers with well-managed,
automated farms, better weather and crop yield predictions for agriculturists, and
information on consumer goods for the general public [3]. To secure food security
and socio-ecological sustainability in the future, greenhouse agriculture is seen as
one of the practical alternatives.
In the nineteenth century, greenhouse farming technology was first made avail-
able for commercial use in France and The Netherlands [4]. The roof is typically
covered with transparent material to maintain the climatic parameters necessary
for plant growth (such as temperature, humidity, illumination, etc.) as well as to
protect the plants from pests, diseases, and unfavorable environmental factors. Two
types of greenhouses are widely used such as glass greenhouses and plastic green-
houses. Commercial glass greenhouses, commonly known as hothouses, are cutting-
edge plant with flower-producing facilities. The glass greenhouse is equipped with
S. Asha Bharathi (B) · B. Meghana · S. Meghana · M. Akshatha · S. Hamsa

Jyothy Institute of Technology, Bengaluru, India
e-mail: asha.bharati@jyothyit.ac.in
@seismicisolation
@seismicisolation
214 S. Asha Bharathi et al.
screening, heating, cooling, and lighting systems, all of which can be managed by a
computer to create the ideal environment for plant growth. The glass used in house-
hold greenhouses is normally 3 mm thick. Depending on the type of greenhouse,
the crop, and environmental controls, greenhouses can boost yield by 10% to 13%
[5]. Crop growth is more reliable on greenhouse agriculture. Farmers do not need
to rely on the natural climatic seasons because greenhouse farming enables them to
establish a microenvironment where they can manage the temperature and humidity.
To achieve and maintain standard conditions, you might need to pay more for green-
house maintenance during the off-season. Crops that cannot tolerate the external
climate can also be grown in greenhouses. The correct parameter control, such as
cooling, heating, lighting, water flow, ventilation, level of CO2 , and other factors
inside the farm, has a significant impact on the output of the crops. This may have an
impact on economic profits and add to the difficulty of the farmers’ decision-making
process. The answer is said to lay in integrating Internet of Things (IoT) technology
with the greenhouse to make it “smart,” or the smart greenhouse. Smart greenhouse
farming may therefore take advantage of these issues and help farmers enhance crop
yield.
A mass system known as the “Internet of Things (IoT)” is interconnected with
several sensors, embedded controllers, decision-making platforms, the Internet, and
cloud servers. The data is automatically fed to the cloud server by the sensors after
being collected. The data is kept on the cloud servers and is accessible from a
distance [6]. This can assist farmers in effectively managing their fields and using
their resources best. Future crop growth within a smart greenhouse is expected to be
automated and remote-controlled, or at least part of the process will be.
This paper has surveyed different IoT platforms, protocols, and sensors. An IoT
system is proposed where it uses the ThingSpeak platform. The data from the sensors
is sent to the platform for analysis, the results are displayed on the website, and data
visualization is provided for the users.
2 Literature Review
Smart greenhouses using IoT technologies have become increasingly popular in

recent years due to their potential to revolutionize the agriculture industry by
improving crop yields, reducing resource consumption, and promoting sustainable
practices. This literature review aims to summarize and analyze recent studies on
smart greenhouses using IoT technologies.
Danita et al. [7] proposed a model that uses a Raspberry Pi3 integrated with various
IoT sensors to measure real-time attributes, which include temperature, humidity, and
soil moisture. Measured parameters are sent to the cloud for further visualization and
analysis. This is made possible through the built-in Wi-Fi function of Raspberry PI3.
Aafreen et al. [8] designed a model using GPRS-enabled sensor nodes to measure
real-time and transmit data to the cloud for analysis. A mobile app was created using
the Blynk IoT platform for messaging and controlling the greenhouse environment.
@seismicisolation
@seismicisolation
Monitoring of Smart Greenhouse Using Internet of Things (IoT) 215
Ameen et al. [9] suggested a model for deploying an Arduino Uno integrated with
the appropriate IoT sensors to measure greenhouse characteristics. The data was sent
to the cloud using a Wi-Fi module for additional processing. A set of guidelines are
established for controlling the greenhouse climate based on the real-time data using
the built-in fuzzy logic capability in ThingSpeak cloud. Jaiswal et al. [10] proposed
a model that uses an Arduino Uno integrated with various sensors was deployed in a
hydroponic environment, which refers to growing without soil. Also stated was the
superiority of hydroponic environments over soil-grown plants. Since food is readily
available in the water, hydroponic plants don’t need much energy to find it. This
conserved energy can be used to increase fruit production and accelerate growth.
3 Methodology
3.1 System Framework
Figure 1 represents the system framework. The system framework is described as

follows.
DHT22 Sensor
The DHT22 is an inexpensive digital temperature and humidity sensor. It measures
the humidity in the air using a thermistor and a capacitive humidity sensor, and it
outputs a digital signal on the data pin without the requirement for analogue input
pins.
Fig. 1 System framework
@seismicisolation
@seismicisolation
Soil Moisture Sensor

The soil moisture sensor module is utilized to measure soil moisture. Two probes
make up the moisture sensor, which measures the moisture content of the soil. The
sensor reads the resistance after passing the current through the soil using these two
probes to determine the soil’s moisture levels.
PH Sensor
A pH sensor is a device that measures the acidity or alkalinity of a solution based on
the concentration of hydrogen ions present.
GSM Module
A GSM module is an electronic device that permits communication between a
computer or mobile device and a GSM network. Typically, a GSM module has a
SIM card slot for connecting to the cellular network. It can be used to send and
receive text messages as well as voice calls.
3.2 System Modeling
The model is intended to quantify and monitor significant factors like temperature,
humidity, soil moisture, and pH level that may have an impact on plant growth. The
ambient temperature, which is a crucial component in determining plant growth, is
measured by temperature sensors. Low temperatures can limit the rate of growth,
while high temperatures can make plants wilt. The amount of moisture in the air,
which is also essential for plant growth, is measured by humidity sensors. High
humidity can raise the risk of fungal illnesses, whereas low humidity might cause
water stress. The amount of water in the soil, which is a crucial factor in determining
plant growth, is measured using soil moisture sensors. The growth of plants can be
negatively impacted by either overwatering or under-watering; therefore, it is critical
to precisely monitor soil moisture levels. The soil’s acidity or alkalinity, which is also
essential for plant growth, can be determined with pH sensors. In order for plants to
grow healthily, the pH of the soil has an impact on how readily available.
This was achieved by deploying the Arduino Uno in the field and integrating
the appropriate sensors for measuring the real-time parameters. Through the GSM
module’s AT instructions, all of these measured parameters are sent to the ThingS-
peak cloud. There are many benefits for using a GSM module in this model over
other wireless communication methods. First off, it provides cellular networks with
a broader range of coverage, allowing users to maintain communications over great
distances, even in off-the-beaten-path areas. Applications like agricultural moni-
toring, where crops may be spread across broad areas or in difficult-to-reach places,
can benefit most from using this. An efficient and cost-effective alternative for IoT-
based applications is to store and process sensor data in the ThingSpeak cloud. The
saved data is visualized and analyzed using ThingSpeak’s built-in MATLAB tools,
@seismicisolation
@seismicisolation
making it simpler to carry out the precise activities required for healthy plant growth.
A website is made to provide users with a better visualization of real-time data.
The datasets consist of 38 different classes, each class consists of more than a
thousand labeled images with 210 × 210 × 3 resolution. The model is composed
of three sequential blocks as shown in Fig. 2, each of which has a convolution layer
with a variety of 3 × 3 filters, a batchnormalization layer, and a maxpooling layer.
Filters in a convolution layer combine with the incoming data to extract features.
The output is then normalized using a batchnormalization layer to make sure it has
a unit variance and zero mean. This will enhance the model’s stability and conver-
gence, which will be useful during training. The output of the preceding layer is then
passed through a maxpooling layer, which takes the maximum value and reduces the
output’s spatial dimension. Each block is made up of the same order of convolution,
batchnormalization, and maxpooling with more filters, allowing the model to observe
more complex features from the input. The output of the final block is converted into
1D vector using flatten layer which can be fed into fully connected layer for classifi-
cation. For classification, two dense layers with different activation layers are used.
The first dense layer, with 256 units, uses the rectified linear unit (ReLU) activa-
tion function, and the second dense layer, with 38 units, uses the Softmax activation
function, which transforms the layer’s output into a probability distribution over all
possible classes.
Fig. 2 Process flow of plant disease detection
@seismicisolation
@seismicisolation
The proposed system provides accurate and timely measurements of the parame-
ters necessary for tracking plant growth and ensuring healthier plant growth. The
system measures parameters such as temperature, humidity, soil moisture, and pH
and displays real-time measurements of these parameters’ content in a user-friendly
interface through the ThingSpeak cloud platform.
The system also provides variations in the parameters over a period of a week or a
day, allowing farmers to track plant growth and make informed decisions regarding
the provision of nutrients and other environmental factors necessary for healthy plant
growth. Additionally, the system detects the soil moisture content and controls the
water pump motor connected to an Arduino, which turns on and off based on the wet
or drought level of the soil.
All of these data are displayed in the Jaivika dashboard, which requires registration
and a login to access. The Jaivika dashboard provides data visualization to the user in
the form of graphs and charts, enabling the user to easily track the real-time updates
of the plant parameters.
Overall, the proposed system provides a comprehensive solution for monitoring
and managing plant growth, offering accurate and timely measurements of key
parameters, real-time updates, and data visualization capabilities. The system can
help farmers make informed decisions about plant growth and promote healthier
crop yields.
Figures 3, 4, 5, and 6 show the real-time measurement of parameters such as
temperature, humidity, soil moisture, and pH level in the form of a graph, which also
provides information about when the data was uploaded along with time and date.
Figures 7, 8, 9 and 10 provide information about the variations of the parameters
over a period of a week or a day formed using the MATLAB built-in function of
ThingSpeak IoT platform.
Fig. 3 Plot of temperature of plants surrounding
@seismicisolation
@seismicisolation
Fig. 4 Plot of humidity of plants surrounding
Fig. 5 Plot of soil moisture
Fig. 6 Plot of pH of soil
Figures 11, 12 and 13 are all about the results that are obtained in the created dash-
board which is named Jaivika dashboard this is the user interface that includes dash-
board, devices, settings, and analysis sections in which the graphs of the measured
parameters are displayed the dashboard section, the specifications of the devices/
sensors used are added in devices section, and finally analysis of the parameters, i.e.,
variations of measured parameters over a period of a week or a day are provided in
the analysis section. Website also provides the option for detecting the plant disease
by uploading the picture of defected plant.
@seismicisolation
@seismicisolation
Fig. 7 Analysis of
temperature of plants
surrounding
Fig. 8 Analysis of humidity

of plants surrounding
The values mentioned in Table 1 are general guidelines, and specific plant species
may have varying requirements. It’s crucial to consider the specific needs of the
plants being grown and adjust the parameters accordingly. Regular monitoring and
observation of plant growth and health can help identify any deviations from optimal
conditions, allowing for timely adjustments and interventions.
@seismicisolation
@seismicisolation
Fig. 9 Analysis of soil

moisture
Fig. 10 Analysis of pH of
soil
@seismicisolation
@seismicisolation
Fig. 11 Jaivika dashboard
Fig. 12 Analysis in Jaivika dashboard
@seismicisolation
@seismicisolation
Fig. 13 Designed web page for plant disease prediction
Table 1 Measured parameter values and effects on plant growth in a greenhouse

Parameter Optimal Optimal Effects of low Effects of high Solutions Measured
daytime night-time levels levels values
range range (average
value)
Temperature 20–30 °C 15–20 °C Slow growth, Water loss, Adjust 26.3 °C
(68–77 (59–68 °F) poor nutrient alteration in ventilation or
°F) absorption, photosynthesis heating
susceptibility to process, systems to
diseases stunted growth, maintain
increased optimal
disease risk temperature
Humidity 50–70% 70–90% Stunted growth, Mold and Use 60.7%
increased fungal growth, humidifiers
susceptibility to reduced or misting
pests and transpiration systems to
diseases and increase
evaporation, humidity.
decreased Provide
nutrient uptake proper
ventilation
and air
circulation to
reduce
humidity
Soil 20–60% 20–60% Less Dehydration, Adjust 24.7%
moisture photosynthesis, wilting, irrigation (1008–20)/
less nutrient nutrient frequency (60–20)
uptake imbalance and duration
based on
plant needs
@seismicisolation
@seismicisolation
5 Conclusion
One of the major challenges in implementing a greenhouse monitoring system using

IoT technology is sensor integration, as it involves connecting and making different
sensors work together, which can be complicated due to compatibility and communi-
cation issues. Additionally, ensuring accurate data synchronization among multiple
sensors is crucial, as it requires precise coordination to collect and time-stamp data
accurately. Regular calibration and maintenance of the sensors are essential to main-
tain accurate measurements, which involves following a maintenance schedule and
conducting timely calibrations. Processing and analyzing the collected data to derive
meaningful insights can be challenging, but implementing data analytics techniques
and visualization tools can aid in data interpretation for effective decision-making.
The suggested system measures crucial metrics for monitoring plant growth, such
as moisture content, and it offers fluctuations in these metrics over time. Using an
Arduino board and a GSM module to transmit this data in real time to the ThingSpeak
cloud platform, the system enables farmers to remotely monitor their crops and make
decisions about the application of nutrients and other environmental factors required
for healthy plant growth. The system includes an automatic water pump that activates
when the moisture content in the soil drops below a threshold value and shuts off
when the moisture content rises over the threshold value, in addition to providing
real-time monitoring. This feature reduces the chance of over- or under-watering,
optimizes water use, and frees farmers from the duty of continuously checking and
adjusting water levels. The user interface provided by the system allows farmers to
readily access the data and make informed decisions on their crops. The interface
presents clear visualizations and alerts when parameters fall outside of desired ranges,
allowing farmers to quickly identify and respond to issues. Comprehensively, the
proposed system offers a scalable and efficient solution to monitoring crop growth,
utilizing modern technologies such as sensors, Arduinos, GSM modules, and cloud-
based platforms like ThingSpeak. The system may be adjusted to match the demands
of particular farms, allowing for greater efficiency, productivity, and sustainability.
6 Future Scope
To inform the farmer or user of the devices’ conditions, the working conditions of the
equipment can be incorporated into the user interface. Based on the weather during
specific seasons, the system can use voice assistance to provide the farmer or user
with seasonal crop growth information or guidelines that can be used in large-scale
agriculture fields.
@seismicisolation
@seismicisolation
References
1. Panhalkar V, Chavan A, Kadam S, Sathe R (2020) Smart greenhouse system using IOT. IJRTI
5(5). ISSN: 2456-3315
2. Kodali R, Jain V, Karagwal S (2016) IoT-based smart greenhouse: 1–6. https://doi.org/10.1109/
R10-HTC.2016.7906846
3. Kumar KG, Rao KN (2019) Autonomous greenhouse using internet of things with ThingSpeak.
Int Res J Eng Technol (IRJET) 06(10). e-ISSN: 2395-0056
4. Rayhana R, Xiao G, Liu Z (2020) Internet of things empowered smart greenhouse. IEEE J
Radio Freq Identification: 1. https://doi.org/10.1109/JRFID.2020.2984391
5. Tapiwa M (2019) The greenhouse concept as used in agriculture, its advantages and disad-
vantages. ACADEMIA, San Francisco, CA, USA. Accessed 19 Dec 2019 [online]. Available
https://www.academia.edu
6. Wiangtong T, Sirisuk P (2018) IoT-based versatile platform for precision farming. In: 2018 18th
international symposium on communications and information technologies (ISCIT), Bangkok,
Thailand, pp 438–441. https://doi.org/10.1109/ISCIT.2018.8587989
7. Danita M, Mathew B, Shereen N, Sharon N, John Paul J (2018) IoT based automated green-
house monitoring system. In: Proceedings of the second international conference on intelligent
computing and control systems (ICICCS 2018), CFP18K74-ART. ISBN:978-1-5386-2842-3
8. Aafreen R, Shamim R, Neyaz SY, Salim Beg M (2014) An IoT based system for telemetry and
control of greenhouse environment. Int J Comput Sci Eng Technol 4:1127–1129
9. Ameen NM, Mahdi Al-Ameri JA (2019) IoT-based shutter movement simulation for smart
greenhouse using fuzzy-logic control. In: 2019 developments eSystems engineering
10. Jaiswal H, Karmali Radha P, Abraham Sampson S, Singuluri R (2019) IoT and machine learning
based approach for fully automated greenhouse. In: 2019 IEEE Bombay section signature
conference (IBSSC)
@seismicisolation
@seismicisolation
Machine Learning Algorithms
for Classifying Land Use and Land Cover
N. R. Asha Rani and M. Inayathulla
1 Introduction
Alterations in land use and land cover (LULC) model have grown in popularity in
recent years since it is a complicated subject including actual, ecological, social, and
economic elements. Physical modifications that affect the surface of the earth, like
alterations to aquatic bodies, soil health, and contamination of the air, are examples
of LULC. Additionally, shifts in land use are frequently linked to the impact of
behavioural patterns on physical alterations on the surface of the earth. Land-cover
alteration is frequently influenced by changes in land use to a certain degree. It has
a complicated relationship between LULC and the natural environment in addition
to human activity. Furthermore, LULC has an impact on the ecological balance.
Changes in the cover of land, for example, might affect the quantity and quality
of water and the character of the surface [20].
The study of LULC, particularly urbanization and growth, has gained prominence
and relevance. Urbanization is currently a worldwide problem that affects people’s
daily existence. Fast economic growth and elevated population percentages have
resulted in significant urbanization and expansion. Urbanization has grown increas-
ingly visible in most nations, particularly in large cities and rapidly growing areas
and territories, taking with it a slew of economic and ecological problems [14].
Numerous machine learning techniques in land usage categorization and
modelling of land use determination procedures were evaluated on various types of
information. Random forest—RF, generative adversarial networks—GANS, support
vector machine—SVM, neural network—NN, and Markov random field are some
of the most prominent approaches [19]. This current paper gives a summary of basic
LULC classification, various simulation approaches and classification methodolo-
gies in machine learning and deep learning algorithms, deep network layout with its
N. R. Asha Rani (B) · M. Inayathulla

Department of Civil Engineering, U.V.C.E., Bangalore, Karnataka, India
e-mail: asharaninr@gmail.com
@seismicisolation
@seismicisolation
228 N. R. Asha Rani and M. Inayathulla
significance, benefits and drawbacks of models based on machine learning, the key
benefits are also emphasized, as are future study prospects.
2 LULC Categorization
Exploiting human beings has caused developments in the last 50 years. The conver-
sion of land used for agriculture to urban areas and the growth of urbanization alter the
overall trend of terrestrial utilization, LULC. Land coverage (LC) defines geographic
discrepancy in the data relating to the exterior of the ground, like soil, flora, and
aquatic, while land utilization (LU) defines the modifications caused by anthro-
pological activity or bodily alterations on the surface of the planet, like as forest
destruction, urbanization, built-up regions, flooding, and foods, among other things.
Change in LULC is an important aspect of satellite imagery because it extracts vital
data, analyses images, and classifies the spectral indications of land covering indi-
cated in Table 1. LULC modifications are classified as follows: (a) initial processing,
(b) identification of changes technique, and (c) accuracy evaluation [1, 25].
3 Land-Use/Land-Cover (LULC) Simulation Approaches
Modelling may be used to simplify complicated geographic frameworks, like the

method of urban landscape transformation. Overall, modelling entails defining
system behaviour and converting it into equations. The model linkages are then
programmed in any computer programming language that may be tailored to the
GIS software. Machine learning models are a key component of artificial intelli-
gence. ML models are computer algorithms that are designed to learn how to execute
a LULC-related activity.
Machine learning models are used to create learning algorithms that do modelling
automatically. As a result, no human aid or expertise is required to simulate future
land-use change. As a result, researchers who utilize ML models should be conversant
with many statistical models as shown in Table 2 [1].
The first application of data from satellites for LULC categorization dates early
during the 1940s as Francis J. Marschner started mapping the whole United States
utilizing aerial photographs to associate land use to the surface of the planet. Years
thereafter, soon after the successful launch of a Space Resources Techniques investi-
gation satellite fitted by hyperspectral spectrometer (MSS) during the month of July
1972 including the beginning of the Landsat programme, investigations employing
images collected from space data to categorize LULC reached an entirely novel level
[2]. Indeed, the creation of the Landsat plug-in as well as the (independent) distri-
bution of information offered innovative difficulties to the area of LULC, such as
multi-modal fusion of data, temporal change in land discovery, and biological uses
@seismicisolation
@seismicisolation
Table 1 The primary three levels of the land cover categorization framework (LCCS) hierarchies [25]
Classification Classification schemes
levels
First level Primarily vegetated areas Primarily non-vegetated areas
Second level Terrestrial primarily vegetated Terrestrial primarily non-vegetated areas Aquatic or regularly Aquatic or regularly flooded
areas flooded primarily primarily non-vegetated areas
vegetated areas
Third level Cultivated and Natural and Cultivated aquatic Natural and semi-natural Artificial Bare Artificial Natural
managed semi-natural or regularly flooded aquatic or regularly surfaces and areas waterbodies, waterbodies,
@seismicisolation
@seismicisolation
terrestrial vegetation areas flooded vegetation associated snow, and ice snow, and ice
areas areas
Machine Learning Algorithms for Classifying Land Use and Land Cover
229
Table 2 Optimized settings for several classifications employed in LULC analysis [1]
S. No. Methods Software used for Optimized parameters
modelling
1 Artificial neural network TerrSet geospatial Hidden layer-1, input layer-1,
monitoring and modelling output layer-1, nodes-6,
system learning rate-0.01,
momentum factor-0.5,
sigmoid constant-1
2 Support vector machine Environment for Kernel type-radial basis
visualizing images (ENVI function, gamma in kernel
5.3) function-1, penalty
parameter-100, pyramid
level-1, pyramid
reclassification
threshold-0.90
3 Fuzzy ARTMAP TerrSet geospatial F1 layer neurons-12, F2 layer
monitoring and modelling neurons-385, map field layer
system neurons-6, a choice
parameter for ARTa-0.01,
learning rate and vigilance
parameter for ARTa-1 and
0.98, learning rate and
vigilance parameter for
ARTb-1 and 1, iteration 3338
4 Spectral angle mapper Environment for Wavelength
visualizing images (ENVI units-micrometres, Y data
5.3) multiplier-1, set maximum
angle (radiance)- single
value, maximum angles
(radians)-0.100
5 Random forest R programming language –
(R 3.5.3)
6 Mahalanobis distance Environment for –
visualizing images (ENVI
5.3)
of satellite imagery. Some early efforts on these subjects are discussed in [21, 25].
Figure 1 depicts the flow of LULC mapping or image classification.
4 Remotely Sensed LULC Categorization Employing

Machine Learning and Deep Learning Algorithms
For huge-scale LULC modelling using multispectral and hyperspectral aerial

pictures, improved artificial intelligence algorithms, in particular deep learning
(DL), were established, as has the combination of satellite imagery into additional
@seismicisolation
@seismicisolation
Machine Learning Algorithms for Classifying Land Use and Land Cover 231
Fig. 1 Flow of image classification [5]
geographic big data. Deep learning outperformed older approaches like RF and
SVM. Nonetheless, there are several challenges in integrating sophisticated arti-
ficial intelligence or deep learning algorithms for effective LULC modelling with
large geographic data [24].
4.1 How DL Techniques and Machine Learning Techniques

Used for LULC Identification
Pixel-based categorization is the conceptual division of pictures that assigns cate-

gories to pixels (like road, grass, or urban area location). The main purpose of pixel-
based categorization is to classify the pixels in a picture correlate to certain percep-
tive objects, hence giving interpretation for the images. The pixel-based method fails
to deliver the desired result due to the substantial amount of wavelength consis-
tency between categories and the variability inside classes. In classic techniques,
imagery from satellite pictures is classified using pixel-based methods, with the pure
label pixel being considered as the natural goal. Object-based classification, on the
contrary hand, is a revolutionary model for categorizing remotely detected pictures
that exceed pixel-based classification [28].
4.2 LULC, Deep Neural Network Architecture
AlexNet. In [18], it is mentioned AlexNet is world’s first profound CNN framework

for picture categorization and identification. AlexNet’s capacity for learning had
been boosted by using several variable optimization algorithms. The AlexNet level
@seismicisolation
@seismicisolation
is being extended by 5–8 levels for several picture database categories, improving
image resolution.
ZfNet. As proposed in [4, 19], ZfNet is a multi-layer de-convolutional artificial
neural network. It was designed to do statistical analyses of network efficiency. ZfNet
revealed that solely a small portion of neurons are functioning, that some neurons
are quiescent in the first layer, and their filter width and stride can be optimized in
the subsequent layer.
VGGNet. As mentioned in [8, 9], it is a simple and detailed CNN architecture
development methodology that decreased the total number of variables and led to
a 19-layer depth and 33-filter structure having the additional benefit of the reduced
computational cost. Inception-VI has been developed having the main goal of giving
outstanding precision at a low processing cost.
GoogleNet. As proposed in [9, 11], the layers of convolution have been substituted
with tiny neural system levels in every level in GoogleNet. The concept of retained
learning in CNN is a very effective training for deep network approach.
ResNet. As discussed in [11, 12] that ResNet has lesser computational difficulty than
the previously suggested network. The answer to the issue of disappearing slopes.
DenseNet. As proposed in [12, 22] stated that DenseNet solved this problem by
reusing cross-layer connections.
Convolutional Block Attention Module. As proposed in [13, 23], the Convolutional
Block Concentration Modules (CBAM) represent a new sort of CNN that relies on
concentration.
CapsuleNet. As proposed in [13, 26], capsule neurons, which have the capacity to
recognize facial features in addition to various associated information, are used in
this approach. CapsuleNet is a capsule network made up of several different capsules.
This denotes an excellent resolution for eyesight tasks.
HRNetV2. As proposed in [13], HRNet offers two primary characteristics. Initially,
an alternating link is established between the high-to-low-resolution inversion
sequence.
The above literature gives brief representation of deep neural network designs
such as VGGNet, GoogleNet, AlexNet, ResNet, and DensNet are increasing in
prominence in image classification and segmentation based on semantics.
4.3 Machine Learning Classification Approaches
This section gives significance of most popular machine learning algorithms.
@seismicisolation
@seismicisolation
Artificial Neural Networks (ANN)

Among all neural network approaches, ANN can be effectively employed in nonlinear
processes like variable retrieval and LULC modifications having the capacity to
operate on huge data analytics. It is now among its most frequently employed non-
parametric in classifying approaches. It doesn’t seem to be predicated on any sort
of presumption of normally distributed data. The backpropagation method, which is
a gradient descent technique, constitutes one of the most essential algorithms that
ANN often employs. Its primary goal is to minimize mistakes between real network
results and practice input or output combinations. The communication system gets an
amount of input/output combinations regularly, and the failure is transmitted through
the layer that produces the result back to the input level [3, 6].
Support Vector Machine (SVM)
SVM is a type of non-parametric controlled algorithm for a learning approach that
was designed to address issues with binary classification in the beginning. It depends
on the notion of structural risk reduction (SRM), which increases and isolates the
portion of the hyper-plane and information details closest to the hyperplane spectral
angle mapper (SAM). It divides points of information into different classifications
utilizing a hyper-spectral plane. These vectors guarantee that the length of the border
is maximized throughout this operation [6].
Fuzzy ARTMAP (FA)
The fuzzy ARTMAP approach relies on similarities of fuzzy subgroup computation
and dynamic and adaptive categorization identification via feature-space exploration.
The fuzzy ARTMAP layout consists of a pair of sections, ARTa and ARTb [3].
Random Forest
Breiman created RF, a revolutionary non-parametric combination algorithmic real-
izing technique. An algorithm known as RF is being extensively used to solve envi-
ronmental issues such as managing water resources and natural disaster control.
Copes with a wide variety of data, such as satellite imagery and mathematical infor-
mation. This is the decision tree-based combination study system that blends huge
ensemble classification and regression networks [7, 10].
5 Benefits and Drawbacks of Models Based on ML
Gradient descent seems to be a methodical approach that seeks to minimize a cost

value. It must be able to calculate the fractional differential of the gradient or slope
variable. At every repetition, coefficients are generated by using a negative value of
the gradient and decreasing the parameters at every stage by a rate of learning (step
size) multiplied by the derivatives, such that a local minimum may be reached after
several iterations [16]. Regression serves as a controlled learning strategy. It may be
@seismicisolation
@seismicisolation
used for predicting and modelling constant variables. Some instances of regression
technique applications include real-estate cost forecasting, forecasting sales, student
test results forecasting, and price fluctuation predicting [16, 17]. Regression with
multiple variables, tabular methods, and partial correlation are used to accomplish
this. It represents difficult real-world situations in a realistic and practical manner [16,
27]. A decision tree provides these benefits: It is appropriate for both classification
and regression problems, has a simple interpretation, can handle categorized and
numerical variables, and can fill in value gaps [16–19].
6 Future Scope
Efforts to build procedures and platforms for compiling, sharing, visualising, and
disseminating big LULC datasets, including its associated biases and mistakes, are
critically needed. Because of the increasing expansion of data, traditional cyberin-
frastructure technologies have reached a few restrictions. When LULC data becomes
exceedingly huge, for example, human eyes struggle to extract significant information
for visualization. It is difficult to convey too much information on a small screen. Due
to lower displays and resolutions, presenting large amounts of data on mobile devices
is considerably more difficult. Traditional ways of creating tiling and pyramids are
inefficient for fast visualization [19, 21].
Computer vision approaches can help in the extraction of significant LULC infor-
mation from remotely sensed imagery [15]. To recognize LULC spatial characteris-
tics and patterns, computer vision approaches with neural networks as the underlying
framework can be employed. Computer vision techniques may also be used to elim-
inate noise from satellite and aerial imaging data and improve its quality for LULC
classification, change detection, and data fusion. For LULC mapping, novel tools and
methodologies that integrate cutting-edge computer vision technology with remotely
sensed images must be developed in the near future [28].
7 Conclusion
LULC assessment is a recently investigated topic in applications of satellite imagery

such as changing the climate, urban planning, disaster mitigation, and environmental
change. The prominence of DL techniques in imagery collection for land use predic-
tions prompted the research. Because of the abundance of funds, the HSI/MSI images
and the Landsat collection are particularly often employed based on image classifica-
tion in LULC. Additionally LULC mapping continues to face obstacles, such as insuf-
ficient data, imprecise and varied information, unbalanced and sparse sample data,
mixed LULC, and interpretability of models and the transferability concerns. Deep
learning or sophisticated machine learning methods.
@seismicisolation
@seismicisolation
References
1. Aburas MM, Ho YM, Ramli MF, Ash’aari ZH (2017) Improving the capability of an integrated
CA-Markov model to simulate spatio-temporal urban growth trends using an analytical hier-
archy process and frequency ratio. Int J Appl Earth Observ Geoinf 59:65–78. https://doi.org/
10.1016/j.jag.2017.03.006
2. Aburas MM, Ho YM, Ramli MF, Ash’aari ZH (2018) Monitoring and assessment of urban
growth patterns using spatio-temporal built-up area analysis. Environ Monit Assess 190:156
3. Arabameri A, Pal SC, Rezaie F, Nalivan OA, Chowdhuri I, Saha A, Lee S, Moayedi H
(2021) Modeling groundwater potential using novel GIS-based machine-learning ensemble
techniques. J Hydrol Regional Stud 36:100848
4. Arun PV, Buddhiraju KM, Porwal A (2019) Capsulenet-based spatial–spectral classifier for
hyperspectral images. IEEE J Sel Top Appl Earth Observ Remote Sens 12:1849–1865 (2019)
5. Chen S, Tian Y (2015) Pyramid of spatial relations for scene-level land use classification. IEEE
Trans Geosci Remote Sens 53:1947–1957. https://doi.org/10.1109/TGRS.2014.2351395
6. Francini M, Salvo C, Vitale A (2023) Combining deep learning and multi-source GIS methods
to analyze urban and greening changes. Sensors 23:3805. https://doi.org/10.3390/s23083805
7. Garosi Y, Sheklabadi M, Conoscenti C, Pourghasemi HR, Van Oost K (2019) Assessing the
performance of GIS-based machine learning models with different accuracy measures for
determining susceptibility to gully erosion. Sci Total Environ 664:1117–1132
8. Huang B, Zhao B, Song Y (2018) Urban land-use mapping using a deep convolutional neural
network with high spatial resolution multispectral remote sensing imagery. Remote Sens
Environ 214:73–86. https://doi.org/10.1016/j.rse.2018.04.050
9. Hu J, Mou L, Schmitt A, Zhu XX (2017) FusioNet: a two-stream convolutional neural network
for urban scene classification using PolSAR and hyperspectral data. In: 2017 joint urban remote
sensing event (JURSE). IEEE, Dubai, United Arab Emirates, pp 1–4
10. Juliev M, Mergili M, Mondal I, Nurtaev B, Pulatov A, Hübl J (2019) Comparative analysis of
statistical methods for landslide susceptibility mapping in the Bostanlik District, Uzbekistan.
Sci Total Environ 653:801–814
11. Kussul N, Lavreniuk M, Skakun S, Shelestov A (2017) Deep learning classification of land
cover and crop types using remote sensing data. IEEE Geosci Remote Sensing Lett 14:778–782
12. Mahdianpari M, Salehi B, Rezaee M, Mohammadimanesh F, Zhang Y (2018) Very deep convo-
lutional neural networks for complex land cover mapping using multispectral remote sensing
imagery. Remote Sensing 10:1119. https://doi.org/10.3390/rs10071119
13. Maxwell AE, Warner TA, Fang F (2018) Implementation of machine-learning classification in
remote sensing: an applied review. Int J Remote Sens 39:2784–2817
14. MohanRajan SN, Loganathan A, Manoharan P (2020) Survey on land use/land cover (LU/LC)
change analysis in remote sensing and GIS environment: techniques and challenges. Environ
Sci Pollut Res 27:29900–29926. https://doi.org/10.1007/s11356-020-09091-7
15. Ray S (2019) A quick review of machine learning algorithms. In: 2019 international conference
on machine learning, big data, cloud and parallel computing (COMITCon). IEEE, Faridabad,
India, pp 35–39
16. Reba M, Seto KC (2020) A systematic review and assessment of algorithms to detect,
characterize, and monitor urban land change. Remote Sens Environ 242:111739
17. Shih H, Stow DA, Tsai YH (2019) Guidance on and comparison of machine learning classifiers
for Landsat-based land cover and land use mapping. Int J Remote Sens 40:1248–1274
18. Simonyan K, Vedaldi A, Zisserman A (2014) Deep inside convolutional networks: visualising
image classification models and saliency maps. http://arxiv.org/abs/1312.6034
19. Talukdar S, Singha P, Mahato S, Shahfahad Pal S, Liou Y-A, Rahman A (2020) Land-use
land-cover classification by machine learning classifiers for satellite observations—a review.
Remote Sensing 12:1135. https://doi.org/10.3390/rs12071135
20. Tang L (2022) Geographic information system and climate big data applied in the research of
bird migration. In: 2022 IEEE international conference on electrical engineering, big data and
algorithms (EEBDA). IEEE, Changchun, China, pp 338–341
@seismicisolation
@seismicisolation
21. Tuia D, Kellenberger B, Beery S, Costelloe BR, Zuffi S, Risse B, Mathis A, Mathis MW,
Van Langevelde F, Burghardt T, Kays R, Klinck H, Wikelski M, Couzin ID, Van Horn G,
Crofoot MC, Stewart CV, Berger-Wolf T (2022) Perspectives in machine learning for wildlife
conservation. Nat Commun 13:792
22. Vali A, Comai S, Matteucci M (2020) Deep learning for land use and land cover classification
based on hyperspectral and multispectral earth observation data: a review. Remote Sensing
12:2495
23. What is machine learning and how does it work? https://www.androidauthority.com/machine-
learning-explained-3074635/. Last accessed 14 May 2023
24. Yan J, Wang L, Song W, Chen Y, Chen X, Deng Z (2019) A time-series classification approach
based on change detection for rapid land cover mapping. ISPRS J Photogramm Remote Sens
158:249–262
25. Yin J, Dong J, Hamm NAS, Li Z, Wang J, Xing H, Fu P (2021) Integrating remote sensing
and geospatial big data for urban land use mapping: a review. Int J Appl Earth Observ Geoinf
103:102514
26. Zhang W, Tang P, Zhao L (2019) Remote sensing image scene classification using CNN-
CapsNet. Remote Sensing 11:494
27. Zhu Z (2017) Change detection using Landsat time series: a review of frequencies, prepro-
cessing, algorithms, and applications. ISPRS J Photogramm Remote Sens 130:370–384
28. Zeiler MD, Fergus R (2014) Visualizing and understanding convolutional networks. In: Fleet D,
Pajdla T, Schiele B, Tuytelaars T (eds) Computer vision—ECCV 2014. Springer International
Publishing, Cham, pp 818–833. https://doi.org/10.1007/978-3-319-10590-1_53
@seismicisolation
@seismicisolation
Comparative Analysis of Machine
Learning Models for Earthquake
Prediction Using Large Textual Datasets
K. R. Niteesh, T. S. Pooja, T. S. Pushpa , P. Lakshminarayana,

and K. Girish
1 Introduction
This study focussed on a comprehensive and comparative study and the analysis of
earthquake prediction using different regression models. The models used in the study
include linear regressor, which is the simplest regression model used. Despite being
a simple regression model, linear regressor is found to be effective in various cases.
Other machine learning algorithms such as decision tree regressor and random forest
regressor are used. These models along with an artificial neural network regressor
were trained on a dataset consisting of roughly 23,000 earthquakes that were recorded
between 1965 and 2021.
The major objective of the study was to use multiple regression models and train
them on the dataset to understand the real-world application in earthquake prediction.
The predictive capabilities and the accuracy of the models were evaluated using mean
squared error (MSE) and R-squared score.
Following an ample amount of experimentation, the models were compared. The
outcomes of the comparative study aid not only in understanding the strengths and
weaknesses of different models in the case of earthquake prediction but also in
K. R. Niteesh (B) · T. S. Pooja · T. S. Pushpa · P. Lakshminarayana · K. Girish

Department of Computer Applications, B.M.S. College of Engineering, Bangalore,
Karnataka 560019, India
e-mail: niteesh.mca20@bmsce.ac.in
T. S. Pooja
e-mail: poojats.mca21@bmsce.ac.in
T. S. Pushpa
P. Lakshminarayana
e-mail: pln.mca@bmsce.ac.in
K. Girish
e-mail: gk.mca@bmsce.ac.in
@seismicisolation
@seismicisolation
238 K. R. Niteesh et al.
understanding the applicability of machine learning and deep learning for earthquake
prediction and in enabling practitioners in opting for the most effective approach for
earthquake prediction.
2 Literature Review
Numerous studies have already explored the application of machine learning

techniques to improve the accuracy of earthquake predictions and magnitude
estimations.
Asim et al. in a study titled “Earthquake magnitude prediction in Hindukush
region using machine learning techniques” used a dataset consisting of earthquakes
that occurred in the Hindu Kush region. The dataset consisted of eight different
features that were used to predict earthquakes as a binary classification problem.
The study used four different models including recurrent neural network, pattern
recognition neural network, random forest, and linear programming boost [1].
Li et al. conducted a study titled “DLEP: A Deep Learning Model for Earthquake
Prediction” explored the application of deep learning for earthquake prediction using
a dynamic loss function [2].
Salam et al. explored the application of a hybrid machine learning model to
predict earthquakes in an article titled “Earthquake Prediction using Hybrid Machine
Learning Techniques.” Four assessment criteria including RMSE, MAE, PMRE,
and SMAPE were used to assess the models. The proposed model of FPA-LS-SVM
outperformed all other models [3].
Debnath et al. categorised earthquakes into three different categories based on
their severity in a study titled “Analysis of Earthquake Forecasting in India Using
Supervised Machine Learning Classifiers.” They also used six different datasets
to train various classical machine learning models such as simple logistic, random
forest, logistic model tree (LMT), ZeroR, and logistic regression. All models achieved
satisfactory results on different datasets [4].
Chittora et al. in a paper titled “Experimental analysis of earthquake predic-
tion using machine learning classifiers, curve fitting, and neural modeling” used
six different machine learning classifiers such as artificial neural network, random
tree, CHAID, discriminant, XGBoost tree, and Tree-AS on six different datasets
collected from various regions in India [5].
Xiong et al. used satellite data to predict earthquakes using machine learning in
a paper titled “Towards advancing the earthquake forecasting by machine learning
of satellite data.” An AdaBoost-based ensemble model was proposed and a dataset
with 1371 earthquakes between 2006 and 2013 was used [6].
Various machine learning approaches have been used already to predict earth-
quakes in the past although, the models were trained on smaller, regional datasets
and lacked generalisability. The studies also highlighted the applicability of various
machine learning and deep learning algorithms for earthquake prediction. The present
@seismicisolation
@seismicisolation
Comparative Analysis of Machine Learning Models for Earthquake … 239
research focuses on building upon the findings of the existing research and enhancing
the accuracy and reliability of earthquake prediction using machine learning.
3 Objectives
The objectives of this research are the following:

1. To compare and evaluate the performance of machine learning models, including
linear, random forest, and decision tree regressors and deep learning architecture,
for earthquake prediction using a large textual dataset.
2. To assess the accuracy and predictive capabilities of these models by analysing
their mean squared error (MSE) and R-squared (R2 ) scores.
3. To provide insights into the strengths and weaknesses of each model in earthquake
prediction tasks based on historical earthquake data spanning from 1965 to 2021,
comprising approximately 23,000 earthquakes.
4 Methodology
Three different classical machine learning models along with an ANN regressor were
trained on the collected dataset.
A dataset consisting of around 23,000 earthquakes recorded by the United States
Geological Survey (USGS). The USGS is an authoritative source and keeps track of
all the earthquakes and other natural calamities that happen around the globe. The
collected data was publicly available and consisted of various features including the
date and time of occurrence of an earthquake, the latitude and longitude position of
the epicentre where the earthquake occurred, the magnitude of the earthquake on the
Richter scale along with a possible percentile error, and the maximum depth of quake
that was caused by a given earthquake in kilometres. The dataset initially consisted
of a total of 21 features [7].
During data pre-processing and analysis, the dataset was explored in detail. Unnec-
essary and empty features including magnitude and depth errors, and latitude and
longitude errors were removed. The frequency distribution of the Richter magnitude
and the depth were visualised and can be seen in Figs. 2 and 3, respectively. It is
worth noting that most of the earthquakes have a Richter magnitude of 5.5 and a
depth of around 1 km. A world map showcasing the earthquakes and their Richter
magnitude was also generated using Python and can be seen in Fig. 1. Time-related
data of the date and the time of the earthquake, which were there in two different
features were combined together to form a single feature vector. This was done
using the Unix time format, which is a widely used date and time representation in
computing. It measures the number of seconds since the 1 January 1970. A combined
feature called “Timestamp” was created. Important features in earthquake prediction
@seismicisolation
@seismicisolation
including the latitude and longitude positions of the earthquake, the Richter magni-
tude, and the maximum depth were also extracted from the original dataset to form
a dataset comprising five features. Timestamp, latitude, and longitude were used as
the input feature vector, whereas the depth and Richer magnitude were the values to
be predicted through regression.
Classical machine learning algorithms such as linear regressor, decision tree
regressor, and random forest regressor were trained on the dataset. Linear regressor is
the simplest of the regression models and helps in understanding the linear relation-
ship between dependent and independent variables in the given dataset. The major
drawback of the linear regressor is that it can’t capture non-linear relationships in the
data and many real-world datasets may not contain linear relationships only. Despite
Fig. 1 A world map showcasing various earthquakes that happened between 1965 and 2021
Fig. 2 Magnitude (Richter)

distribution
@seismicisolation
@seismicisolation
Fig. 3 Depth (mtr)

distribution
this, linear regression is found to be a very effective algorithm in many cases. On

the other hand, decision tree regressor and random forest regressor overcome the
drawback of the linear regression model by capturing non-linear relationships in the
dataset. Random forest regressor uses multiple decision tree regressors and enables
ensemble learning as well.
In order to assess the predictive capabilities of the models used, two widely used
performance metrics, mean squared error (MSE) and R-squared (R2 ) scores, were
used. MSE measures the mean of the squared differences between the actual and
predicted values, whereas R2 score measures the proportion of the variance in actual
and predicted values. A regression model is said to be performing well if it achieves
a lower MSE and a higher R2 score.
These models along with an artificial neural network with three hidden layers
were also trained on the dataset. The artificial neural network can capture more
innate details and patterns in the dataset. The artificial neural network used consisted
of two layers each with 32 neurons and another layer with 16 neurons. These layers
used ReLU as the activation function. The model also had two dropout layers, each
with a dropout rate of 50% that were stacked up after the second and third dense
layers to avoid overfitting. The output layer constituted of 2 neurons for the prediction
of the values. The model was compiled using MSE and Adam as the optimiser. The
model had a total of 1746 parameters and was trained for 1000 epochs using fivefold
stratified cross-validation with a batch size of 32 in the manner specified in Figs. 4
and 5. During the training process, the weights of the lowest validation loss were
stored using model checkpoints which enables reproducibility of the results.
Fig. 4 ANN architecture

using Visualkeras
@seismicisolation
@seismicisolation
Fig. 5 ANN architecture
@seismicisolation
@seismicisolation
The machine learning models also were trained using a fivefold stratified cross-
validation technique which splits the given data into five different folds and during one
iteration, four-folds will be used for training and one-fold will be used for evaluating
the model. This ensures the model’s robustness by using all the data for both training
and testing but, at different iterations.
5 Implementation
All the selected models were trained using a fivefold stratified cross-validation tech-
nique. This was followed by extensive data analysis, feature extraction, and data
pipelining.
The regression models were evaluated using two key metrics: mean squared error
(MSE) and R-squared (R2 ) score. The MSE values for the linear regressor, decision
tree regressor, and random forest regressor were found to be 7453.5, 2922.61, and
1317.19, respectively. Additionally, the R2 scores for these models were determined
to be − 0.0019, − 0.249, and 0.3761, respectively.
The testing R2 scores of the models were evaluated to determine their predictive
capabilities against the testing fold of the dataset. The mean values of R2 scores were
taken following the fivefold cross-validation. For the random forest regressor, the
mean R2 score was found to be 0.3765. The decision tree regressor achieved a mean
R2 score of − 0.2325, while the linear regressor demonstrated a mean R2 score of
− 0.001976. The ANN, with its deep learning architecture, displayed an accuracy
score of 0.8948, indicating its ability to effectively capture the complex relationships
between the input feature vector and the target variables.
The models were also evaluated using the digitisation technique with a threshold
of [5, 70] to find the accuracy scores. Digitisation creates indices of the bins to
which the values belong, almost creating the effect of labelling was used. The accu-
racy scores were 76.1%, 95.5%, 94.5%, and 96.5% for linear regressor, random forest
regressor, decision tree regressor, and the ANN, respectively. These accuracy scores
provide insights into the performance of the regression models, with higher values
indicating better predictive capabilities. The random forest regressor and the ANN
exhibited relatively higher accuracy scores, suggesting their effectiveness in accu-
rately predicting the earthquake magnitudes and depths. The results are tabulated in
Table 1.
Table 1 Detailed results of testing

Model R2 score Accuracy (%) MSE
Linear regression 0.0041 76.1 7453.5
Random forest regression 0.3728 95.5 1317.19
Decision tree regression 0.06 94.5 2922.61
ANN 0.8948 96.5 2046.87
@seismicisolation
@seismicisolation
6 Conclusion
This study majorly focused on evaluating various machine learning models and a
deep learning architecture for regression analysis in order to predict earthquakes.
The study also focused on providing a comparative analysis of the selected models
which were compared based on different evaluation metrics. A large dataset with
the earthquakes recorded between 1965 and 2021 was used to train three classical
machine learning models of the linear regressor, decision tree regressor, and random
forest regressor, all of which exhibited satisfactory performance. The dataset was
also used to train an ANN regressor which exhibited superior performance to all
other models with a prediction accuracy of 96.5%. In terms of R2 score and MSE
as well, the proposed ANN model outperformed all the other models. The findings
of this research help in understanding the applicability of various machine learning
models in predicting earthquakes.
References
1. Asim KM, Martínez-Álvarez F, Basit A et al (2017) Earthquake magnitude prediction in

Hindukush region using machine learning techniques. Nat Hazards 85:471–486. https://doi.
org/10.1007/s11069-016-2579-3
2. Li R, Lu X, Li S, Yang H, Qiu J, Zhang L (2020) DLEP: a deep learning model for earthquake
prediction. In: 2020 international joint conference on neural networks (IJCNN), Glasgow, UK,
2020, pp 1–8. https://doi.org/10.1109/IJCNN48605.2020.9207621
3. Abdul Salam M, Abdelminaam SD (2021) Earthquake prediction using hybrid machine learning
techniques. Int J Adv Comput Sci Appl 12:654–665. https://doi.org/10.14569/IJACSA.2021.012
0578
4. Debnath P, Chittora P, Chakrabarti T, Chakrabarti P, Leonowicz Z, Jasinski M, Gono R, Jasińska E
(2021) Analysis of earthquake forecasting in India using supervised machine learning classifiers.
Sustainability 13(2):971. https://doi.org/10.3390/su13020971
5. Chittora P, Chakrabarti T, Debnath P et al (2022) Experimental analysis of earthquake prediction
using machine learning classifiers, curve fitting, and neural modeling, 01 Sept 2022. PREPRINT
(Version 2) available at Research Square. https://doi.org/10.21203/rs.3.rs-1896823/v2
6. Xiong P, Tong L, Zhang K, Shen X, Battiston R, Ouzounov D, Iuppa R, Crookes D, Long C,
Zhou H (2021) Towards advancing the earthquake forecasting by machine learning of satellite
data. Sci Total Environ 771:145256. ISSN: 0048-9697. https://doi.org/10.1016/j.scitotenv.2021.
145256
7. Latest Earthquakes by the United States Geological Survey. https://earthquake.usgs.gov/earthq
uakes/map/?extent=5.7909,-147.39258&extent=60.97311,-42.62695
@seismicisolation
@seismicisolation
Assessment and Planning of Solid Waste
Management System for the Ichamati
Riverside Area of Pabna City
Ashik Iqbal , Imtiaz Ahmed Emu, Fahreen Hossain, Sirajum Monira Popy,
Rashed Uz Zzaman, and Md. Helal Ahmmed
1 Introduction
Bangladesh is a land of rivers, and Jamuna, Padma, Ichamati, Baral, and Atrai rivers
flow through the Pabna district [1]. Ichamati River originating from the Ganges River
flowing through Pabna city, Atghoria, Santhia, and Bera falls into the Hurasagar River
and finally makes its way to Jamuna. The 10 km portion of the 44 km long Ichamati
River flows through the city area [2]. The maximum width of Ichamati is 140 m with
an average width of 120 m, but the width is 30 m in the Pabna city. The river in
the city area has turned into a narrow canal in the last three decades. Once mighty
Ichamati River in Pabna is now lifeless. The dumping of domestic and commercial
waste is one important cause of the disconnection and disruption of flow of the river.
The population has been increasing at a very steep rate in Pabna for last two
decades due to urbanization and industrialization [3]. The increasing number of
people is generating huge amount of solid wastes in the city [4]. Due to severe
constraints, absence of appropriate technology, lack of awareness, and ineffective
legislation, the entire system is becoming a threat to the city [5, 6]. To ensure a
clean, hygienic, and environmental-friendly city, the appropriate management of
solid waste is essential [7–9]. In Pabna Pourashava, almost 52 tons of waste are
generated every day, and 31 tons of waste are collected by the municipality [10, 11].
Pabna municipality is the sole organization in sustainable waste management of the
A. Iqbal (B) · F. Hossain · S. M. Popy · R. U. Zzaman · Md. H. Ahmmed

Institute of Water and Flood Management, Bangladesh University of Engineering and
Technology, Dhaka, Bangladesh
e-mail: iqbal.ashik.ce@gmail.com
I. A. Emu
Local Government Engineering Department, Dhaka, Bangladesh
Md. H. Ahmmed
Harold Hamm School of Geology and Geological Engineering, University of North Dakota,
Grand Forks, USA
@seismicisolation
@seismicisolation
246 A. Iqbal et al.
area. However, due to low efficiency, much raw, untreated waste is being dumped
in the riverside area or in rivers [12, 13]. It is also causing riverbed siltation and
waterlogging in nearby areas. Sustainable solid waste management is a concept that
emphasizes the need to consider the long-term future as well as the present [14, 15].
Riverside community dumps their waste in the Ichamati River and along the bank
which results in clogging the river and so the water flow has been blocked. Even if the
government takes the initiative to revive the river and ensure the water flows, it will
be challenging to resuscitate the urban segment due to the amount of waste deposited
every day. Excavation is required to remove the waste; however, without an adequate
solid waste management system, the riverbed will end up being clogged again in the
future. Few pieces of literature investigated waste generation and management issues
in Pabna city, but no work assessed the impact of waste dumping in the Ichamoti
River and associated solutions. In the research gap, the objective of the study is to
explore the present solid waste dumping scenarios along the Ichamati riverbank and
to propose a waste management plan that will be sustainable to adopt the free flow
in the river.
2 Materials and Methodology
2.1 Study Area
Pabna city area is an “A” Category Pourashava administration since 1989. It is situ-
ated 161 km North-West of Dhaka and 110 km East of Rajshahi. Being a flood-
plain the lands in the area are very fertile, and so agriculture and farming are the
main occupation of the peri-urban people [11]. The city people are involved in busi-
ness, government service, and industry sector [11]. The percentage of occupations
people practicing are agriculture 34%, farming 22.77%, day laborer 4.46%, trans-
port workers 2.18%, weavers 2.85%, commerce 13.27%, service 7.26%, and others
13.21% [11]. The location of Pabna city in Bangladesh is shown in Fig. 1a, and the
Ichamati River in the city area is shown in Fig. 1b. For this study, three distinct areas
have been selected which are marked in the figure below. The SP Bungalow Mor
area (marked as 1) in Wards 1&2, Police Line Bridge (marked as 2) in Wards 7&10,
and Rupkotha Bridge (marked as 3) in Wards 9&10 were selected for the study.
2.2 Methods
Data collection
A community-based participatory approach and tool were used to identify the present
condition of solid waste generation, dumping, and management. The socio-technical
approach along with GIS toll was used to plan a sustainable and community-friendly
@seismicisolation
@seismicisolation
Assessment and Planning of Solid Waste Management System … 247
Fig. 1 Study area map a Pabna city and b Ichamati River with study locations
solid waste management plan. The methodological framework comprises field survey
and participatory rural appraisal (PRA) tools, i.e., focus group discussion (FGD) and
key informant interview (KII), and then making a solution for solid waste manage-
ment system based on the field findings [16]. It was a four-day field trip, where day
1 consisted of reconnaissance surveys and field investigations. The other three days
were used for the questionnaire survey, FGD, and KII.
For the quantitative data, a questionnaire survey was adopted, and the solid waste
generation and management data were collected from the KII with the municipality.
The qualitative data was collected from KII and FGD with local people, government
officials, and councilors. To collect the data, random surveys were conducted with an
open-ended questionnaire as indicated in Table 1. At the beginning, a reconnaissance
survey was done to understand the various issues in the field, and then with four
days of in-depth field survey, the data was collected. The focal group discussion
(FGD) was done with the people living surrounding the river about the present waste
disposal and collection situation. The key informant interview (KII) was done as an
in-depth interview with commercial and residential persons shown in Fig. 2. Four
FGDs were done with the workers, women, housewives, and elderly people. A total
of 84 questionnaire surveys were conducted. Three KII was conducted with the waste
collector, ward councilor, and municipal authority. The FGD and interview persons
were selected who are related to waste dumping.
Solid Waste Management Planning
The field observation (KII, FGD, and survey) brought out the gaps and associ-
ated problems, and then possible sustainable solutions have been identified. In this
socio-technical approach, the social findings and consultations were integrated with
technical planning to make it more community-friendly and sustainable. Also, the
suitability analysis for solid waste management and a waste dumping station was
proposed considering the biodiversity of the Ichamati River.
@seismicisolation
@seismicisolation
248 A. Iqbal et al.
Table 1 Data collection and purpose

Data collection method Number Purpose and respondent
Field reconnaissance 3 Field reconnaissance was conducted at SP
Bungalow Mor, Police Line Bridge, and
Rupkotha Bridge area to identify the potential
areas which were impacted due to waste
disposal. Different groups were also identified
who were associated with waste generation,
disposal, and management
Questioner survey 84 The questionnaire survey focused on the
generation, collection, and disposal of waste.
The use and availability of the dumping bins
were investigated. An open-ended questionnaire
was used
Focal group discussions (FGD) 4 FGDs were conducted to identify the people’s
perception of waste dumping and management
and the impacts of waste in the Ichamati River.
Also, information on point sources, non-point
sources, disposal sites, and transportation of
waste was collected from people and the waste
collector
Key informant interviews (KII) 4 The waste collection, manpower data, dumping
system, and provision of services of the
municipality were investigated in the KII. The
liabilities and responsibilities of the
municipality in waste management were
discussed. The possible solid waste
management system was also discussed to
incorporate the government’s perception of the
planning
Suitability analysis GIS-based process used to determine the appropriateness of

solid waste dumping in riverside following a multi-factor analysis guideline [17,
18]. The buffering process is achieved through the reclassification of regions from
the rivers, where parameters are the hydrological network, roads, railways, and the
central business district [18, 19]. After a thorough analysis of the suitable sites, the
main dumping site for the Pabna municipality was chosen. The layers of previously
buffered parameters were superimposed, and a map clearly showed suitable areas for
locating the dump site. The plan also suggested the role of different organizations in
integrated management.
@seismicisolation
@seismicisolation
Fig. 2 Field survey and FGD in the Ichamati riverside community of Pabna
3.1 Existing Solid Waste Dumping and Management

Condition
Two types of waste collection systems are in Pabna city. In a few areas, waste collec-
tors visit houses and collect directly. And in some areas, some bins are installed by the
municipality, people are expected to dump waste there, and then, the municipality
will collect it from those bins. Some bins are installed in the Ichamati riverside;
however, people do not utilize them properly instead they dump them in the river
brink near their houses. Much of the river in the city area is full of waste and so no
water flows (Fig. 3). Permanent disposal of solid waste is in Fakirpur having 3 Acre
Khash land located 4.5 km from the river.
The type of waste dumped in Ichamati River is mostly organic. As S.P Bungalow
is a residential area and the Police Line is developing a commercial area, most of
the waste dumped is organic (food, garden and lawn clippings, animal and plant-
based material, paper, cardboard, and timber). Rupkotha Bridge area being a highly
commercial area, the organic waste is 57.1%, and the other 42.9% is plastic waste.
The social survey focused on six important issues, (i) where the residents dump
their waste, (ii) whether they used designated garbage disposal containers, open
spaces, or any spot near their home, (iii) if there are available bins, (iv) do they
@seismicisolation
@seismicisolation
250 A. Iqbal et al.
Fig. 3 Existing solid waste dump scenario in Ichamati River and riverside, and municipality waste
collection van
dump alongside them or elsewhere, and (v) whether the bins are too far or close to
their house? This information illustrated in Fig. 4 provided an understanding of their
waste disposal behavior and their perception of waste dumping. There is an existing
door-to-door municipal waste collection mechanism, but some areas were neglected
by this service.
The residents of S.P. Bungalow area dump their waste in a container nearby, and
their dumping bin is not very far (Fig. 4a and c). However, the municipality does not
collect waste from the area. Rupkotha Bridge is a highly commercialized area, and
the municipality regularly collects waste from this area, so they do not dispose of
their waste elsewhere. Police Line is a new commercial area, and there is not much
support from the government. The municipality does not collect waste from there on
a regular basis, their waste dumping bins are very far from their location, and hence,
they dispose of their waste in open spaces. In all three areas, residents prioritize
littering and its effects on human health as priority concerns of waste, but not much
regarding the effect on the environment and so they never consider Ichamati River
to save from waste.
According to Municipal Authority, there are a total of 320 dustbins. Among them
120 are concrete bins of 60 kg capacity and 200 plastic bins of 20 kg capacity. Among
them, many plastic bins are not functional, and as a result, the inhabitants dump their
waste in open spaces nearby the bins which finally end up in the river. The location
of the waste bins of the municipality and the point and non-point sources of waste
in the Ichamati riverside area is shown in Fig. 5. There are several point and non-
point sources of solid waste in the riverside area, but there lacks the bin in those
location. Though most of them are in the Rupkotha Bridge area and S.P Bungalow
area but the non-point sources are also significant in number. These wastes caused the
riverbed elevated decreasing the water carrying capacity of the river, thus resulting
waterlogging in surrounding areas in monsoon.
@seismicisolation
@seismicisolation
Fig. 4 People’s perception of the existing waste dumping practice
3.2 Discussion of the Findings
In critical analysis of the information collected, the lack of dustbins, unwilling-

ness of community, improper collection system, and monthly subscription for waste
collection through the municipality are found the main reason for the dumping of
waste in the riverbed. These reasons were also identified by Saha [12]. Whereas the
shortcomings in management systems of the municipality are lack of manpower,
shortage of fund allocation, lack of monitoring of the existing system, unwillingness
which echoed the finding of Atauzzaman et al. [10]. It has become a vicious chain,
where the municipality fails to collect the waste and the waste finds it resting place
at the bottom of the river which ultimately causes the clogging of the river. Another
@seismicisolation
@seismicisolation
252 A. Iqbal et al.
Fig. 5 a Existing waste bins of Pabna municipality and b solid waste dumping points in the riverside
area
interesting finding is that the garbage collectors are unwilling to collect waste from
the poorer neighborhoods. Because there is no proper payment system established,
the households would pay according to their affordability. The well-off households
would pay more, so the garbage collectors would rather collect their waste, and not
of the poor. With no other options, the poorer households would dump their waste
in or along the river.
4 Proposed Solid Waste Management System
The reduce, reuse, and recycle (3R) is a very excellent option for Pabna as practiced
by some poor families. If there is livestock in any family, the organic waste is used
as food for the cows and goats. As the maximum waste generation of Police Line
Bridge is organic waste (as food waste and related) so the waste can be used for
feeding the waste to livestock (such as cows, and goats) because of the availability
of livestock and food waste as well in the area [20]. The municipality can train other
people to reuse and recycle their wastes at home, and the rest can be collected by the
municipality.
The dustbin of the municipality in the riverside needs to be converted from plastics
to concrete as they are firm and provide a better space. People also preferred one
dustbin centered for 15–20 households so that people do not feel the bins far away
from the houses. Solvent families mostly preferred a regular door-to-door collection
system with the involvement of the Ward Councilor as the responsible monitoring
person. Concrete dustbins are recommended because of their stability instead of
@seismicisolation
@seismicisolation
movable plastic dustbins which already failed. Regular collection of waste should
be ensured by the municipality. The municipality should ensure a regular door-
to-door waste collection system in the Rupkotha Bridge area. Separate dustbins are
required for separate waste (e.g., organic waste, plastic waste, etc.). The municipality
can undertake a public–private-partnership (PPP) strategy for door-to-door waste
collection where private organizations can make a profit and the municipality can
monitor the quality of service.
The waste collection cost should be revised with different rates for different
income houses, but it should be fixed. The cost should not be more than BDT 30 for
poorer families in Police Line Bridge and SP Bungalow, BDT 50 can be for low-
income people in Rupkotha Bridge, and the wage can be 80–100 taka for wealthier
families.
4.1 Suitability Analysis and Proposed Landfill Station
The buffer and suitability analysis of waste dumping stations is proposed considering
the Convention on Biological Diversity (CBD). Not only the clogged riverbed but
also the biodiversity of the river needs to be considered while planning to revive the
river for the future [21, 22]. Considering the center of the river in the Police Line
Bridge the protected area buffer map is generated (Fig. 6a). The first ring denotes an
area with a radius of 0.5 km, outside of which is moderately suitable for the dumping
bins, and in second circle, it is 1 km, and dumping bins outside of the protected land
is suitable. Anthropological interaction with the river except fishing in this 0.5 km
radius and development activities in the 1 km radius area needs to be controlled for
sustainable river management.
The existing dumping station has 7.02 acres of land requited for landfill; however,
it is not enough for the 100 metric tons of garbage generated per day. The Khash
Fig. 6 a Protected area for the biological diversity, b suggested dumping locations and shortest
routes for solid waste collection
@seismicisolation
@seismicisolation
254 A. Iqbal et al.
Land near Meril bypass outside of the Ichamati River buffer zone is suggested for
the dumping station. Short route selection from the waste collection points to the
dumping station is conducted by using GIS software and illustrated in Fig. 6b. This
shortest path can be transported to dump the collected wastes.
4.2 Organizational Role for Integrated Management
The integration and communication among the stakeholders are essential to uptake
the responsibility for the sustainable solid waste management system. The role of
the government, private sector, and the community is identified and proposed for
sustainable management.
Government must uptake the role of improving existing guidelines, rules, and
regulations and most importantly ensuring the execution of the rules in Pabna city
[20]. Provide facilities and opportunities to local authorities to develop dustbin infras-
tructure. Public understanding can be increased on the importance of waste manage-
ment for sustainable rivers by government’s awareness programs. The private sector’s
role is the engagement and investment in 3R-related projects, providing 3R-related
training to the local people. The government must be the pioneer in establishing a
useful PPP system for waste collection where the private sector will be the executor.
Community’s role is to practice reducing and recycling waste, formation of aware-
ness groups and making everyone responsible for the protection of the river as they
are the direct beneficiary of the river.
A committee can be formed, with the above-mentioned representatives who will
oblige to their designated duties. This process must include stakeholders who would
be relevant in long-term sustainable development, such as waste producers, waste
management experts, councilors, and environmental scientists who can provide
training on the importance of resuscitating the river.
5 Conclusion
A systematic effort is necessary to improve the factors of waste generation, dumping,

collection, and reuse including policy and legal frameworks, institutional arrange-
ments, financial provisions, and awareness. The major issues in this area are inad-
equate waste management, disposal, and collection system and a lack of awareness
of the public. The reduce, reuse, and recycle (3R) is an excellent option in Pabna
as practiced by some poor families. To overcome the waste management issues,
this study suggests a low-cost regular from door-to-door waste collection system
by the Pabna municipality by PPP, reducing costs for low-income people, and relo-
cating the dumping station from the riverbank. Formulating an independent solid
waste management plan will not bring the change needed, however, an integrated
system involving the local inhabitants and their willingness to bring about a change
@seismicisolation
@seismicisolation
in behavior is necessary to save the river. In future studies, the importance of differen-
tiating types of waste can be included. As a way forward, future work can investigate
the efficient separation and disposal mechanism for the chemical composition of the
wastes in the landfill.
Acknowledgements Alhamdulillahi Rabbil Alamin. We praise and glorify Allah Subhanahu wa

taala, as he is ought to be praised and glorified. This research was funded by “Interdisciplinary Field
Research Methodology in Water Management” course of IWFM, BUET. Authors like to thank the
“Feasibility Study for Resuscitation of Ichamati River in Pabna” project team of IWFM, BUET and
to BWDB, LGED, DAE officials, and UNO of Pabna.
References
1. Iqbal A, Mondal MS, Khan MSA, Hakvoort H, Veerbeek W (2022) Flood propagation processes
in the Jamuna River floodplain in Sirajganj. In: Water management: a view from multidisci-
plinary perspectives. Springer, Switzerland, pp 45–67. https://doi.org/10.1007/978-3-030-957
22-3
2. Alamgir M, Bidlingmaier W, Glawe U, Martens J, Sharif LA, Visvanathan C, Stepniewski W
(2007) Safe and sustainable management of municipal solid waste in Khulna city of Bangladesh.
In: Sardinia 2007. Eleventh international waste management and landfill symposium, S.
Margherita di Pula, Cagliari, Italy, 1–5 Oct 2007
3. Rafi F, Iqbal A, Islam R, Sheonty S, Armin R (2021) Drinking-water quality assessment of
public tube-wells and their spatial distribution in the Rangpur city of Bangladesh. In: Climate
change and water security. Lectures notes in civil engineering, vol 178. Springer, Switzerland.
https://doi.org/10.1007/978-981-16-5501-2_28
4. Ahmed MF, Rahman MM (2000) Water supply & sanitation: rural and low-income urban
communities. ITN-Bangladesh, Center for Water Supply and Waste Management, BUET,
Dhaka, Bangladesh
5. Alam O, Qiao X (2020) An in-depth review on municipal solid waste management, treatment
and disposal in Bangladesh. Sustain Cities Soc 52:101775. https://doi.org/10.1016/j.scs.2019.
101775
6. Bari QH, Hasan KM, Haque ME (2012) Solid waste recycling in Rajsahi city of Bangladesh.
Waste Manag 32(2012):2029–2036. https://doi.org/10.1016/j.wasman.2012.05.036
7. Parvez M, Sadat N, Tasnim F, Nejhum IJ (2021) Identifying the causes of waterlogging
on people’s perception towards a resilient community: a case study on Pabna Municipality,
Bangladesh. Ecofemin Clim Change 2(3):110–126. https://doi.org/10.1108/EFCC-11-2020-
0033
8. Alamgir M, Ahsan A, Bari QH, Upreti BN, Bhatttari TN, Glawe U, Visavanathan C, Sthapit
L (2005) Present scenario of municipal solid waste and its management. In: Alamgir M,
McDonald C, Roehl KE, Ahsan M (eds) Integrated management & safe disposal of MSW
in LDACs: a feasibility study. KUET, Bangladesh, pp 135–228
9. Loucks DP (2000) Sustainable water resources management. Water Int 25(1):3–10. https://doi.
org/10.1080/02508060008686793
10. Atauzzaman M, Khatun MT, Hossain S (2020) Solid waste management system in Pabna
Municipality of Bangladesh: a case study. Recent Trends Civ Eng Technol 10(10). https://doi.
org/10.37591/rtcet.v10i1.3554
11. BBS (2011) Bangladesh population census 2011. National Report, Bangladesh Bureau of
Statistics, Government of Bangladesh, Pabna
12. Saha AK (2013) Environmental aspects of solid waste management: a case study of Pabna city
in Bangladesh. IJASETR 2(2):1–8
@seismicisolation
@seismicisolation
256 A. Iqbal et al.
13. Shekdar AV Tanaka M (2004) Recycling of solid waste in some Asian countries. In: Proceedings
of global symposium on recycling, waste treatment and clean technology, at Madrid, Spain,
26–29 Sept 2004
14. World Commission on Environment and Development (WCED) (1987) Our common future.
The Brundtland Report, Oxford University Press, Oxford, UK, p 383
15. Iqbal A, Swapnil FJ (2017) Sustainable green materials for building construction. In: Interna-
tional conference on mechanical, industrial and materials engineering (ICMIME2017) 28–30
Dec 2017, RUET, Rajshahi, Bangladesh
16. Iqbal A, Hasan MH, Rasheduzzman M, Paul SR, Hamid R (2023) Environmental and socio-
economic impact assessment in Hazaribag area of Dhaka for tannery relocation. Int J Soc Sci
Econ Rev 5(1):29–39. https://doi.org/10.36923/ijsser.v5i1.184
17. Jerie S, Zulu S (2017) Site suitability analysis for solid waste landfill site location using
geographic information systems and remote sensing: a case study of Banket Town Board,
Zimbabwe. Rev Soc Sci 2:19–31. https://doi.org/10.18533/rss.v2i4.99
18. Lunkapis GJ (2010) GIS as decision support tool for landfills sitting. University Malaysia
Sabah, Locked Bed, 2073, 88999. http://www.gisdevelopment.net/application/urban/products/
ma04133pf.html
19. Pawandiwa CC (2013) Municipal solid waste disposal site selection: the case of Harare.
University of the Free State, Bloemfontein. http://hdl.handle.net/11660/1566
20. Mamun AHMM (2014) An approach towards sustainable city environment: critical analysis of
municipal solid waste management of Pabna municipality of Bangladesh from the perspective
of resource recovery and environmental conservation. Sci Res J (SCIRJ) 10(12):33–37
21. Iqbal A, Mondal MS, Veerbeek W, Khan MS, Hakvoort H (2023) Effectiveness of UAV-based
DTM and satellite-based DEMs for local-level flood modeling in Jamuna floodplain. J Flood
Risk Manag e12937. https://doi.org/10.1111/jfr3.12937
22. Iqbal A, Jahan I, Wasiew QA, Emu IA, Chowdhury D (2023) From existing conventional
building towards LEED certified green building: case study in Bangladesh. Front Built Environ
9:1194636. https://doi.org/10.3389/fbuil.2023.1194636
@seismicisolation
@seismicisolation
Flood Inundation Mapping and Flood
Intimation Using IoT for KRS Dam
D. Ashwini, R. Revuprasad, R. G. Ranjith Kumar,

C. S. Mallikarjun Swamy, and G. P. Bharath
1 Introduction
The most common natural disaster that occurs frequently is the flood. By far,
accounting for 67% of all events were recorded. India observes flood due to exor-
bitant downpour which then, at that point, brings about flood of waterways, lakes
and dams, making irremediable harm individuals’ business, abundance and essential
framework. Out of the total 329 million hectares (mha), more than 40 mha are annu-
ally prone to flooding due to a combination of internal factors, such as river density
and topographic relief, as well as the precipitation that caused the flood.
The storage structure that is built across the river to store water in the upstream
and release it to the downstream when demand arises is called a dam, and it is critical
to the country’s economy. The majority of applications for the retained water include
aquaculture, industries, human consumption, and irrigation. The flood caused by the
dam breaking is much higher than the flood caused by heavy precipitation [1].
A dam can fail for a variety of reasons, the most common of which are seepage
failure, overtopping failure, foundation failure, construction failure, and material
failure. As a result, it is necessary to conduct dam break analysis to study the flood
wave’s behavior and prepare inundation maps as part of an emergency action plan to
evacuate the population during a flooding event.
By employing 2D approach modeling, the U.S. Army Corps of Engineers’ HEC-
RAS (Hydrologic Engineering Center River Analysis System) is used in this paper
to conduct a comprehensive analysis of the Krishnarajasagara dam in order to study
the effects on downstream areas and create inundation maps in the event that the dam
breaches as a result of overtopping. This kind of analysis of dam failures and their
inundation maps would help the public get enough warnings before the catastrophic
event.
D. Ashwini (B) · R. Revuprasad · R. G. Ranjith Kumar · C. S. Mallikarjun Swamy · G. P. Bharath

Department of Civil Engineering, MIT Mysore, Karnataka, India
e-mail: ashusujadhan@gmail.com
@seismicisolation
@seismicisolation
258 D. Ashwini et al.
The Internet of things (IoT) depicts actual articles (or then again gatherings of such
articles) with sensors, handling capacity, programming and different advancements
that associate and trade information with different gadgets and frameworks over the
Web or different interchanges organizations. Due to the fact that each device only
requires a network connection and the ability to be individually addressable, the term
“Internet of things” has been regarded as a fallacy [2–4].
The field has developed because of the combination of various advances, including
universal registering, ware sensors, progressively strong installed frameworks, as
well as machine learning [5]. The Internet of things is made possible by tradi-
tional fields like embedded systems, wireless sensor networks, control systems, and
automation (including home and building automation) [6]. In the purchaser market,
IoT innovation is generally inseparable from items relating to the concept of the
“brilliant home”, including gadgets and machines (like lighting apparatuses, indoor
regulators, home security frameworks, cameras, and other home machines) that help
one or more normal biological systems and can be controlled through gadgets related
with that biological system, for example, cell phones and shrewd speakers.
With the use of hydraulic model, HEC-RAS in analyzing the dam breach of
high-hazard potential dam—Krishnarajasagara located at Mandya [1, 7], dam break
simulation and analysis based on given geometry data. The analysis result shows the
maximum water surface at each river station after the dam breaks were animated
with the help of Arc Map. The proper analysis of the hazards associated with dam
failure will assist in land use planning and in developing emergency response plans
to help mitigate catastrophic loss to human life and property [8, 9]. The flood profiles
for different flood intensities with different return periods can be plotted at any given
cross-section of river. Also, such flood profile can be plotted for entire length of river
reach [10]. It can be suggested that necessary and urgent solutions must be provided
for safety of lives and properties [11].
In the present study, the flood plain zone, at the downstream finish of KRISH-
NARAJASAGARA Repository close to Mandya, has been planned, and involving
DEM in Bend GIS in the wake of choosing yield area in Smack has been carried out.
2 Study Area and Data Used
The Krishnarajasagara dam, also known as the Kannambadi katte, is a barrier built
across the river Kaveri, also known as the “Dakshina Ganga” (Ganges of the south). It
saves the life of the southern part of Karnataka. With the assistance of renowned chief
architect of the dam, Sir M. Visvesvaraya, the dam was constructed around 1924 to
initially meet the requirements of the Mysore and Mandya region for irrigation and
drinking water supply.
The KRS dam is a type of gravity dam made of stone masonry and surki mortar.
It has a length of 2621 m, a maximum height of 44.66 m between the top of the
dam and the foundation in the lowest section of the river bed, and a bottom width
of 33.88 m at this level, covering 10,961 km2 . The dam has a total storage capacity
@seismicisolation
@seismicisolation
Flood Inundation Mapping and Flood Intimation Using IoT for KRS Dam 259
Fig. 1 Location map of the study area
of 1400 MCum. The estimated design flood discharge is 9911 cumecs. In order to
empty the KRS dam of excess water and water that has been stored there, 152 gates
have been installed at various dam levels with varying capacities. Figure 1 shows the
location selected for the research [1].
The following data are utilized for the purpose of analysis in the current study:
• SRTM—30 m-resolution DEM from NASA and the USGS (https://dwtkns.com/
srtm30m).
• From the dam authority also the characteristics of the dam and river in relation
to topography, such as the full supply level, catchment area, and design flood,
among other things were collected.
• The HEC-RAS programming created by the US Armed force corps of architects.
3 Methodology
A water-driven model created by U.S Armed force Corps of Engineers, HEC-

RAS programming has been utilized to picture the flood immersion maps [12]. For
studying flood inundation maps and dam break analysis, numerous free and paid
models are currently on the market. They differ in terms of their user interface,
input information requirements, accuracy, and other characteristics. However, for
best outpouring computation and flood inundation mapping, all models use the same
or comparable calculations.
@seismicisolation
@seismicisolation
Fig. 2 Workflow process adopted using HEC-RAS
The current work comprises different moves toward creating required immersion
maps. The HEC-RAS software work flow shown in Fig. 2 was used to accomplish
the necessary goal [1].
4 HEC-RAS Model Setup and Execution
4.1 Background
The Hydrologic Engineering Center (HEC) of the U.S. Army Corps of Engi-
neers created the hydraulic model known as HEC-RAS [12]. HEC-RAS is a
broadly acknowledged instrument for the floodplain assurance because of its easy
to use believability additionally helps with performing different other activities, for
example, dam break examination, water quality examination, residue transport inves-
tigation, and so forth. 2-D hydrodynamic flow routing has been made possible by
recent developments in HEC-RAS in the unsteady flow analysis [7, 9, 12, 13].
@seismicisolation
@seismicisolation
Fig. 3 Geometric file
4.2 Geometric File and Hydraulic Input
A water-powered design, for example, dam alongside the channel is made by giving
the required water driven inputs. The absolute number of doors gave for this situation
represents 152 numbers. There are four groups of gates, each with a different height
and dimension. Figure 3 shows the geometric details of upstream and downstream
of KRS dam.
4.3 Boundary Conditions
In order to carry out the calculations required to run the model under unsteady flow
simulation, upstream (U/S) and downstream (D/S) boundary conditions are required.
The flow hydrograph was used as the analysis’s upstream boundary condition in
this study, while the normal depth of 0.001 was used as the downstream boundary
condition. The HEC-RAS software is used to study the flow hydrograph. Figure 4
shows the flow hydrograph.
@seismicisolation
@seismicisolation
Fig. 4 Flow hydrograph
4.4 Unsteady Flow Simulation
The unsteady flow simulation is carried out after the necessary boundary condition
has been established, as well as the plan and geometry data files. For the current
review, hourly release information was utilized. The 2D flow was simulated using
the diffusion wave equation after a 1-h warm-up period. The hydrograph and mapping
output interval was set to one hour, and the chosen computational interval was 1 min.
4.5 Implementation of IoT
The term “Internet of things” (IoT) refers to physical objects—or groups of such
objects—equipped with software, processing power, sensors, and other technologies
for transferring data to and from other systems and devices via the Internet or other
communications networks. The reality is viable flood cautioning frameworks and
other water level checking frameworks incorporate a mix of IoT sensor innovations.
4.6 Different Level Considered for Floods
High-frequency radar and electromagnetic waves are directed from radar sensors to
the substance being monitored, like water, by means of probes. Here, level readings
@seismicisolation
@seismicisolation
are provided by radar level sensors based on how long it takes the radar pulse to
return after it has been sent.
4.7 Different Level Considered for Floods
Arduino software has been used to write the program for flood alert message. Imple-
menting IoT-based flood alert systems may face challenges such as the cost of sensor
deployment and maintenance, ensuring network connectivity in remote areas, and
the need for data privacy and security measures to protect sensitive information.
Figure 5 shows the demonstration model.
Flood hazard management, peril moderation, and the preparation of emergency action
plans (EAPs) all depend on the creation of flood maps. These will, in turn, assist the
dam authorities in developing the appropriate basic framework, which aids areas
with vulnerable populations that may require protective measures during a disaster.
Fig. 5 Demonstration model
@seismicisolation
@seismicisolation
Fig. 6 Depth map
These plans should be made before the occasion of an event, considering most cynical
situation circumstances.
The HEC-RAS software’s dam flood intimation modeling due to overtopping
failure produced a number of inundation maps, including depth map, velocity map,
water surface elevation map, and flood arrival time map, all of which aid in assessing
the risk of flooding. Figures 6 and 7 depict the maps of the inundation caused by
overtopping failure in relation to the aforementioned parameters.
It is important to note that the velocity of water in a river is not uniform across
the entire cross-section. The velocity is typically highest near the surface and along
the central part of the channel, while it is slower near the riverbed and banks.
5.1 Flood Alert Using IoT
The components of a typical IoT-based flood alert system may include sensors,
communication network (IoT devices are connected through wireless communi-
cation networks such as Wi-Fi, cellular networks, or dedicated IoT protocols like
LoRaWAN), data processing and analysis (the collected data is processed and
analyzed using algorithms and machine learning techniques), alert generation and
visualization and reporting.
Using Arduino code, the sensor sends the flood alert message.
@seismicisolation
@seismicisolation
Fig. 7 Velocity map
Out of the entire length of the Arduino code a part is shown in Fig. 8, and Fig. 9
shows the flood alert received by the device stringed.
Fig. 8 Part of the code
Fig. 9 Flood alert received
@seismicisolation
@seismicisolation
6 Conclusions
Dam failure is unavoidable, but its management with appropriate measures is certain
to reduce fatalities and property damage. The use of the hydraulic model HEC-RAS
in the analysis of the dam breach at the high-hazard potential dam Krishnarajasagara
in Mandya is one such measure that was demonstrated in this paper. Because of its
immense stockpiling limit and huge home in the downstream piece of the dam, it
was important to direct the exploration and gauge the conceivable immersion due
to overtopping disappointment which would not be concentrated on exhaustively in
this undertaking.
The current study demonstrates that the HEC-RAS model can be used to create
flood inundation maps, allowing the dam authority, owners, stakeholders, engineers,
and others to take appropriate measures. From the analysis on the Google map, it
was seen that more than 40 villages would tend to be affected, Narasipura was found
to be the last point due to time constraints as per the study. By the time the message
reaches the dam authority, it would have also reached those in the government and
the people suspected to be affected so that they can all take immediate action based
on the overlay analysis on the Google Maps constraint. The study on the dam break
contributes to the improvement of commuters’ resilience to such a catastrophic event
as well as the reduction of potential impacts through rapid disaster recovery.
In conclusion, IoT-based flood alert systems have the potential to revolutionize
flood warning and response capabilities. By leveraging real-time data, advanced
analytics, and seamless communication, these systems can enhance preparedness,
minimize damage, and save lives in flood-prone areas.
References
1. Madhusudhan MS, Vandana P, Tahir Arif Peer Manoj G (2020) Dam analysis of Krish-
narajasagara Dam due to overtopping failure by HES-RAS. Int Res J Eng Technol (IRJET)
7(8)
2. Oussama D, Djemili L, Bouchehed H, Tachi SE (2015) A GIS based approach for the prediction
of the dam break flood hazard—a case study of Zarzedas reservoir Skidda, Algeria. J Water
Land Dev 27:15–20
3. USACE (2002) HEC-RAS river analysis system, hydraulic reference manual. U.S Army Corps
of Engineers (USACE). Hydrological Engineering Center, Davis, CA. Accessed 7 Jul 2020
4. Xiong Y (2011) A dam break analysis using HEC-RAS. J Water Resour Prot 3(6):370–379
5. Yochum SE, Goertz LA, Jones PH (2008) Case study of the big bay dam failure: accuracy and
comparison of breach predictions. J Hydraul Eng ASCE 134(9):1285–1293
6. Using HEC-RAS for Dam Break Studies, TD-39, US Army Corps of Engineers, Institute for
Water Resources, Hydrologic Engineering Centre, 609 Second Street, Davis, CA 95616, USA,
p 62
7. Subrahmanyam VSR, Nikhitha A, Sai Prakash Ch, Ramya S (2019) Dependable flow and flood
control of water dam. Int J Eng Res Technol (IJERT) 6 (12)
8. Razack R (2014) Dam break analysis using GIS application. Int J Eng Res Technol (IJERT)
3(3)
@seismicisolation
@seismicisolation
9. Guidelines for Mapping Flood Risks Associated with Dams. Dam Rehabilitation and
Improvement Project (DRIP), CWC (2018)
10. Hajeri SS, Shivapur AV, Venkatesh B (2016) Flood plain mapping and dam break analysis for
Neerasagar reservoir. Int J Eng Res Technol (IJERT) 3(6)
11. Lukmam S, Otum JA, Adie DB, Ismail A, Oke IA (2011) A brief assessment of a dam and its
failure prevention. J Fail Anal Prev 11(2)
12. 2D Modelling User Manual. HEC-RAS River Analysis System. U.S Army Corps of Engineers,
Hydrologic Engineering Centre, Version 5.0 (2016)
13. Asnaashari A, Meredith D, Scruton M (2014) Dam breach inundation analysis using HEC-
RAS and GIS—two case studies in British Columbia, Canada. In: Canadian Dam Association
Annual Conference, Banff, Alberta, Canada, Oct, pp 1–8
14. Hicks F, Peacock T (2005) Suitability of HEC-RAS for flood forecasting. Can Water Resour J
30(2):159–174
@seismicisolation
@seismicisolation
Monitoring Climate Hazards, Rice
Production Risks and Management
Practices in Bharathapuzha River Basin
(BRB), Palakkad, Kerala
P. Dhanya and K. Jayarajan
1 Introduction
Humans have realized that climate change is not a future matter of concern. It
is impacting all walks of life especially the sensitive economic sectors. Agricul-
ture ecosystems face high economic and non-economic losses from disasters and
sneaking climatic changes and variabilities. Climatic change, especially the more
evident extremities, could affect cropping systems in several ways such as quantity
and quality of crops, productivity, photosynthesis and transpiration rates, moisture
availability, etc. [1]. Due to growing demand from the population, government has
targeted higher food grain production for the year, 2023, despite concerns of higher
probability of El Nino during 2023. Even in the monsoon long range forecasts and risk
communication, 2023 issued by Indian Meteorological Department (IMD), sustain-
able agriculture is a critical concern in a monsoon dependent country like India. The
sixth assessment report of the United Nations Intergovernmental Panel on Climate
Change (IPCC) has predicted an overall rise in demand for irrigation water by 2080,
causing a further decline in rain-fed agriculture. Rain plays a major role in sustaining
rain-fed crop production in developing countries. Water for agriculture is becoming
increasingly limited due to rising demands from other sectors, especially in India, if
adaptation actions are not in place [2]. It is now imperative to increase farm produc-
tivity. The rice farm sector in Kerala is highly climate sensitive. Demand for both
surface and groundwater irrigation is high in the rice bowls of Kerala such as Palakkad
and Alappuzha. Adopting climate smart measures is one of the key solutions to tide
over the challenge.
Accelerating these transitions from the conventional farm systems to a more
climate smart agriculture (CSA) is a necessity in the context of sustainable devel-
opment [3]. Providing early warning and weather forecast services is one of the key
P. Dhanya (B) · K. Jayarajan

Department of Geography, Government College Chittur, Palakkad, Kerala, India
e-mail: dhanya.eptri@gmail.com
@seismicisolation
@seismicisolation
270 P. Dhanya and K. Jayarajan
areas of climate smart agriculture. Such predictions, which have a time range of a few
months to a few decades, are very important to society, and making them better is an
exciting scientific issue. Apart from that Sendai Framework on Disaster Risk Reduc-
tion (SFDRR), policy makers and planners have shifted their attention to multi-hazard
early warning systems [4, 5]. Most of the times, it is not only one disaster that create
chaos in a place, but a combination of disasters. It is most likely that extreme precip-
itation incidences will become more frequent as a result of global climate change.
Drastic changes in rainfall pay ways for developing new crop pests, diseases, crop
failures and poverty. As a result, accurate understanding of the spatial and temporal
distribution of rainfall at finer resolutions is vital for disaster preparation, drought, or
flood forecasting. Inadequate data, monitoring infrastructure, scarce resources, and
limited adaptive capacities are significant factors of increasing flood and drought
vulnerability [6]. Satellite remote sensing is a powerful tool to monitor and estimate
rainfall events at high spatial and temporal resolutions. There is a dearth for fund
for establishing adequate amounts of rain gauges and ground monitoring stations or
radars in a developing country like India. Climate Hazard Group Infrared Precipi-
tation with Station Data (CHIRPS) is one of the latest high-resolution quasi-global
satellite-based rainfall datasets used for rainfall measurements. There are numerous
gridded precipitation datasets developed throughout the world, and researchers make
use of it after validating it with observed data for various river basins Kumar et al.
and Wang et al. [7, 8].
In many tropical countries, availability of rains has become highly erratic. Mini
cloud bursts (MsCB) are reported in the state of Kerala. Severe rains with high impacts
have become more frequent. Incidences such as total quantity of normal rain for a
month, occurring within one or two days and two- or three-days rains would happen
within few hours in a day [9]. Kerala received high intensity rainfall during August
month in 2018 and 2019, leading to flash floods. It is reported that rainfall exceeding
50 mm in 2 h has been reported from many places on the August 8, 2019. Satellite-
derived rainfall and cloud microphysical parameters further reveal the uniqueness
of the 2019 MsCB event. Another phenomenon that recently got attention is the
increasing frequency and intensity of cyclones formed in Arabian Sea. In 2021, there
was heavy rains and associated floods reported in Kerala during North East Monsoon
period, October and November, severe rains caused landslides and swollen several
rivers throughout the state, roads have been washed away, residences have been
damaged, and trees have been uprooted. Palakkad, the present study region, is one
the least rainfall availing and heat wave prone district of the state.
As per IPCC, the future climate will be warmer than the present, dry areas may
get drier, and land degradation will be more. Hence, strengthening the existing moni-
toring and weather forecast services is recommended for vulnerable areas. There are
weather research and forecasting systems (WRF), regional climate model widely
used to forecast/downscale micro-meteorological weather variables. In India, Indian
Meteorological Department (IMD) provides biweekly extended, medium and short-
range forecasts that are state specific. Medium- and short-term forecasts available
at district and block wise are helpful for the farming communities. There are only
very few data points/rain gauge stations in the basin at present. This research article
@seismicisolation
@seismicisolation
Monitoring Climate Hazards, Rice Production Risks and Management … 271
is an attempt to study the distribution of rainfall and existing risk management tools
available with the rice farmers of Bharathapuzha basin, Palakkad. Both rain-fed
and irrigate rice farming is happening in the basin. There is lack of recent research
studies for understanding the distribution of rainfall in this basin. Hence, this study
was undertaken, using the satellite data for the period 1989–2020. This spatiotem-
poral rainfall variability analysis supports not only farmers or agriculturists, but
hydrologists, geologists, engineers working on surface or groundwater irrigation
and policymakers to manage available water resources efficiently. Adoption of risk
aversion strategies is highly recommended for a sensitive economic sector such as
agriculture and disaster-prone areas. Evaluating the existing risks and hazard anal-
ysis is preliminary step forward any adaptation planning. Monitoring helps to build
capacities in the disaster-prone areas. Where there are fewer monitoring stations,
satellite remote sensing helps to get clear picture of rainfall distribution of a region.
Climate Hazard Group Infrared Precipitation with Station Data (CHIRPS) is one
of the latest high-resolution quasi-global satellite-based rainfall datasets used for
rainfall measurements.
2 Data and Methodology
2.1 Study Area Profile
The Bharathapuzha River originating from the Anamalai hills of Western Ghats is the
life line for the rice farmers in the basin. The basin extends between Western Ghats
and the Palakkad gap on its east, narrowing down to the Arabian seacoast at the west.
This river flows through Palakkad Gap. It flows through highly elevated high lands
in the east, mid lands and low lands in the west. This river has many names such as
Bharathapuzha, Ponnani River, Nila, Perar and Kuttippuram River. It has a length
of 209 km and size of 6186 km2 of which 4400 km2 falls in the state of Kerala and
the rest in Tamil Nadu state of India. Farmers depend upon the summer, Southwest
Monsoon rainfall for the commencement of Kharif ‘Viruppu’ cultivation. Western
Ghats plays a critical role in providing ample rainfall in the basin, only the eastern
and northeastern side of the basin falls in the rain shadow region. There are various
dams built in the basin mainly Malampuzha, and the Monsoon rainfall affects the
inflow and outflow and irrigation water from the dam and hence rice productivity.
This basin is divided into various subbasins Chittur Puzha, Kalpathy puzha, Gayathri
puzha and Thootha puzha (Fig. 1). Malampuzha dam is exclusively extensively used
for irrigation purpose. Irrigation is used mainly for Rabi/Mundakan cultivation.
@seismicisolation
@seismicisolation
Fig. 1 Showing the study area, Bharathapuzha Riven Basin and watersheds
2.2 Rainfall Analysis
Observed rainfall data from India Meteorological Department (IMD) was used to
understand ground reality in rainfall distribution. These gridded data was analyzed
146 and interpolated in ArcGIS 10.5 to understand the spatial pattern. There were
only five grid points falling with in the basin, and hence, neighboring seven grid
points were also taken to get clearer spatial analysis. In order to understand the
usability of satellite remote sensing products in rainfall analysis in places having
fewer monitoring stations, CHIRPS dataset was also used. It is a global rainfall
dataset covering 50°S–50°N (and all longitudes) with high resolution (0.05°) and
is blended with in-situ station data to create a gridded rainfall time series [10]. The
CHIRPS dataset was released by the Climate Hazards Group at the University of
California, Santa Barbara. The different temporal resolution datasets are publicly
available: ftp://ftp.chg.ucsb.edu/pub/org/chg/products/CHIRPS.
2.3 Statistical Analysis of Rainfall Distribution
Rainfall pattern analysis was performed using precipitation concentration index

(PCI). There is a direct relationship between variability and coefficient of variation,
i.e., where there is more variability in rainfall, PCI will be higher.
@seismicisolation
@seismicisolation
Table 1 PCI class values

Sl. No. Spatial distribution PCI classes
1 Uniform distribution ≤ 10
2 Moderate distribution > 10 to < 15
3 Irregular distribution > 15 to < 20
4 Strongly irregular distribution-high ≥ 20
12
(Pi)2
PCI =
i=1
( Pi)2
where Pi is the quantity of rainfall on ith month, used to calculate annual PCI values,
precipitation concentration index (PCI), defined by Li et al. and Oliver et al. [11,
12], is a powerful indicator for temporal precipitation distribution. PCI is generally
used for evaluating seasonal precipitation changes to investigate the heterogeneity
of monthly rainfall data in spatially different areas. The calculation is described in
Table 1.
3.1 Understanding the Rainfall Distribution in the Basin

Using IMD and CHIRP Data
The preliminary analysis was to understand the spatial homogeneity and hetero-
geneity in the pattern of rainfall distribution in BPB. As this was taken as a scoping
study to find criteria for establishing automatic weather stations in the high, medium
and low rainfall receiving areas of the basin. Results shows that both IMD and
CHIRPS data showed eastern side of the basin as comparatively dry and western
and south western part receiving more rainfall distribution during 1989 to 2020.
IMD data showed that areas near the head waters of Chittur puzha and Kalpathi
puzha subbasins (Chittur, Palakkad and Mannarkkad blocks) have received around
700 mm of mean total rainfall, whereas west portion of the basin, tail end of the
stream, namely Nila and Thoothapuzha subbasins (Pattambi, Thrithala and Shornur
blocks), has received around 2600 mm of annual rainfall. Both maps showed east and
south east side of the basin receiving better rainfall than the eastern side of the basin.
The mean rainfall distribution of the BPB basin during the period 1989–2020 was
1749 mm. BPB receives rainfall majorly during Southwest Monsoon and north east
monsoon season with 1165.5 mm and 380.4 mm, respectively. Rains during summer
also contributes to maintain the river flow and summer rice (‘Podi Vitha’ in local
language) vegetable cultivation. The mean summer rainfall is around 217.8 mm, and
winter is the driest season of all with only 13.7 mm of rainfall on an average. The
rainfall distribution during the major rice growing period ‘Kharif’ was also analyzed,
@seismicisolation
@seismicisolation
both IMD and CHIRPS data followed similar pattern, and tail end of the basin has
received more rainfall than the head portion.
CHIRP has slightly over predicted the overall rainfall distribution in the basin but
has captured the seasonality in distribution as indicated in Figs. 3, 4, 5 and 6.
3.2 Precipitation Concentration Index (PCI)
Precipitation concentration index (PCI) is generally used for evaluating seasonal

precipitation changes to investigate the heterogeneity of monthly rainfall in the basin.
Results shows that eastern side of the basin is comparatively dry with 700 mm
of rainfall and west portion of the basin gets around 2600 mm of annual rainfall.
Precipitation concentration index (PCI) is generally used for evaluating seasonal
precipitation changes to investigate the heterogeneity of monthly rainfall in the basin.
As indicated in Fig. 7, PCI value is less than 10 in winter; hence, the precipitation
is uniformly distributed throughout the basin; however, PCI over 40 in July, June
and October months representing a significant irregular distribution throughout the
basin. PCI value is less than 10 in winter; hence, the precipitation is uniformly
distributed throughout the basin; however, PCI over 40 in July, June and October
months representing a significant irregular distribution throughout the basin.
3.3 Agromet Advisory Services for Rice Farmers in the Basin
Rice farming in Kerala is mainly concentrated on two districts, namely Palakkad and
Alappuzha, and they are known as the rice granaries/rice bowl of the state. Small
and marginal rice farmers had an opinion that extreme rainfall events and floods and
droughts have posed havoc in rice production especially during Rabi seasons. Early
warning dissemination has helped the rice farmers to a great extent for risk reduction.
In the recent years, the year, 2016, was a drought year, and 2018, 2019 and 2021 had
floods in the basin. Floods have created havoc to human life and crop production,
and the bridge near Pattambi, Shornur, etc., was flooded as Bharathappuzha River
was overflowing in August 2018 and 2019. Traffic was disrupted completely for a
day. Seed loss and crop damage were also reported extensively. There needs to be
denser network of for river flow monitoring stations at sub-watersheds. Majority of
the farmers are producing rice organically; however, the use of modern machineries
such as power tillers, combined harvesters have become common in the area. It is
essential to revive the dwindling area under traditional rice varieties in Kerala. It
was noteworthy that flood and drought disasters have demonstrated the ability of
native traditional rice varieties such as Kuruva/Kuruka, Njavara and Kattamodan to
endure flood and drought conditions, respectively. Majority of the farmers surveyed
and responded were having 15 years or more experience in rice cultivation (Fig. 2).
More than 50% of the surveyed farmers were highly experienced farmers. The overall
@seismicisolation
@seismicisolation
access to agro-advisory services was collected using a semi-structured questionnaire

survey. The different sources to which the respondents were seek the information
on climate change and adaptation measures have been collected and furnished in
Table 2.
Fig. 2 Total mean annual rainfall for the period 1989–2020 (data: IMD)
Fig. 3 Total mean annual rainfall for the period 1989–2020 (data: CHIRPS)
@seismicisolation
@seismicisolation
Fig. 4 Distribution of Southwest Monsoon rainfall using IMD data
Fig. 5 Distribution of Southwest Monsoon rainfall using CHIRP data
From Table 2, it can be seen that barely half of the respondents (45.6%) had
moderate level of access to agro-advisory sources followed by rare/low level of
access (40.5%) and high level of access to agro-advisory sources (13.40%). It could
be concluded that majority of the respondents had only either medium level of
access or rare access to weather warnings and agro-advisory sources at present.
@seismicisolation
@seismicisolation
Monthwise Rainfall Distribution in the BPB (1989-

2020)
800
rainfall (mm)
600
400
200
10 12 14
Months (Jan-Dec)
Fig. 6 Grid-wise distribution of monthly rainfall in the BPB basin (there were in total 12 IMD grid
points, out of which five were falling with in the basin)
Precipitation Concentration Index-BPB
60
40
PCI
20 PCI 10
PCI 7
PCI 4
PCI 1
10 11 12
13
0-20 20-40 40-60
Fig. 7 Precipitation concentration index of various grids falling in the basin
Table 2 Overall frequency distribution of respondents (from both ground and online survey)
according to their access to weather warnings and agro advisory services
Sl. Categories Access to weather alerts ground Access to weather alerts online
No. survey (%) survey (%)
1 Often or 14.0 23.4
frequently
2 Moderately 45.6 42.2
3 Rarely 40.5 34.4
Total 100.0 100
@seismicisolation
@seismicisolation
This might be due to the fact that the penetration or dissemination of weather warn-
ings needs to be strengthened among the stakeholders. Similarly, it was found that
there is an ample scope to enhance mass media exposure or android phone usage
among the respondent rice farmers in the rural areas. Many were found to not even
possessing the ordinary phones. In order to reduce carbon footprints in rice farming
and minimize severe negative impacts from extreme weather events, a combination
of strategies and solution are recommended both ecosystem-based and technology
solutions such as ICT-based weather early warning and nature-based solutions such
as organic pest and insect management, rain water harvesting, farm pond rejuve-
nation, adopting Azolla for fertilizer, adoption of native tolerant/resilient varieties,
alternative wetting and drying in comparatively drier areas in the basin, etc. Aware-
ness and sensitization campaigns are conducted, and brochures are distributed with
the help of Krishibhavans.
This is a preliminary assessment that is essential to understand the impacts of
rainfall on river bank flooding or river flow regime changes as a future course of
study. Similar studies are reported by Singh et al. and Patakamuri et al. [13, 14].
Spatiotemporal rainfall variability analysis helps hydrologists, geologists, engineers
working on surface or groundwater irrigation and policymakers manage available
water resources more efficiently. Adequate data enables dam officials to co-relate
rain fall precipitation at multiple watershed locations, monitor rising water levels in
reservoirs during heavy precipitation events, and mitigate loss and damage. India is
making significant progress in disaster monitoring and mitigation; however, ensuring
access to impact-based early warnings to all still has a long way to go. Knowing
its criticality in DRR, Early Warnings for All initiative (EW4All) was launched
by World Meteorological Organization, the International Federation of Red Cross
and Red Crescent Societies and the International Telecommunication Union during
COP27 [15]. The main purpose of this is to improve monitoring, access and minimize
loss and damage from climate induced disasters. More funding support is needed in
this line to establish monitoring stations and Doppler radar facilities, etc. (UNDRR,
2021).
4 Conclusions
It is significant to link scientific advancements as services for the benefits of the

societies. Spatiotemporal rainfall variability analysis is a critical input for key stake-
holders. In order to reduce carbon footprints in agriculture and minimize severe
negative impacts from extreme weather events, a combination of solution, both tech-
nology solutions such as ICT-based weather early warning and nature-based solutions
such as organic farming, adoption of tolerant varieties and crop diversification are
required.
@seismicisolation
@seismicisolation
References
1. IPCC (2019) https://www.ipcc.ch/srccl/cite-report/. Accessed Mar 2022

2. IWMI (2010) https://www.iwmi.cgiar.org/Publications/Water_Issue_Briefs/PDF/Water_
Issue_Brief_10.pdf. Accessed 2 Nov 2022
3. Denton F, Halsnæs K, Akimoto K, Burch S, Morejon CD, Farias F, Jupesta J, Shareef A,
Schweizer-Ries P, Teng F, Zusman E (2022) Accelerating the transition in the context of
sustainable development. In: Shukla PR, Skea J, Slade R, Al Khourdajie A, van Diemen R,
McCollum D, Pathak M, Some S, Vyas P, Fradera R, Belkacemi M, Hasija A, Lisboa G, Luz S,
Malley J (eds) IPCC, 2022: climate change 2022: mitigation of climate change. Contribution of
Working Group III to the Sixth Assessment Report of the Intergovernmental Panel on Climate
Change. Cambridge University Press, Cambridge, UK and New York, NY, USA. https://doi.
org/10.1017/9781009157926.019
4. https://www.gfdrr.org/en/publication/designing-inclusive-accessible-early-warning-systems-
good-practices-and-entry-points. Accessed 22 Apr 2023
5. ILO (International Labor Organization) (2022) The importance of early warning systems
in disaster risk reduction. https://www.ilo.org/global/topics/employment-promotion/recovery-
and-reconstruction/WCMS_858123/lang--en/index.htm?
6. Abbas G, Ahmad S, Ahmad A, Nasim W, Fatima Z, Hussain S, ur Rehman MH, Khan MA,
Hasanuzzaman M, Fahad S, Boote KJ, Hoogenboom G (2017) Quantification the impacts of
climate change and crop management on phenology of maize-based cropping system in Punjab,
Pakistan. Agric For Meteorol 247:42–55
7. Kumar M, Hodnebrog Ø, Daloz AS, Sen S, Badiger S, Krishnaswamy J (2021) Measuring
precipitation in Eastern Himalaya: ground validation of eleven satellite, model and gauge
interpolated gridded products. J Hydrol 599:126252
8. Wang F, Yang H, Wang Z, Zhang Z, Li Z (2019) Drought evaluation with CMORPH satel-
lite precipitation data in the Yellow River Basin by using gridded standardized precipitation
evapotranspiration index. Remote Sens 11:485
9. Vijaykumar P, Abhilash S, Sreenath AV, Athira UN, Mohanakumar K, Mapes BE, Chakrapani
B, Sahai AK, Niyas TN, Sreejith OP (2021) Kerala floods in consecutive years—its association
with mesoscale cloudburst and structural changes in monsoon clouds over the west coast of
India. Weather Clim Extremes 33:100339. https://doi.org/10.1016/j.wace.2021.100339
10. Funk C, Peterson P, Landsfeld M, Pedreros D, Verdin J, Shukla S, Husak G, Rowland J,
Harrison L, Hoell A, Michaelsen J (2015) The climate hazards group infrared precipitation with
stations—a new environmental record for monitoring extremes. Sci Data 2:150066. https://doi.
org/10.1038/sdata.2015.66
11. Oliver JE (1980) Monthly precipitation distribution: a comparative index. Prof Geogr
32(3):300–309. https://doi.org/10.1111/j.0033-0124.1980.00300.x
12. Li C, Zhang H, Singh VP, Fan J, Wei X, Yang J et al (2020) Investigating variations of precip-
itation concentration in the transitional zone between Qinling Mountains and Loess Plateau
in China: implications for regional impacts of AO and WPSH. PLoS ONE 15(11):e0238709.
https://doi.org/10.1371/journal.pone.0238709
13. Patakamuri SK, Muthiah K, Sridhar V (2020) Long-term homogeneity, trend, and change-point
analysis of rainfall in the arid district of Ananthapuramu, Andhra Pradesh State, India. Water
12:211
14. Singh DS, Prajapati SK, Singh P, Singh K, Kumar D (2015) Climatically induced levee break
and flood risk management of the Gorakhpur region, Rapti River basin, Ganga Plain, India. J
Geol Soc India 85:79–86
15. World Meteorological Organization (2022) Early warnings for all. https://public.wmo.int/en/
earlywarningsforall. Accessed 23 Feb 2023
@seismicisolation
@seismicisolation
Investigating Spatio-Temporal Trends
and Anomalies in Long-Term
Meteorological Variables to Determine If
Maharashtra is an Emerging Warming
State in India
Aman Srivastava , Rajib Maity , and Venkappayya R. Desai
1 Introduction
As per the Intergovernmental Panel on Climate Change (IPCC), the ongoing period
of the twenty-first century (2001–2020) recorded a 0.99 °C higher global surface
temperature than 1850–1900. The rise in the global surface temperature is primarily
attributed to the increase in well-mixed anthropogenic Greenhouse Gases (GHGs).
Consequently, a higher temperature increase has been recorded on land (1.59 °C) than
in the ocean (0.88 °C) [1, 2]. One of the direct implications of global warming has been
observed as the intensified hydrological cycle coupled with erratic spatio-temporal
precipitation (or rainfall) pattern and distribution. The rainfall has likely increased
globally since 1950, wherein the post-1980s period witnessed a comparatively faster
rate of increase [1]. The rise in global temperature due to anthropogenic GHG emis-
sions has dramatically increased relative humidity, resulting in a more significant
atmospheric water-holding capacity [3–6]. In the context of anthropogenic global
warming, the Clausius–Clapeyron Equation has further indicated an expected rise of
> 7% in the water-holding capacity of the atmosphere corresponding to a 1 °C rise
in global temperature [7]. Consequently, the widespread increase in flood events due
to unprecedented rainfall in magnitude, duration, and intensity has exacerbated the
situation [8, 9]. Furthermore, the intense short-duration rainfall has been followed
by long durations of no-rainfall conditions. Such erratic rainfall patterns mostly lead
A. Srivastava · R. Maity (B)

Department of Civil Engineering, Indian Institute of Technology (IIT) Kharagpur, Kharagpur,
West Bengal 721302, India
e-mail: rajib@civil.iitkgp.ac.in
V. R. Desai (B)
Indian Institute of Technology (IIT) Dharwad, Dharwad, Karnataka 580001, India
e-mail: rajib@civil.iitkgp.ac.in
@seismicisolation
@seismicisolation
282 A. Srivastava et al.
to meteorological droughts, and subsequently, other drought types, such as hydro-

logical, agricultural, and socio-economic droughts, are also observed [10–12]. The
recent decade has witnessed floods and droughts cyclically at a regional/local scale
and sometimes concurrently in the same study location, where rapid urbanization
significantly affected the surface water–groundwater interaction [13–19]. Remark-
ably, the implications of global climate change anomalies due to abrupt changes in the
aforementioned meteorological variables and other hydroclimatological variables,
such as evaporation, evapotranspiration, solar radiation, wind speed, vapor pressure
deficit, etc., are urging current researchers to further probe into understanding the
nuances behind regional climate change [20–24]. This study, therefore, considered
the temperature (T, which is the mean temperature), relative humidity (RH), and rain-
fall (P) for investigating climate change aspects in one of the similar subcontinental
regions of India.
With its multifaceted geography, India is no longer exempt from the negative
consequences of global climate change, especially regarding the pattern and distri-
bution of T, RH, and P at spatial (i.e., regional) scales [25]. This study considered
the Indian subcontinent for climate change study, since India has the second-largest
human population (having > 1.4 billion in 2022, accounting for 18% of the world’s
population) to feed with merely 4% of available water resources [26, 27]. Further-
more, India is an agriculturally-dominant nation with a 16% contribution to GDP,
wherein > 55% of the population engages in the agrarian professions on 9% of the
world’s arable land (53% of India’s geographical area), such that > 56% is rainfed [28,
29]. As far as Indian summer monsoon rainfall is concerned, studies have shown a
slightly decreasing trend since 1901 on a national scale. Western states/regions have
shown a greater decreasing trend, while northern and northeastern states/regions
have shown a marginally greater increasing trend [30, 31]. As far as temperature is
concerned, 2001–2020 has been recorded as the warmest period since 1901, with a
significant positive trend in temperature. The rising rate of temperature from 1871
to 1997 remained at 0.57 °C/century and from 1971 to 2003 at 0.22 °C/decade [28].
On a spatial scale, southern, eastern, and western regions experienced comparatively
greater adverse impacts due to the rise in annual mean temperature and uneven fall
in the rainfall pattern as against the other regions of India. More specifically, the
western states such as Rajasthan, Gujarat, and Maharashtra remained water-stressed
and mostly drought-prone since historical times with occasional events of flash floods
[32, 33]. As a result, the present study focuses on one of these states—Maharashtra,
to better account for the influence of T, RH, and P on its climate.
The state of Maharashtra is the third-largest in terms of area (9.4% of India) and
the second-largest in terms of population (9.3% of India), where agriculture (12.3%
net sown area of India) is the primary occupation for over 53% of the state popula-
tion [34]. Any change in the behavior of meteorological variables within the state
can exert significant pressure on the Indian climate system and economy, which are
interlinked. In conjunction with this, a few studies have conducted trend tests on long-
term meteorological variables at the state level and many studies at the district or
regional level to extricate the variability in Maharashtra’s climate. Singh et al. [35]
have statistically analyzed the spatio-temporal trends and variability [using Mann
@seismicisolation
@seismicisolation
Investigating Spatio-Temporal Trends and Anomalies in Long-Term … 283
Kendall (MK), Sen’s slope (SS), and other related tests] for the rainfall time series
data of 118 years (1901–2018) for entire Maharashtra. Their study found signifi-
cantly decreasing trends for winter and summer rainfall, while both increasing and
decreasing trends were recorded for monsoon and post-monsoon rainfall. The results
remained more profound when the trend post-twenty-first century was compared
to those for the twentieth century. Similar findings for Indian summer monsoonal
rainfall for various regions within Maharashtra state were obtained by Mahajan
and Dodamani [24, 36, 37], and [38]. However, unlike rainfall, studies on trend
assessment for temperature and relative humidity still need to be documented for the
entire Maharashtra state. More specifically, these studies remained limited to region-
specific investigations, as a result of which, their local inferences on climate could
only be extrapolated to some of the states. For example, in the Konkan subdivision
(i.e., western Maharashtra), [39] conducted trend tests using MK and SS methods,
especially for maximum and minimum temperature along with rainfall for the period
from 1981 to 2016. Specifically, for maximum temperature, their results found it to
increase significantly in the south and central Konkan at a rate ranging from 0.03
to 0.09 °C/year. At the same time, it decreased considerably in northern Konkan at
a rate ranging from − 0.1 to − 0.19 °C/year. For the Karjat region in the Raigad
district of the Konkan subdivision, they analyzed maximum and minimum tempera-
ture, rainfall, maximum and minimum relative humidity, wind speed, and sunshine
hours using MK and SS tests for the period 1991 to 2015. Specifically, for relative
humidity, their study found maximum relative humidity to decrease non-significantly,
while minimum relative humidity to increase significantly. Similar studies have been
conducted for the Madhya Maharashtra, Marathwada, and Vidarbha subdivisions
of Maharashtra [40–43]. In general, all these studies indicated profound changes in
the spatio-temporal trends of meteorological variables in the local climate of Maha-
rashtra, especially post-2000. Nevertheless, a clear picture of Maharashtra’s climate
has not been documented, given the temperature and relative humidity variabilities.
The lack of investigations demanded extending the entire state’s spatio-temporal
studies of meteorological variables. Another observation from the literature survey
is that non-parametric statistical tests, such as MK and SS, are used widely for trend
analysis. Hence, the present study employed them for the entire state’s T, RH, and P
trend analysis.
Considering the published literature on meteorological studies in Maharashtra, it
is necessary to bridge this research gap. In conjunction with climate change, observa-
tions from the previous findings raise a major research question—whether the Maha-
rashtra state is an emerging rapidly warming state in India. Apparently, Maharashtra’s
state economy is compounded by the magnitude of agricultural efficiency and by the
extent of water resources availability [44]. Unwarranted climate change due to a
hotter climate could complicate both agricultural and water resources management
of the state and thus can severely impede the economy. Therefore, policymakers
in the recent decade by the Government of Maharashtra, in particular, and India
in general, have sought research advisories on subdivision-specific sustainable and
climate-resilient pathways. In the purview of the present literature survey, the lack of
comprehensive studies on Maharashtra’s historical climate is an additional research
@seismicisolation
@seismicisolation
gap. These research limitations were observed, particularly regarding the influence
of T, RH, and P on the overall climate of Maharashtra. Although investigations on
the impact of any single meteorological variable on the climate in certain regions of
Maharashtra and the state as a whole have been conducted, it is essential to proclaim
here that such studies lacked corollaries in specific while drawing conclusive pictures
of the climate change facets at state-level. Since the assessment was mostly conducted
for one meteorological variable, their findings remained biased toward the climate
change dimensions. Studies on understanding the influence of multiple meteorolog-
ical variables on disease outbreaks in Maharashtra (on a comprehensive basis) have
been conducted in the recent decade [45–48]. Similarly, a need was felt to extend such
investigations considering multiple meteorological variables, for studying climate
change impact in Maharashtra state as a whole.
The motivation behind the present investigation is a need for studies on the compar-
ative analysis among the four subdivisions of Maharashtra in terms of T, RH, and P.
Therefore, this study aims to holistically comprehend the inter-relationships among
T, RH, and P variables to know whether Maharashtra is an emerging rapidly warming
state in India. Hence, the primary objectives of this study are: (1) to conduct region-,
month-, and year-wise spatio-temporal analysis of three meteorological variables,
viz., temperature (T), relative humidity (RH), and rainfall (P) for the study period
1981–2020 in the India Meteorological Department (IMD) categorized four subdivi-
sions, viz., Konkan, Madhya Maharashtra, Marathwada, and Vidarbha, covering the
entire Maharashtra state; and (2) to determine possible anomaly in spatio-temporal
components of aforesaid long-term meteorological variables at sub-divisional scale
against each IMD categorized seasons [winter, summer, south-west (SW) monsoon,
and north-east (NE) monsoon) to ascertain the ‘warming climate’ dimension for the
state. The findings from this study could provide suitable frameworks for State Action
Plans on Climate Change (SAPCC) and related missions for Maharashtra State.
2 Materials and Methods
2.1 Study Area Description
Maharashtra state is a part of the semi-arid region of western peninsular India.

Geographically, it is located at latitudinal coordinates between 15 and 23°N and
longitudinal coordinates between 72 and 83°E (Fig. 1). This study has considered
all 34 districts of the state spread across an area of 307,713 km2 . This state’s expan-
sion from north to south is 700 km, and east to west is 800 km. The state has a
massive geographical significance in India, given that it is the third largest state after
Rajasthan and Madhya Pradesh. In fact, the Maharashtra state’s area is equivalent to
that of countries like Poland, Oman, Italy, and the Philippines. States like Madhya
Pradesh border it in the north, Chhattisgarh in the east, Telangana in the southeast,
Karnataka in the south, and Goa in the southwest. While the Arabian Sea borders
@seismicisolation
@seismicisolation
in the west, occupying a coastline of 840 km. The population in the state is around
125.7 million (population density of about 365 per km2 ); 54% live in urban areas
and 46% in rural areas. The state has been characterized into four subdivisions by
IMD viz., Zone A—Konkan located in the western region of Maharashtra and is
essentially having a coastal environment; Zone B—Madhya Maharashtra situated
toward the west; Zone C—Marathwada located toward the east; Zone D—Vidarbha
located in the far-eastern region of Maharashtra, which also represents Central India.
Maharashtra has the maximum number of subdivisions when compared to any Indian
state.
Hydrometeorologically, the state experiences a hot moist, sub-humid to humid
climate with tropical monsoons and dry summers. The summer season comprises
March, April, and May (MAM), and the summer monsoon, also called the south-
west (SW) monsoon, includes June, July, August, and September (JJAS). The
post-monsoon season, also called the Northeast (NE) monsoon, occurs in October,
November, and December (OND), while January and February (JF) mark the winter
season. The annual minimum temperature of the state is 15 °C [standard devia-
tion (SD) = 0.97 °C], while the maximum is 30 °C (SD = 1.25 °C). The average
Fig. 1 Study area—the Maharashtra state in western India, with its four meteorological subdi-
visions along with their constituent districts (Inset shows the location of Maharashtra state and
India)
@seismicisolation
@seismicisolation
annual mean temperature between 1981 and 2020 for Konkan, Madhya Maharashtra,
Marathwada, Vidarbha, and Maharashtra state as a whole is 22.51 °C, 21.00 °C,
21.85 °C, 22.73 °C, and 21.88 °C, respectively (Fig. 2a). For the same subdivi-
sions and study period, relative humidity is 86.43%, 77.02%, 68.55%, 70.83%, and
75.36%, respectively (Fig. 2b), and annual average rainfall (i.e., mean annual precip-
itation) is 957.71 mm, 429.77 mm, 339.28 mm, 496.27 mm, and 534.01 mm, respec-
tively (Fig. 2c). As evident from Fig. 2, the aforementioned meteorological variables
show medium to high heterogeneity for the various subdivisions of the state. At the
same time, considering the equation of a fitted straight line (y = mx + c; where y
is the predicted score, m is the slope of the fitted line, x is the year, and c is the
y-intercept), these variables show an increasing trend (i.e., m is positive) since 1981.
Land use-wise, agriculture is the most extensive land use covering 56% of the
total geographical area, followed by forest cover of 16.9% [34]. The major portion
of India’s two essential river basins, viz., Krishna and Godavari, are also located in
Maharashtra, which plays a critical role in multi-purpose water management in the
state. Despite this, the state has a limited area under irrigation, low natural fertility
of soils, and large areas prone to recurrent drought. Due to this, Maharashtra’s agri-
cultural productivity is generally low compared to the national averages for various
crops. Altogether, conducting hydroclimatological studies for states like Maharashtra
becomes imperative when past literature additionally reports on the state’s changing
climate amidst aggravating water scarcity issues.
2.2 Data Types and Acquisition
The climate datasets were collected from the repository of the fifth genera-
tion of the European Centre for Medium-Range Weather Forecasts (ECMWF)
reanalysis product (ERA5, https://www.ecmwf.int/en/forecasts/datasets/reanalysis-
datasets/era5, accessed April 2022). The dataset contains observational data for a
period of 40 years (1981–2020). The data records include T, RH, and P variables.
The details of each dataset obtained from ECMWF-ERA5 are provided in Table 1.
The ERA5 data was obtained on a 0.25 × 0.25-degree grid resolution. However,
the grid resolutions were further increased to a 0.05 × 0.05-degree resolution to
achieve a better data representation and associated statistical analysis. This study
has considered month-wise, season-wise, and year-wise records of climate variables.
Each dataset across the four subdivisions was spatially averaged and converted to
the above scale before further processing.
2.3 Methods and Methodology
The present study has primarily focused on studying T, RH, and P variables for
the IMD-categorized four seasons, viz., winter (JF), summer (MAM), SW monsoon
@seismicisolation
@seismicisolation
Maharashtra state Linear (Maharashtra state) y = 0.0413x - 60.661

24.0
23.5
Mean Temperature (T; oC)
23.0
22.5
22.0
21.5
21.0
20.5
1981 1984 1987 1990 1993 1996 1999 2002 2005 2008 2011 2014 2017 2020
Time (yearly)
(a)
Maharashtra state Linear (Maharashtra state) y = 0.0729x - 70.435
80
78
Relative Humidity (RH; %)
76
74
72
70
68
1981 1984 1987 1990 1993 1996 1999 2002 2005 2008 2011 2014 2017 2020
Time (yearly)
(b)
Maharashtra state Linear (Maharashtra state) y = 7.6042x - 13849
2000
1800
Annual Mean Rainfall (P; mm)
1600
1400
1200
1000
800
600
400
200
0
1981 1984 1987 1990 1993 1996 1999 2002 2005 2008 2011 2014 2017 2020
Time (yearly)
(c)
Fig. 2 Annual variation in meteorological variables viz., a temperature (T), b relative humidity
(RH), and c rainfall (P) for the study period 1981–2020 in Maharashtra state, India
@seismicisolation
@seismicisolation
Table 1 Details of the climatological dataset obtained from the fifth generation of the Euro-
pean Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis Product (ERA5) for
the Maharashtra state of India for the study period 1981–2020
Dataset Variable Spatial Converted Vertical/ Unit Unit
resolution spatial pressure (ERA5) converted
resolution level
ERA5-Land Total 0.25° × 0.05° × Surface mm/day mm/month
monthly precipitation 0.25° 0.05°
averaged data (P)
2-m 0.25° × 0.05° × 2-m above K °C
temperature 0.25° 0.05° surface
(T)
ERA5 Relative 0.25° × 0.05° × 1000 hPa Percent –
monthly humidity 0.25° 0.05° (%)
averaged data (RH)
on pressure
levels
(JJAS), and post-monsoon (OND) for the study period 1981–2020 on a decadal,
yearly, seasonally, monthly, state, and a sub-divisional basis. Trend analysis for each
variable was evaluated in two steps viz., first, by employing a non-parametric Mann
Kendall (MK) Test [49, 50] for testing the presence of a monotonic decreasing or
increasing trend; second, by employing Sen’s Slope Test [51] for estimating the
rate and direction of change. The results were tested at a significance level (α) =
0.05 (5%), wherein variations at the spatial and temporal scales mentioned above
were analyzed. Furthermore, the trend analysis was conducted for the 40-years study
period (viz, 1981–2020), for the months January to December (where each month
showed a 40-year average value), and for the four subdivisions (i.e., Zone A to
Zone D) along with that for the entire state. Month-wise trend analysis allowed
for estimating the significance of trends in T, RH, and P variables at the scales
described above. These results were used to ascertain whether Maharashtra is an
emerging rapidly warming state. Moreover, the study attempted to establish a corre-
lation between the three variables using the Pearson Correlation Coefficient (PCC).
Correlation analysis allowed quantifying the degree of influence of relative humidity
(RH) and rainfall (P) on the states as well as its sub-divisional temperatures (T).
Besides, this study also estimated the anomalies by comparing the last decadal (i.e.,
2011–2020) values with those for the entire study period (1981–2020). The findings
from the decadal comparison could help ascertain how drastic the changes occurred
amidst recent modernization. The flow of this study is further summarized in Fig. 3.
The statistical tests (MK and SS) are described in the below-mentioned subsections.
@seismicisolation
@seismicisolation
Time-series & scatter plots for correlation Temperature (T) Relative Humidity (RH) Rainfall (P)
Mann Kendall’s Test Trend Analysis (for 40 years) Sen’s Slope Test
Maharashtra State Konkan Madhya Maharashtra Marathwada Vidarbha
Season-wise Analysis (1981-2020)
Winter (JF) Summer (MAM) South-West monsoon (JJAS) Post-monsoon (OND)
Mann Kendall’s Test Trend Analysis (40 years/season) Sen’s Slope Test
Is Maharashtra a rapidly warming state?
Winter (JF) Summer (MAM) South-West monsoon (JJAS) Post-monsoon (OND)
Season-wise anomaly detection: comparing 1981-2020 with 2011-2020
Temperature (T) Relative Humidity (RH) Rainfall (P)
Fig. 3 Flowchart of the methodology for ascertaining the study hypothesis. (Note JF, MAM, JJAS,
and OND represent the initial letter of each month from January to December)
2.3.1 Mann Kendall’s (MK) Test
The non-parametric Mann Kendall’s test is a rank-based method that detects the trend
in a time series following any probability distribution. Thus, it is used widely and
mainly for consistently increasing or decreasing variable trends. The MK test was
considered the most preferable since the present study dealt with consistently rising
trends for T, RH, and P variables. The MK Statistic (S) for the time series X(a) (for
each variable) was defined using Eq. 1, where N is the number of observations or
data points, and sign [.] is the signum function, as described in Eq. 2. The sign [.] and
the magnitude of S allowed for estimating the intensity and direction of the trend.
N −1
N

S= sign X a − X (a) (1)
a=1 a =a+1

X (a )−X (a)
sign X a − X (a) = |X (a )−X (a)| , i f x = 0
0, ifx = 0
⎧
⎨ 1, i f X a − X (a) > 0
= 0, i f X a − X (a) = 0 (2)
⎩
−1, i f X a − X (a) < 0
The present study considered the Null Hypothesis as ‘Time-series do not have
a trend,’ and Alternate Hypothesis as ‘Time-series have a trend.’ These time series
were tested at the significant level (α) of 5%. In the case of no trend (i.e., acceptance
of null hypothesis), the S statistic followed a distribution with zero mean value,
@seismicisolation
@seismicisolation
for which the variance [Var (S)] was defined using Eq. 3, such that ‘t’ represents
the number of tied groups. In contrast, Oi represents the number of data points in
‘t.’ However, in the case of no tied groups [considering that the present study is
investigating long-term hydroclimatic variables (N > 10), which may be lacking
tied groups], Eq. 3 was revised as Eq. 4. Once all the values were identified, the
test statistic (Ts) was estimated using Eq. 5, wherein Ts was assumed to follow the
standard normal distribution. It is to be noted that the null hypothesis was rejected
(i.e., there is a presence of trend in the time series) when the | Ts | > Z (α/2), where Z
(α/2) is the standardized normal variate for the non-exceedance probability of (1 −
α/2) × 100%. Finally, the p-value was estimated to conclude the significance and
non-significance of the trend in the times series, such that p-value < 0.05 represented
the former and p-value > 0.05 represented the latter case.
1 t
Var(S) = N (N − 1)(2N + 5) − Oi (Oi − 1)(2Oi + 5) (3)
18 i=1
N (N − 1)(2N + 5)
V ar (S) = (4)
18
S − sign(S)
Ts = √ (5)
Var(S)
2.3.2 Sen’s Slope (SS) Estimator
The non-parametric Sen’s Slope is another method for trend analysis for a long-
term hydroclimatic time series. Here, the magnitude of the linear trend is estimated
wherein the slope at all pair of points provides the rate and direction of change for
that time series. Long-term hydroclimatic time series may contain outliers; however,
the SS test estimates slopes of a trend accurately, given that it is insensitive to outliers.
Therefore, considering this benefit alongside its widespread applications, the present
study employed the SS Test for analyzing the time series for T, RH, and P variables.
For a linear model f (c), the SS estimator can be defined by following Eqs. 6–8,
where Qc is the slope and C is the constant, X j and X k represent the magnitude of
the sample at the jth and kth time such that j is the succeeding year of data, while
k is the preceding year of data; N is the total number of data points in the series. In
the SS Test, if there are n number of sample values, it implies that the time series
(say for X j ) comprises of N = [(n − 1)]/2 slope estimates of Qc . This study ranked
N number slope values of Qc from the smallest to the largest and thereby estimated
the slope, which is the median of these N values of Qc (as shown in Eq. 8).
fc = Q c + C (6)
@seismicisolation
@seismicisolation
X j − Xk
Qc = , where j > k and series runs from 1, 2, 3, . . . .., N (7)
j −k

Q N +1 , if N is odd
Qc = 1 2 (8)
2
Q N2 + Q N 2+2 , if N is even
3.1 Subdivision-Wise Analysis of Annual Variability

(1981–2020) in the Meteorological Variables
The statistics behind the annual variability of T, RH, and P variables for the Maha-
rashtra state and the four subdivisions are shown in Table 2. As far as the state’s
temperature is concerned, the mean annual temperature (Tannual) considering the
40-years study period is observed as ~ 21.9 °C with a standard deviation (SD) of ~
0.7 °C. Findings from the MK Test show a significant trend in the Tannual time series
(with a p-value < 0.05 and thus MK = 1) at a confidence interval of 95%, thereby
indicating strong evidence for the defined alternate hypothesis. Coherently, the SS
estimator shows a positive value (Qc = 0.03) indicative of the positive trend in the
time series for Tannual, such that the Tannual in the state has been increasing since
1981 at the rate of 0.03 °C/year (which can be validated from Fig. 2a). While from the
four meteorological subdivisions of the Maharashtra state, the maximum Tannual is
recorded for Vidarbha (~ 22.7 °C) with a second-highest SD (~ 0.8 °C), a significant
positive trend, and an increasing rate of 0.04 °C/year since 1981. In contrast, the
minimum Tannual is recorded for Madhya Maharashtra (~ 21 °C) with the least SD
(~ 0.5 °C), a significant positive trend, and an increasing rate of 0.028 °C/year. It
can be inferred that the difference in the Tannual attributes between maximum and
minimum is not too far from the Tannual characteristics for the entire Maharashtra
state, such that the trend is identified as significant and positive for each of the four
subdivisions. Therefore, the Tannual analysis further demands monthly, seasonal,
as well as spatial analysis of temperature (Tseasonal; as detailed in the subsequent
sections).
For the relative humidity of the Maharashtra state (Table 2), the mean annual
relative humidity (RHannual), considering the 40-year study period, is observed as
~ 75.4%, with an SD of ~ 2.4%. Findings from the MK Test show an insignificant
trend in the RHannual time series (p-value slightly > 0.05 and thus MK = 0) at a
confidence interval of 95%, thereby indicating strong evidence for the defined null
hypothesis at least at the state level. Contrarily, the SS estimator shows a positive
value (Qc = 0.074) indicative of the positive trend in the RHannual time series,
indicating that the RHannual in the state has been increasing since 1981 at the rate
of 0.074%/year (which can be validated by Fig. 2b). While for the four subdivisions
@seismicisolation
@seismicisolation
Table 2 Subdivision-wise trend analysis for temperature, relative humidity, and rainfall variables
for the study period 1981–2020 at a yearly scale using Mann Kendall’s (MK) Test (providing p-
value) and Sen’s Slope (SS or Qc ) Estimator along with their annual mean and standard deviation
(SD) values
Variable State/ Mean SD MK p-value SS
Subdivision
Temperature Maharashtra 21.88 0.68 1 0.00002 0.030
(°C) State
Konkan 22.51 0.51 1 0.00009 0.028
Madhya 21.00 0.64 1 0.00006 0.028
Maharashtra
Marathwada 21.85 0.87 1 0.00010 0.044
Vidarbha 22.73 0.83 1 0.00001 0.040
Relative Maharashtra 75.36 2.39 0 0.06737 0.074
humidity (%) State
Konkan 86.43 1.07 0 0.08256 0.026
Madhya 77.02 2.36 1 0.03113 0.076
Maharashtra
Marathwada 68.55 3.93 0 0.19592 0.088
Vidarbha 70.83 3.23 0 0.12124 0.085
Rainfall (mm) Maharashtra 1362.90 212.83 1 0.02453 7.024
State
Konkan 2445.84 393.57 1 0.04634 12.758
Madhya 1096.48 202.33 1 0.01919 7.750
Maharashtra
Marathwada 864.94 173.17 1 0.03300 5.661
Vidarbha 1266.99 254.15 0 0.09569 5.260
of the Maharashtra state, the maximum RHannual is recorded for Konkan (~ 86.4%)
with the least SD (~ 0.8 °C) and an insignificant positive trend, having the lowest
increasing rate of 0.026%/year since 1980. The minimum RHannual is recorded for
Marathwada (~ 68.6%) with the highest SD (3.9%) and an insignificant positive trend,
with the highest increasing rate of 0.088%/year since 1980. It can be inferred that the
difference in the RHannual attributes between maximum and minimum is far from the
RHannual characteristics for the Maharashtra state, such that the trend is identified as
insignificant (except for Madhya Maharashtra; having MK = 1) but positive for each
subdivision (SS > 0). Given this, the relative humidity analysis cannot be ignored for
monthly, seasonal, and spatial-scale analysis (RHseasonal; detailed in the subsequent
sections). Hence, this study conducted similar investigations for the relative humidity
variable also, similar to the temperature variable.
As far as the rainfall of the Maharashtra state is concerned (Table 2), the mean
annual rainfall (Pannual) considering the 40-years study period is observed as ~
1363 mm with an SD of ~ 213 mm. Findings from the MK Test show a significant
@seismicisolation
@seismicisolation
trend in the Pannual time series (p-value < 0.05 and thus MK = 1; similar to Tannual,
but unlike RHannual) at a confidence interval of 95%, thereby indicating strong
evidence for the defined alternate hypothesis. Coherently, the SS Estimator shows
a positive value (Qc = 7.024) indicative of the positive trend in the Pannual time
series, such that the Pannual in the state is increased since 1981 at the rate of ~
7 mm/year (which can be validated from Fig. 2c). Among the four subdivisions of
the Maharashtra state, the maximum Pannual is recorded for Konkan (~ 2446 mm)
with the highest SD (~ 394 mm), a significant positive trend, and an increasing rate
of 0.04 mm/year since 1980. In comparison, the minimum Pannual is recorded for
Marathwada (~ 865 mm) with the least SD (~ 173 mm), a significant positive trend,
and the second-least increasing rate of 5.66 mm/year since 1980. The difference
in the Pannual attributes between maximum and minimum is way too far from the
Pannual characteristics of the Maharashtra State. Furthermore, the trend is identified
as significant and positive for each subdivision (however, an insignificant trend is
observed only for Vidarbha, contrary to other subdivisions). In fact, if we broadly
and comparatively classify these subdivisions as high, moderate, and low Pannual
classes, Konkan lies in the high class, Madhya Maharashtra and Marathwada lie in
the moderate class. In contrast, Vidarbha lies in the low class. Hence, conducting the
Pannual analysis for monthly, seasonally, and spatial scales (Pseasonal; detailed in
the subsequent sections) will ascertain the erratic rainfall pattern.
To summarize, the preliminary investigation here reveals that Maharashtra as a
whole is experiencing a positive trend in Tannaul, RHannual, and Pannual, such that
a significant trend is observed for Tannual and Pannual. Subdivision-wise, all the
variables have shown a positive trend for all the subdivisions of Maharashtra. While
for the Tannual and Pannual, all the subdivisions have additionally demonstrated a
significant trend (Vidarbha is an exception for Pannual). However, for RHannual,
merely Madhya Maharashtra has shown a significant trend. Konkan is experiencing
the highest Pannual with maximum RHannual, while Vidarbha is experiencing the
highest Tannual. More or less, the eastern subdivisions of Maharashtra (Vidarbha
and Marathwada) are experiencing more warming conditions, low rainfall, and rela-
tive humidity conditions than the western subdivision of Konkan. While Madhya
Maharashtra is experiencing conditions at the interface of the eastern and western
subdivisions of Maharashtra, both the influence of the coastal environment of Konkan
and the Deccan plateau environment of Marathwada and Vidarbha can be observed in
Madhya Maharashtra. In general, the above findings affirm the appropriate variable
selection for the present investigation, thereby demanding further probing to quan-
tify the influence of these variables on Maharashtra’s climate at spatial and varying
temporal (i.e., decadal, annual, seasonal, and month-wise) scales.
@seismicisolation
@seismicisolation
3.2 Season-Wise and Annually-Based Spatio-Temporal

Analysis of Meteorological Variables During 1981–2020
The time series and spatial plots of meteorological variables for the period 1981–2020
on a seasonal basis (Tseason, RHseason, and Pseason) and annual basis (Tannual,
RHannual, and Pannual) for the Maharashtra State and its four subdivisions viz.,
Konkan, Madhya Maharashtra, Marathwada, and Vidarbha are shown in Figs. 4 and
5, respectively. The sections below draw variable-wise inferences on the same. The
analysis has been conducted by recording the changes across 40 years (1981–2020).
Furthermore, the study period has been partitioned into the initial three decades
(1981–2010) and the last or the recent decade (2011–2020) to account for abrupt
changes, if any, in recent times against the long-term mean.
3.2.1 Temperature (T)
Considering the annual mean temperature (Tannual), the Tannual for the Maharashtra
State slightly increased from 21.5 to 21.7 °C between 1981 and 2010. However, a
drastic increase of 1 °C (from 21.7 to 22.7 °C) can be observed during 2011–2020.
As far as the subdivisions are concerned, maximum Tannual has been recorded for
Vidarbha, experiencing an increment from 22.3 to 22.5 °C between 1981and 2010
and then to 23.8 °C during 2011–2020 (a sudden rise of 1.3 °C). Marathwada follows
Vidarbha, experiencing an increment from 21.5 to 21.6 °C and then to 23 °C during
the periods mentioned above (a sudden rise of 1.4 °C) (Fig. 4a).
Fig. 4 Subdivision-wise comparative analysis between seasonal variation and annual variation in
meteorological variables, viz., temperature (a–e), relative humidity (f–j), and rainfall (k–o), for the
study period 1981–2020
@seismicisolation
@seismicisolation
Fig. 5 Spatial variability of meteorological variables in the last 40 years (1981–2020) [as shown in
(I)] and 10 years (2011–2020) [as shown in (II)] across four seasons, viz, winter, summer, south-west
monsoon, and post-monsoon; for temperature (a–d), relative humidity (e–h), and rainfall (i–l)
Similar observations are recorded for season-wise mean temperature (Tseason)

for the Maharashtra state and its subdivisions (here, Tseason corresponds to Twinter,
Tsummer, Tmonsoon, and Tpost-monsoon). Considering the entire state, the Twinter
records the least temperature among all seasons. It slightly increased from 17.5
to 17.6 °C between 1981 and 2010 and further to 18.1 °C during 2011 to 2020.
Naturally, Tsummer values record the maximum temperature among all the seasons.
It slightly increased from 23.9 to 24.2 °C between 1981 and 2010 and drastically
increased to 25.4 °C (a sudden rise of 1.2 °C) during 2011–2020. After Tsummer,
@seismicisolation
@seismicisolation
the Tmonsoon values experienced the maximum temperature. It slightly increased

from 23.7 to 23.9 °C between 1981 and 2010 and increased rapidly to 24.7 °C
from 2011–2020. The values of Tpost-monsoon record the second-least temperature
among all seasons. It barely increased from 19 to 19.1 °C between 1981 and 2010 and
drastically increased to 20.6 °C (a sudden rise of 1.5 °C) during 2011–2020. Given
the season-wise spatial variations in the Tseason between a maximum value lying
in the 1981–2010 period (primarily located in the 2001–2010 period) and 2011–
2020 [Fig. 5 (I and II) a–d] for the winter season, the maximum rise in the Twinter
is recorded for Marathwada (a rise of 1 °C = from 17.3 to 18.3 °C between the
aforementioned periods), followed by Vidarbha (an increase of 0.9 °C = from 17.2
to 18.2 °C). However, the maximum Twinter is observed in the Konkan (18.6 °C
during 2011–2020). For the summer season also, the maximum rise in the Tsummer
is recorded for Marathwada (a 1.7 °C rise = from 24.9 to 26.6 °C), followed by
Vidarbha (a 1.6 °C rise = from 26.1 to 27.7 °C). However, the maximum Tsummer
is observed in Vidarbha only (27.7 °C during 2011–2020). For the SW monsoon
season, both Marathwada and Vidarbha subdivisions experience a similar rise in
Tmonsoon (i.e., an increase of 0.9 °C), which is from 23.8 to 24.7 °C and from
25.4 to 26.3 °C, respectively, indicating that Vidarbha is turning out to be the hotter
between them. On the other hand, Tpost-monsoon values show a sharp increase in
Marathwada (a rise of 2.1 °C from 18.2 to 20.3 °C), followed by Vidarbha (wherein
an increase of 1.8 °C from 18.6 to 20.5 °C is observed). Numerically, the maximum
Tpost-monsoon values are recorded for the Konkan subdivision (Fig. 4a–e).
To summarize, the rise in the mean temperature across all seasons remains low
during the period 1981–2010; however, the post-2010 period witnessed a signifi-
cant increase in temperature (> 1 °C within a decade) for all four seasons. More
specifically, the Tsummer and Tpost-monsoon values record a comparatively greater
rise compared to Tmonsoon and Twinter. From this, we can infer that for Maha-
rashtra state, the summer and SW monsoon seasons (i.e., the period from March
to September) have become relatively warmer, while the post-monsoon and winter
seasons (i.e., from October to February) have become slightly cooler. As far as the
spatial variation is concerned, among the four subdivisions of Maharashtra, Marath-
wada and Vidarbha record the maximum rise in annual as well as season-wise mean
temperature and have become more warmer since the 1980s. On the other hand,
Konkan experiences the highest post-monsoon temperature. Moreover, the effect
of the warming climate in the Konkan and Madhya Maharashtra subdivisions is
comparatively less.
3.2.2 Relative Humidity (RH)
On analyzing the annual relative humidity (RHannual), for the Maharashtra state, an
increase from 73.2 to 76.5% between 1981 and 2010 is noticed, followed by a slight
decrease of 1% (from 76.5 to 75.5%) during 2011–2020. At the sub-divisional scale,
the maximum RHannual has been recorded for Konkan, experiencing an increase
from 85.8 to 87.1% between 1981 and 2010, followed by a slight decrease to 86.5%
@seismicisolation
@seismicisolation
during 2011–2020 (an overall 0.7% increase from 1981 to 2020). Madhya Maha-
rashtra follows the Konkan subdivision, experiencing an increase from 74.7 to 78.2%
between 1981 and 2010, followed by a slight decrease to 77.3% during 2011–2020
(an overall 3.6% increase from 1981–1990 to 2011–2020) (Fig. 4f). Interestingly,
RHannual remains mostly consistent (with minimum fluctuations) during the study
period. During the same time, eastern Maharashtra [consisting of Marathwada and
Vidarbha] experiences a lower RHannual than western Maharashtra [Fig. 5 (I and II)
e–h].
Similar observations are also noticed in season-wise relative humidity (RHseason)
in Maharashtra state as well as at the sub-divisional scale (here, RHseason corre-
sponds to RHwinter, RHsummer, RHmonsoon, and RHpost-monsoon). Considering
the entire state, the RHwinter records the second-least relative humidity among all
four seasons; it moderately increases from 65.1 to 69% between 1981–1990 and
2001–2010 and further increases to 70.4% during 2011–2020, thereby recording an
overall increase of 8.1%. The RHsummer records the least relative humidity among
all the seasons; it moderately increases from 54.8 to 59.6% between 1981–1990 and
2001–2010 and then slightly decreases to 58.4% during 2011–2020, thereby still
recording an overall increase of 6.5%. The RHmonsoon experiences the highest rela-
tive humidity; it slightly increases from 89.4 to 90% between 1981–1990 and 2001–
2010 and then slightly decreases to 88.3% during 2011–2020, thereby recording an
overall decrease of 1.2%. The RHpost-monsoon records the second-highest relative
humidity among all seasons; it moderately increased from 75.3 to 80.2% between
1981–1990 and 2001–2010 and then declined to 78.8% during 2011–2020, thereby
still recording an overall increase of 4.7% (Fig. 4f-j). Given the season-wise spatial
variations in the RHseason between a maximum value lying in the 1981–2010 period
(primarily located in the 2001–2010 period) and 2011–2020 [Fig. 5 (I and II) e–h],
for the winter season, the maximum rise in the RHwinter is recorded for Vidarbha (a
rise of 1.8% = from 69.2 to 71% between the periods mentioned above) followed by
Madhya Maharashtra (a rise of 1.8% = from 69.2 to 70.5%); however, the maximum
RHwinter is observed in the Konkan (80.9% during 2011–2020). For the summer
season, a fall is observed for RHsummer across all the subdivisions, wherein the
maximum fall is recorded for Marathwada (a fall of 3.2% = from 42.5 to 39.3%),
followed by Madhya Maharashtra (a fall of 1.2% = from 64.6 to 63.4%); however,
the maximum RHsummer is observed in the Konkan only (84.6% during 2011–
2020). For the SW monsoon season also, a fall is observed for RHsummer across
all the subdivisions, wherein the maximum fall is recorded for Marathwada (a fall
of 2.3% = from 87.1 to 84.9%) followed by Madhya Maharashtra (a fall of 2% =
from 92 to 90%); however, as usual, the maximum RHsummer is observed in the
Konkan subdivisions only (92.8% during 2011–2020). While for the post-monsoon
season, RHpost-monsoon records fall for all the subdivisions except Konkan. The
maximum fall is recorded for Vidarbha (a fall of 3.8% = from 83 to 79.2%) followed
by Marathwada (a fall of 3.3% = from 79.8 to 76.5%); however, the maximum
RHsummer is observed in the Konkan subdivisions only (83.6% during 2011–
2020). In general, RHannual experiences an overall maximum increase of 5.3%
in Marathwada followed by 3.6% in Madhya Maharashtra; RHwinter experiences
@seismicisolation
@seismicisolation
an overall maximum increase of 18.6% in Vidarbha followed by 16.3% in Marath-

wada; RHsummer experiences an overall maximum increase of 5.3% in Marathwada
followed by a 3.6% in Madhya Maharashtra; contrarily, RHmonsoon experiences a
general reduction, wherein maximum decrement of 1.5% is observed in Madhya
Maharashtra and Konkan; while RHpost-monsoon experiences an overall maximum
increase of 7.5% in Madhya Maharashtra followed by a 6.4% in Marathwada.
To summarize, the relative humidity across all seasons remained comparable
between 1981–2010 and 2011–2020. Within these periods, a slight increment in
RHannual and RHseason is observed from 1981 to 2010, followed by a slight decrease
from 2011 to 2020. This slight increase or decrease between 1981–2010 and 2011–
2020 is always at most 10%. However, moderate changes can be recognized when
comparing 1981–1990 to 2011–2020. For example, Vidarbha experiences an 18.6%
rise in RHwinter, followed by a 16.3% rise in Marathwada. Remarkably, RHannual
and RHseason remain mostly consistent (with minimal fluctuations) across the study
period. In contrast, eastern Maharashtra (Vidarbha and Marathwada) experiences a
lower RHannual as compared to western Maharashtra (Konkan and Madhya Maha-
rashtra) [Fig. 5 (I and II) e–h]. On a spatial and seasonal scale (when comparing 1981–
2010 to 2011–2020), even though the highest RHseason is recorded for Konkan for
all the seasons, the Vidarbha and Marathwada (having comparatively lower magni-
tudes of RHseason) record the maximum rise in the RHseason. Furthermore, when
comparing 1981–1990 to 2011–2020, Marathwada and Madhya Maharashtra (having
comparatively lower and moderate magnitudes of RHseason, respectively) record the
maximum rise or fall in the RHseason. These are the same subdivisions witnessing
the higher Tseason, while Madhya Maharashtra and Konkan are experiencing moder-
ately lower Tseason amidst higher RHseason. The indication of a possible correlation
between mean temperature and relative humidity in the subdivisions (as evident from
the above discussion) demands further attention while probing into the influence of
these variables on changing climate.
3.2.3 Rainfall (P)
Considering the year-wise (annual) mean rainfall (Pannual), the Pannual for the
Maharashtra state records a consistent increment. Pannual during 1981–1990 is aver-
aged to 1269.8 mm that, increases to 1321.2 mm in 1991–2000, then to 1401.1 mm
during 2001–2010, and further to 1459.5 mm in 2011–2020, thereby recording an
overall increase of 14.9% between 1981–1990 and 2011–2020. As far as the subdivi-
sions are concerned, maximum Pannual has been recorded for Konkan experiencing
a consistent increment: 2319.7 mm during 1981–1990, 2312.4 mm during 1991–
2000, 2540 mm during 2001–2010, and 2611.3 mm during 2011–2020, thereby an
overall increase of 12.6% between 1981–1990 and 2011–2020. Magnitude-wise, the
Konkan subdivision is followed by Vidarbha, Madhya Maharashtra, and then Marath-
wada, each experiencing an overall increase from 1228.9 to 1354.3 mm, 973.6 to
@seismicisolation
@seismicisolation
1178.9 mm, and 772.2 to 913.7 mm between 1981–1990 and 2011–2020, respec-
tively. However, the maximum increase in the Pannual is observed for Madhya Maha-
rashtra (21.1%), followed by Marathwada (18.3%), Konkan (12.6%), and Vidarbha
(10.2%) (Fig. 4c).
Similar observations are recorded for season-wise mean rainfall (Pseason) for
the Maharashtra state and its subdivisions (here, Pseason corresponds to Pwinter,
Psummer, Pmonsoon, and Ppost-monsoon). Considering the entire state and the study
period between 1981–1990 and 2011–2020, the Pwinter records the second-least
rainfall magnitude among all seasons; it drastically reduces from 10.2 to 7.6 mm,
recording a fall of 25.7%. The Psummer records the least rainfall magnitude among
all seasons; it slightly reduces from 41.6 to 38.5 mm, recording a fall of 7.5%. The
Pmonsoon experiences the highest rainfall; it moderately increases from 1126.4 to
1285.3 mm, recording a rise of 14.1%. The Ppost-monsoon experiences the second-
highest rainfall; it drastically increases from 91.6 to 128.1 mm, recording the highest
rise of 39.9%. Given the season-wise spatial variations in the Pseason (considering
1981–1990 and 2011–2020) [Fig. 5 (I and II) i–l], for the winter season, the maximum
Pwinter is recorded for Vidarbha [drastically decreases from 30 to 20.8 mm (30.8%
fall) between the periods as mentioned above] followed by Marathwada [drasti-
cally reduces from 10.5 to 7.1 mm (32.6% fall)]. Comparatively, very low rainfall
magnitude is recorded for Konkan and Madhya Maharashtra during winter. Besides,
summer receives more rainfall than winter but less than the post-monsoon season. The
maximum Psummer is recorded for Vidarbha [moderately decreases from 47.2 to 38
mm (19.6% fall)] followed by Marathwada [slightly decreases from 37.1 to 34.2 mm
(7.6% fall)] and Madhya Maharashtra [slightly increases from 37 to 37.7 mm (2%
rise)]. For the SW monsoon season, the maximum Pmonsoon is recorded for Konkan
[moderately increases from 2159.8 to 2392.9 mm (10.9% rise)] followed by Vidarbha
[moderately increases from 1059.5 to 1207.1 mm (13.9% rise)]. While Marath-
wada and Madhya Maharashtra experience comparatively lesser rainfall with drastic
changes (21.9% and 19.5%, respectively). While, for the post-monsoon season,
Ppost-monsoon is again dominated by Konkan [drastically increases from 115 to
185.2 mm (61.1% rise)] followed by Madhya Maharashtra [drastically increases
from 80.4 to 119.8 mm (48.9% rise)] (Fig. 4k–o).
To summarize, a consistent increase in the rainfall magnitude (Pannual) is
observed when the entire Maharashtra state is considered between 1981–1990 and
2011–2020. However, at the seasonal scale, both Pwinter and Psummer record
drastic and slight decrements in rainfall, respectively. While both Pmonsoon and
Ppost-monsoon record moderate and drastic decrements in rainfall, respectively.
Subdivision-wise, Konkan is observed to receive the highest magnitude of rain-
fall during Pmonsoon and Ppost-monsoon (followed by Vidarbha in Pmonsoon and
Madhya Maharashtra in Ppost-monsoon). In contrast, Vidarbha and Marathwada
are observed to dominate during Pwinter and Psummer. Moreover, decadal varia-
tions can be observed in the rainfall pattern such that more rain occurred during
1991–2000 and 2011–2020 than in 1981–1990 and 2001–2010. More or less, central
Maharashtra (i.e., Marathwada and Madhya Maharashtra) is witnessing compara-
tively less Pannual and Pseason, as compared to Konkan and Vidarbha [representing
@seismicisolation
@seismicisolation
other ends (west and east) of Maharashtra state]. In fact, [52], in their investigation
of changes in rainfall magnitude in Konkan, Madhya Maharashtra, and Marathwada,
indicated a 10–20% change by 2100. Given the temperature and relative humidity,
Sects. 3.2.1 and 3.2.2 summarized that eastern Maharashtra is experiencing higher
temperatures while western Maharashtra is experiencing higher relative humidity.
Coherently, past studies such as [53] and Korade and [54] also reported the dominance
of warming scenarios over the state of Maharashtra. Evidently, the Konkan region’s
highest humidity and moderate temperature (under the influence of the coastal envi-
ronment) may be a possible cause behind the highest rainfall magnitudes. The studies
conducted by [39] reported similar inferences for the Konkan subdivision. Therefore,
the highest temperature and lowest humidity in central and eastern Maharashtra may
be a probable cause behind comparatively lower rainfall. These inferences demand
further cross-validation, in that the increasing and decreasing trend of meteorological
variables at a seasonal scale should be scaled down to a monthly scale and thereby
requires to be tested statistically (refer to Sect. 3.3). Furthermore, the anomaly that
exists therein is to be identified spatially and temporally (refer to Sect. 3.4) to ascertain
any possible correlation between T, RH, and P.
3.3 Statistical and Correlational Analysis of Month-Wise

Meteorological Variables for 1981–2020
Monthly mean magnitudes of T, RH, and P variables (Tmonth, RHmonth, and

Pmonth) for the period 1981–2020 along with its statistical properties (MK, p-value,
and SS) for Maharashtra state and its four subdivisions viz., Konkan, Madhya Maha-
rashtra, Marathwada, and Vidarbha are shown in Figs. 6, 7, and 8. The sections
below (Sect. 3.3.1, 3.3.2 and 3.3.3) draw variable-wise inferences across the water
year (from 1st June to 31st May of the following calendar year). The last subsection
(Sect. 3.3.4) demonstrates a possible correlation between T, RH, and P variables
using the Pearson correlation coefficient to ascertain any causal effect.
3.3.1 Temperature (T)
Considering the Tmonth, which represents the mean of a particular month based
on 40 years database, the Maharashtra state records the maximum Tmonth during
summer (TMay is the peak attaining 26.6 °C), thereby experiencing a continuous fall
from May to January of the following year (TJan is the valley point reaching 16.6 °C)
(Fig. 6a). From February onwards, a continuous rapid increase in Tmonth is observed
until May. Imperatively, except for January, all other months show a significant trend
at a 95% confidence interval such that their p-value < 0.05. The positive value of
SS (or Qc ) for all months indicates an increasing trend. Particularly, the summer
@seismicisolation
@seismicisolation
Mean (Maharashtra) MK (0 or 1) p-value SS
30 0.07
Temperature (oC) and MK

25 0.06
0.05
p-value and SS
20
0.04
15
0.03
10
0.02
5 0.01
1 1 1 1 1 1 1 1 1 1 1
0
0 0.00
Jan Feb Mar Apr May Jun Jul Aug Sept Oct Nov Dec
Time (monthly)
(a)
Mean (Konkan) MK (0 or 1) p-value SS Mean (Madhya Maharashtra) MK (0 or 1) p-value SS
30 0.1200 30 0.07

25 0.1000 25 0.06
0.05
p-value and SS
p-value and SS
20 0.0800 20
0.04
15 0.0600 15
0.03
10 0.0400 10
0.02
5 0.0200 5 0.01
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
0 0 0 0
0 0.0000 0 0.00
Jan Feb Mar Apr May Jun Jul Aug Sept Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sept Oct Nov Dec
Time (monthly) Time (monthly)
(b) (c)
Mean (Marathwada) MK (0 or 1) p-value SS Mean (Vidarbha) MK (0 or 1) p-value SS
30 0.08 30 0.09
0.08
0.07
25 25
0.06 0.07
p-value and SS
p-value and SS
20 20 0.06
0.05
0.05
15 0.04 15
0.04
0.03
10 10 0.03
0.02 0.02
5 5
1 1 1 1 1 1 1 1 1 1 1 0.01 1 1 1 1 1 1 1 1 1 1 0.01
0 0 0
0 0.00 0 0.00
(d) (e)
Fig. 6 Monthly trend analysis for temperature (red in color) for the study period 1981–2020 at a
monthly scale using Mann Kendall’s (MK) test (providing p-value) and Sen’s Slope (SS) estimator
along with their monthly mean values for the subdivisions a Maharashtra state, b Konkan, c Madhya
Maharashtra, d Marathwada, and e Vidarbha
and post-monsoon seasons are witnessing a rapid increment in the Tmonth such
that the maximum rate of increase can be observed for November [TNov is rising
significantly at the highest rate of (@) 0.06 °C/year] followed by April (TApr is
rising significantly @ 0.05 °C/year). While the winter season immediately following
summer and undergoes rapid increment in Tmonth (for example, TFeb is rising
significantly @ 0.04 °C/year) except for January, which is experiencing the least
rate of increase (TJan is rising insignificantly @ 0.01 °C/year) among all months.
In general, the least rate of increase in the Tmonth is observed for the SW monsoon
season, which is marginally lesser than other seasons, ranging significantly at 0.03 °C/
year.
On a spatial scale, all the Maharashtra subdivisions followed a similar trend of
attaining maximum Tmonth during May and minimum Tmonth during December,
January, and February, marking the latter as the coldest months. For Konkan, TMay
@seismicisolation
@seismicisolation
100 1.00
Relative Humidity (%) and MK

80 0.80
p-value and SS
60 0.60
40 0.40
20 0.20
0 1 1 1 0 0 0 0 0 0 0 1
0 0.00
-20 -0.20
Time (monthly)
(a)
100 1.00 100 1.00


80 0.80 80 0.80
p-value and SS
p-value and SS
60 0.60 60 0.60
40 0.40 40 0.40
20 0.20 20 0.20
0 0 0 0 1 1 1 1 1 0 0 1 0 1 1 0 0 0 0 0 0 0 0 1
0 0.00 0 0.00
-20 -0.20 -20 -0.20
(b) (c)
100 1.00 100 1.00

80 0.80 80 0.80
p-value and SS
60 0.60 60 0.60
Axis Title
40 0.40 40 0.40
20 0.20 20 0.20
0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0
0 0.00 0 0.00
-20 -0.20 -20 -0.20
(d) (e)
Fig. 7 Monthly trend analysis for relative humidity (light blue in color) for the study period 1981–
2020 at a monthly scale using Mann Kendall’s (MK) test (providing p-value) and Sen’s Slope (SS)
estimator along with their monthly mean values for the subdivisions a Maharashtra state, b Konkan,
c Madhya Maharashtra, d Marathwada, and e Vidarbha
is 25.9 °C followed by 25.5 °C for TJun, while TJan is 18 °C followed by 18.9 °C for
TFeb. The maximum rate of increase in the Tmonth is recorded for April (TApr rising
significantly @ 0.05 °C/year) followed by November (TNov rising significantly @
0.04 °C/year), thereby all other months rising significantly @ 0.03 °C/year, while
the minimum rate is observed for December (TDec rising insignificantly @ 0.02 °C/
year) and January (TJan least rising insignificantly @ 0.004 °C/year) (Fig. 6b). For
Madhya Maharashtra, TMay is 24.8 °C followed by 24.3 °C for TJun, while TJan is
15.9 °C followed by 16.3 °C for TDec (both maximum and minimum Tmonth are
higher and lower than Konkan, respectively). The maximum rate of increase in the
Tmonth is recorded for April (TApr rising significantly @ 0.06 °C/year) followed
by November (TNov significantly rising @ 0.05 °C/year), thereby all other months
(rising significantly @ 0.03 °C/year) and minimum rate is observed for January
(TJan rising insignificantly @ 0.0008 °C/year) (all the rising rates are comparable to
@seismicisolation
@seismicisolation
400 4.00
350 3.50
Rainfall (mm) and MK

300 3.00
p-value and SS
250 2.50
200 2.00
150 1.50
1
100 1.00
50 0.50
0 0 0 0 0 0 0 0 0 0 0
0 0.00
-50 -0.50
Time (monthly)
(a)
900 6.00 350 3.50
750 5.00 300 3.00


250 2.50
600 4.00
p-value and SS
p-value and SS
200 2.00
450 3.00
150 1.50
300 2.00
100 1.00
150 1.00
50 0.50
0 0.00 0 0.00
-150 -1.00 -50 -0.50
(b) (c)
240 4.00 400 4.00

3.50 350 3.50
200
300 3.00
3.00
160 250 2.50
p-value and SS
p-value and SS
2.50
200 2.00
120 2.00
150 1.50
80 1.50
100 1.00
1.00
40 50 0.50
0 0 0 0 0 0 0 0 1 0 0 0
0 0 0 0 0 0 0 0 1 0 0 0 0.50 0 0.00
0 Jan Feb Mar Apr May Jun Jul Aug Sept Oct Nov Dec
0.00 -50 -0.50
-40 -0.50 -100 -1.00
(d)
(e)
Fig. 8 Monthly trend analysis for rainfall (dark blue in color) for the study period 1981–2020 at a
monthly scale using Mann Kendall’s (MK) test (providing p-value) and Sen’s Slope (SS) estimator
along with their monthly mean values for the subdivisions a Maharashtra state, b Konkan, c Madhya
Maharashtra, d Marathwada, and e Vidarbha
Maharashtra state and higher than Konkan). Apart from TJan, it is TDec that has an
insignificant positive trend (Fig. 6c). For Marathwada, TMay is 27.6 °C, followed by
25.8 °C for TApr. In comparison, TDec is 16 °C, followed by 16.3 °C for TJan (both
maximum and minimum Tmonth are higher and lower than the attributes for the
Maharashtra state, Konkan, and Madhya Maharashtra, respectively). The maximum
rate of increase in the Tmonth is recorded for May (not April) and November (both
TMay and TNov rising significantly @ 0.07 °C/year), followed by April and February
(both TApr and TFeb rising significantly @ 0.06 °C/year), thereby all other months
(rising significantly @ ranging between 0.05 °C/year and 0.03 °C/year) and minimum
rate is observed for January (TJan rising insignificantly @ 0.02 °C/year) (all the
rising rates are greater than Maharashtra state, Konkan, and Madhya Maharashtra)
(Fig. 6d). For the last subdivision—the Vidarbha, TMay is 29.5 °C followed by
@seismicisolation
@seismicisolation
27.7 °C for TJun, while TDec is 15.8 °C followed by 16 °C for TJan (both maximum
and minimum Tmonth are higher and lower than the attributes for the Maharashtra
state and all the other subdivisions, respectively). The maximum rate of increase
in the Tmonth is recorded for November (not April and not May) (TNov rising
significantly @ 0.09 °C/year), followed by December and then October and May
(TDec, TOct, and TMay rising significantly @ 0.07 °C/year, @ 0.06 °C/year, and @
0.06 °C/year, respectively). Rest all other months are rising significantly @ ranging
between 0.05 °C/year and 0.03 °C/year, whereas the minimum rate is observed for
January (TJan rising insignificantly @ 0.02 °C/year). To infer, all the rising rates
of Vidarbha are greater than Maharashtra state and all the other subdivisions. Apart
from TJan, TJun has an insignificant positive trend (Fig. 6e).
Evidently, the findings mentioned above affirm the inferences drawn from
Sect. 3.2.1 that the Maharashtra state (if considered at a glance) and its subdivi-
sions are approaching a warming climate such that the winter seasons are becoming
less cooler, and summer and post-monsoon seasons are becoming warmer. More
specifically, the Vidarbha and Marathwada subdivisions (or eastern Maharashtra)
are not only experiencing higher temperatures but comparatively and significantly
greater rising trends than the average rates of the Maharashtra state and then the
Konkan and Madhya Maharashtra (or western Maharashtra).
3.3.2 Relative Humidity (RH)
Considering the RHmonth, which represents the mean of a particular month based
on 40 years database, the Maharashtra state records the maximum RHmonth during
the SW monsoon, specifically August and September (RHAug is the peak attaining
92.1%, followed by 92% in RHSep), thereby experiencing a continuous fall from
September to April of the following year (RHApr is the valley point attaining 54.6%)
(Fig. 7a). From April onwards, a continuous rapid increase in RHmonth is observed
until September. Imperatively, only four months, viz., February, March, April, and
December, could show a significant trend at a 95% confidence interval such that their
p-value < 0.05, while the remaining months show an insignificant trend. More or less,
this finding indicates that winter and summer seasons have recorded significant trends
against insignificant trends in the SW monsoon and post-monsoon seasons. Further-
more, the positive value of SS (or Qc ) for all months in the post-monsoon, winter,
and summer seasons indicates an increasing trend. In contrast, the negative values
of SS are observed for SW monsoonal months. Nevertheless, the rate increment or
decrement is marginal and ranges between 0.22%/year and − 0.05%/year.
On a spatial scale too, all the subdivisions of Maharashtra followed a similar
trend of attaining maximum RHmonth during SW monsoonal months (such as
August and September) and minimum RHmonth during summer months (such as
March and April), marking the former to cause the maximum rainfall and the latter
to minimum rainfall, thereby leading to acute dryness (……). For Konkan, the
maximum RHmonth is recorded for RHAug (94.7%), followed by RHSep (94.4%)
and RHJul (94.3%), while the minimum RHmonth is recorded for RHDec (77.3%),
@seismicisolation
@seismicisolation
followed by RHFeb (78.8%) and RHJan (80%). Except for the months from May
to September and December, all other months have insignificant trends. Summer
and SW monsoon months showcase negative trends, while post-monsoon and winter
months follow positive trends. As usual, the rate increment or decrement is marginal
and ranges between 0.17%/year and -0.05%/year (Fig. 7b). For Madhya Maha-
rashtra, the maximum RHmonth is recorded for RHSep (93%), followed by RHAug
(92.7%) and RHJul (91.5%). In comparison, the minimum RHmonth is recorded for
RHMar (55%) followed by RHApr (59.2%) and RHFeb (63.5%). All other months
have insignificant trends except for February, March, and December. Like Konkan,
summer and SW monsoon months showcase negative trends, while post-monsoon
and winter months follow positive trends. As usual, the rate increment or decre-
ment is marginal and ranges between 0.24%/year and -0.08%/year (Fig. 7c). For
Marathwada, the maximum RHmonth is recorded for RHSep (89.8%) followed by
RHAug (89.6%) and RHJul (87.3%). In comparison, the minimum RHmonth is
recorded for RHApr (33%), followed by RHMar (40.3%) and RHMay (46.1%).
Except for February and March, all other months have insignificant trends. Unlike
Konkan and Madhya Maharashtra, the months in the SW monsoon showcase negative
trends, while the post-monsoon, winter, and summer months (except May) follow
positive trends. As usual, the rate increment or decrement is marginal and greatly
ranges between 0.36%/year and − 0.14%/year (Fig. 7d). For the last subdivision—
the Vidarbha, the maximum RHmonth is recorded for RHAug (91.1%), followed
by RHSep (90.4%) and RHJul (88.6%). In comparison, the minimum RHmonth is
recorded for RHMay (38.3%) followed by RHApr (39.2%) and RHMar (49.7%).
Except for April, all other months have insignificant trends. Like Marathwada, a
couple of months in the SW monsoon (August and September) showcase negative
trends, while post-monsoon, winter (except January), and summer months follow
positive trends. As usual, the rate increment or decrement is marginal and greatly
ranges between 0.36%/year and − 0.07%/year (Fig. 7e).
Evidently, the aforementioned findings affirm the inferences drawn from
Sect. 3.2.2. The very low positive and negative SS values indicate that the change
in the relative humidity across the study period and across the seasons remained
comparable. Moreover, the trends mostly remained insignificant (MK = 0) for most
months (not only for Maharashtra state but also for its subdivisions). In particular,
the Vidarbha and Marathwada subdivisions (or eastern Maharashtra) experience a
greater range of change between maximum and minimum values than Konkan and
Madhya Maharashtra. Nevertheless, the fluctuations are marginal, and the rates are
statistically insignificant. Magnitude-wise, western Maharashtra has higher relative
humidity. For example, in Konkan, relative humidity ranges from 94.4 to 78.8%,
followed by Madhya Maharashtra (93 to 55%). Whereas for eastern Maharashtra,
the relative humidity is lower. For example, in Vidarbha, it ranges from 91.1 to 38.3%,
followed by Marathwada (89.8 to 33%). Eastern Maharashtra is experiencing higher
temperatures against lower relative humidity than moderate to lower temperatures
against higher relative humidity in western Maharashtra. Although the present inves-
tigation attempted to define the possible correlation between temperature and relative
@seismicisolation
@seismicisolation
humidity (in Sect. 3.5), it also directs future studies to be conducted on understanding
the influence of these variables on the historical climate.
3.3.3 Rainfall (P)
Considering the Pmonth, which represents the mean of a particular month based on
40 years database, the Maharashtra state records the maximum Pmonth during the SW
monsoon season (TJul) is the peak attaining 400.3 mm), followed by post-monsoon,
winter, and summer, thereby marking SW monsoon as the principal rainfall season for
Maharashtra (Fig. 8a). From March onwards, a continuous rapid increase in Pmonth is
observed until July, which begins to fall from July to February. Imperatively, except
for September, all other months show an insignificant trend at a 95% confidence
interval such that their p-value is greater than 0.05. The positive value of SS (or
Qc ) for all months (except for January and August) indicates an increasing trend.
Particularly, the months of the SW monsoon season are witnessing a rapid increment
in the Pmonth such that the maximum rate of increase can be observed for September
[TSep is rising significantly at the highest rate of (@) 3.3 mm/year] followed by July
(PJul is rising significantly @ 2.1 mm/year). In general, the rate of increase in the
Pmonth is observed to be marginal for post-monsoon, winter, and summer seasons,
ranging insignificantly between 0.95 mm/year and − 0.05 mm/year.
On a spatial scale, too, all the subdivisions of Maharashtra followed a similar trend
of attaining maximum Pmonth during SW monsoonal months (particularly July and
August) and minimum RHmonth during summer months. For Konkan, the maximum
Pmonth is recorded for PJul (796.4 mm), followed by PAug (635.6 mm) and PJun
(504.4 mm), while the minimum or negligible Pmonth is recorded for December to the
following April (ranging from 0.8 to 5.8 mm). Except for September and October, all
other months have insignificant trends. Months in post-monsoon, winter, and summer
seasons showcase nearly negligible trends, ranging insignificantly between zero and
1.93 mm/year. At the same time, the SW monsoon follows positive trends, as high
as 3.6 mm/year in June to 5.1 mm/year in September (Fig. 8b). The increment rates
for Konkan are comparatively higher than the mean rate of the Maharashtra state.
For Madhya Maharashtra, the maximum Pmonth is recorded for PJul (307.2 mm),
followed by PAug (260.6 mm) and PJun (208.7 mm). In comparison, the minimum
or negligible Pmonth is recorded for December to the following April (ranging from
1.5 to 10.7 mm), while May month received 25 mm of the rainfall. Except for
September, all other months have shown insignificant trends. Similar to Konkan,
months in post-monsoon, winter, and summer seasons showcase nearly negligible
trends; ranging insignificantly between − 0.01 mm/year and 0.1 mm/year, while SW
monsoon follows positive trends, as high as 2.2 mm/year in July to 2.7 mm/year in
September (Fig. 8c). The rates of increment are comparable to the Maharashtra state.
For Marathwada, the maximum Pmonth is recorded for PJul (213.1 mm), followed by
PAug (202.2 mm) and PJun (157.6 mm), while the minimum or negligible Pmonth is
recorded for December to the following April (ranging from 3.4 to 10.2 mm), while
October, November, and May month received 71.4 mm, 17.4 mm, and 15.7 mm of the
@seismicisolation
@seismicisolation
rainfall, respectively. Except for September, all other months have shown insignifi-
cant trends. Similar to Konkan and Madhya Maharashtra, months in post-monsoon,
winter, and summer seasons showcase nearly negligible trends, ranging insignifi-
cantly between − 0.05 mm/year and 0.37 mm/year, while SW monsoon follows posi-
tive trends, as high as 2.4 mm/year in July to 3.5 mm/year in September (Fig. 8d). The
rates of increment are comparable to the Maharashtra state. For the last subdivision—
Vidarbha, the maximum Pmonth is recorded for PJul (370.3 mm), followed by PAug
(351.2 mm) and PJun (200.6 mm). In contrast, the minimum or negligible Pmonth is
recorded for December to the following April (ranging from 7.7 to 13.7 mm), while
October, November, and May received 71.4 mm, 17.4 mm, and 15.7 mm of rain-
fall, respectively. Except for September, all other months have shown insignificant
trends. Similar to Konkan, Madhya Maharashtra, and Marathwada, months in post-
monsoon, winter, and summer seasons showcase nearly negligible trends, ranging
insignificantly between − 0.05 mm/year and 0.36 mm/year, while SW monsoon
follows positive trends, as high as 3.3 mm/year in July to 3.5 mm/year in September
(Fig. 8d). The rates of increment are comparable to the Maharashtra state.
Evidently, the findings mentioned above affirm the inferences drawn from
Sect. 3.2.3. The SW monsoon marks the principal monthly rainfall period for the
state of Maharashtra, which is additionally experiencing the highest increase in the
rate of rainfall between 1981 and 2020 for the state and all of its subdivisions.
However, the rates for the other months are mostly negligible. Moreover, the trend is
mostly insignificant for every month (except September). In particular, the Madhya
Maharashtra, Marathwada, and Vidarbha subdivisions are experiencing a compara-
tively lower magnitude of rainfall, such that the former two are experiencing an acute
water crisis due to a lack of erratic rainfall. Magnitude-wise, Konkan receives the
highest monthly rainfall (500 to 800 mm), followed by Vidarbha (200 to 370 mm),
and Madhya Maharashtra (200 to 300 mm), while Marathwada receives the least
(150 to 300 mm).
3.3.4 Correlation Analysis Between Meteorological Variables
Considering the findings from the previous subsections, the Konkan is experi-
encing high rainfall mainly due to high relative humidity and moderate temperature
conditions yielding a suitable climate. Conversely, Marathwada and Vidarbha are
witnessing a lower magnitude of rainfall due to higher temperatures and lower rela-
tive humidity. The climate of Madhya Maharashtra lies at the interface of Konkan and
eastern Maharashtra, such that moderate temperature and relative humidity condi-
tions exist. Since the results also pointed to the insignificance of the trend on a monthly
scale for rainfall and relative humidity, the influence of both these variables on the
temperature of the study sites and the entire climate is subject to future research.
Nevertheless, preliminary correlation analysis has been conducted to address this
inference and depict the possible association between these variables (Fig. 9).
@seismicisolation
@seismicisolation
Fig. 9 Analysis of meteorological variables, viz., a temperature, b relative humidity, and c rainfall,
is shown through time series (left) and scatter plots (right). In the scatter plots, the x-axis and y-axis
represent the corresponding meteorological variables. The dotted red lines show the 1:1 line (line
of perfect correlation), and the solid black line shows the best-fit lines for the scatter plots
The time series and scatter plots for each of the three meteorological variables
are shown in Fig. 9. Three possible combinations are used: temperature with rain-
fall (Fig. 9a), temperature with relative humidity (Fig. 9b), and relative humidity
with rainfall (Fig. 9c). Visual interpretation indicates a corresponding decrease in
the temperature values with the increase in rainfall magnitude (Fig. 9a_left). For
example, between 2010 and 2020, the temperature in the study site remains the
highest compared to any decade, wherein rainfall magnitude is comparatively lower.
Hence, the PCC value is obtained less (~ 0.44), indicating that the peak points of
temperature and rainfall do not coincide, especially when they attain their highest
values (Fig. 9a_right). Visual inspection depicts a negligible correlation between
temperature and relative humidity, such that even after the drastic increment in the
temperature from 1981 to 2020, hardly any major change can be observed for relative
humidity (Fig. 9b_left). Hence, the PCC value is much less (~ 0.09), indicating a
rare coincidence of peak temperature with relative humidity (Fig. 9b_right). On the
contrary, a very high correlation is observed between relative humidity and rainfall,
such that both variables have witnessed a slight increase since 1981 (spatial disparity
exists) (Fig. 9c_left). Hence, the PCC value is much greater (~ 0.72), indicating that
the high magnitude of the rainfall may be primarily driven due to corresponding high
relative humidity (Fig. 9c_right). As was inferred previously, since the SS values for
relative humidity are very low than rainfall, the increment rate is observed to be lower
than that of rainfall.
To summarize, the correlation’s strength is moderate between temperature and
rainfall and poor with relative humidity. While the same was very strong between
@seismicisolation
@seismicisolation
relative humidity and rainfall. However, spatial disparity indicates that parts of
eastern Maharashtra experiencing high temperatures have moderate to low rainfall
and relative humidity. Similarly, parts of western Maharashtra experiencing moderate
temperatures have moderate to high rainfall and relative humidity. Reasonably, for
the Maharashtra state, rainfall and relative humidity jointly govern the tempera-
ture, so an inverse correlation exists between them. Besides, the fluctuations in the
rainfall remain the highest as compared to other variables (considering their stan-
dard deviation values). Although annual trends remain significant, monthly rainfall
trends and relative humidity are mostly insignificant. In contrast, as the significance
of the trends for temperature remained consistent at decadal, annual, seasonal, and
monthly scales (concerning findings from Sect. 3.3.1), this study strongly proclaims
the warming of Maharashtra. A similar analogy for the rising temperature, declining
relative humidity, and thereby decreasing and intensifying rainfall has been drawn
by recent studies. Khan et al. [55] concluded in their investigation that an increase
in temperature significantly reduces the amount of precipitation over the region of
Bangladesh, however, insignificant changes were recorded for relative humidity (as
observed in the present case). Similar findings on interrelation among meteorological
variables mentioned above have been reported [24, 35–39, 56]. These inferences can
further be ascertained by studying the anomaly (the following section) by comparing
the recent decade (2011–2020) to the entire study period (1981–2020).
3.4 Subdivision-Wise and Season-Wise Spatio-Temporal

Anomaly
The spatio-temporal anomaly has been estimated for T, RH, and P variables by
subtracting their mean magnitude for 1981–2020 from the mean magnitude for 2011–
2020. The large difference is indicative of a tremendous change (high positive or
negative anomaly) in the particular variable in the recent decade (2011–2020), while a
small difference will indicate the consistency of the variable (low positive or negative
anomaly) across the study period (1981–2020). Figures 10 and 11 show temporal
and spatial anomalies for the three variables mentioned above across Maharashtra
state and its four subdivisions: Konkan, Madhya Maharashtra, Marathwada, and
Vidarbha. The sections below draw variable-wise inferences across the water year
and Maharashtra.
3.4.1 Temperature
Considering the Maharashtra state (refer to Fig. 10a), both summer and post-monsoon
seasons (i.e., before and after the SW monsoon) are experiencing a rapid rise in
Tmonth anomaly, specifically TApr and TMay from summer and TOct, TNov, and
@seismicisolation
@seismicisolation
Fig. 10 Subdivision-wise anomaly average of meteorological variables between 2011–2020 and

1981–2020 for a temperature, b relative humidity, and c rainfall across winter, summer, south-west
monsoon, and post-monsoon
Fig. 11 A spatial anomaly of meteorological variables between 2011–2020 and 1981–2020 for
temperature (a–d), relative humidity (e–h), and rainfall (i–l)
@seismicisolation
@seismicisolation
TDec from post-monsoon. The highest anomaly is recorded for TMay (1.29 °C),
followed by TOct (1.24 °C), TNov (1.13 °C), and TApr (1.12 °C). Every month
generally shows a positive anomaly, including winter and SW monsoon. The lowest
anomaly is recorded for TJan (0.19 °C), followed by TJul (0.61 °C), TAug (0.67 °C),
and TJun (0.68 °C). The reasoning behind the rise and fall in the temperature anomaly
may be better clarified when the anomaly in relative humidity and rainfall is described
(Sects. 3.4.2 and 3.4.3).
Coherent to the temperature anomaly in Maharashtra, a similar pattern is followed
for all subdivisions with varying magnitudes on spatial scales (refer to Figs. 10a and
11a–d). For Konkan, the highest anomaly is recorded for summer and post-monsoon
months [TOct (0.83 °C), followed by TMay (0.76 °C) and TApr (0.70 °C); ~ 60%
of Maharashtra’s magnitude]. In contrast, the lowest anomaly is recorded for winter
months (ranging from −0.21 to 0.45 °C), followed by SW monsoon months (ranging
from 0.45 to 0.52 °C) (~ 70% of Maharashtra’s magnitude). Konkan is experiencing
a moderate increment in the temperature anomaly (from − 0.21 to 0.83 °C), and
thus, the consideration that ’Konkan is approaching a warming climate’ cannot be
ruled out. For Madhya Maharashtra, the highest anomaly is recorded for summer
and post-monsoon months [TMay (1.26 °C), followed by TOct (1.25 °C) and TApr
(1.09 °C); comparable to Maharashtra’s magnitude]. Whereas, the lowest anomaly is
recorded for winter months (ranging from null to 0.79 °C), followed by SW monsoon
months (ranging from 0.61 to 0.79 °C) (again comparable to Maharashtra’s magni-
tude). Since a moderately high (positive) increment in the temperature anomaly
(from 0 to 1.26 °C) is experienced, the consideration that ’Madhya Maharashtra
is approaching a warming climate’ is evident from the present investigation. For
Marathwada, the highest anomaly is recorded for summer and post-monsoon months
[TMay (1.77 °C), followed by TNov (1.63 °C), TOct (1.60 °C), TDec (1.50 °C), and
TApr (1.48 °C); ~ 40% greater than Maharashtra’s magnitude]. At the same time,
the lowest anomaly is recorded for winter months (ranging from 0.62 to 1.25 °C),
followed by SW monsoon months (ranging from 0.72 to 0.88 °C) (~ 30% greater
than Maharashtra’s magnitude). Since a high (positive) increment in the tempera-
ture anomaly (from 0.62 to 1.77 °C) is experienced, the consideration that ’Marath-
wada is approaching a warming climate’ becomes evident. For the last subdivision—
Vidarbha, the highest anomaly is recorded for summer and post-monsoon months
[TOct (1.56 °C), followed by TNov (1.55 °C), TMay (1.49 °C), TApr (1.40 °C), and
TDec (1.38 °C); ~ 35% greater than Maharashtra’s magnitude]. In comparison, the
lowest anomaly is recorded for SW monsoon months (ranging from 0.67 to 1.92 °C),
followed by winter months (ranging from 0.49 to 1.05 °C) (~ 20% greater than Maha-
rashtra’s magnitude). Since a high (positive) increment in the temperature anomaly
(from 0.67 to 1.56 °C) is experienced, considering that ’Vidarbha is approaching a
warming climate’ is evident.
Previous sections’ findings on the temperature in decadal, annual, seasonal, and
monthly time series and spatial analysis for the Maharashtra state and its subdivisions
are affirmed by the results of temperature anomaly. The latter apparently deduced
that 2011–2020 is the warmest in Maharashtra. Among all seasons, the summer
encounters a comparatively greater anomaly in that the summer months have become
@seismicisolation
@seismicisolation
more hotter than in previous decades. Eastern subdivisions are experiencing higher
temperature trends and anomalies, especially in Marathwada. Nevertheless, Madhya
Maharashtra and Konkan too are experiencing gradually rising temperatures with
significant trends. This study thus arrives at the comprehensive supposition that
Maharashtra is an emerging warming state of India.
3.4.2 Relative Humidity
Considering the Maharashtra state (refer to Fig. 10b), post-monsoon, winter, and part
of summer months (excluding May) are experiencing a moderate rise in RHmonth
anomaly, specifically at the interface of winter and summer (RHFeb, RHMar, and
RHApr). A negative anomaly is observed for the SW monsoon months, indicating
that the relative humidity has declined in the recent decade compared to the entire
study period for the monsoonal period. The highest positive anomaly is recorded for
RHFeb (2.73%), followed by RHMar (2.69%), RHApr (1.61%), and RHJan (1.03%).
The highest negative anomaly is recorded for RHMay (− 2.91%), followed by RHSep
(− 1.17%), RHAug (− 1.12%), and RHJul (− 1.01%). Nevertheless, the anomaly is
slight (ranging between − 2.91% and 2.73%), indicating that in the recent decade, the
relative humidity in Maharashtra is not much influenced, instead mostly consistent.
But the highest negative anomaly during RHMay reasonably clarifies why TMay has
become more hotter in Maharashtra.
Coherent to the relative humidity anomaly in Maharashtra, a similar pattern is
followed for all subdivisions with varying magnitudes on spatial scales (refer to
Figs. 10b and 11e–h). For Konkan, the highest positive anomaly is recorded for
post-monsoon, winter, and part of summer months [RHDec (2.25%), followed by
RHJan (1.57%) and RHFeb (1.39%)]. The anomaly for post-monsoon and winter
months is higher in Konkan than in Maharashtra, indicating that non-monsoonal
seasons have become more humid in the recent decade than in 1981–2020. Hence,
this may have become a plausible reason behind the high rainfall magnitudes in these
seasons. However, the exception comes in the summer months when the anomaly for
RHmonth becomes negative such that it continuously falls from RHMar (− 0.35%) to
RHJun (− 1.53%). The highest negative anomaly is recorded for RHJun (− 1.53%),
which remains negative until RHSep (− 0.98%). Therefore, considering the recent
decade, rainfall magnitude is decreasing given declining relative humidity in the SW
monsoon. Thus, the consideration that ’Konkan is approaching a warming climate,’
specifically during part of summer (May), SW monsoon, and post-monsoon, may be
attributed to the declined relative humidity in the recent period.
For Madhya Maharashtra, the highest positive anomaly is recorded for winter,
part of summer months, and post-monsoon [RHFeb (3.10%), followed by RHMar
(2.91%) and RHJan (2.14%)]. Like Konkan, the anomaly for seasons mentioned
above is higher in the Madhya Maharashtra subdivision than in Maharashtra state,
indicating that non-monsoonal seasons have become more humid in the recent
decade. Hence, this may have become a plausible reason behind the considerable
rainfall magnitudes in these seasons. However, the exception comes in the summer
@seismicisolation
@seismicisolation
months when the anomaly for RHmonth becomes negative such that it continuously
increases from the highest negative anomaly month RHMay (− 3.22%) to RHSep (−
1.24%). Therefore, considering the recent decade, rainfall magnitude is decreasing
given declining relative humidity. Thus, the consideration that ’Madhya Maharashtra
is approaching a warming climate,’ specifically during summer and SW monsoon,
may be attributed to the declined relative humidity.
For Marathwada, the highest positive anomaly is recorded at the interface between
late winter and early summer [RHMar (3.41%), followed by RHFeb (3.05%) and
RHApr (1.28%)]. The anomaly for the seasons mentioned above is much higher than
in Maharashtra state, as well as Konkan and Madhya Maharashtra. However, unlike
the latter two subdivisions, the exception comes in the summer months when the
anomaly for RHmonth becomes negative such that it continuously decreases from the
highest negative anomaly month RHMay (− 6.11%) to RHJan (− 0.41%). Hence, this
may have become a plausible reason behind the considerably low rainfall magnitudes
in part of summer, SW monsoon, and post-monsoon seasons. Thus, the consideration
that ’Marathwada is approaching a warming climate,’ specifically during the seasons
mentioned above, may be attributed to the declining relative humidity.
For the last subdivision—Vidarbha, the highest positive anomaly is recorded at
the interface between late winter and early summer [RHApr (3.68%), followed by
RHMar (3.53%) and RHFeb (2.32%)]. The anomaly for the abovementioned seasons
is much higher than in Maharashtra, Konkan, and Madhya Maharashtra but compa-
rable to Marathwada. However, in the summer months, the anomaly for RHmonth
becomes negative such that it continuously decreases from the highest negative
anomaly month RHMay (− 1.80%) to RHJan (− 0.82%). Hence, this may have
become a plausible reason behind the considerably low rainfall magnitudes in part
of summer, SW monsoon, and post-monsoon seasons. Thus, the consideration that
’Vidarbha is approaching a warming climate,’ specifically during the aforementioned
seasons, may be attributed to the declining relative humidity.
Previous sections’ findings on the relative humidity in decadal, annual, seasonal,
and monthly time series and spatial analysis for the Maharashtra state and its subdi-
visions are affirmed by the results of the relative humidity anomaly. The latter appar-
ently deduced that 2011–2020 is experiencing both positive and negative anomalies.
The positive anomalies are mostly observed during the winter and part of summer,
while the negative anomalies remain dominant across the water year, especially
during late summer (May) and the SW monsoon. Among all months, May, known
for recording maximum temperature, has become hotter than in previous decades.
Furthermore, the decline in the relative humidity across SW monsoon and post-
monsoon seasons may have resulted in a reduced rainfall magnitude across the
recent decade than the entire study period. Spatially, eastern subdivisions are expe-
riencing higher negative anomalies than western subdivisions, ascertaining that the
former experiences comparatively lower rainfall magnitudes than the latter. However,
Madhya Maharashtra (part of western Maharashtra) also demonstrates one of the
lowest rainfall-receiving subdivisions indicating that except for Konkan, the entire
Maharashtra is experiencing a decline in relative humidity and, thereby, rainfall. All
these factors might have influenced the temperature because it has risen significantly
@seismicisolation
@seismicisolation
and sharply in the recent decade. Hence, the comprehensive supposition could be
that Maharashtra is an emerging warming state in India.
3.4.3 Rainfall
Considering the Maharashtra state (refer to Fig. 10c), the SW monsoon season
is experiencing a rapid positive rise in Pmonth anomaly, including October from
the post-monsoon season. The highest anomaly is recorded for PSep (37.3 mm),
followed by PJul (35.7 mm) and POct (21.6 mm). Apart from the months mentioned
above, all other months showcase nearly minor anomalies, indicating that only a
few changes have been recorded during the recent decade. An exception comes for
May and November when considerable negative anomalies are recorded (PMay =
− 8.8 mm and PNov = − 7.8 mm). The inference indicates that the peak of the
summer month (May, in response to temperature) records the fall in the percentage
of relative humidity and rainfall magnitude. These recent changes may have altered
the temperature trends for the Maharashtra state, such that the rise is the maximum
(in May) across any decade since 1981.
Coherent to the rainfall anomaly in Maharashtra, a similar pattern is followed for
all subdivisions with varying magnitudes on spatial scales (refer to Figs. 10c and
11i-l). For Konkan, the highest anomaly is recorded for the SW monsoon season
and October [PSep (68.8 mm), followed by POct (42.2 mm) and PJul (40.5 mm)].
In contrast, the lowest anomaly is recorded for May and November (PMay =
− 17.3 mm and PNov = − 8.6 mm). Imperatively, both positive and negative
anomalies are 1.5–2 times Maharashtra’s magnitude. These findings indicate that
Konkan is experiencing heavy rainfall in the SW monsoon due to a considerable
rise in rainfall anomaly in the recent decade. However, it is simultaneously expe-
riencing abrupt fall during late summer and early post-monsoon seasons. These
inferences indicate that non-monsoonal seasons now receive lesser rainfall mainly
due to moderate relative humidity declines and moderate temperature rises, while
the monsoon season receives more significant rainfall. Thus, the consideration that
’Konkan is approaching a warming climate’ can be revised as ’Konkan is grad-
ually approaching a warming climate,’ such that the monsoon has become more
intense and shorter, while the non-monsoon period is gradually becoming hotter. For
Madhya Maharashtra, the highest anomaly is recorded for the SW monsoon season
and October [PJul (32.5 mm), followed by PSep (26.8 mm) and POct (20.1 mm)].
In contrast, the lowest anomaly is recorded for May and November (PNov = −
8.3 mm and PMay = − 6.0 mm). Both positive and negative anomalies are approx-
imately 80–90% of Maharashtra’s magnitude. These findings indicate that Madhya
Maharashtra is experiencing more rainfall in the SW monsoon due to a considerable
rise in rainfall anomaly in the recent decade. However, it is simultaneously experi-
encing a slight fall during late summer and early post-monsoon seasons. These infer-
ences indicate that non-monsoonal seasons receive slightly lesser rainfall, mainly
due to high relative humidity declines and high-temperature rises, against a signif-
icant increase in monsoonal rainfall. Thus, considering that ’Madhya Maharashtra
@seismicisolation
@seismicisolation
is approaching a warming climate’ is more or less acceptable, given the changing

temperature, relative humidity, and rainfall anomaly. For Marathwada, the highest
anomaly is recorded for some months in the SW monsoon season only and not in
October, as is observed for western Maharashtra [PJul (29.5 mm), followed by PSep
(27.5 mm)]. Whereas, unlike west Maharashtra, the lowest anomaly in Marathwada
is recorded for August of the SW monsoon itself, followed by May and November
(PAug = − 8.3 mm, PMay = − 5.8 mm, and PNov = − 5 3 mm). Both moderate
positive and slight negative anomalies are observed, such that the former is lower
than and the latter is greater than Maharashtra’s magnitude. These findings indi-
cate that Marathwada is experiencing moderately more rainfall in the SW monsoon
(except August) due to a considerable rise in rainfall anomaly in the recent decade.
However, it is simultaneously experiencing a slight fall during all other seasons,
such that its negative magnitude is far greater than Maharashtra’s magnitude. These
inferences indicate that non-monsoonal seasons receive lesser rainfall, mainly due
to high relative humidity declines and high-temperature rises, against a moderate
increase in monsoonal rainfall. Thus, considering that ’Marathwada is approaching
a warming climate’ is acceptable, given the abruptly changed temperature, relative
humidity, and rainfall anomaly. For the last subdivision—Vidarbha—the rainfall
anomaly remains comparable to both the Marathwada subdivision and Maharashtra
state. The highest anomaly is recorded for some months in the SW monsoon season
only [PSep (46.8 mm), followed by PJul (37.7 mm)]. The lowest anomaly is recorded
for May and November (PMay = − 8.2 mm and PNov = − 5.0 mm). Both high posi-
tive and slight negative anomalies are observed, such that the former is greater than
and the latter is comparable to Maharashtra’s magnitude. These findings indicate that
Marathwada is experiencing more rainfall in the SW monsoon (except August) due
to a high rise in rainfall anomaly in the recent decade. However, it is simultaneously
experiencing a slight fall during all other seasons. These inferences indicate that
non-monsoonal seasons receive lesser rainfall, mainly due to high relative humidity
declines and high-temperature rises, against a considerable increase in monsoonal
rainfall. Thus, similar to Marathwada, considering that ’Vidarbha is approaching a
warming climate’ is acceptable, given the abruptly changed temperature, relative
humidity, and rainfall anomaly.
To summarize the temperature, relative humidity, and rainfall anomaly for Maha-
rashtra and its subdivisions, evidently, this investigation strongly proclaims that
’Maharashtra state is an emerging rapidly warming state of India.’ The findings
on the behavior of the meteorological variables given the decadal, annual, seasonal,
and monthly scales, alongside their result on statistical significance, get validated
while analyzing variable anomalies. The latter signified the abrupt changes in the
recent decade (2011–2020) compared to the entire study period (1981–2020), such
that the temperature anomaly recorded the highest positive magnitude, wherein rela-
tive humidity and rainfall anomaly appeared as influencing variables behind the
warming climate. On the spatial scales, anomalies indicate that temperature rise
is greater for eastern Maharashtra, while rainfall rise is greater for western Maha-
rashtra. However, the study identified a warming climate for all subdivisions in either
@seismicisolation
@seismicisolation
case. Monsoonal seasons are preceded and proceed with declining rainfall magni-
tude such that summer and non-monsoonal seasons are becoming more hotter and
less cooler while the monsoon season is receiving more rainfall than ever. All these
findings are in coherence with the results obtained from the past literature. Dash et al.
[57] identified similar observations for India, wherein they found that the relative
humidity and rainfall are severely affected such that extreme temperature events are
becoming phenomenal due to a warming climate. The warming climate in the state
of Maharashtra is also documented by Niu et al. [58], Korade and [54], and [53]. In
general, these studies claim that atmospheric warming is responsible for the rise in
moisture content in the atmosphere, which subsequently results in extreme rainfall
[23, 59–64]. Parallelly, the study by Li et al. [23] concluded that warming at the local
scale decreases the amount of precipitation by reducing the moisture content from
the atmosphere. Furthermore, investigations over the regions of Iran showed signif-
icantly rising minimum and mean temperature trends and falling relative humidity
trends. However, no clear rising/decreasing trend for maximum temperature and
relative humidity was observed in Iran case studies [63, 65]. Besides all, multiple
modeling studies focused explicitly on temperature, relative humidity, and rainfall
also indicated the strong interlinkages and dependencies of these variables with each
other and their future predictions [66–68]. To this end, this study suggests that future
researchers conduct more detailed investigations of the Maharashtra state. Since
such studies may allow reasoning out the cause-effect of a warming climate amidst
extreme events resulting from meteorological disturbances, there is emerging scope
for better quantifying climate change aspects in Maharashtra.
4 Conclusion
Maharashtra state covers an area of almost 10% of India’s geographical area of

3,280,000 km2 , encompassing a maximum of four meteorological subdivisions
among all the Indian states. The state has massive geographical significance coupled
with a direct socio-economic contribution to India. Given recurrent reports and
research being published on changing climate of Maharashtra, this study found
it imperative to undertake a fresh hydroclimatological investigation. Hence, the
present study hypothesized Maharashtra and its four meteorological subdivisions,
viz., Konkan and Madhya Maharashtra (toward the west) and Marathwada and
Vidarbha (toward the east), as emerging warming regions in India. This hypoth-
esis was tested by comprehensively investigating the spatio-temporal trends (using
MK and SS tests) and anomalies (between 1981–2020 and 2011–2020) of three
meteorological variables, viz., rainfall (P), surface temperature (T), and relative
humidity (RH). The time series and scatter plots were developed and analyzed to
determine a possible correlation between meteorological variables. The analysis was
conducted via visual interpretation and using the Person correlation coefficient with
the consideration of three variable combinations, viz., (i) temperature with rainfall,
@seismicisolation
@seismicisolation
(ii) temperature with relative humidity, and (iii) relative humidity with rainfall. Find-
ings were generated across four seasons, viz., winter (JF), summer (MAM), SW
monsoon (JJAS), and post-monsoon (OND) for the study period 1981–2020 on a
temporal basis (decadal, yearly, seasonally, and monthly) and spatial basis (state and
its subdivisions). Based on the results and discussion, the following conclusions are
drawn:
• Temporal findings from 1981 to 2020 (40 years period at a glance) indicate that
Maharashtra and its subdivisions are experiencing a statistically significant and
positive trend for mean temperature. However, when relative humidity is tested,
findings reveal insignificant (except Madhya Maharashtra) but positive trends for
Maharashtra and its subdivisions. Whereas, except Vidarbha, all other subdivi-
sions and Maharashtra state show significant and positive trends for rainfall. More
or less, the study validates that temperature, relative humidity, and rainfall are the
appropriate meteorological variables to consider while investigating Maharashtra
as an emerging rapidly warming state of India.
• Furthermore, subdivision-wise findings from 1981 to 2020 (40 years period at
a glance) indicate that Konkan is experiencing the highest mean annual rainfall
and relative humidity percentage, while Vidarbha is experiencing the highest
mean temperature. In fact, both Vidarbha and Marathwada (eastern Maharashtra)
indicate rapid warming since 1981, which is observed to be more pronounced in
the twenty-first century. In contrast, to a moderate extent, Konkan and Madhya
Maharashtra (western Maharashtra) indicate the dominance of high rainfall and
relative humidity conditions. These preliminary findings paved the way for more
detailed spatial and temporal investigations (their conclusions are in the fourth
coming points).
• The Maharashtra state’s temperature records a rise from 21.5 to 22.7 °C (=
1.2 °C rise) between 1981–1990 and 2011–2020. Season-wise, summer records
a maximum temperature increase, from 23.9 to 25.4 °C (= 3.9 °C rise), followed
by the SW monsoon, from 23.7 to 24.7 °C (= 1 °C rise). On the spatial scales and
during the same period, the Vidarbha subdivision records the maximum tempera-
ture increase, from 22.3 to 23.8 °C (= 1.5 °C rise), followed by Marathwada, from
21.5 to 23 °C (= 1.5 °C rise). On the seasonal scales and during the same period,
Vidarbha records a maximum temperature increase during summer, from 25.8 to
27.7 °C (= 1.9 °C rise), followed by Marathwada, from 24.7 to 26.6 °C (= 1.9 °C
rise). While during the SW monsoon season, Marathwada records a maximum
temperature increase, from 23.5 to 24.7 °C (= 1.2 °C rise), followed by Vidarbha,
from 25.2 to 26.3 °C (= 0.9 °C rise). The study apparently concludes that eastern
Maharashtra is witnessing both a comparatively higher mean annual temperature
rise and the seasonal mean temperature rise than western Maharashtra.
• In the case of both season-wise relative humidity and mean annual relative
humidity, between 1981–2010 (three decades together) and 2011–2020 (the most
recent decade), its trend and magnitude remain comparable. An insignificantly
rising trend was observed during 1981–2010, followed by an insignificantly falling
trend during 2011–2020. Even though eastern Maharashtra is experiencing a lower
@seismicisolation
@seismicisolation
relative humidity percentage, the rising rates across the study period are more
remarkable than in western Maharashtra. The study concludes that the abrupt rise
in temperature in Vidarbha, Marathwada, and to a moderate extent, Madhya Maha-
rashtra is causing relative humidity percentage to fall (or be at a lesser magnitude),
resulting in fewer rainfall magnitudes.
• The mean annual rainfall in Maharashtra state is increasing consistently from
1269.8 mm in 1981–1990 to 1495.5 mm in 2011–2020 (a rise of 14.9%). Season-
wise and between the same periods, the SW monsoon receives the highest rainfall,
recording an increase from 1126 to 1285.3 mm (a rise of 14.1%), followed by the
post-monsoon season, recording a drastic increase from 91.6 to 128.1 mm (a rise
of 39.9%). Between the same periods, the maximum rain-receiving subdivision is
Konkan, which is also experiencing an increment from 2319.7 to 2611.3 mm
(a rise of 12.6%), followed by Vidarbha, recording an increase from 1228.9
to 1354.3 mm (a rise of 10.2%). However, the maximum rise in the rainfall is
recorded for low rain-receiving subdivisions, viz., Madhya Maharashtra (21.1%),
followed by Marathwada (18.3%). Season-wise and between the same periods,
the maximum rain-receiving subdivision during the SW monsoon is Konkan
[moderately increases from 2159.8 to 2392.9 mm (10.9% rise)] followed by
Vidarbha [moderately increases from 1059.5 to 1207.1 mm (13.9% rise)]. While
the same for the post-monsoon season is again Konkan [drastically increases from
115 to 185.2 mm (61.1% rise)] followed by Madhya Maharashtra [drastically
increases from 80.4 to 119.8 mm (48.9% rise)]. The study apparently concludes
that although eastern Maharashtra is witnessing a higher rise in rainfall magni-
tudes, the quantity of rainfall in the west being received is reasonably higher.
As a result of lower relative humidity and lower rainfall magnitudes in the east,
Vidarbha and Marathwada are experiencing more hotter summers and less cooler
winters.
• The PCC values are obtained maximum for relative humidity with rainfall,
followed by temperature with rainfall, and least for temperature with relative
humidity, indicating the strongest correlation between relative humidity and rain-
fall but a poor correlation of both with temperature. To conclude, the reason
behind the high rainfall in western Maharashtra can be attributed to its high rela-
tive humidity conditions. However, the low rainfall magnitude and low relative
humidity percentage in eastern Maharashtra may be a plausible reason behind
reasonably high-temperature conditions. Hence, rainfall and relative humidity
may be considered for the Maharashtra state as governing variables behind
moderate to high-temperature increases.
• Analysis of temperature anomaly indicates that Maharashtra and its subdivisions
are experiencing a positive anomaly for all months, more specifically during
summer and post-monsoon months. Maharashtra state’s temperature anomaly
ranges from 0.19 to 1.29 °C. On the spatial scales, the maximum positive anomaly
is recorded for Marathwada, ranging from 0.62 to 1.77 °C (~ 30% higher than
Maharashtra’s magnitude), followed by Vidarbha, ranging from 0.67 to 1.56 °C
(~ 20% higher). In fact, findings from relative humidity and rainfall anomalies
indicate that their magnitude has fallen in the recent decade, especially during
@seismicisolation
@seismicisolation
summer and winter, thereby appearing as the reason behind increased tempera-
ture anomaly. Consequently, the study could validate that summer has become
hotter and winter has become less cooler. Evidently, the state and its subdivisions
(especially from eastern Maharashtra), more or less, are approaching a warming
climate in the recent decade (2011–2020) compared to 1981–2020; this validates
the hypothesis that Maharashtra is an emerging rapidly warming state of India.
Acknowledgements This work is partially supported by the Ministry of Earth Sciences (MoES)
under the Government of India. The funding for the student (AS) was supported by Prime Minister’s
Research Fellowship (PMRF/2401746/21CE91R03) under the Ministry of Education, Government
of India. Technical assistance received from Leena Khadke is gratefully acknowledged.
Author’s Contribution All authors contributed to the study’s conception and design. AS collected
the data and conducted the analysis, while RM and VRD guided the material preparation, data
collection, and analysis. AS wrote the first draft of the manuscript, while RM and VRD commented
on previous versions. All authors read and approved the final manuscript.
Conflict of Interest The authors declare that they have no known competing financial interests or
personal relationships that could have appeared to influence the work reported in this study.
Data Availability Statement The datasets generated during the current study and acquired from
the sources will be made available from the corresponding author upon reasonable request.
References
1. IPCC (2021) Summary for policymakers. In: Climate change 2021: the physical science basis.
Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental
Panel on Climate Change
2. Wang Y, Dong P, Hu W, Chen G, Zhang D, Chen B, Lei G (2022) Modeling the climate suit-
ability of northernmost mangroves in China under climate change scenarios. Forests 13(1):64.
https://doi.org/10.3390/f13010064
3. Byrne MP, O’Gorman PA (2018) Trends in continental temperature and humidity directly linked
to ocean warming. Proc Natl Acad Sci 115(19):4863–4868. https://doi.org/10.1073/pnas.172
2312115
4. Gunawardhana LN, Al-Rawas GA, Kazama S (2017) An alternative method for predicting
relative humidity for climate change studies. Meteorol Appl 24(4):551–559. https://doi.org/10.
1002/met.1641
5. Srivastava A, Maity R, Desai VR (2022a) Assessing global-scale synergy between adaptation,
mitigation, and sustainable development for projected climate change. In: Chatterjee U et al
(eds) Ecological footprints of climate change. Springer Climate. https://doi.org/10.1007/978-
3-031-15501-7_2
6. Tung YS, Wang CY, Weng SP, Yang CD (2022) Extreme index trends of daily gridded rainfall
dataset (1960–2017) in Taiwan. Terr Atmos Oceanic Sci 33(1):1–16. https://doi.org/10.1007/
s44195-022-00009-z
7. Trenberth KE, Dai A, Rasmussen RM, Parsons DB (2003) The changing character of
precipitation. Bull Am Meteor Soc 84(9):1205–1218. https://doi.org/10.1175/BAMS-84-9-
1205
@seismicisolation
@seismicisolation
8. Rusca M, Messori G, Di Baldassarre G (2021) Scenarios of human responses to unprecedented

social-environmental extreme events. Earth’s Future, 9(4), e2020EF001911. https://doi.org/10.
1029/2020EF001911
9. Ummenhofer CC, Meehl GA (2017) Extreme weather and climate events with ecological
relevance: a review. Philos Trans Royal Soc B Bio Sci 372(1723):20160135. https://doi.org/
10.1098/rstb.2016.0135
10. Pande CB, Al-Ansari N, Kushwaha NL, Srivastava A, Noor R, Kumar M et al (2022) Forecasting
of SPI and meteorological drought based on the artificial neural network and M5P model tree.
Land 11(11):2040. https://doi.org/10.3390/land11112040
11. Winsemius HC, Jongman B, Veldkamp TI, Hallegatte S, Bangalore M, Ward PJ (2018) Disaster
risk, climate change, and poverty: assessing the global exposure of poor people to floods and
droughts. Environ Dev Econ 23(3):328–348. https://doi.org/10.1017/S1355770X17000444
12. Yang TH, Liu WC (2020) A general overview of the risk-reduction strategies for floods and
droughts. Sustainability 12(7):2687. https://doi.org/10.3390/su12072687
13. Arora NK (2019) Impact of climate change on agriculture production and its sustainable
solutions. Environ Sustain 2(2):95–96. https://doi.org/10.1007/s42398-019-00078-w
14. De Silva MMGT, Kawasaki A (2018) Socioeconomic vulnerability to disaster risk: a case study
of flood and drought impact in a rural Sri Lankan community. Ecol Econ 152:131–140. https://
doi.org/10.1016/j.ecolecon.2018.05.010
15. Dhanuka A, Srivastava A, Khadke L, Kushwaha NL (2023) Smart geometric design of high-
ways using HTML programming for sustainable and climate resilient cities. In: Chatterjee
U, Bandyopadhyay N, Setiawati MD, SarkarS (eds) Urban commons, future smart cities
and sustainability. Springer Geography. Springer, Cham. https://doi.org/10.1007/978-3-031-
24767-5_39
16. Elbeltagi A, Raza A, Hu Y, Al-Ansari N, Kushwaha NL et al (2022) Data intelligence and
hybrid metaheuristic algorithms-based estimation of reference evapotranspiration. Appl Water
Sci 12(7):152. https://doi.org/10.1007/s13201-022-01667-7
17. Kalantari Z, Ferreira CSS, Keesstra S, Destouni G (2018) Nature-based solutions for flood-
drought risk mitigation in vulnerable urbanizing parts of East-Africa. Current Opin Environ
Sci Health 5:73–78. https://doi.org/10.1016/j.coesh.2018.06.003
18. Raikes J, Smith TF, Jacobson C, Baldwin C (2019) Pre-disaster planning and preparedness for
floods and droughts: a systematic review. Int J Disaster Risk Reduction 38:101207. https://doi.
org/10.1016/j.ijdrr.2019.101207
19. Srivastava A, Jain S, Maity R, Desai VR (2022b) Demystifying artificial intelligence amidst
sustainable agricultural water management. In: Zakwan M, Wahid A, Niazkar M, Chatterjee
U (eds) Water resource modeling and computational technologies. Current directions in water
scarcity research, vol 7. Elsevier. https://doi.org/10.1016/B978-0-323-91910-4.00002-9
20. Elbeltagi A, Srivastava A, Kushwaha NL, Juhász C, Tamás J, Nagy A (2023a) Meteorological
data fusion approach for modeling crop water productivity based on ensemble machine learning.
Water 15(1):30. https://doi.org/10.3390/w15010030
21. Elbeltagi A, Srivastava A, Al-Saeedi AH, Raza A, Abd-Elaty I, El-Rawy M (2023b) Forecasting
long-series daily reference evapotranspiration based on best subset regression and machine
learning in Egypt. Water 15(6):1149. https://doi.org/10.3390/w15061149
22. Elbeltagi A, Srivastava A, Deng J, Li Z, Raza A, Khadke L, Yu Z, El-Rawy M (2023c)
Forecasting vapor pressure deficit for agricultural water management using machine learning
in semi-arid environments. Agric Water Manag 283:108302. https://doi.org/10.1016/j.agwat.
2023.108302
23. Li X, Zhang K, Gu P, Feng H, Yin Y, Chen W, Cheng B (2021) Changes in precipitation extremes
in the Yangtze River Basin during 1960–2019 and the association with global warming,
ENSO, and local effects. Sci Total Environ 760:144244. https://doi.org/10.1016/j.scitotenv.
2020.144244
24. Mahajan DR, Dodamani BM (2015) Trend analysis of drought events over upper Krishna basin
in Maharashtra. Aquat Procedia 4:1250–1257. https://doi.org/10.1016/j.aqpro.2015.02.163
@seismicisolation
@seismicisolation
25. Singh C, Deshpande T, Basu R (2017) How do we assess vulnerability to climate change in
India? A systematic review of literature. Reg Environ Change 17(2):527–538. https://doi.org/
10.1007/s10113-016-1043-y
26. NIH—National Institute of Hydrology (2020). Hydrology and water resources information
system for India, Water Resources Systems Division, National Institute of Hydrology. http://
117.252.14.242/rbis/India_Information/Water%C2%A0Budget.htm. Accessed 3 Apr 2022
27. NWM—National Water Mission (2018) State-water budgeting report. Ministry of Jal
Shakti, Department of Water Resources, Government of India. http://nwm.gov.in/sites/default/
files/A%20note%20on%20State%20Water%20Budgeting%2019.3.2018.pdf . Accessed 3 Apr
2022
28. Goyal MK, Surampalli RY (2018) Impact of climate change on water resources in India. J
Environ Eng 144(7):04018054. https://doi.org/10.1061/(ASCE)EE.1943-7870.0001394
29. Singh D, Tsiang M, Rajaratnam B, Diffenbaugh NS (2014) Observed changes in extreme wet
and dry spells during the South Asian summer monsoon season. Nat Clim Chang 4(6):456–461.
https://doi.org/10.1038/nclimate2208
30. Kumar V, Jain SK, Singh Y (2010) Analysis of long-term rainfall trends in India. Hydrol Sci
J 55(4):484–496. https://doi.org/10.1080/02626667.2010.481373
31. Mondal A, Khare D, Kundu S (2015) Spatial and temporal analysis of rainfall and temperature
trend of India. Theoret Appl Climatol 122(1):143–158. https://doi.org/10.1007/s00704-014-
1283-z
32. Dhawan V (2017) Water and agriculture in India. In: Background paper for
the South Asia expert panel during the Global Forum for Food and Agricul-
ture (Vol. 28). https://www.oav.de/fileadmin/user_upload/5_Publikationen/5_Studien/170
118_Study_Water_Agriculture_India.pdf. Accessed 3 Apr 2022
33. Mehta L (2005) The politics and poetics of water: The naturalisation of scarcity in Western
India. Orient Blackswan.
34. GoM—Government of Maharashtra (2020) Economic survey of Maharashtra 2019–
20. Directorate of Economics and Statistics, Planning Department, Government of
Maharashtra, Mumbai. https://www.maharashtra.gov.in/Site/upload/WhatsNew/ESM_2019_
20_Eng_Book.pdf. Accessed 3 Apr 2022
35. Singh RN, Sah S, Das B, Vishnoi L, Pathak H (2021) Spatio-temporal trends and variability of
rainfall in Maharashtra, India: Analysis of 118 years. Theoret Appl Climatol 143(3):883–900.
https://doi.org/10.1007/s00704-020-03452-5
36. Guhathakurta P, Saji E (2013) Detecting changes in rainfall pattern and seasonality index vis-
à-vis increasing water scarcity in Maharashtra. J Earth Syst Sci 122(3):639–649. https://doi.
org/10.1007/s12040-013-0294-y
37. Ingle ST, Patil SN, Mahale NK, Mahajan YJ (2018) Analyzing rainfall seasonality and trends
in the North Maharashtra region. Environ Earth Sci 77(18):1–12. https://doi.org/10.1007/s12
665-018-7837-0
38. Mandale VP, Jedhe SH, Khadtare MY (2019) Spatio-temporal trends of rainfall and rainy days
in the Marathwada Region of Maharashtra State. Clim Change 5(17):55–61. http://www.dis
coveryjournals.org/climate_change/current_issue/v5/n17/A7.pdf
39. Jedhe SH, Kadam US, Mane MS, Mahale DM, Nandgude SB, Thokal RT (2018) Trends of
rainfall and temperature in Konkan region of Maharashtra. J Agrometeorol 20(1), 80–83. http://
krishi.icar.gov.in/jspui/handle/123456789/48447
40. Dawoodi HH (2021) Rainfall prediction in North Maharashtra region using support vector
machine. Turk J Comput Mathematics Educ 12(7):1501–1505. https://doi.org/10.17762/tur
comat.v12i7.2957
41. Kelkar SM, Kulkarni A, Rao KK (2020) Impact of climate variability and change on crop
production in Maharashtra, India. Curr Sci 118(8):1235–1245. https://www.currentscience.ac.
in/Volumes/118/08/1235.pdf
42. Masroor M, Rehman S, Avtar R, Sahana M, Ahmed R, Sajjad H (2020) Exploring climate
variability and its impact on drought occurrence: evidence from Godavari Middle sub-basin
India. Weather Clim Extremes 30:100277. https://doi.org/10.1016/j.wace.2020.100277
@seismicisolation
@seismicisolation
43. Masroor M, Avtar R, Sajjad H, Choudhari P, Kulimushi LC, Khedher KM et al (2022)

Assessing the influence of land use/land cover alteration on climate variability: an analysis
in the Aurangabad district of Maharashtra state India. Sustainability 14(2):642. https://doi.org/
10.3390/su14020642
44. Shah SH, Harris LM, Johnson MS, Wittman H (2021) A ‘drought-free’ Maharashtra? politi-
cising water conservation for rain-dependent agriculture. Water Altern 14(2):573–596. https://
www.water-alternatives.org/index.php/alldoc/articles/vol14/v14issue2/628-a14-2-6/file
45. Kaicker N, Imai KS, Gaiha R (2020) Severity of the Covid-19 Pandemic in India. The Case of
three States: Maharashtra, Jharkhand and Meghalaya. GDI Working Paper, 2020–047. https://
doi.org/10.2139/ssrn.3709831
46. Pol SS, Rajderkar SS, Dhabekar PD, Bansode Gokhe SS (2021). Effect of climatic factors like
rainfall, humidity and temperature on the dengue cases in the metropolitan city of Maharashtra.
Int J Community Med Public Heal, 8(2):672–677. https://doi.org/10.18203/2394-6040.ijcmph
20210220
47. Sasane SA, Jadhav AS, Barik RK, Kumar GK, Raghavswamy V (2019) Application of spatial
technology in malaria information infrastructure mapping with climate change perspective
in Maharashtra, India. MAUSAM 70(4):787–806. https://metnet.imd.gov.in/mausamdocs/170
411.pdf
48. Saunik S, Shil P, Das SN, Rajankar SP, Khare O, Hosalikar KA, Kabir Y (2021) Predictions
of disease spikes induced by climate variability: a pilot real time forecasting model project
from Maharashtra, India. In: Kumar MD, Kabir Y, Hemani R, Bassi N (eds) Management of
irrigation and water supply under climatic extremes. Global Issues in Water Policy, vol 25.
Springer, Cham. https://doi.org/10.1007/978-3-030-59459-6_9
49. Kendall MG (1975) Rank correlation methods. Edition 4. Charles Griffin, London
50. Mann HB (1945) Nonparametric tests against trend. Econ J Econometric Soc 245–259. https://
doi.org/10.2307/1907187
51. Sen PK (1968) Estimates of the regression coefficient based on Kendall’s tau. J Am Stat Assoc
63(324):1379–1389. https://doi.org/10.1080/01621459.1968.10480934
52. Jena P, Azad S, Rajeevan MN (2015) Statistical selection of the optimum models in the CMIP5
dataset for climate change projections of Indian monsoon rainfall. Climate 3(4):858–875.
https://doi.org/10.3390/cli3040858
53. Dhorde AG, Korade MS, Dhorde AA (2017) Spatial distribution of temperature trends and
extremes over Maharashtra and Karnataka States of India. Theoret Appl Climatol 130(1):191–
204. https://doi.org/10.1007/s00704-016-1876-9
54. Korade MS, Dhorde AG (2016) Trends in surface temperature variability over Mumbai and
Ratnagiri cities of coastal Maharashtra, India. Mausam 67(2):455–462. https://doi.org/10.
54302/mausam.v67i2.1352
55. Khan MHR, Rahman A, Luo C, Kumar S, Islam GA, Hossain MA (2019) Detection of changes
and trends in climatic variables in Bangladesh during 1988–2017. Heliyon 5(3):e01268. https://
doi.org/10.1016/j.heliyon.2019.e01268
56. Ali H, Fowler HJ, Mishra V (2018) Global observational evidence of strong linkage between
dew point temperature and precipitation extremes. Geophys Res Lett 45(22):12–320. https://
doi.org/10.1029/2018GL080557
57. Dash SK, Jenamani RK, Kalsi SR, Panda SK (2007) Some evidence of climate change in
twentieth-century India. Clim Change 85(3):299–321. https://doi.org/10.1007/s10584-007-
9305-9
58. Niu X, Wang S, Tang J, Lee DK, Gutowski W, Dairaku K et al (2015) Projection of Indian
summer monsoon climate in 2041–2060 by multiregional and global climate models. J Geophys
Res Atmos 120(5):1776–1793. https://doi.org/10.1002/2014JD022620
59. Allan RP, Barlow M, Byrne MP, Cherchi A, Douville H, Fowler HJ et al (2020) Advances in
understanding large-scale responses of the water cycle to climate change. Ann N Y Acad Sci
1472(1):49–75. https://doi.org/10.1038/s41598-020-70816-2
60. Cox DT, Maclean IM, Gardner AS, Gaston KJ (2020) Global variation in diurnal asymmetry
in temperature, cloud cover, specific humidity and precipitation and its association with leaf
area index. Glob Change Biol 26(12):7099–7111. https://doi.org/10.1111/gcb.15336
@seismicisolation
@seismicisolation
61. Fowler HJ, Lenderink G, Prein AF, Westra S, Allan RP, Ban N et al (2021) Anthropogenic
intensification of short-duration rainfall extremes. Nat Rev Earth Environ 2(2):107–122. https://
doi.org/10.1038/s43017-020-00128-6
62. Hardwick JR, Westra S, Sharma A (2010) Observed relationships between extreme sub-daily
precipitation, surface temperature, and relative humidity. Geophys Res Lett 37(22). https://doi.
org/10.1029/2010GL045081
63. Kousari MR, Ekhtesasi MR, Tazeh M, Saremi Naeini MA, Asadi Zarch MA (2011) An inves-
tigation of the Iranian climatic changes by considering the precipitation, temperature, and
relative humidity parameters. Theoret Appl Climatol 103(3):321–335. https://doi.org/10.1007/
s00704-010-0304-9
64. Singh P, Kumar V, Thomas T, Arora M (2008) Changes in rainfall and relative humidity in
river basins in northwest and central India. Hydrol Processes Int J 22(16):2982–2992. https://
doi.org/10.1002/hyp.6871
65. Kousari MR, Asadi Zarch MA (2011) Minimum, maximum, and mean annual temperatures,
relative humidity, and precipitation trends in arid and semi-arid regions of Iran. Arab J Geosci
4(5):907–914. https://doi.org/10.1007/s12517-009-0113-6
66. Bhimala KR, Gouda KC, Himesh S (2021) Evaluating the spatial distribution of WRF-simulated
rainfall, 2-m air temperature, and 2-m relative humidity over the urban region of Bangalore
India. Pure Appl Geophys 178(3):1105–1120. https://doi.org/10.1007/s00024-021-02676-4
67. Kharin VV, Zwiers FW, Zhang X, Wehner M (2013) Changes in temperature and precipitation
extremes in the CMIP5 ensemble. Clim Change 119(2):345–357. https://doi.org/10.1007/s10
584-013-0705-8
68. Simmons AJ, Willett KM, Jones PD, Thorne PW, Dee DP (2010) Low-frequency variations in
surface atmospheric humidity, temperature, and precipitation: Inferences from reanalyses and
monthly gridded observational data sets. J Geophys Res Atmos 115(D1). https://doi.org/10.
1029/2009JD012442
@seismicisolation
@seismicisolation
Structural Analysis
@seismicisolation
@seismicisolation
A Study on Transient and Modal Analysis
of a GFRP Bridge Deck Under the Action
of Heaviest Main Battle Tank Currently
Available in India (T-90S Bhishma)
Arpita Mandal and Hiranmoy Barman
1 Introduction
The use of fiber-reinforced polymer (FRP) composites becomes an integral part of

construction industry because of its superior material properties like superior material
properties like high resistance to fatigue, corrosion free, enhanced durability, huge
heat resistance, lower maintenance and life-cycle costs, etc. The relative lightness
makes it possible to enhance the portability of the structures made by FRP only.
To improve the mobility of the military forces to the impregnable areas and also to
provide a temporary access for rescue force to the disaster area when the existing
transportation system is damaged or destroyed, lightweight portable bridges are very
much useful.
Yeh et al. subjected to present a portable bridge structure with highly customiz-
able joints that integrate bolts, welding, and adhesives for easy manufacture and rapid
assembly as well as effective structural design methods for increasing joint bonding
strength and tried to reduce the deflection-to-span ratio [1]. Smits reviewed the use of
FRP composite material and challenges for both architects and structural engineers
[2]. Votsis et al. designed and fabricated a 51.3 m span Wilcott FRP suspension foot-
bridge and performed a finite element analysis on ANSYS. The early investigation
yielded data for the bridge’s dynamic properties, vibration tests, damping measures,
and many other things [3]. Tomasz et al. designed and made a hybrid composite
bridge successful. According to the analysis, the estimated first natural frequency of
3.98 Hz was higher (i.e., better) than the suggested value (3 Hz) for that composite
bridge [4]. Yan et al. demonstrate a hybrid FRP-aluminum space truss string bridge
based on a pre-tightened tooth connection, applying a total of 170 kN and identified
a barely decreased bending stiffness [5]. Thakur and Pahwa made a study on tran-
sient analysis of bridge deck slab under the action of moving load [6]. Drygala et al.
A. Mandal (B) · H. Barman

Narula Institute of Technology, Kolkata, India
e-mail: mandal.arpita6@gmail.com
@seismicisolation
@seismicisolation
328 A. Mandal and H. Barman
made a numerical analysis of the serviceability of footbridges on ABAQUS software

using THA approach [7]. Mandal et al. made an experimental and numerical study
on modal analysis of composite laminated shells with cut-out [8]. Sadiyah modeled
a structure of Portable River Bridge using fiber-reinforced polymer [9]. Mandal et al.
made a study on dynamic behavior of composite plates with arbitrary cut-out [10].
Saleem et al. provided an outline of recent researches on prefabricated bridge deck
systems [11]. Zou aims to review and evaluate FRP-concrete hybrid sections in the
context of bridge engineering [12]. Zhang proposed a prefabricated fiber-reinforced
polymer (FRP)-concrete hybrid beam (HB) [13].
From an extensive literature review, it becomes very much clear that in civil
structural industry; most of the uses of FRP are for retrofitting purpose. Most of the
published literatures are either on foot bridges or bridges constructed using steel or
concrete composite. Number of available analytical and experimental studies on FRP
road bridges in available literature is very less. Particularly dynamic analysis with
real-life loading is scarce in present literature.
2 Methodology
2.1 Geometry and Units
In this present study, a 16 m long bridge has been taken under consideration. The
proposed bridge superstructure consists of four main girders and a deck. The stiffened
laminated deck, which has dimensions of 3500 × 16,000 mm, has a thickness of
88.632 mm (see Fig. 1b). Depth of each girder is 550 mm (see Fig. 1). Each girder’s
thickness after creating a solid model is 16.172 mm. Honeycomb has been used for
the core. With the help of Fig. 1, it is possible to get a total idea about the geometry of
the structure. The primary task of the main girder is to support the imposed load on the
bridge. In addition, the function of the deck is to provide a passage for the human and
vehicle (T-90S Bhishma) to travel. For building the FRP composite model in ANSYS
Composite PrepPost (ACP), shell elements have been taken under consideration.
The model has meshed at lesser than 150 mm (as per requirement), which is
researched and concluded to supply a balance between computing time and accuracy
of the result as well as to defining the boundary conditions, connections, and trans-
porting the loads. Fabric—S-Glass UD (unidirectional) with a thickness of 0.142 mm,
resin epoxy, and honeycomb used for the core. The properties of same are indicated
in Table 2. It is necessary to provide the fabric and its orientation for each layer of
the stack up (Table 1).
@seismicisolation
@seismicisolation
A Study on Transient and Modal Analysis of a GFRP Bridge Deck … 329
Fig. 1 Detailing of the bridge structure. a 3D view of the bridge b cross-sectional view of bridge
c cross-sectional view of girder d cross-sectional top view of deck core e side view of bridge
@seismicisolation
@seismicisolation
Table 1 Fabric orientation

Lamina layer Lamina-oriented (degree)
for layers of the stack up
Inner lamina Outer lamina
1–6 0 0
7–12 45 45
13–18 90 90
19–24 − 45 − 45
25–30 − 45 − 45
31–36 90 90
37–42 45 45
43–48 0 0
Table 2 Engineering properties of epoxy S-Glass UD and honeycomb structure

Engineering properties Epoxy S-Glass UD Honeycomb
Young’s modulus X direction (Pa) 5.e × 1010 1 × 106
Young’s modulus Y direction (Pa) 8× 109 1 × 106
Young’s modulus Z direction (Pa) 8.e × 109 2.55 × 108
Poisson’s ratio—XY 0.3 0.49
Poisson’s ratio—YZ 0.4 1 × 10–3
Poisson’s ratio—XZ 0.3 1 × 10–3
Shear modulus—XY (Pa) 5 × 109 1
Shear modulus—YZ (Pa) 3.8462 × 109 3.7 × 107
Shear modulus—XZ (Pa) 5× 109 7 × 107
Density (kg m−3 ) 2000 80
Stress limit—tensile X direction (Pa) 1.7 × 109 0
Stress limit—tensile Y direction (Pa) 3.5 × 107 0
Stress limit—tensile Z direction (Pa) 3.5 × 107 0
Stress limit—compressive X direction (Pa) −1× 109 5.31 × 106
Stress limit—compressive Y direction (Pa) − 1.2 × 108 0
Stress limit—compressive Z direction (Pa) − 1.2 × 108 0
Stress limit—shear XY (Pa) 8× 107 –5.31 × 106
Stress limit—shear YZ (Pa) 4.6154 × 107 1.21 × 106
Stress limit—shear XZ (Pa) 8 × 107 2.24 × 106
2.2 Properties of the Used Resin Epoxy
Density (kg/m3 ) = 1160.

Young’s modulus (Pa) = 3.78 × 109 .
Poisson’s ratio = 0.35.
@seismicisolation
@seismicisolation
Bulk modulus (Pa) = 4.2 × 109 .

Shear modulus (Pa) = 1.4 × 109 .
Tensile yield strength (Pa) = 5.46 × 107 .
2.3 Load Calculation
The imposed load used for the simulation is the weight of the heaviest main battle
tank currently available in India. T-90S [14] combat is India’s current heaviest main
battle tank. The projected weight of this tank is 48 tons.
Weight of tank = 48 tons.
Weight of tank in kg = 48 × 1000 = 48,000 kg.
Weight of tank in Newton = 48,000 × 9.81 = 470,880 N.
Imposed load = 470,880 N.
Provisions from IRC 6 (2010) have been taken to apply the moving load on the
bridge deck (see Fig. 2). The method divides the axle weights in half and applies
two-wheel loads to either side of the vehicle’s.
Fig. 2 Class AA tracked vehicle (IRC:6-2010)
@seismicisolation
@seismicisolation
2.4 Modal Analysis
In practical field to ensure the safety of any structure against vibration, it is very
much essential to make study on the natural frequencies of the structure. To improve
the working condition, it is important to adjust the natural frequency of the structure
to avoid resonance due to unavoidable frequency range caused by any working load.
Our goal is to determine the FRP bridge structure’s first ten natural frequencies and
corresponding mode shapes. For this present investigation, minimum frequency range
0 Hz and maximum frequency range 100 Hz have been taken under consideration.
2.5 Transient Analysis of FRP Bridge
The transient analysis is simulated under the moving loads of 48 tons. As load moves
from starting point to endpoint of 16 m spans, considered time to cross the bridge
is 15 s. Due to load on deck deflection, stress and strain developed in that time
period. The applied loading is (tracked vehicle) 48 tons. As load moves on slab
deflection, stress and strain develop in it. To identify fiber failure and inter-fiber
failure, Puck fracture criteria have been used in this study. The Puck fracture criteria
for unidirectional fiber/polymer composites are widely accepted for physically based
prediction of failure and post-failure degradation behavior.
The failure plot for a certain failure criterion description shows the critical safety
components for first-ply failure. Each component and layer are given a safety factor
evaluation, and the critical value across the thickness of the layup is then mapped onto
the reference shell mesh. The failure plot for an envelope solution functions similarly
and is a superposition of multiple failure plots for solutions. As an alternative, the
safety factors can be shown for each analysis. A failure criterion definition must be
specified before generating a failure plot.
Failure Mode Measures
• IRF = The inverse margin to failure is defined by the inverse reserve factor (IRF).
The load divided by the IRF is the failure load. IRF > 1 discloses failure.
• MoS = The margin to failure is described by the margin of safety (MoS). MoS
defined as (1/IRF − 1). MoS < 0 discloses failure.
3 Result
3.1 Transient Analysis Result
The transient analysis is conducted for 16 m span lengths under moving load (48
tons tracked load). Sample pictures of loading on span are shown in Fig. 3.
@seismicisolation
@seismicisolation
Fig. 3 Loading on span a first loading (time 0–3 s) b second loading (time 3–6 s) c third loading
(time 6–9 s) d fourth loading (time 9–12 s) e fifth loading (time 12–15 s)
To identify fiber failure and inter-fiber failure, Puck fracture criteria have been
used in this study. The design and finite element analysis of the honeycomb core
epoxy-carbon laminated deck–girder model obtained a safety factor result of 4 (see
Fig. 4d).
3.2 Modal Analysis Results
In this present design and analysis, natural frequencies are restricted within 100 Hz.
Natural frequencies of first ten modes obtained from numerical studies are presented
in Table 3. The mode shapes corresponding to first ten modal frequencies of the
modeled FRP bridge deck are shown in Fig. 5.
@seismicisolation
@seismicisolation
Fig. 4 a Safety factors b equivalent von Mises stress c composite failure tool (from ANSYS R19.2)
d composite failure result (from ANSYS R19.2)
Table 3 Modal analysis result GFRP (in Hz)

Mode 1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th
Modal 10.84 14.41 20.21 28.88 30.81 33.51 37.09 48.66 52.28 56.29
frequencies
@seismicisolation
@seismicisolation
Fig. 5 Mode shapes for GFRP bridge deck. a mode 1 b mode 2 c mode 3 d mode 4 e mode 5
f mode 6 g mode 7 h mode 8 i mode 9 j mode 10
@seismicisolation
@seismicisolation
4 Discussion and Conclusions
• In this present study, we use glass fiber-reinforced polymer (GFRP) composite as

primary material. Due to the excellent strength-to-weight ratio structures made
by FRP can be lighter than other structures made by steel and alloys. FRPs are
ideal material for advance engineering as it offers excellent tensile strength, harsh
environments tolerance, and extremes of temperature resistance. Although further
technological development is essential to make the process cost-effective and
ensure the extreme durability of the final FRP as in terms of potential applications
in construction field, this material array is astounding.
• A fiber-reinforced polymer girder–deck bridge model has been developed using
commercial software ANSYS R19.2 for the practical field application. The honey-
comb epoxy-S glass laminated model for the bridge structures has been modeled.
Wrapping of carbon-epoxy laminates with honeycomb core lamina significantly
improves the deck and girders behavior under tracked loading conditions.
• The bridge deck reaction under moving vehicles is investigated for practical field.
Glass fiber-reinforced polymer (GFRP) bridge deck slab subjected to moving load
of T-90S Bhishma (heaviest battle tank currently available in India) is modeled
and analyzed here by using ANSYS 19.2.
• The transient analysis is simulated under the moving loads of 48 tons (T-90S
Bhishma) which takes 15 s time to cross the 16 m long bridge.
• The design and finite element analysis of the honeycomb core epoxy-carbon
laminated deck–girder model obtained a safety factor result of 4 (see Fig. 4).
• As the modal analysis was performed (using ANSYS 19.2) to find out the natural
frequencies of the FRP bridge deck.
• The first ten natural frequencies of the bridge fall in the range of 10.837–56.291 Hz.
References
1. Yeh FY, Chang KC, Liu KY, Hung HH, Chou CC, Liu T, Sun PF, Pan WY, Sung YC, Yin SH,
Chiu YT. A novel composite emergency bridge for disaster rescue. In: 15th World Conference
on Earthquake Engineering (15WCEE), Lisbon, Portugal (2012).
2. Smits J (2016) Fibre-reinforced polymer bridge design in the Netherlands. Architectural
challenges toward innovative, sustainable, and durable bridges. Engineering 2(4):518–527
3. Votsis RA, Stratford T, Chryssanthopoulos MK, Tantele E (2017) Dynamic assessment of an
FRP suspension footbridge through field testing and finite element modeling. Steel Compos
Struct 23(2):205–215
4. Tomasz S, Rajchel M, Kalet D, Własak L (2017) The first Polish road bridge made of
FRP composites. Struct Eng Int 27(2):308–314. https://doi.org/10.2749/101686617X14881
932436339
5. Yan X, Zhao Q, Zhao D, Ke M, Zeng ZG, Zhang D (2018) Experimental investigation of
bending stiffness of a novel 54m FRP space truss string bridge. IOP Conf Ser: Mater Sci Eng
452. https://doi.org/10.1088/1757-899X/452/2/022103
6. Thakur C, Pahwa S (2019) A study on transient analysis of bridge deck slab under the action
of moving load. Int J Sci Res Dev 7(4):170–173
@seismicisolation
@seismicisolation
7. Drygala IJ, Polak MA, Dulinska JM (2019) Vibration serviceability assessment of GFRP
pedestrian bridges. Eng Struct 184:176–185
8. Mandal A, Ray C, Haldar S (2019) Experimental and numerical modal analysis of composite
laminated shells with cut-out. Int J Marit Eng Trans RINA 161:A1–A12. https://doi.org/10.
3940/rina.ijme.2019.a1.495
9. Sadiyah A, Prasetya AF, Alfianto E (2020) Modelling structure of portable river bridge using
fibre-reinforced polymer (FRP). J Phys: Conf Ser 1445(1):012–029
10. Mandal A, Ray C, Haldar S (2020) Experimental and numerical free vibration analysis of
laminated composite plates with arbitrary cut-outs. Inst Eng (India): Ser C 101:281–289. https://
doi.org/10.1007/s40032-019-00537-7
11. Saleem MA, Zafar MN, Saleem MM, Xia J (2021) Recent developments in the prefabricated
bridge deck systems. Case Stud Constr Mater 15:e00750. https://doi.org/10.1016/j.cscm.2021.
e00750
12. Zou X, Lin H, Feng P, Bao Y, Wang J (2021) A review on FRP-concrete hybrid sections for
bridge applications. Compos Struct 262:113336
13. Zhang P, Qi Y, Zou X, Feng Y, Sheikh SA (2023) Flexural performance of prefabricated
FRP-concrete hybrid beam with in-situ-cast UHPC pockets. Thin-Walled Struct 185:110616
14. ANSYSACP guide. https://pdfcoffee.com/ansys-composite-preppost-users-guide-7-pdffree
@seismicisolation
@seismicisolation
Investigation of Crack Propagation of Fly
Ash-Based Geopolymer Concrete Using
Digital Image Segmentation Approach
Aashish Lamichhane , Gaurav Basnet , Amrit Panta , Shankar Shah ,

and Nishant Kumar
1 Introduction
Concrete has several benefits as a building material, including its toughness, adapt-
ability, and durability. A type of concrete is called geopolymer concrete (GPC) if it is
manufactured with industrial waste materials like fly ash, slag, and other elements. It
is created through a chemical reaction between by-products and an alkaline activator
solution. When making concrete, fly ash is frequently utilized as an additional cemen-
titious element [1]. Fly ash is a fine powder resultant of crushed burnt coal in power
plants. It is a kind of coal ash that is extracted from coal-fired power stations’ exhaust
gasses. Fly ash can improve the strength, workability, and durability of concrete since
it contains numerous elements like silica, iron, and alumina [2]. Fly ash can also be
added in concrete to lessen waste disposal and carbon emissions from power plants
[3]. Ongoing research studies have demonstrated that substituting fly ash for some
of the cement can increase the strength, longevity, and workability of concrete while
lowering its carbon footprint [3]. However, using too much fly ash will diminish
early strength and delay setting time, which can have a bad effect on how well the
concrete performs [4]. Raheel et al. [5] discussed the optimal usage of fly ash in
concrete along with laboratory research on the strength improvement of concrete
including fly ash based on stoichiometric approach. According to this study, fly ash
reacted with the calcium hydroxide of cement producing binder gel. This enhanced
the strength, lowered permeability, and made the concrete denser. A fly ash content
of roughly 30% of cement was found ideal in his study.
Concrete cracks are a prevalent problem in construction and can happen for a
number of reasons, such as settlement, thermal expansion, contraction, shrinkage,
and overloading [6]. The occurrence of cracks can be perceived based on the concrete
constituents and its handling while manufacturing. Fly ash as a finest particle being
A. Lamichhane (B) · G. Basnet · A. Panta · S. Shah · N. Kumar

Civil Engineering Department, Sharda University, Greater Noida, Uttar Pradesh, India
e-mail: aasisleolc@gmail.com; 2019002460.aashish@ug.sharda.ac.in
@seismicisolation
@seismicisolation
340 A. Lamichhane et al.
kept in the mix definitely steer the crack propagation in concrete. Golewski et al.
[7] stated that the fly ash particle imparts their pozzolanicity in the C-S-H phase of
concrete hydration strengthening the interfacial alteration zone among coarse aggre-
gate and the mix paste. This improves the compactness, which affects the behavior
of crack propagation [8, 9].
Concrete shrinkage cracks, thermal cracks, and structural fractures can all be
identified and categorized using sophisticated image processing methods including
edge detection, segmentation, and feature extraction [10]. Automated crack detection
and classification systems that can precisely identify and classify cracks in concrete
buildings can be created using machine learning approaches [11]. Moreover, image
processing can be used to assess the breadth and length of cracks in concrete to deter-
mine their severity and the likelihood that the structure would fail or sustain further
damage. Fan et al. [12] studied an image processing technology (IPT)-based quick
method for regulating the mechanical behaviors of geopolymer concrete (GPC). The
results reveal that the IPT and numerical simulation-based techniques can properly
predict the mechanical characteristics of GPC with a fault of less than 10%. Wang
et al. [13] developed an image preprocessing system merging numerous adaptive cate-
gorizing and contrast improvement based on the concrete crack image processing
technology. The case analysis findings demonstrated that the detection algorithm’s
data processing accuracy could fulfill the real engineering criteria for concrete bridge
crack detection by processing data to a precision of 0.02 mm. Asmari et al. [14]
published a paper with a goal to segment images of fly ash particles captured with a
micro computed tomography (CT) imaging instrument.
This paper utilizes the Otsu threshold image segmentation approach as a binary
thresholding method to detect the crack in concrete cylinder. The associated priori
parameters in this segmentation approach are the difference in the intensity of the
object and the background [15]. The cracks’ white intensity compared to the cylin-
ders’ black background intensity provided the thresholding point. The correlation is
done between the crack propagation and the existing fly ash content to establish a
most optimal and economical category range of fly ash replacement in the concrete
mix.
2 Methodology
In order to observe the crack propagation correspondence to the optimum content

of fly ash substitution with cement, laboratory experiments were carried out with
varying different percentages of fly ash in a casted concrete cylinder specimen. The
cracks on concrete cylinders were processed using image processing of segmentation.
@seismicisolation
@seismicisolation
Investigation of Crack Propagation of Fly Ash-Based Geopolymer … 341
Table 1 Test data of

Specifications Results
materials
Specific gravity of cement 3.15
Specific gravity of fly ash 2.2
Specific gravity of coarse aggregate 2.74
Specific gravity of fine aggregate 2.65
Water absorption of coarse aggregate 0.5
Water absorption of fine aggregate 1
Free (surface) moisture of coarse aggregate 1.5
Free (surface) moisture of fine aggregate 4
Grading zone of coarse aggregate Zone II
Grading zone of fine aggregate Zone II
2.1 Material Specifications for Specimen Creation
The specimens were prepared using ordinary Portland cement (OPC 43 following to
IS 269), fly ash (corresponding to IS 3812), coarse aggregates (20 and 10 mm), fine
aggregates, and water. The fly ash content was varied according to the percentages of
5%, 15%, 25%, 30%, 40%, and 50% by weight of cement. The different preliminary
tests of materials were conducted to prepare the mix design and the validation of
quality materials. The tests conducted were initial and final tests of cement, setting
time, specific gravity of cement, specific gravity of fly ash, specific gravity of fine and
coarse aggregates, and water absorption of fine and coarse aggregates as mentioned
in Table 1.
2.2 Creation of Specimens
The design mix was done using IS 10262: 2019, and the grade of concrete prepared
was M-25. Accor Tech Pan Mixture of 55L capacity was used to prepare concrete
mixture. The concrete mix was casted as a cylinder of dimension 150 mm × 300 mm.
The number of 120 cylinders was casted. IS 1199-1959 as a reference had been used
to carry out the workability test. A standard cone of a diameter of 10 cm in the upper
rim, 20 cm in the lower rim, and 30 cm in height had been used to test workability
(slump) for the prepared concrete mixture. The workability (slump) results were
noted for comparisons. The mix was then poured to cylindrical mold which was
vibrated for roughly around 10 s. Mold was placed in safe side after vibration which
was demolded after 24 h. The specimens thus casted were cured for about 28 days
under the normal room temperature. The specimen preparation is indicated in Figs. 1,
2, and 3.
@seismicisolation
@seismicisolation
Fig. 1 Casting process
Fig. 2 Workability test
@seismicisolation
@seismicisolation
Fig. 3 Curing process
2.3 Nomenclature of Specimens
Out of 120, 60 cylinders were prepared for finding the ultimate loading (100%) at
collapsing. And the other 60 cylinders underwent 90% of the same ultimate load
for the sole purpose of crack generation without concrete collapsing. For both the
loading conditions, 10 each cylinder was kept at each fly ash variation casting class.
Specimens with 5% fly ash replacement undergoing ultimate load were titled as
5Aa100 to 5Aj100 (10 cylinders). For the same replacement, cylinders undergoing
90% loading was titled as 5Aa90 to 5Aj90 (10 cylinders). Likewise, for every casting
class, detailed nomenclature and quantity of cement and fly ash is tabulated in Table 2.
2.4 Compressive Strength Tests
The compression load is transmitted through the entire cross-section of the element,
causing it to deform and potentially fail under excessive loading conditions. It is
typically expressed in units of force, such as Newtons (N) or Kilonewton (kN).
Compression testing machine is used to find the compressive strength of specimen.
The compressive strength of cylinder was calculated in reference by following stan-
dard code IS 516-1959. It involves precise sample preparation, testing apparatus,
and a standardized testing procedure to obtain accurate and reliable results. The 60
specimens were placed in the compression testing machine as shown in Fig. 4 in
such a way that the load was applied continuously at a constant rate of 140 kg/cm2
per minute, until the cylinder failed. The maximum load/peak load on the specimen
before failure was recorded. Moreover, 90% of the ultimate load was also applied
to the rest 60-cylinder specimens for the image acquisitions of the cracks. Every
specimen had the stress–strain curve generated as in Fig. 5, showcasing the ultimate
stress.
@seismicisolation
@seismicisolation
Table 2 Fly ash variation with nomenclature and quantity details of cement and fly ash
Casting class Quantity nomenclature Cement (kg/ Fly ash (kg/
Ultimate load with 10 90% of ultimate m3 ) per m3 ) per
specimens load with 10 cylinder cylinder
specimens
A (5% FA (5Aa100, (5Aa90, 5Ab90… 2.55 0.13
Substitution) 5Ab100…0.5Aj100) 5Aj90)
B (15% FA (15Ba100, (15Ba90, 2.28 0.403
Substitution) 15Bb100…15Bj100) 15Bb90…
15Bj90)
C (25% FA (25Ca100, (25Ca90, 2.02 0.67
Substitution) 25Cb100…25Cj100) 25Cb90…
25Cj90)
D (30% FA (30Da100, (30Da90, 1.88 0.807
Substitution) 30Db100…0.30Dj100) 30Db90…
30Dj90)
E (40% FA (40Ea100, (40Ea90, 1.614 1.076
Substitution) 40Eb100…0.40Ej100) 40Eb90…
40Ej90)
F (50% FA (50Fa100, (50Fa90, 1.345 1.345
Substitution) 50Fb100…0.50Fj100) 50Fb90…
50Fj90)
Fig. 4 Compressive strength

test
@seismicisolation
@seismicisolation
Fig. 5 Stress–strain curve
2.5 Image Segmentation Procedure
Image processing procedure consisted of several steps in this research. Initial process
started with optimal lighting condition provided to capture the images of concrete
cylinder. The raw images were captured using pre mounted digital camera. The
images were extracted into the computer for further processing. Image processing
was used with evaluated Python libraries like OpenCV. The images were scaled and
restored to convert it into greyscale images. The segmentation procedure was used to
extract the features for crack detection. Then morphological processing was carried
out to have the clear representation of useful features [16]. The cracks were detected,
and recognition was done using blending with the original image as the final output.
Image Acquisition. Plywood box was prepared with dimension of 0.3 m × 0.15 m ×
0.24 m. It was wrapped with white laminated wooden sheets on every side internally.
LED tube light was placed so as to provide lights in every corner of the box. The
intensity of the light was varied using a connected regulator. Camera was just placed
in front of the opening side of cardboard at the height of 15 cm. The clear distance
between lens and face of cylinder was 53 cm. The digital camera Cannon EOS 200D
paired with the Cannon EFS 18-55 mm lens of a 58 mm filter and 24.1-megapixel
resolution was used to generate the high-resolution raw image samples keeping ISO
value of 250.
Images were captured from multiple projection views, extracting 12 images from
one specimen. The sample of captured image is shown in Fig. 6. Altogether 720
images were captured and stored for the process of image processing from 60
specimens in the research.
Image Processing. The camera produced the image with the width of 6000-pixels
and height of 4020 pixels as shown in Fig. 7. The acquired images were resized to a
@seismicisolation
@seismicisolation
Fig. 6 Plywood Box Studio

for image acquisition
size of 800 × 534 pixels using OpenCV’s function (cv2.resize()) [17]. The size was
scaled down to decrease the file size for all the acquired images to make it easy for
the further processing without degrading the quality of images. The digitized images
were further processed to remove noises and other unwanted features. After that, the
images were enhanced by Python imaging library to adjust the brightness, contrast,
and sharpness followed by restoration process. Then, the RGB format was converted
to grayscale for simpler differentiation of crack in the concrete. The greyscaling was
done using OpenCV function cv2.cvtColor(). Likewise, the segmentation process
was used to divide the images into multiple segments to filter out the region of interest.
The binary threshold segmentation separated the features from the background using
a function cv2.threshold(). Morphological operators such as erosion and dilation were
applied to the segmented image to remove noise, smooth out edges, and extract useful
features such as straight lines and curves [18].
Canny edge detection technique segmented the structural information in the image
using OpenCV function as cv2.Canny() [19]. At the end, the featured cracks were
detected using cv2.findcontours() and cv2.contourarea(). In this technique, specific
threshold was provided as a pixel length. If the features were above the specified
threshold, then the procedure would identify it as a crack [20–23].
Fig. 7 Outline of image processing
@seismicisolation
@seismicisolation
3 Results and Analysis
3.1 Compressive Strength Data
This paper is based upon the digital image segmentation process in differentiating
the crack propagation. In order to validate those results, compressive strength data
were also studied to find an optimum percentage of fly ash replacement.
The specimen class of D as shown in Fig. 8, with fly ash substitution of 30%, had
sustained the maximum compressive force of 464.1 kN, which was the maximum
out of all the specimen class. The compressive strength was somewhat gradually
increasing from class A to D, where the fly ash substitution was being increased at
the rate of 5%. This stage is favorable as the strength is maintained.
But, as the research is based on finding the most economical substitution of fly
ash, it can be stated that we can replace the fly ash till 30% (class D) for getting
the maximum strength and being very economical, which was also concluded by
Golewski. He specified that the fly ash particle imparts their pozzolanicity in the C-
S-H phase of concrete hydration strengthening the interfacial transition zone between
coarse aggregate and the mix paste, thus improving the compactness [7].
Moreover, the strength was severely dropping as the replacement went very high as
in the class E and F. This may be due to the limited chemical reaction between the fly
ash and cement. The fly ash particles cannot fully replace the binding properties of the
cement. Additionally, the large quantity of fly ash increases the water demand of the
mix, resulting in higher water-to-cement ratio and thus decreasing the strength. The
overall results show the potential optimal percentage of fly ash when the substitution
is 30%. The same optimal has been reached using the digital images of cracks, and
thus, the classification is done as per the range.
Fly Ash comparison with Slump & Compressive Strength

COMPRESSIVE STRENGTH (N/mm2)
150 28.00
25.13 25.62 25.94
140 26.25 26.00
130 140 24.00
SLUMP (mm)
134 22.00
120 126
20.00
110 120 107 18.00
100
18.11 90 16.00
90 14.00
80 13.32 12.00
70 10.00
5 15 25 30 40 50
PERCENTAGE OF FLY ASH SUBSTITUTION
Slump Compressive Strength
Fig. 8 Fly ash relation with compressive strength and slump
@seismicisolation
@seismicisolation
Fig. 9 Image histogram of Class D and F
3.2 Image Histogram Data
The segmentation procedure output was analyzed with Otsu thresholding. It gener-
ated the histogram data for each image to analyze the crack pixels. Graphical repre-
sentation of pixel intensity distribution of an image is known as image histogram. It
is a useful tool for image processing and analysis as it can provide information about
image brightness, contrast, and overall quality [13].
In this study, the histogram showed the frequencies of occurrence of white/black
pixels. The white/black pixel intensity was defined as: if the pixel intensity in the
segmented greyscale image exceeded the threshold value, it was set to white (255);
otherwise, it was set to black (0). The white pixel intensities indicated the crack
present in the specimens. The histogram of Class D, with the white intensity level of
3164, as shown in Fig. 9, was resulted as the optimum class for the fly ash replacement
as crack white pixels were comparatively lesser. The histogram of Class F, with the
white pixel intensity of 7929.3, was showing the extreme formation of cracks. These
histograms are few of the ones taken out of the entire image process, which show
average value of its class.
The horizontal bar graph of the analyzed histograms between the fly ash
percentage and white pixel (cracks) of each class average is compared in Fig. 10.
3.3 Classification of Cracks
Based on the frequency of the white pixel intensity 255 (cracks), the classification
range of category has been made as shown in Table 3. With the category classification,
the classes of fly ash substitution have been resulted as shown in Table 4.
@seismicisolation
@seismicisolation
Fly Ash Variation Vs Histograms's White Pixel (Cracks)

Fly Ash Percentage Class Class F 7923.3
Class E 6635
Class D 3164
Class C 3699.2
Class B 5525
Class A 4839
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Frequency of Occurance of White Intensity Pixel
Fig. 10 Fly ash variation versus histogram’s white intensity (crack) pixels
Table 3 Classification of
Category/range Minimum pixel Maximum pixel
category in terms of white
pixel (crack) frequency Low 2931 4624
Medium 4625 6318
High 6319 8243
Table 4 Classifying
Class Category
specimens in terms of
category defined in Table 3 Class A Medium
Class B Medium
Class C Low
Class D Low
Class E High
Class F High
4 Conclusions
The research work was carried out in concrete cylinders containing variations of fly
ash by weight of cement (as classes). Out of total 120 concrete cylinders, 60 cylinders
were utilized in determining the ultimate strength for each class specimen, and the
rest 60 cylinders underwent for the photo acquisitions under controlled setup. The
Otsu threshold image segmentation detected the cracks based on the digital images,
and the histograms were generated for the analysis.
• The class “D” where the substitution of fly ash was 30% has been resulted as
the most economical for fly ash replacement when analyzing the compressive
strength. Although the strength was largely maintained at lower percentage of fly
ash replacement as well, yet the most economical range was looked after in this
presented study.
@seismicisolation
@seismicisolation
• The various histogram generated based on different variations of fly ash replace-
ment displayed different crack propagation with the resulted pozzolanic reactivity
of the fly ash fine particles within the concrete mix.
• Additionally, based on the analysis of image processing, the range between 25
and 30%, falling under the “low crack” category based upon the histogram study
has been concluded as the best and most desirable replacement range of fly ash.
• In a nut shell, from the experimental and image processing, it can be concluded
that the optimal category of fly ash replacement is within the percentage of 25–
30%. It showcases the binding capability of fly ash just like the cement. However,
when fly ash gets abundant, it cannot fully provide the finesse of cement, hence
showing that it cannot entirely replace cement.
References
1. Fauzi A et al (2016) Study of fly ash characterization as a cementitious material. Procedia Eng
148:487–493
2. Sun B et al (2022) A review: reaction mechanism and strength of slag and fly ash-based
alkali-activated materials. Constr Build Mater 326:120–133
3. Vargas J et al (2015) Effective carbon emission reductions from using upgraded fly ash in the
cement industry. J Clean Prod 103:948–959
4. Yasmin H et al (2018) Influence of activation of fly ash on the mechanical properties of concrete.
Constr Build Mater 172:728–734
5. Muhammad R et al (2020) A stoichiometric approach to find optimum amount of fly ash needed
in cement concrete. SN Appl Sci 2:1–9
6. Safiuddin M et al (2018) Early-age cracking in concrete: causes, consequences, remedial
measures, and recommendations. Appl Sci 8(10):17–30
7. Golewski G (2021) Validation of the favorable quantity of fly ash in concrete and analysis of
crack propagation and its length—using the crack tip tracking (CTT) method—in the fracture
toughness examinations under Mode II, through digital image correlation. Constr Build Mater
296:12–23
8. Oderji S et al (2019) Fresh and hardened properties of one-part fly ash-based geopolymer
binders cured at room temperature: effect of slag and alkali activators. J Clean Prod 225:1–10
9. Hefni Y et al (2018) Influence of activation of fly ash on the mechanical properties of concrete.
Constr Build Mater 172:728–734
10. Dorafshan S et al (2018) Comparison of deep convolutional neural networks and edge detectors
for image-based crack detection in concrete. Constr Build Mater 186:1031–1045
11. Gehri N et al (2020) Automated crack detection and measurement based on digital image
correlation. Constr Build Mater 256:119–383
12. Fan L et al (2023) A rapid approach for determining the mechanical properties of geopolymer
concrete based on image processing technology (IPT). Constr Build Mater 378:131–165
13. Wang Y et al (2019) Research on crack detection algorithm of the concrete bridge based on
image processing. Procedia Comput Sci 154:610–616
14. Asmari A et al (2017) Segmenting fly ash particles using image processing techniques. In:
2017 international mechanical engineering congress and exposition. ASME, Tampa, pp 1–8
15. Goh T et al (2018) Performance analysis of image thresholding: Otsu technique. Measurement
114:298–307
16. Dougherty E (2020) Digital image processing methods, 1st edn. CRC Press, Rochester
17. Devi R (2017) Geometric transformations and thresholding of images using Opencv-Python.
GRD J Eng 2(11):49–52
@seismicisolation
@seismicisolation
18. Zhao X et al (2017) Canny edge detection based on Open CV. In: 13th IEEE international
conference on electronic measurement & instruments. IEEE, pp 10650–10668
19. Kumar S et al (2023) Morphological operations and histogram analysis of SEM images using
Python. Indian J Eng Mater Sci 29(16):796–800
20. Talab A et al (2016) Detection crack in image using Otsu method and multiple filtering in
image processing techniques. Optik 127(3):1030–1033
21. Waris M et al (2022) An alternative approach for measuring the mechanical properties of hybrid
concrete through image processing and machine learning. Constr Build Mater 328:126–138
22. Valença J et al (2017) Assessment of cracks on concrete bridges using image processing
supported by laser scanning survey. Constr Build Mater 146:668–678
23. Dogan G et al (2017) Concrete compressive strength detection using image processing based
new test method. Measurement 109:137–148
@seismicisolation
@seismicisolation
Building Surface Crack Detections Using
Deep Convolutional Neural Network
(DCNN) Architectures
Rajashri Khanai, Basavaraj Katageri, Dattaprasad Torse,

and Rajkumar Raikar
1 Introduction
Surface fractures are a major problem in concrete construction. Building inspections

are carried out to assess the tensile strength and rigidity of the structure. Impor-
tant crack detection features of building inspection include locating the cracks and
determining the building’s condition. Concrete constructions are often inspected via
visual inspection of their surfaces. However, this procedure is time-consuming, labor-
intensive, and dangerous for inspectors to use. Additionally, the engineers conducting
the examination depend on their expertise for their skill level, competency, and
experience. Unmanned aerial vehicles and machine learning techniques have been
investigated as automated solutions for crack detection evaluation to get over these
restrictions. Automation in evaluations has many advantages, including lower costs,
greater efficiency, less labor requirement, and lower danger of workplace accidents
[1]. The condition evaluation is also made using a computer algorithm, making it
more unbiased and trustworthy. Numerous studies have looked into the feasibility of
automating the assessment of concrete structures using image processing techniques
(IPT) and DL approaches [2] employing an edge detection method based on the IPT
to detect cracks in concrete bridge pictures automatically. Edge detection methods
were shown to be more effective in identifying cracks than edge detection algorithms,
with the best accuracy rating being 86%.
Structural damage assessment has received more attention recently as the safety
of ageing structures, and civil infrastructure has become more crucial. The presence
of structural exterior damage is a structural safety factor that is directly related to
the occurrence of cracks in a structure. In order to assess the status of fractures
in structures, labor has been utilized to monitor them [3]. Because enormous civil
constructions may be difficult to access in addition to experts’ skills, traditional
R. Khanai (B) · B. Katageri · D. Torse · R. Raikar

KLE Technological University Belagavi Campus, Belagavi, India
@seismicisolation
@seismicisolation
354 R. Khanai et al.
monitoring techniques do not provide high degrees of objectivity. As a result, there

is a growing need for automated crack identification and quantification technologies
[4].
2 Convolutional Neural Network Architectures
To learn spatial hierarchies of data automatically and adaptively, convolutional

neural networks employ the backpropagation method, which consists of many layers,
including convolution layers, pooling layers, and fully connected layers [5].
The depth of Deep CNNs (DCNNs) is what gives them their strength. The red,
green, and blue colors of a picture are processed simultaneously by DCNNs using a
three-dimensional neural network. Compared to feed-forward neural networks, this
requires fewer neurons to process images.
Images are used as input by DCNN to train the classifier. The network uses a
novel mathematical method called “convolution” to replace matrix multiplication.
The following advantages of the CNN architecture are listed [6]. CNN works well
because it shares parameters and uses special convolution and pooling methods. The
CNN models are now well-known worldwide and may be used on any device. They
can be used in a variety of businesses to do important tasks like face recognition,
data analysis, weather detection, image recognition, and property identification. By
entering the data at each level and slightly modifying the CNN for a particular
purpose, one can extract the essential properties and their weights displayed by the
pre-trained CNN [7].
For numerous image processing applications, CNNs have a variety of models
including MobileNetV2, DenseNet-121, and VGG architectures.
3 Methodology
The noise level data of Deepawali and pre-Deepawali days of four places in Delhi
were initially analyzed using classifier learner and regression learner on MATLAB
platform. Due to the small amount of data, the predictions were not promising. Hence,
python algorithms were employed to build the models. In the current study, different
regression algorithms were employed. They are linear regression, Gaussian Naïve
Bayes Regression, Multinomial Naïve Bayes Regression, Bayesian Ridge Regres-
sion, Bayesian Ridge Regression, Logistic Regression, and Quantile Regression.
These algorithms were run in the Google Colaboratory. The data available were only
72 in number. By using random number technique, random noise pollution data were
generated up to 150 and 300. Using these three data sets, the prediction is performed.
According to the survey indicated above, while each damage detection strategy
has advantages and disadvantages of its own, a general technique that can consider all
@seismicisolation
@seismicisolation
Building Surface Crack Detections Using Deep Convolutional Neural … 355
aspects of structural damage detection is still a research topic of interest. The short-
comings of previously published damage detection methods are therefore addressed
in our work.
3.1 Dataset
Pictures of diverse concrete surfaces, both with and without fractures, can be seen
in the files. For image classification, the visual data are split into two categories:
negative (without crack) and positive (with crack). There are 20,000 images in each
class, for a total of 40,000 images with 227 227 pixels and RGB channels [8].
This information was gathered from the Mendeley Data-Crack Detection website,
which was created by Alar Frat Zgenel. Figure 1a, b depict the samples of 20,000
each of positive means crack and negative means non-crack photos. 70:30 images
are considered for training and testing.
3.2 Deep Learning Model for Crack Detection
Using traditional ML techniques, the classification task must be completed in a series

of sequential steps, including pre-processing, feature extraction, careful feature selec-
tion, learning, and classification. Furthermore, the performance of ML approaches
is significantly impacted by feature selection. Biased feature selection may result
in inaccurate class distinction. In contrast to conventional ML methods, DL can
automate the learning of feature sets for a variety of applications [5]. Learning and
classification can be accomplished simultaneously in DL as shown in Fig. 2.
3.3 MobileNetV2 Architecture
The MobileNetV2 architecture is made to let CNNs operate effectively on mobile

platforms. The fluctuation of the residue with the remaining coupling between the
layers [9] is more significant. Figure 3 illustrates the MobileNetV2’s procedural flow.
According to the confusion matrix in Fig. 4, the MobileNetV2 model identified
6058, 5913, 24, and 5 photos as True Positive (TP), True Negative (TN), False Positive
(FP), and False Negative (FN) out of 12,000 test images. The model’s efficacy is
99.81%, as shown in Table 1.
@seismicisolation
@seismicisolation
a
Positive Negative Positive
Negative Positive Positive
Fig. 1 a Positive and negative images. b Positive and negative images
Fig. 2 Deep learning model Classification

Input Images
for crack detection Deep Learning
Deep Learning Model
3.4 Architecture of DenseNet-121
Every convolutional layer in a conventional feed-forward CNN, aside from the basic
one, receives input from the layer before it and generates an output feature map,
which is then sent to the layer after it [10]. Consequently, as illustrated in Fig. 5,
@seismicisolation
@seismicisolation
Fig. 3 Flowchart of
MobileNetV2
Fig. 4 Confusion matrix for

test data MobileNetV2
architecture
Table 1 Performance parameters of MobileNetV2

Precision Recall F1 score Support Test loss Test accuracy
Positive 01.00 01.00 01.00 6058 0.04124 99.81%
Negative 01.00 01.00 01.00 5913
Accuracy 01.00 01.00 01.00 12,000
@seismicisolation
@seismicisolation
Fig. 5 Concept of DenseNet configuration
there are “L” direct connections for “L” layers, one between each layer and the layer
below it.
However, if CNN has more layers or becomes deeper, the “vanishing gradient”
problem starts to show itself. This implies that when the communication channel
between the input and output layers lengthens, some information may “vanish” or
“get lost,” reducing the effectiveness of the network’s training.
By altering the typical CNN architecture and streamlining the connectivity
between layers, DenseNets alleviates this issue. Convolutional networks with densely
linked layers (DenseNet) have layers that are all directly coupled to one another. For
“L” laid systems, there are L(L + 1)/2 direct connections.
Additionally, it has proven to be startlingly accurate. DenseNet concatenates the
output of the preceding layer and adds it to the output of the following layer. Connec-
tivity, DenseBlocks, Growth Rate, Bottleneck layers, and Connectivity make up
DenseNet.
Instead of being averaged, the feature maps from all the previous levels are
combined and used as inputs in each layer. DenseNets require fewer parameters
than a similar regular CNN, allowing for feature reuse because duplicate feature
mappings are eliminated. As a result, the lth layer receives as input the feature maps
of all layers before it, x0 , . . . , xl−1 .

xl = Hl x0 , x1 , . . . , xl−1
To make this easier, DenseNets are divided into DenseBlocks, where the size
of the feature maps inside a block is fixed but the number of filters between them
changes. The layers that separate the blocks are called transition layers, and they
reduce the number of channels to half of what they were before.
The above equation defines Hl as a composite function that executes batch normal-
ization (BN), a rectified linear unit (ReLU), and a convolution in that sequence for
each layer.
where x0 , x1 , . . . , xl−1 , is the output from all layers before layer l (0, …, l − 1)
or the concatenation of feature maps. Hl’s numerous inputs are consolidated into a
single tensor to simplify implementation.
Figure 6 depicts a deep DenseNet with three dense blocks. The transition layers
between two adjacent blocks implement down-sampling (i.e., change the size of
@seismicisolation
@seismicisolation
Fig. 6 Deep DenseNet structure
Fig. 7 Layers of DenseNet-121
the feature maps) through convolution and pooling procedures, while the size
of the feature maps within the dense block remains the same to support feature
concatenation.
The layers of DenseNet-121 are as follows:
1: Convolution of 7 × 7; 58: Convolution of 3 × 3; 61: Convolution 1 × 1;
AvgPool, Layer 1, and 4, respectively shown in Fig. 7.
In conclusion, DenseNet-121 has 120 Convolutions and 4 AvgPool.
Because all levels, including those in the same dense block and transition layers,
distribute their weights over numerous inputs, deeper layers can utilize attributes that
were extracted earlier.
Since they offer a lot of duplicate information, the layers in the second and third
dense blocks produce the output of the transition layers with the fewest weights.
Furthermore, even though the last layers take the weights from the entire dense
block, it is likely that higher-level features are produced further into the model
because there seems to be more emphasis on final feature mappings in trials.
The layers in the second and third dense blocks provide the output of the transition
layers with the least weight since they provide a large amount of redundant infor-
mation. The DenseNet model resulted into TP, TN, FP, and FN are 6078, 5910, 4,
and 8 images respectively out of 12,000 test images as depicted in confusion matrix
shown in Fig. 8. The efficiency of the model is 99.51% presented in Table 2.
@seismicisolation
@seismicisolation
Fig. 8 Confusion matrix for

test data DenseNet-121
architecture
Table 2 Performance parameters of DenseNet

Positive 01.00 01.00 01.00 6082 0.01813 99.51%
Negative 01.00 01.00 01.00 5918
Accuracy 01.00 01.00 01.00 12,000
3.5 VGG16 Architecture
The VGG network was first conceptualized by Karen Simonyan and Andrew
Zisserman in 2013, and their actual model—which was based on that concept—
was submitted for the 2014 ImageNet Challenge. They named it VGG in recognition
of the Visual Geometry Group department at the University of Oxford, to which they
belonged [8].
The VGG stands out because it uniformly distributes three 3 × 3 filters. A 5 × 5
effective receptive field is produced by two consecutive 3 × 3 filters. Like this, three
3 × 3 filters are combined to create a 7 by 7 receptive field. In this strategy, a larger
receptive area can be represented by a combination of several 3 3 filters.
As a result, decision-making processes grow more sophisticated. The network
would then be able to converge more quickly as a result.
One of the best vision model designs ever developed is this one. The 16 weighted
layers are referred to as VGG16’s 16 layers. Figure 9 depicts this huge network,
which includes 138 million parameters.
VGG16 has some restrictions:
Training takes a lengthy time (with an Nvidia Titan GPU, the original VGG model
required 2–3 weeks). The size of the VGG16 trained imageNet weights is 528 MB.
Exploding gradients are a result of the 138 million parameter problem.
@seismicisolation
@seismicisolation
Fig. 9 VGG16 architecture
The DenseNet model resulted into TP, TN, FP, and FN are 6073, 5904, 9, and 14
images respectively out of 12,000 test images as depicted in confusion matrix shown
in Fig. 10. The efficiency of the model is 99.76% presented in Table 3.
Fig. 10 Confusion matrix

for test data using VGG16
architecture
Table 3 Performance parameters of VGG16

Positive 01.00 01.00 01.00 6082 0.00840 99.76%
Negative 01.00 01.00 01.00 5918
Accuracy 01.00 01.00 01.00 12,000
@seismicisolation
@seismicisolation
3.6 VGG19 Architecture
Images are categorized using a deep CNN dubbed VGG. Two Conv3 × 3 layers (64),
two MaxPool levels of Conv3 × 3 (128), four MaxPool layers of Conv3 × 3 (256),
four MaxPool layers of Conv3 × 3 (512), two MaxPool layers of Fully Connected
(4096), and one Fully Connected layer (1000) make up the layers of the VGG19
model. VGG19 and VGG16 have remarkably comparable architectures. The three
more convolutional layers for the VGG16 network are shown in Fig. 11.
According to the confusion matrix presented in Fig. 12, the VGG19 model
produced TP, TN, FP, and FN as 5892, 6074, 24, and 10 pictures, respectively, out
of 12,000 test images. The model’s efficacy is 99.62%, as shown in Table 4.
The comparison of all four models is shown in the following Table 5 and
mobilenetV2 has better accuracy of 99.81%.
Fig. 11 VGG19 architecture
Fig. 12 Confusion matrix

for test data using VGG19
architecture
@seismicisolation
@seismicisolation
Table 4 Performance parameters of VGG19

Positive 01.00 01.00 01.00 6082 0.01353 99.62%
Negative 01.00 01.00 01.00 5918
Accuracy 01.00 01.00 01.00 12,000
Table 5 Comparison of 4
Test loss Test accuracy (%)
models
MobilenetV2 0.04124 99.81
DenseNet-121 0.01813 99.51
VGG16 0.00840 99.76
VGG19 0.01353 99.62
4 Conclusions
In the present study, DCNNs models are explored on 40,000 images with two cate-
gories of data. The surface crack detection is evaluated using Tensorflow by fine-
tuning a few DCNN models such as VGG19, VGG16, MobileNetV2, and DenseNet.
The four models are trained with 70% (28,000 images comprising positive and nega-
tive images) of the data and tested with 30% of the data. The mobileNetV2 has
achieved test accuracy of 99.81% and test loss of 0.04124. The model VGG16 has
given test loss of 0.00840 and test accuracy of 99.76%.
References
1. Latifi R et al (2023) A brief overview on crack patterns, repair and strengthening of historical
masonry structures. Materials 16(5):1–22
2. Laxman KC et al (2023) Automated crack detection and crack depth prediction for reinforced
concrete structures using deep learning. Constr Build Mater 370
3. Kim J et al (2023) Learning structure for concrete crack detection using robust super-resolution
with generative adversarial network. Struct Control Health Monit 1–16
4. Dais D et al (2021) Automatic crack classification and segmentation on masonry surfaces using
convolutional neural networks and transfer learning. Autom Constr 1–18
5. Cha Y-J, Choi W, Büyüköztürk O (2017) Deep learning-based crack damage detection using
convolutional neural networks. Comput-Aided Civil Infrastruct Eng 32(5):361–378
6. Inam H et al (2023) Smart and automated infrastructure management: a deep learning approach
for crack detection in bridge images. Sustainability 15(3):1–35
7. Katageri B, Khanai R, Raikar RV, Torse DA, Pai K (2023) Surface crack detection using deep
learning framework for civil structures. Struct Integrity Life 23(1):31–37
8. Ai D et al (2023) Computer vision framework for crack detection of civil infrastructure—a
review. Eng Appl Artif Intell 14(1):1–26
@seismicisolation
@seismicisolation
9. Mohan A, Poobal S (2018) Crack detection using image processing: a critical review and
analysis. Alexandria Eng J 57(2):787–798
10. Nguyen A et al (2023) ASR crack identification in bridges using deep learning and texture
analysis. Structures 50
@seismicisolation
@seismicisolation
Using Construction Sequence Analysis
to Mitigate Risk and Prevent Failures
Danette M. Gonsalves and Sumitra S. Kandolkar
1 Introduction
Usually, building analysis is in a single-step, linear elastic method called conven-

tional analysis (CA) wherein it is assumed that all loads act on the building only
once its construction is complete. Engineers thus determine structural responses
such as axial force, bending moment, shear force and displacements and accordingly
design structures. However, as the construction proceeds storey-by-storey, the frame
loads get applied in stages wherein their dead load is supported by the completed
portion of the structure only. Column shortening and undesirable deflection due to the
self-weight of the completed structural members will inevitably occur. The unbuilt
members do not impact the distribution of structural responses within the constructed
storeys. CSA is a nonlinear, static analysis that considers this behaviour. In both cases
of analyses, the structure’s performance differs significantly with increasing storey
height [1–4], and thus, careful designing of multi-storey structures is essential. CSA
considers the stages of load application on the frame, total number of days required
and the order of casting structural members. The lack of knowledge on its impor-
tance in analysis deems CSA less popular, but with advanced software capabilities,
designers can simulate nonlinear analysis, especially in high-rise structures.
Providing floating columns in buildings improves the aesthetics but causes a
discontinuity between the storey it rests on, and the storey beneath it. Since the storey
shears are unable to transfer to the ground via the shortest load path available, the
building’s usability and stability during an earthquake is endangered. Thus, IS 1893
(Part 1): 2016 [5] discourages the inclusion of floating columns in buildings. Shear
D. M. Gonsalves (B)
PG student, Goa College of Engineering, Goa, India
e-mail: danves1999@gmail.com
S. S. Kandolkar
Professor, Goa College of Engineering, Goa, India
e-mail: sumitra@gec.ac.in
@seismicisolation
@seismicisolation
366 D. M. Gonsalves and S. S. Kandolkar
walls constructed from the foundation to the topmost-storey level provide lateral
stiffness to the structure and reduce the structural responses of members. Usually in
a normal loading state, the shear walls are less stressed as they resist lateral loads
and columns resist gravity loads. The exterior columns of buildings are subjected
to almost one-half of gravity load as compared to an interior column with identical
sectional area, thus creating an inequality in the applied gravity load-to-section area
ratios between the two column types [6, 7]. The result is an axial force redistribution
among vertical members, leading to differential elastic shortening along the building
height and deteriorating its serviceability by severely damaging horizontal members
such as beams and slabs. Following this, slabs may develop a slight tilt, thereby
inducing additional internal forces and bending moments in rigidly connected beams
between the vertical members. The result is damaged core-walls and non-structural
elements, such as partitions, curtain walls and pipes [8]. The amount of column
shortening is regulated by its stiffness, and the concrete grade used. The relative
stiffness of two adjacent spans influences the distribution of bending moments in
the frame [9]. The stiffness of a member is inversely proportional to its length but
directly proportional to its elastic modulus ‘E’ where f ck is the concrete characteristic
strength in N/mm2 and ‘E’ expressed in N/mm2 is given in Eq. (1) as:
√
E = 5000 f ck (1)
In this paper, the effects of using M60 and M40 concrete grades were compared
using CA and CSA. The objectives of this study are to analyse construction sequences
to identify potential safety hazards, analyse the structural responses of members
and conclude the effectiveness of adopting this design approach, thereby making
CSA a proactive tool to mitigate risks, prevent failures and improve performance of
structures.
2 Methodology
The preliminary checks on member sizes were done according to load cases and
combinations as per the Indian Standard codes, and the first analysis model was
executed for conventional analysis with dead load only. For the same total dead load,
the program formulates an iterative, nonlinear static load case that assigns vertical
loads from the ground up to that specific storey. The components are grouped as
floors, so the software identifies all the steps required to complete the procedure.
The program stores the results of structural responses until the last constructed
stage, which function as the initial parameters to analyse the consecutive stage, thus
ensuring that residual stresses are assessed at every storey level and the sequential
effects are accurately visualised.
@seismicisolation
@seismicisolation
Using Construction Sequence Analysis to Mitigate Risk and Prevent Failures 367
3 Finite Element Modelling
A typical elevation of a modelled structure as determined by CA seen in Fig. 1a. A

sample three-storey structure illustrates the method of CSA as in Fig. 1b. Table 1
presents the details of columns, and Table 2 presents other structural details adopted
in the analysis.
Fig. 1 Methods of analyses: a conventional analysis; b construction stage analysis
Table 1 Column dimensions and their locations

Column type Base to 10th storey 11th to 20th storey 21st to 30th storey
External, floating (4A, 500 mm × 700 mm 500 mm × 500 mm 500 mm × 400 mm
4H)
Internal 500 mm × 700 mm 500 mm × 600 mm 500 mm × 400 mm
Columns supporting the 600 mm × 750 mm 600 mm × 600 mm 600 mm × 500 mm
transfer girder
Table 2 Other structural details

Member/material property Dimensions/details
External beam 400 mm × 700 mm
Internal beam 300 mm × 500 mm
Transfer girder at storey 2 500 mm × 1000 mm
Slab 150 mm thick
Internal wall 115 mm thick
External wall 230 mm thick
Grade of concrete M60, M40, M50-shear wall, M30-slabs
Grade of steel Fe 550
Typical storey height 3.2 m
Height between ground floor and 1st floor 2.7 m (for stilt parking)
Height between foundation and ground levels 1.5 m
Shear wall 200 mm thick
Diaphragm Rigid type
@seismicisolation
@seismicisolation
Table 3 Loads considered

Material Density/intensity
Density of concrete 25 kN/m3
Floor finish, roof treatment 1 kN/m2
External wall load 10.84 kN/m
Parapet wall load 4.4 kN/m
Internal wall load 5.85 kN/m
Live load on floors 3 kN/m2
Live load on roof 0.75 kN/m2
Table 4 Loading combinations according to IS 456: 2000

Loads Loading combinations
Dead load, live load 1.5 (DL + LL)
Dead load, live load, earthquake/wind load in X or Y direction 1.2 (DL + LL ± (EL/WL))
Dead load, earthquake/wind load in X or Y direction 1.5 (DL ± (EL/WL))
Dead load, earthquake load in X or Y direction 0.9DL ± 1.5EL
Various dead and live load considerations given in Table 3 were as per IS: 875
(Part 1)-1987 and IS: 875 (Part 2)-1987 [10]. A storey’s infill structural walls were
assumed to be constructed upon completion of the frame of that storey. There was no
access provided to the flat roof. IS 875 (Part 3): 2015 [11] was used to perform wind
analysis on all the models, along with seismic analysis as per IS 1893 (Part 1): 2016
[5]. The building in question is a reinforced concrete residential building situated in
the crowded region of Bangalore on medium soil. To obtain safe member sizes, the
model was analysed as per IS 456: 2000 [12] with all the loading combinations as
given in Table 4.
The beams and columns are modelled as frame elements, and the slabs are
modelled as membranes. The wall frame loads are loaded as super-imposed dead
loads on the beams. To compare the results of CA and CSA, only the wall loads and
the members’ self-weights were considered, as it is assumed that during construction,
the structure will be affected by only these two load types. Only after the building is
fully constructed and occupied will the live loads and lateral loads begin affecting it.
This sequential load case ignores the effects of time-dependent material parameters.
3.1 Storey-Wise Response Plots
ETABS v20 was used to model two 30-storey buildings identical in plan, as under:
1. Case A: all beams were of M40 concrete, and all columns were of M60 concrete.
2. Case B: all beams were of M60 concrete, and all columns were of M40 concrete.
@seismicisolation
@seismicisolation
3.2 Response of Transfer Girder
Six buildings with identical plans were modelled in ETABS v20 to examine the
storey effects of the building constructed with 5, 10, 15, 20, 25 and 30 storeys. These
models will depict the variation of bending moment and deflection in the transfer
girder.
In this study, the objective was to establish a baseline for comparing the structural
responses of two identical 30-storey buildings under different material configura-
tions. Figure 2 shows the building modelled in ETABS v20 by considering two
cases: Case A and Case B. The structural responses of both buildings were eval-
uated based on parameters of the maximum displacement and bending moment in
the transfer girder, the axial force and deformation in columns B2 and C1, and the
variation of bending moment and shear force in the beam. The following discussion
presents the results obtained for each case and highlights the key differences noted
by a ratio δ, calculated by Eq. (2) as:
Structural response by CSA

δ= (2)
Structural response by CA
The analysis of Case A revealed the following results:

1. Overall Structural Behaviour: The structure exhibited relatively higher stiffness
due to stronger M60 concrete columns, thus resulting in improved structural
stability.
Fig. 2 Building in ETABS: a plan view with members marked for study; b 3D view
@seismicisolation
@seismicisolation
Fig. 3 Storey-wise plot of external column C1 for: a axial deformation; and b ratio δ
Fig. 4 Storey-wise plot of internal column B2 for: a axial deformation; and b ratio δ
2. Column Responses: When the external and internal columns of higher concrete
grade are loaded, they undergo lower axial deformations as shown by Figs. 3a
and 4a, respectively; while carrying higher axial loads, as seen in Figs. 5a and
6a. Figures 3b and 4b show constantly overlapping plots, unlike Figs. 5b and 6b.
3. Beam Responses: Even though the beams were constructed of a lower grade than
the columns, the shear forces at the left end presented in Fig. 7a and the right
end in Fig. 7b, of the beam almost matched the shear forces of the beam in Case
B. However, the bending moments of the beam as seen in Fig. 8a were constant
from the lowest to the mid storeys and then slightly increased towards the top
storeys and then faced a drop in value at the top-most storey. In Fig. 8b, the ratio
δ between results of CSA and CA indicates that there is a lesser deviation from
unity at lower storeys.
4. Transfer Girder Responses: The observed maximum deflection of the girder in
Fig. 9a is on the higher side as it is composed of concrete with a lower strength.
The deflection in the transfer girder increases with the addition of more floors. The
sagging moment seen in Fig. 10a is lower while the hogging moment presented
in Fig. 10b did not show much change in comparison with Case B.
@seismicisolation
@seismicisolation
Fig. 5 Storey-wise plot of external column C1 for: a axial force; and b ratio δ
Fig. 6 Storey-wise plot of internal column B2 for: a axial force; and b ratio δ
Fig. 7 Storey-wise shear force plot for interior beam 4-EF at: a left end; and b right end
The analysis of Case B yielded the following results:

1. Overall Structural Behaviour: The building exhibited slightly less stiffness and
rigidity compared to Case A due to the use of M40 concrete columns. However,
@seismicisolation
@seismicisolation
Fig. 8 Storey-wise plot of interior beam 4-EF for: a sagging moment; and b ratio δ
Fig. 9 Plot of transfer girder displacement for a group of storeys: a Case A; and b Case B
the overall structural performance remained satisfactory and met the required
safety criteria.
2. Column Responses: Since the plots of external column in Figs. 3 and 5, and
internal column in Figs. 4 and 6, are of a weaker concrete grade, their capacity
to resist axial load is lower, making them experience higher axial deformations.
3. Beam Responses: Although the beams are constructed of a higher concrete grade
than the columns, the shear forces values at the left end in Fig. 7a and the right
end of the beam in Fig. 7b run parallel to those in Case A. However, the sagging
moments of the beam seen in Fig. 8a were like Case A. The ratio δ shows that
there is a slight deviation from unity, as displayed in Fig. 8b.
4. Transfer Girder Responses: The maximum deflection plot as observed in Fig. 9b
is slightly lesser as it is composed of concrete with a higher strength. Like Case
A, the deflection in the transfer girder increased when more floors were added.
The sagging moment seen in Fig. 10a is higher, while the hogging moment shown
by Fig. 10b did not display much change in comparison with Case A.
Based on the results obtained from the analysis of the two cases, the following
observations can be made:
1. Overall Structural Behaviour: Case A, with stronger columns (M60 concrete),
exhibited better stiffness and rigidity compared to Case B. However, both
@seismicisolation
@seismicisolation
Fig. 10 Transfer girder

moments in storeys:
a sagging moment; and
b hogging moment
cases met the required safety criteria and demonstrated satisfactory structural
performance.
2. Column Responses: Case A demonstrates fewer axial deformations as in Figs. 3
and 4, and higher axial load capacity seen in Figs. 5 and 6 as compared to Case B,
thus implying that using stronger concrete yields favourable column responses.
3. Beam Responses: The shear forces of the beam in Case A at the left end and the
right end match those in Case B as seen in Fig. 7a, b. When analysed by CSA, the
sagging moments in Case A and Case B share the same values; but if analysed
by CA, the sagging moments in Case B are more than those in Case A, as seen
in Fig. 8.
4. Transfer Girder Responses: Case A exhibited higher maximum deflection
compared to Case B as shown in Fig. 9, along with lower sagging moments given
in Fig. 10a, while the hogging moments displayed in Fig. 10b did not show much
deviation from Case B. This indicates that the use of weaker concrete contributed
to increased deflection in the girder, along with lesser moment capacity of the
section.
@seismicisolation
@seismicisolation
5. If transfer girders are not analysed for the order of construction, the analysis
infers all loads on it are transferred by truss action. Thus, CSA procures higher
values of moments and deflection in the transfer girder in comparison with CA.
5 Conclusions
Two buildings identical in plan and member sizes were analysed for the conventional
loads and sequential loads. The bending moments and shear forces in beams and axial
force and axial deformation in columns were compared for concrete grades M60 and
M40 along with ratios to depict the performance of both analyses to conclude that:
1. In comparison with lower concrete grade, columns and beams of higher concrete
grade deform and deflect lesser and are better at resisting more loads, shear and
moments.
2. In CSA, the cumulative weight at every floor increases by adding more storeys;
this may be because upper floors have more inbuilt member stiffness than lower
floors.
3. Buildings are made to withstand a variety of load combinations, and since the
dead load contribution to the critical load combination typically ranges from 20
to 50% depending on various scenarios, the relative impact of CSA on design
reinforcement diminishes. Thus, the requirement for design reinforcements has
not changed despite the significant differences in response quantities for dead
load under CA and CSA.
3. Compared to columns, beams are more susceptible to sequential loading.
4. Even though axial load might be lower after considering sequential effects, this
should not be interpreted as the end of the process. Structures must be well-
designed for every preliminary stage, not just the final one.
6. In CSA, external columns are loaded with the weight of the newly constructed
floor before the adjacent internal columns are constructed. Thus, due to the unbal-
anced loading, the external column experiences increased axial deformation. It
may also be attributed to differential settlement between external and internal
columns related to the sequence followed. In CA, the building is built all at once,
with uniform loading. As a result, there is less differential settlement between
columns overall than there would be with sequential construction.
7. As the lowest storey is fixed, it supports the greatest load, while the column
length is the least. The opposite occurs at the topmost storey. Thus, the axial
deformations at the extremities almost match. However, at the building’s mid-
height, the axial load and column length have a significant value; thus, the axial
deformation attains a maximum value, giving rise to a parabolic curve of CSA.
However, the deformations in CA increase with the building height.
@seismicisolation
@seismicisolation
References
1. Banugariya ND, Solanke SS (2019) Analysis of multi-storey structures with respect to sequen-
tial analysis. In: Smart technologies for energy, environment and sustainable development:
select proceedings of ICSTEESD 2018. Springer, Singapore, pp 367–379
2. Das GG, Praseeda KI (2016) Comparison of conventional and construction stage analysis of a
RCC building. Int J Sci Technol Eng 3(3)
3. Dinar Y, Rasel MM, Chowdhury MJA, Ashraf MA (2014) Chronological construction sequence
effects on reinforced concrete and steel buildings. Int J Eng Sci 3:52–63
4. Pham TB (2022) Construction sequence analysis of reinforced concrete high-rise buildings. J
Sci Tech-Sect Special Constr Eng 5(02)
5. IS 1893 (Part 1): 2016 Indian standard criteria for earthquake resistant design of structures.
Bureau of Indian Standards, New Delhi
6. Choi CK, Kim ED (1985) Multistory frames under sequential gravity loads. J Struct Eng
111(11):2373–2384
7. Choi CK, Chung HK, Lee DG, Wilson EL (1992) Simplified building analysis with sequential
dead loads—CFM. J Struct Eng 118(4):944–954
8. Jayasinghe MTR, Jayasena WMVPK (2004) Effects of axial shortening of columns on design
and construction of tall reinforced concrete buildings. Pract Period Struct Des Constr 9(2):70–
78
9. Chakrabarti SC, Nayak GC, Agarwala SK (1978) Effect of sequence of construction in the
analysis of multistoryed building frame. Build Environ 13(1):1–6
10. BIS, IS: 875 (Part 1 and Part 2) (1987) Indian Standard code of practice for design loads (other
than earthquake) for buildings and structures. Bureau of Indian Standards, New Delhi
11. IS 875 (Part 3): 2015 Indian standard code of practice for design loads (other than earthquake)
for buildings and structures. Bureau of Indian Standards, New Delhi
12. IS 456: 2000 Indian standard plain and reinforced concrete—code of practice. Bureau of Indian
Standards, New Delhi
@seismicisolation
@seismicisolation
Seismic Study on Step Back Buildings
and Step Back Setback Buildings
by Providing Bracing in the Soft Storey
Sinju Jose and Binu M. Issac
1 Introduction
Past earthquakes have severely damaged both infrastructure and human lives. Earth-
quakes should be of major concern as it may cause significant damage to the society
if they are of high magnitude and intensity [1]. Due to the nature of terrain on the
hilly regions, it has become necessary to construct most of structures on slope. This
may lead to variation in the heights of columns at different floors. Due to the sloping
nature of ground some columns in the building behave as short columns and some
will behave as long columns. During earthquakes, the shorter columns become stiff
and attracts more lateral load. This can lead to severe damage because of significant
change in the response under earthquakes. This is known as short column effect.
Hence, seismic analysis of RCC framed structures in slopes is of great significance.
It is also an accepted fact that the behaviour of RCC short column structure under
lateral loading in hilly regions with and without infill varies [2]. The buildings in the
hilly regions are generally very irregular and asymmetric. Due to the irregularity and
asymmetry which is associated with the buildings located in hilly areas, the centre
of mass of buildings and their centre of rigidity do not coincide resulting in exces-
sive shear and torsion [3]. Step back, step back and setback, and setback are three
common building configurations adopted in hilly regions [4]. Due to the nonuniform
distribution of mass and stiffness in irregularly shaped buildings, the analysis is more
difficult.
In case of high rise buildings, stiffness is more important when compared to
strength. Moment resisting frames and braced frames have been commonly used
as lateral load resisting structural elements in different buildings. Moment resisting
frames provide ductility through yielding, but due to its flexibility, they do not satisfy
the stiffness criteria [10]. There are several ways of providing braces to increase
S. Jose (B) · B. M. Issac

Amal Jyothi Collage of Engineering, Kanjirapally, Kerala, India
e-mail: sinjujosethadathil2014@gmail.com
@seismicisolation
@seismicisolation
378 S. Jose and B. M. Issac
the seismic resistance of buildings. The different bracing systems include typical
diagonal bracing, X bracing, chevron bracing and V bracing configurations, which
connect the brace concentric to beam–column joint [5, 7].
Infill walls are the non-structural walls which are built in between the columns of
a building or structure [6]. During an earthquake, the infill walls can have different
effects on the overall structural response of the building. Infill walls can increase
the rigidity and stiffness of the building, thereby resulting in an improvement in
its seismic performance. The strength and stiffness of the infill walls can resist the
external forces and reduce the deformation of the structure. The infill walls can
add to the overall damping of the structure by generating frictional forces, damping
vibrations and dissipating the energy. This effect leads to a reduction in the structural
acceleration and the corresponding damaging effects [16, 17]. Due to the sloping
nature of the ground, infills are not provided on the ground level storey of step back
buildings and step back setback buildings. Hence, it reduces the stiffness of the
bottom storey of the buildings and it acts as soft storey.
For an accurate prediction of the building’s behaviour, the software programmes
take into account the various geometries and load circumstances. Dynamic anal-
ysis is an important factor of structural engineering design since it determines how
a structure responds to dynamic loads like earthquake, wind and blast. ETABS is a
finite-element-based software that can conduct dynamic analysis of a structure, which
includes modal analysis, response spectrum analysis and time history analysis [8].
ETABS performs response spectrum analysis by applying the standard response spec-
trum curves for various seismic zones specified in the building code [11]. Response
spectrum analysis helps to understand the structural response to dynamic loads and
provides information on the maximum response of the building for various periods
of vibration [12]. Results can be obtained in terms of displacement and storey drift.
1.1 Building Configurations
Step back and step back and setback building configurations are considered for this
study. Step back building configuration is given to buildings resting on sloped ground
as shown in Fig. 1. Stepping back of building towards hill slope results in unequal
column heights which causes severe stiffness irregularity. Step back setback building
configuration is the combination of step back and setback building. This design
approach provides visual interest to the building and often allows for improved views
from the upper floors. Additionally, the setbacks can provide shading and reduce the
overall volume of the building, making it more visually pleasing and less imposing
on the surrounding environment. Figure 2 shows the elevation of step back setback
building.
@seismicisolation
@seismicisolation
Seismic Study on Step Back Buildings and Step Back Setback Buildings … 379
Fig. 1 Step back building

configuration
Fig. 2 Step back setback

building configuration
2 Methodology
The present study deals with the comparison of step back buildings and step back
setback buildings with and without bracing in the soft storey. The building model
considered is 30 m high rise reinforced concrete vertical irregular building with
G + 9 storeys. Different types of bracings used were X bracing, chevron bracing,
eccentric forward bracing and eccentric backward bracing. The beams and columns
were modelled as frame elements. The slabs were modelled as thin shell element. A
@seismicisolation
@seismicisolation
live load of 2.5 kN/m2 is considered on the typical floor and 1 kN/m2 at the roof. The
analysis of the RCC structure was carried out using the guidelines of IS 456:2000,
and dynamic analysis called response spectrum analysis was done in ETABS 2016
software following the Guidelines of IS 1893:2016 [18]; the results were compared
between both type of structures, and its displacement and storey drift were computed.
Figure 3 shows the plan of structure, Fig. 4 shows section and elevation of step back
building (UBF), Fig. 5 shows step back building with X bracing (XBF) and chevron
bracing (CBF), Fig. 6 shows step back building with eccentric backward bracing
(EBBF) and bracing (CBF), Fig. 7 shows section and elevation of step back setback
building (SS UBF), Fig. 8 shows SS UBF with X bracing and chevron bracing, and
Fig. 9 shows SS UBF with eccentric backward and eccentric forward bracing.
Fig. 3 Plan of building
Fig. 4 Section and elevation of step back building (UBF)
@seismicisolation
@seismicisolation
Fig. 5 Step back building with X bracing (XBF) and chevron bracing (CBF)
Fig. 6 Step back building with eccentric backward bracing (EBBF) and bracing (CBF)
Fig. 7 Section and elevation of step back setback building (SS UBF)
@seismicisolation
@seismicisolation
Fig. 8 SS UBF with X bracing and chevron bracing
Fig. 9 SS UBF with eccentric backward and eccentric forward bracing
2.1 Properties of Structure
The details of step back buildings and step back setback buildings are given in Table 1.
3 Results
Response spectrum analysis was conducted on different building models, and results
were obtained in terms of storey displacement and storey drift. Different displace-
ment values and storey drift ratios are expressed in graphical form. The displacements
of step back buildings in X direction are shown in Fig. 10, and displacement in Y
@seismicisolation
@seismicisolation
Table 1 Properties of
Particulars Structures
structure
Occupancy type Commercial
Plan dimension 28 m × 12 m
Storey height 3m
Thickness of slab 130 mm
Response reduction factor (R) 5
Importance factor (I) 1
Seismic zone (Z) V
Columns 300 mm × 600 mm
Beams 270 mm × 400 mm
Grade of concrete M25
Bracing ISMB500
direction is shown in Fig. 11. Unbraced frame (UBF) has the highest displacement
in X and Y directions followed by eccentric backward braced frame (EBBF) and
eccentric forward braced frame (EFBF). X braced frame and chevron braced frame
has the least displacement in X and Y directions. Effect of different types of brac-
ings varies, and it prevents the structure from twisting or bending, which can cause
displacement or even collapse. Reduction of displacement is essential in buildings
as it helps prevent structural damage during earthquakes and increases the building’s
seismic resilience.
Storey drift of step back buildings and step back setback buildings in X and
Y directions are shown in Fig. 12. UBF have the highest storey drift ratio when
compared to XBF and CBF.
From the analysis of displacement values, the maximum displacement of
240.82 mm is found in UBF in Y direction. The maximum value of displacement in
Y direction is 33 mm. The minimum value of displacement is obtained for XBF and
CBF which is 22 mm and 19.8 mm, respectively. The displacement of EFBF and
EBBF are reduced to 27.7 mm 37.8 mm in Y direction.
Fig. 10 Displacement of 40
step back building in X
STORY HEIGHT M
direction
20
0
0 100 200
DISPLACEMENT MM
UBF EFBF EBBF XBF CBF
@seismicisolation
@seismicisolation
Fig. 11 Displacement of UBF XBF CBF EFBF EBBF

step back building in Y 40
STORY HEIGHT M
direction
20
0
0 100 200 300
DISPLACEMENT MM
Story Drift In x direction Story drift in Y direction

UBF CBF XBF CBF UBF XBF
EFBF EBBF EFBF EBBF
40 40
STORY HEIGHT IN M
STORY HEIGHT IN M
30 30
20 20
10 10
0 0
0 0.05 0.1 0 0.05 0.1
STORY DRIFT IN MM STORY DRIFT IN MM
Fig. 12 Storey drift of step back building in X direction and storey drift of step back building in Y
direction
From the analysis of drift values, UBF has the maximum inter-storey drift ratio
of 0.0739 in Y direction. The minimum value of drift is obtained for XBF and CBF
in Y direction which was 0.0021. The highest percentage reduction in the drift value
is 97% for XBF and CBF in Y direction. In X direction, the highest value of inter-
storey drift ratio is 1.015. The storey drift in UBF exceed the requirements of the
IS code, whereas bracings bring the values down to acceptable levels. Therefore,
bracing plays an important role in step back constructions.
Figure 13 shows the displacement of step back setback buildings with different
types of bracings in both X and Y directions. Step back setback unbraced frame
(SS UBF) has the highest displacement followed by step back setback eccentric
forward braced frame (SS EFBF) and step back setback eccentric backward braced
frame (SS EBBF). Step back setback buildings with X braced frame (SS XBF) and
chevron braced frame shows the least displacement (SS CBF). Storey drift of step
back setback buildings in X and Y directions are shown in Fig. 14. SS UBF has the
highest storey displacement, and SS XBF and SS CBF have the least displacement.
The displacement of step back setback buildings without bracing in the soft storey
is reduced as compared to step back buildings without bracing. The minimum value
of displacement and storey drift is obtained for SS XBF and SS CBF. The highest
@seismicisolation
@seismicisolation
Displacement in x direction Displacement in y direction

40
40
Storey height m
Storey height m
30
20 20
10
0 0
0 100 200 0 100 200
Displacement mm Displacement mm
SS UBF SS XBF SS CBF SS UBF SS XBF SS CBF
SS EFBF SS EFBF SS EFBF SS BFBF
Fig. 13 Displacement of step back setback building in X direction and displacement of step back
setback building in Y direction
Storey drift in X direction Storey drift in Y direction

SS UBF SS XBF SS CBF SS UBF SS XBF
SS EFBF SS EBBF SS CBF SS EFBF
SS EBBF
Story height (m)
40
Storey height m
40
30
20 20
10
0
0.00E+00 4.00E-02 0
0 0.02 0.04 0.06
Storey drift ratio Story drift ratio
Fig. 14 Storey drift of step back setback building in X direction and storey drift of step back setback
building in Y direction
percentage of reduction in displacement is 95 and 93%. Also, highest percentage of

reduction in storey drift is 97 and 96% for SS XBF and SS CBF models.
The highest value of storey drift is seen in step back building and step back
setback building without any bracing and the value exceeded IS code specifications.
The storey drift in SS XBF and SS CBF is reduced by 97 and 96% when compared
to SS UBF. It is evident that utilizing braces has a positive impact on buildings.
4 Conclusion
From the analysis of parameters like storey displacement, inter-storey drift ratio and
torsional irregularity, the following conclusions are made:
@seismicisolation
@seismicisolation
The maximum displacement of step back building is 240.82 mm, and it has reduced
by 91% when chevron bracings were provided. Also, the inter-storey drift ratio has
reduced by 96% when chevron bracings were provided at external bays. The displace-
ment of SS UBF is 135.9 mm and has reduced by 95% when X bracings were provided
at both internal and external bays. Step back setback building configurations are more
effective than step back building configuration. The displacement of step back setback
buildings without bracing in the soft storey is reduced to 43% when compared to step
back buildings without bracing. Seismic performance of buildings can be improved
by providing bracings in the soft storey of step back setback buildings.
References
1. Maheri MR, Kousari R, Razazan M (2003) Pushover tests on steel X-braced and knee-braced
RC frames. Eng Struct 25(13):1697–1705
2. Mohammad Z, Baqi A, Arif M (2017) Seismic response of RC framed buildings resting on hill
slopes. Procedia Eng 173:1792–1799
3. Khanal B, Chaulagain H (2020, October) Seismic elastic performance of L-shaped building
frames through plan irregularities. Structures 27:22–36. Elsevier
4. Khadiranaikar RB, Masali A (2015) Seismic performance of buildings resting on sloping
ground. In: Advances in structural engineering. Springer, New Delhi, pp 803–813
5. Nassani DE, Hussein AK, Mohammed AH (2017, August) Comparative response assessment
of steel frames with different bracing systems under seismic effect. Structures 11:229–242.
Elsevier
6. Borsaikia AC, Dutta A, Deb SK (2021) Evaluation of participation of masonry infill walls
in the linear and nonlinear behaviour of RC buildings with open ground storey. J Build Eng
44:103263
7. Formisano A, Massimilla A, Di Lorenzo G, Landolfo R (2020) Seismic retrofit of gravity
load designed RC buildings using external steel concentric bracing systems. Eng Fail Anal
111:104485
8. Khadri SN, Chaitanya BK, Krishna CBR, Nitesh KS (2021) Analysis of structure on slopes
infill shear wall at different locations. Mater Today Proc
9. Pradeep S, Selvan SS, Satyanarayanan KS (2020) Experimental study on the behaviour of RCC
in-filled frame resting on sloped ground level. Mater Today Proc
10. Rahimi A, Maheri MR (2020) The effects of steel X-brace retrofitting of RC frames on the
seismic performance of frames and their elements. Eng Struct 206:110149
11. Al Agha W, Umamaheswari N (2021) Analytical study of irregular reinforced concrete building
with shear wall and dual Framed-Shear wall system by using Equivalent Static and Response
Spectrum Method. Mater Today Proc 43:2232–2241
12. Jammula D, Konda R (2021) Seismic analysis of G+9 framed structures made of reinforced
concrete resting on hill slopes. Mater Today Proc
13. Mashhadiali N, Saadati S, Mohajerani SAM, Ebadi P (2021) Hybrid braced frame with
buckling-restrained and strong braces to mitigate soft story. J Constr Steel Res 181:106610
14. Sanches R, Tao J, Fathieh A, Mercan O (2021) Investigation of the seismic performance of
braced low-, mid-and high-rise modular steel building prototypes. Eng Struct 234:111986
15. Zaidi SA, Naqvi T, Ibrahim SM (2021) Study on the effects of seismic soil-structure interaction
of concrete buildings resting on hill slopes. Mater Today Proc 43:2250–2254
16. Satyanarayanan KS, Varshini RA, Pradeep S (2019) Feasibility studies on proactive disaster
prevention in frames of sloped regions. Mater Today Proc 14:315–322
@seismicisolation
@seismicisolation
17. Baniahmadi M, Vafaei M, Alih SC (2022) Cyclic response of reinforced concrete frames
partially infilled with relatively weak masonry wall. J Build Eng 46:103722
18. IS 1893-2016 (Part 1) Criteria for earthquake resistant design of structure, general provisions
and buildings. Bureau of Indian Standards, New Delhi
@seismicisolation
@seismicisolation
Analysis of Multistory Steel Framed
Structure with Different Infills Subjected
to Seismic Loading
Anjan Kumar, A. R. Pradeep, M. Vijayanand, and H. Siddesha
1 Introduction
Earthquake is a disturbance that occurs below the ground surface which causes the
vibration results in the damage or collapse of the structure leading to injury or loss
of life of the occupants [5]. Concrete is known as the universal building material
and has been widely used for construction of buildings. Steel is one of the useful
material for construction of building [1]. The past researches have shown that the
strength of steel is ten times more than that of concrete. It can resist the extreme
forces in buildings [3]. The time of construction of steel building is less compared
to RCC structure, making steel structure a good alternative for RCC structure [3].
Infill walls have been widely used in most of the buildings. They are used to cover
the gaps between the columns [4]. The past researches have shown that buildings
with infills perform well during an event of earthquake compared to building with
no infills (bare framed structures) [16].
A. Kumar (B)
PG Student Department of Civil Engineering, Sri Siddhartha Institute of Technology, SSAHE,
Tumakuru 572105, India
e-mail: anjankumaranjan10031999@gmail.com
A. R. Pradeep
Assistant professor Department of Civil Engineering, Sri Siddhartha Institute of Technology,
SSAHE, Tumakuru 572105, India
M. Vijayanand
Assistant professor School of Architecture, MSRIT, Bengaluru 560054, India
H. Siddesha
Associate professor Department of Civil Engineering, Siddaganga Institute of Technology,
Tumakuru 572105, India
@seismicisolation
@seismicisolation
390 A. Kumar et al.
1.1 Infills
Infills are non-load bearing walls that covers the perimeter of the building built with
three-dimensional framework structure. Beams can carry the weight of the infills [3].
The common function of infill wall is to bear its own weight and that of the cladding
[10]. In this project, concrete block masonry infill and burnt brick masonry infill
have been used.
Solid Concrete Block Masonry Infill. Concrete block masonry also known as
concrete masonry unit has been commonly used in the construction field. These
blocks are composed of cement and aggregates. The cement–aggregate ratio in these
blocks is 1:6. A sum of 60% fine aggregates and 40% coarse aggregates will be
used. The common size of these blocks is (39 cm × 19 cm × (30 cm or 20 cm
or 10 cm)). There are two types of concrete blocks; they are solid concrete blocks
and hollow concrete blocks. Solid concrete blocks are widely used; they are heavy
in weight and are manufactured using dense aggregates. For large-scale masonry
works, solid concrete blocks are preferred. Compared to bricks, they are available
in large sizes. Hence, the construction time of solid concrete block masonry is less
than brick masonry [14]. The properties of the solid concrete block masonry infill,
the density and Young’s modulus values, have been given in Table 1.
Burnt Brick Masonry Infill. Burnt clay bricks are one of the oldest and commonly
used building material which are generally used in masonry walls. There are mainly
four types of burnt clay bricks; they are first-class bricks, second-class bricks, third-
class bricks, and fourth-class bricks in which first-class bricks are the most superior
quality of brick while the fourth-class brick is the worst one. These bricks are made
up of sand and clay. These bricks are rectangular in shape and reddish in color. As
per Indian standard, the size of these bricks is (19 cm × 9 cm × 9 cm) [11]. The
properties of the burnt brick masonry infill, the density and Young’s modulus values,
have been given in Table 2.
Table 1 Properties of solid

S. No Density (kN/m3 ) Young’s modulus (Mpa)
concrete block masonry
1 17.65 3600
Table 2 Properties of burnt

S. No Density (kN/m3 ) Young’s modulus (Mpa)
brick masonry infill
1 18.85 15,800
@seismicisolation
@seismicisolation
Analysis of Multistory Steel Framed Structure with Different Infills … 391
1.2 Seismic Analysis Methods
Figure 1 shows the methods of seismic analysis. In the current project, an attempt
is made by the analysis of equivalent static analysis which is a linear static analysis
method and response spectrum analysis which is a linear dynamic analysis method
for all steel framed model cases.
Equivalent Static Analysis. The equivalent static analysis is a linear static analysis
method in which the dynamic load which is resulting from the vibration caused due
to earthquake at the base of the structure which is distributed vertically to create
a load combination which will envelope the maximum shear at any level. These
distributed forces are used to determine the story displacements and internal forces
such as bending moment and shear force at various structural components. The results
obtained from the analysis will be used for design of global and local structure [7].
Response Spectrum Analysis. The response spectrum is the plot of peak or stead
state response (velocity, acceleration, or displacement) of series of oscillators of
varying natural frequencies which are focused into motion by same base vibration.
The response spectrum analysis is a linear dynamic analysis method which is used
to determine structural response to short, nondeterministic, and transient dynamic
events. This method is used to evaluate dynamic effect of ground motion. In this
method, base shear will be calculated using time period and mass participation of
the structure. In practice, the base shear for static seismic loads and base shear for
dynamic seismic loads will be calculated [7].
Fig. 1 Seismic analysis methods
@seismicisolation
@seismicisolation
392 A. Kumar et al.
2 Structural Modeling and Analysis
In the current project, a (G + 9) story steel building having total height 40 m and
height of each story 4 m with a plan (Fig. 2) of (16 × 16) m is considered. Three
models which are steel bare frame model as shown in Fig. 3, steel frame model
with solid concrete block infill as shown in Fig. 4, and steel frame model with burnt
brick masonry infill as shown in Fig. 5 are taken for analysis. The models have been
analyzed as per the Indian standard code.
Fig. 2 Plan
Fig. 3 Steel bare framed

model (SBM)
@seismicisolation
@seismicisolation
Fig. 4 Steel framed model

with solid concrete block
masonry infill (SSCM)
Fig. 5 Steel framed model

with burnt brick masonry
infill (SBBM)
3 Section Check
The design check is done after the analysis in E-tabs software to find out whether
the selected steel sections for beams and columns are safe under the gravity loading
and seismic loading conditions. From Fig. 6, it can be seen that the selected steel
sections for the beam and column are safe in earthquake zone-V loading conditions.
@seismicisolation
@seismicisolation
394 A. Kumar et al.
Fig. 6 Design check
4 Material Properties
The analysis is carried out in E-tabs software which is a finite element analysis
software. M25 grade of concrete is used in slab, and Fe345 grade of steel is used for
column and beam steel sections. The parameters considered for the study are.
1. RC slab of thickness 150 mm is used in the steel framed models.
2. ISWB 250 steel I-section is used for steel beam sections.
3. ISHB 450-2 steel I-section with 12 mm thick and 320 mm width flats at top and
bottom is used for steel column sections.
4. The 230 mm thick burnt brick masonry infill and solid concrete block masonry
infill are used.
5. The dead load will be calculated automatically by the E-tabs software.
6. The live load of 4 kN/m2 is applied for the slabs which is as per the code IS 875
(part 2)-1987.
7. The earthquake load is taken as per the code IS 1893 (part 1): 2016.
8. The steel framed models are analyzed for type II (medium Type) soil condition.
9. The importance factor = 1 is considered for the study.
10. The response reduction = 5 is considered for the study.
11. The steel framed models are analyzed for earthquake zone V loading conditions.
12. Fixed base condition is used for the analysis of the steel framed models in the
E-tabs software.
13. Welding connection is used for the connection of the steel frames.
@seismicisolation
@seismicisolation
5 Analysis and Results
5.1 Base Shear
The maximum expected lateral force which occurs at the base of the structure due
to earthquake ground motion is known as base shear. The structure is analyzed with
gravity loading and static earthquake loading method. The manual calculation of base
shear of all three models has been done as per IS 1893 (part 1): 2016 clause No. 7.6.1
by using Eq. (1). The base shear obtained from E-tabs software for static earthquake
loading case and response spectrum loading case along X and Y directions have been
given in Table 3.
VB = Ah W (1)
where
Ah Design horizontal acceleration coefficient value.
W Seismic weight of the building.
1. Steel bare frame model (SBM)
VB = 0.036 × 16787 = 604.332 kN.
2. Steel framed model with solid concrete block masonry infill (SSCM)
VB = 0.054 × 26610.2 = 1436.95 kN.
3. Steel framed model with burnt brick masonry infill (SBBM)
VB = 0.054 × 27278.2 = 1473 kN
From the above-obtained values of base shear obtained from manual calculation and
from E-tabs indicated in Table 3, there is a small difference between these values;
for steel bare frame model (SBM), there is 0.16% difference; for steel framed model
with solid concrete block masonry (SSCM), there is 0.06% difference; and for steel
framed model with burnt brick masonry infill (SBBM), there is 0.067% difference
which are in acceptable limit. From Fig. 7, it can be observed that the steel framed
Table 3 Base shear along X and Y direction in E-tabs

Structure type Ex. (kN) Ey. (kN) RSx. (kN) RSy. (kN)
SBM 604.98 604.98 603.69 605.07
SSCM 1445.89 1445.89 1445.8 1445.15
SBBM 1482.4 1482.4 1481.7 1481.7
@seismicisolation
@seismicisolation
396 A. Kumar et al.
Fig. 7 Base shear 2000
Base shear (kN)

comparison
1000
0
Ex Ey Rsx Rsy
SBM SSCM SBBM
model with burnt brick masonry infill (SBBM) has the maximum value of base shear
and the steel bare framed model (SBM) has the least value of the base shear.
5.2 Story Displacement
The maximum lateral displacement which occurs in a story with respect to base of the
structure is known as story displacement. The maximum story displacement which
occurs at the top story along X and Y directions for all three models obtained from
equivalent static method and response spectrum method has been given in Table 4.
From Fig. 8, it can be seen that the steel framed with burnt brick masonry infill
(SBBM) has the least value of the story displacement and the steel bare framed model
(SBM) has the highest value of story displacement.
Table 4 Story displacement along X and Y directions in E-tabs

Structure type Ex. (mm) Ey. (mm) RSx. (mm) RSy. (mm)
SBM 68.946 117.58 54.26 92.67
SSCM 8.064 8.226 6.94 7.064
SBBM 2.634 2.64 2.29 2.29
Story displacement (mm)
Fig. 8 Story displacement

comparison 150
100
50
0
Ex Ey Rsx Rsy
SBM SSCM SBBM
@seismicisolation
@seismicisolation
5.3 Shear Force Diagram
Shear force is the force which is acting on one part of the body in a specific direction
and on another part of the body in the opposite direction. The diagram which repre-
sents the variation of shear force along the length of the member is known as shear
force diagram. The shear force diagram for all the three steel framed models as seen
in Figs. 9, 12, and 15 for gravity loading (dead + live) combination has been shown
below.
In the steel bare framed model (SBM), the maximum shear force in columns
occurs in the top story (10th story) and the maximum values of positive shear force
(17.8249 kN) and negative shear force (− 17.8249 kN) have been shown in Figs. 10
and 11.
In the steel framed model with solid concrete block masonry infill (SSCM),
the maximum shear force in columns occurs in the top story (10th story) and the
Fig. 9 Shear force diagram

of steel bare framed model
(SBM)
Fig. 10 Maximum +ve

shear force in column for
steel bare framed model
Fig. 11 Maximum −ve

steel bare framed model
@seismicisolation
@seismicisolation
398 A. Kumar et al.

of steel framed model with
solid concrete block masonry
infill (SSCM)
Fig. 13 Maximum +ve shear force in column for steel framed model with solid concrete block
masonry infill
Fig. 14 Maximum −ve shear force in column for steel framed model with solid concrete block
masonry infill
maximum values of positive shear force (24.1421 kN) and negative shear force (−
24.1421 kN) have been shown in Figs. 13 and 14.
In the steel framed model with burnt brick masonry infill (SBBM), the maximum
shear force in columns occurs in the top story (10th story) and the maximum values
of positive shear force (24.1421 kN) and negative shear force (− 24.1421 kN) have
been shown in Figs. 16 and 17.
5.4 Bending Moment Diagram
The reaction created in the structure when a force or moment is applied to the structure
is known as bending moment. The diagram which represents the variation of bending
@seismicisolation
@seismicisolation

of steel framed model with
burnt brick masonry infill
(SBBM)
Fig. 16 Maximum +ve

steel framed model with

steel framed model with
moment along the length of the member is known as bending moment diagram. The
bending moment diagram for all the three steel framed models as seen in Figs. 18,
21, and 24 for gravity loading (dead + live) combination has been shown below.
In the steel bare framed model (SBM), the maximum bending moment in columns
occurs in the top story (10th story) and the maximum values of positive bending
moment (38.8677 kN m) and negative bending moment (− 27.9758 kN m) have
been shown in Figs. 19 and 20.
In the steel framed model with solid concrete block masonry infill (SSCM), the
maximum bending moment in columns occurs in the top story (10th story) and the
maximum values of positive bending moment (52.1415 kN-m) and negative bending
moment (− 38.3915 kN-m) have been shown in Figs. 22 and 23.
In the steel framed model with burnt brick masonry infill (SBBM), the maximum
bending moment in columns occurs in the top story (10th story) and the maximum
values of positive bending moment (52.9272 kN-m) and negative bending moment
(−37.7139 kN-m) have been shown in Figs. 25 and 26.
@seismicisolation
@seismicisolation
400 A. Kumar et al.
Fig. 18 Bending moment

diagram of steel bare framed
model (SBM)
Fig. 19 Maximum +ve

bending moment for steel
bare framed model

bare framed model

diagram of steel framed
model with solid concrete
block masonry infill (SSCM)
@seismicisolation
@seismicisolation
Fig. 22 Maximum +ve

framed model with solid
concrete block masonry infill

framed model with solid
concrete block masonry infill

diagram of steel framed
model with burnt brick block
masonry infill (SBBM)
Fig. 25 Maximum +ve

framed model with burnt
6 Conclusion
The following conclusions are made from the obtained results and are in the
acceptable limits and as per the references which matches the referred journals.
1. From the design check, it can be observed that all the selected steel section
for beams and columns are safe under the gravity loading and seismic loading
condition; hence, the structure is safe during the event of an earthquake.

framed model with burnt
@seismicisolation
@seismicisolation
402 A. Kumar et al.
2. Use of infills in the models resulted in increase in base shear and decrease in
story displacement compared to bare framed model.
3. The base shear for steel framed model with solid concrete block masonry infill
was more than the base shear of steel bare framed model, and the steel framed
model with burnt brick masonry infill has the maximum value of base shear.
4. The story displacement for steel framed model with solid concrete block masonry
infill was less than the story displacement of steel bare framed model, and the
steel framed model with burnt brick masonry infill has the least value of story
displacement for both equivalent static and response spectrum case.
References
1. Pradeep AR, Rangaraj C, Jayaramappa N (23 Nov 2022) Analysis of infilled steel frames
subjected to lateral loading. In: ICAMP-International conference, St. Martin’s Engineering
College Dhulapally, Secunderabad, Telangana State, India, 08th–09th July 2022 published in
Materials today conference proceedings
2. Alex AA, Chaithra S (July 2022) Seismic analysis of steel building with bracings. Int J Res
Appl Sci Eng Technol (IJRASET) 10(VII)
3. Rakesh HT, Pradeep AR, Rangaraj C (Feb 2022) Analysis of multi-storey steel structure with
different infills and steel bracings. Int Res J Eng Technol 9
4. Sanjay HS, Pradeep AR, Kiran T, Jayaramappa N (Jan 2022) Analysis of multistory steel
structure with different infills. Int Res J Eng Technol 9
5. Binu KS, Raman A (June 2022) Analysis of steel framed structure with different structural
steel braces using E-TABS. Int J Eng Res Technol (IJERT)
6. Yeshodhara TS, Pradeep AR, Jayaramappa N (April 2021) Analysis of multistoried infilled
steel structure subjected to lateral loading. Int Res J Eng Technol 8
7. Lavitha V, Gokul PV (2021) Comparision of behaviour of RCC and steel structure using ETABS
software. Int J Eng Res Technol (IJERT). In: ICART-2021 conference proceedings
8. Ashwini LK, Gowda NP, Padma VP, Parneetha M, Ramya B (6 June 2019) Pushover analysis
on reinforced concrete building using E-TABS. Int Res J Eng Technol 6
9. Shah S, Patel R (April 2019) Pushover analysis of steel structure. Int J Eng Res Technol (IJERT)
8(04)
10. Kaley P, Baig MA (July 2018) Pushover analysis of steel framed building. J Civil Eng Environ
Technol 4
11. Joseph A, Joshma M (June 2017) Influence of infills on fundamental time period of RC
structures. Int Res J Eng Technol 4
12. Reddy ChB, Vasungi K (5 May 2017) Capacity spectrum method on RCC and composite frame
structures. Int J Civil Eng Technol (IJCIET) 8(5)
13. Dinar Y, Hossain MdI, Biswas RK, Rana MdM (4 February 2017) Descriptive study of pushover
analysis in RCC structure of Rigid Joint. IOSR J Mech Civil Eng (IOSR-JMCE) 11(1)
14. Radovanovic Z, Grebovic RS, Dimovska S, Serdar N, Vatin N, Murgul V (2015) The mechanical
properties of masonry walls—analysis of the test results. ScienceDirect Elsevier
15. Pradeep AR, Rangaraj C, Jayaramappa N (2022) Analysis of infilled steel frames subjected to
lateral loading. In: Materials Today Conference Proceedings 23 Nov 2022. https://doi.org/10.
1016/j.matpr.2022.11.169
16. Pradeep AR, Rangaraj C (2022) Presented paper on “infilled steel frames subjected to seismic
loads”. Lecture series notes, 14 Oct 2022 Springer, p 123. https://doi.org/10.1007/978-981-19-
3371-4_11
@seismicisolation
@seismicisolation
17. Pradeep AR, Rangaraj C, Jayaramappa N (2022) Behaviour of Steel frame subjected to lateral
load. Indian J Sci Technol 15(28):1381–1388. https://doi.org/10.17485/IJST/v15i38.1266
18. Pradeep AR, Rangaraj C, Jayaramappa N (2019) Analysis of steel structure with different infills
subjected to lateral loading. JETIR May 2019 6(5):130–134. ISSN-2349-5162
@seismicisolation
@seismicisolation
Application of Internet of Things (IoT)
in Seismic Performance Evaluation of 3D
Printed Structure
Bhumika Jay, G. M. Basavanagowda, Archana Kumari,

Aishwarya Chauhan, and Bhavana Prasad
1 Introduction
The modern-day manufacturing can be broadly categorized into two main methods—
additive manufacturing and subtractive manufacturing. Additive manufacturing
builds objects by adding material layer by layer, whereas subtractive manufac-
turing removes material to create parts. An example of additive manufacturing is
3D printing, which involves the construction of a three-dimensional object from a
CAD or digital 3D model. This can be achieved through various processes such as
material deposition, joining, or solidification under computer control.
Research on 3D printing began in the early 1950s, but the additive manufacturing
equipment and materials were developed in the 1980s. Charles Hull [1, 2] introduced
the technology of Additive Manufacturing (AM) in 1984, designing the first 3D
printer and introducing the STL (Stereo Lithography) format. After this, further
research and advancements led to the development of more efficient 3D printers
with advanced features.
3D printing has various applications in different industries such as manufacturing,
medical, industrial, humanitarian, and sociocultural sectors. One of the main advan-
tages of 3D printing in the construction industry is the reduced cost of construction
compared to conventional methods.
The Internet of Things (IoT) refers to a system where internet-connected objects
or things can collect and transfer data wirelessly without the need for human interven-
tion. These things are physical objects with unique identifiers, embedded systems,
and the ability to transfer data through a network [3].
The development of IoT has been facilitated by the convergence of various tech-
nologies, including real-time analytics, machine learning, ubiquitous computing,
commodity sensors, and embedded systems. This has brought together different fields
B. Jay (B) · G. M. Basavanagowda · A. Kumari · A. Chauhan · B. Prasad

Ramaiah Institute of Technology, Bangalore 560054, India
e-mail: bhumikajay23@gmail.com
@seismicisolation
@seismicisolation
406 B. Jay et al.
such as wireless sensor networks, control systems, automation, home and building
automation, and others to enable the Internet of Things.
The characteristics of materials used in Additive construction by extrusion (ACE)
have been studied [4] for formulating a set of robust and sustainable binders suitable
for laboratory-scale 3D printing. The scaled-down model study of a prototype (G +
4) building using similitude relations to know its seismic behaviour under laboratory
conditions for real-time earthquakes has been investigated [5].
Previous studies have suggested the strength design equations for different failure
mechanisms of a 3D printed concrete buildings using Equivalent lateral force method
[6].
The seismic wave simulation studies using 3D printed Los Angeles model have
been carried out to know the effect of sedimentary basins and their edges on the high
frequency earthquake ground motion [7].
Even though there are quite a few studies available on the analysis of 3D printed
building subjected to seismic loadings, use of IoT in measurement of displacement
and acceleration values still needs further research. Hence, the present study is
focused on measuring the response of buildings using the concept of IoT.
2 Methodology
2.1 Similitude Laws
Experimental testing of a prototype structure can be a difficult task due to several

factors such as complexity, high cost, and unavailability of testing platforms. As a
solution, researchers often scale down the prototype to a smaller size that can be
tested on a laboratory shake table with limited payload capacity. Scaling factors are
applied to ensure that the scaled-down model is proportional to the original prototype.
In this study, a scaling factor of 1:30 was adopted as the dimensions obtained through
scaling down the model using 1:10 or 1:20 were too large for the available shake table
of 500 mm × 500 mm dimension available in the laboratory of Ramaiah Institute of
Technology, Bangalore, India.
There are three types of similarities: Geometric, Kinematic, and Dynamic. To
achieve complete similarity between the model and prototype, all three types of
similarity must be maintained. Meymand Similitude relations [8, 9] (see Table 1) are
used in the present study as they have been widely used by many researchers in the
past to study soil-structure interaction through laboratory modelling and shake table
tests.
The geometric and material properties of the prototype structure and the scaled-
down model after using Meymand Similitude Relations are shown in Table 2.
@seismicisolation
@seismicisolation
Application of Internet of Things (IoT) in Seismic Performance … 407
Table 1 Meymand
Parameters Scaling relationship
similitude relations √
Frequency 1/ λ
√
Time λ
Modulus of elasticity Ʌ
√
Shear wave velocity λ
Acceleration 1
Mass density 1
Stiffness λ2
Force λ3
Length Ʌ
Stress λ
Strain 1
EI λ5
Table 2 Geometric and material properties of building frame prototype and scaled-down model
S. No Properties Prototype structure Scaled-down model
1 No. of storeys 5 (G + 4) 5
2 Storey height 3m 100 mm
3 Material used Concrete PLA (Polylactic acid)
4 Modulus of elasticity 3500 MPa 2910 MPa
5 Bay width (In both directions) 4m 135 mm
6 Slab thickness 0.15 m 5 mm
7 Size of column 0.45 m × 0.45 m 20 mm × 5 mm
8 Size of beam 0.45 m × 0.45 m –
2.2 3D Printing Process
The first step of 3D Printing involves creating a digital file or blueprint of the desired
object using a 3D modelling software such as AutoCAD, SolidWorks, or Maya. Once
the design is complete, it must be converted into a 3D printing file format, such as
.STL, .OBJ, or .3MF, with .STL being the most commonly used format due to its
compatibility with most 3D printers.
Next, the design must be sliced into hundreds of horizontal layers using special-
ized software that tells the 3D printer exactly what to do step-by-step. This process
is known as slicing and generates a new file called G-CODE, a numerical code
programming language used in computer-aided manufacturing to control automated
machine tools like 3D printers and CNCs.
The final product is printed using a process called Fused Deposition Modelling
(FDM) or Fused Filament Fabrication (FFF), which involves melting plastic filament
@seismicisolation
@seismicisolation
408 B. Jay et al.
Fig. 1 Printing of model using 3D printer with PLA material
and depositing it layer by layer onto a build platform through a heated extruder,
following the automated G-CODE instructions (see Fig. 1). As each layer of filament
hardens and bonds to the previous layer, the final product takes shape.
2.3 Internet of Things
ADXL335 Accelerometer is a small, low power, complete 3-axis accelerometer with

signal conditioned voltage outputs. It can measure the static acceleration of gravity
in tilt-sensing applications as well as dynamic acceleration resulting from motion,
shock, or vibration. In this project, the accelerometer is being used to measure the
dynamic acceleration of the scaled-down 3D printed model when it is subjected to
2001 Bhuj Earthquake loads. The connections for this measurement are completed
using Arduino UNO Microcontroller, NodeMCU, Jumper Wires, Breadboard, USB
connector, and Laptop with Arduino IDE application.
The programming required for successful functioning of the entire IoT system
is coded in Arduino IDE application. Two programmes are run-the first to measure
and obtain the results of the movement of the model upon being subjected to known
seismic waves, and the second to export these results into the public cloud ThingSpeak
for further use. The data is downloaded from ThingSpeak and this is followed by
data handling and analysis to obtain the desired results.
@seismicisolation
@seismicisolation
Fig. 2 Close view of: a conventional accelerometer, b ADXL335 acceleration
The conventional accelerometers and ADXL335 accelerometers are connected at

different levels of the model (see Fig. 2). A comparative analysis of the performance
of the two accelerometers has been done.
2.4 Shake Table Test
For the Shake Table Test of the scaled-down model, the Bhuj Earthquake motion is
given as the input motion at its base. The conventional accelerometers were placed
at alternate floor levels along with ADXL335 accelerometers.
The specifications of the Shake Table used for testing are shown in Table 3.
The base of the scaled-down mode shown in Fig. 3a is subjected to 2001 Bhuj
Earthquake ground motion.
Table 3 Specifications of
S. No Motion Simple harmonic
shake table
1 Payload 50 kg
2 table size 500 mm × 500 mm
3 Displacement 0–100 mm
4 Frequency 0–10 Hz
5 Frequency Control 0.05 Hz (±5%)
@seismicisolation
@seismicisolation
410 B. Jay et al.
Fig. 3 a Plan, b elevation, c 3-dimensional view, and d deflection shape of the selected structure
2.5 Numerical Analysis of Model Using ETABS
To verify the results of scaled-down model, the studies are compared with an analyt-
ical study that is carried out on the prototype of G + 4 model. ETABS Software is
used to model and analyze the prototype structure.
The plan, elevation, and 3-dimensional view of the prototype structure are
presented in Fig. 3. Figure 3d represents the deflection shape of the structure.
The details of various types of loads considered in the present study are presented
in Table 4 [10–12].
Table 4 Loads and seismic

S. No Content Description
parameters considered for
analysis 1 Floor finish 1 kN/m2
2 Live load 3 kN/m2
3 Seismic zone II
4 Zone factor 0.1
5 Importance factor 1
@seismicisolation
@seismicisolation
3.1 Experimental Studies
For the prototype building, fixed base analysis was carried out by subjecting the
model to 2001 Bhuj Earthquake ground motion. The accelerometers were placed
at different floor levels to measure the respective acceleration values. The input
earthquake motion is represented in Fig. 4.
Experimental Results Obtained from Accelerometers
The values of maximum storey acceleration at different floor levels are obtained by
shake table analysis using conventional and ADXL335 accelerometers (see Table 5).
Figure 5 represents the acceleration values measured through conventional
accelerometers and ADXL335 accelerometers. It is clear that acceleration values
increased with an increase in the number of storeys. Also, the comparison of conven-
tional accelerometer results with ADXL335 accelerometer results indicates the
capability of the latter to accurately predict the response.
Further, it was observed that there is amplification of acceleration values at the
top floor. The acceleration value at the base is 0.85 g and that in the top storey is
1.16 g. The acceleration values measured in the scaled-down model are considered
to be the same in the prototype buildings [13]. Hence the scaling factor is not used in
Fig. 4. 2001 Bhuj Earthquake ground motion
Table 5 Experimental results obtained from accelerometers

S. No Floor Conventional accelerometer ADXL335 accelerometer
1 G 0.863 g 0.850 g
2 G+1 Not Measured 0.930 g
3 G+2 0.970 g 0.980 g
4 G+3 Not Measured 1.040 g
5 G+4 1.245 g 1.160 g
@seismicisolation
@seismicisolation
412 B. Jay et al.
5
Conventional Accelerometers
ADXL335 Accelerometer
No. of Stories 4
0
0.50 0.75 1.00 1.25 1.50
2
Acceleration (m/s )
Fig. 5 Comparison of acceleration values measured at different floor levels
order to get the acceleration of the prototype building. However, the scaling factor for
the displacement values measured in the scaled model is multiplied with the scaling
factor to get the displacement of the prototype buildings [13].
3.2 Numerical Analysis from ETABS
Maximum Storey Displacement and Storey Drift

Figure 6a shows that the maximum storey displacement value observed for Bhuj
Earthquake. The maximum value is found to be approx. 11.2 mm which is observed
at the top (G + 4) floor. Figure 6b shows the maximum storey drift value of 95 at the
second floor level.
Results obtained from ETABS Analysis
The values of maximum storey displacement and drift at different floor levels are
obtained by the numerical analysis of the model using ETABS (refer Table 6).
Figure 7 shows the comparison of displacement values between experimental test
results and analytical study. The results are in close agreement with each other.
@seismicisolation
@seismicisolation
Fig. 6 a Maximum storey displacement and b maximum storey drift of structure subjected to Bhuj
Earthquake
Table 6 Results obtained from ETABS analysis

S. No Floor Displacement in mm
Scaled model Prototype (Experimental) Prototype (Numerical)
1 G 0.60 0.60*30 = 18 15
2 G+1 1.47 1.47*30 = 44 40
3 G+2 2.27 2.27*30 = 68 65
4 G+3 3.07 3.07*30 = 92 88
5 G+4 3.73 3.73*30 = 112 108
Fig. 7 Comparison of displacement along the height of structure
@seismicisolation
@seismicisolation
414 B. Jay et al.
4 Conclusion
Based on the experimental test results and numerical analysis carried out on the
prototype model and on the scaled-down model, the following conclusions can be
drawn:
1. The ADXL335 Accelerometers would be effectively used to measure the
acceleration values compared to conventional accelerometers.
2. The values of displacement obtained from the scaled-down model are in close
agreement with the values obtained from numerical studies.
3. The amplification of acceleration values as well as the displacement values is
observed at the top floor in the scaled-down model.
The acceleration values serve as an important parameter while designing the
multistorey buildings. In absence of large-scale shake table, scaled-down model
studies serve as an alternate method to estimate the behaviour of prototype buildings.
References
1. Gokhare VS, Raut DN, Shinde DK (2017) A review paper on 3D-printing aspects and various
processes used in the 3D-printing. Int J Eng Res Technol (IJERT) 6(6)
2. U.S. Patent 4,575,330 (“Apparatus for production of three-dimensional objects by stereolithog-
raphy”)
3. Ali ZH, Ali HA, Badawy MM (2015) Internet of Things (IoT): definitions, challenges and
recent research directions. Int J Comput Appl 128(1). ISSN 0975-8887
4. Bentz DP, Jones SZ, Bentz IR, Peltz MA (2019) Towards the formulation of robust and sustain-
able cementitious binders for 3D additive construction by extrusion, Chap. 15. In: 3D concrete
printing technology. Elsevier
5. Prakash P, Basavangowda GM (2022) Seismic analysis of 3D printed structures. In: Kolhe
ML, Jaju SB, Diagavane PM (eds) Smart technologies for energy, environment and sustainable
development, vol 1. Springer proceedings in energy. Springer, Singapore. https://doi.org/10.
1007/978-981-16-6875-3_1
6. Delavar MA, Chen H, Sideris P (2023) Seismic design methodology for 3D printed concrete
buildings. Cityscape 25(1):177–197
7. Park S, Shin C, Kim Y, Clayton RW (2022) Seismic wave simulation using a 3D printed model
of the Los Angeles Basin. Sci Rep 12(1):1–8
8. Casaburo A, Petrone G, Franco F, De Rosa S (June 2019) A review of similitude methods for
structural engineering. Appl Mech Rev
9. Balawi S, Shahid O, Al Mulla M (2015) Similitude and scaling laws—static and dynamic
behaviour beams and plates. Elsevier
10. IS: 456-2000, Plain and reinforced concrete—Code of practice. Bureau of Indian Standards,
New Delhi, India
11. IS: 875(Part 3): 2015, Design loads for buildings and structures. Bureau of Indian Standards,
New Delhi, India
12. IS: 1893 (Part 1): 2016, Criteria for earthquake resistant design of structures: general provisions
and building. Bureau of Indian Standards, New Delhi, India
13. Maymand P, Reimer M, Seed R (2000) Large scale shaking table tests of seismic soil-pile
interaction in soft clay. In: Proceedings of 12th world conference on earthquake engineering,
New Zealand, vol 5(0915)
@seismicisolation
@seismicisolation
Experimental and Numerical Study
on Hysteretic Behavior of Laminated
Rubber Bearing Under Quasi-Static
Loading and Its Performance
on Secondary System
Bharat Chalise
1 Introduction
Base isolation is an innovative and cutting-edge passive control mechanism of seismic

control which shifts the structure’s period away from the predominant period of
ground motion. Datta [2] mentioned the laminated rubber bearing (LRB), and lead
core rubber bearings are used extensively in practice. Of the friction type, elastic
sliding bearings, friction pendulum systems (FPS), resilient friction systems (R-
FBI), and pure friction (P-F) systems are much popular. Zhou et al. [14] studied high
damping rubber bearings to simplify its nonlinear behavior using equivalent lineariza-
tion principle. Ma et al., [9] carried out experiment for the failure mode of small,
laminated rubber bearings for seismic isolation of nuclear components, secondary
structures. Result shows the bearings as good option among seismic isolation devices
for the seismic enhancement of the selective safety-related secondary structures of
operating nuclear plants. Several other studies have been carried out in the past to
analyze the effectiveness of the laminated rubber bearing to isolate primary as well
as secondary structures [1, 3–6, 13].
The performance of the laminated bearing is influenced by various factors such as
geometric properties, types of material, and loading history [8]. Matsagar and Jangid
[7] concluded the response of base-isolated structure is significantly influenced by
the shape of hysteresis loop of isolator. In this study, two dynamic properties viz.
stiffness and damping property of the bearings were identified experimentally and
applied in the secondary structure to investigate the linear response. Similarly, using
experimental results, nonlinear properties of the bearing were found numerically.
B. Chalise (B)
Indian Institute of Technology Delhi, New Delhi, India
e-mail: naturead@gmail.com
@seismicisolation
@seismicisolation
416 B. Chalise
Bearing with linear and nonlinear properties was used to isolate the secondary struc-
ture. Response of the secondary system under linear and nonlinear scenario was
analyzed.
2 Experimental Setup
Typical test setup was fabricated and assembled in the Heavy Structures Laboratory
(HSL) at Indian Institute of Technology Delhi (IITD) as shown in Fig. 2. A 25 kN
capacity hydraulic jack was used for the loading purpose. LVDT with stroke capacity
of + 5 mm was used to measure the displacement. Separate setup was fabricated for
shear test and vertical stiffness test of the bearing.
Quasi-static manual loading through hydraulic pump connected with jack was
used. The hydraulic jack was mounted in such a way that can push the loading plate
of the system when applying the load on it. A load cell of capacity 10 ton (100 kN)
with rating 3.06 mV/V was connected in between hydraulic pump and the load plate.
LVDT with stroke length + 5 mm was connected on the opposite side of loading
system, on top plate of bearing. There was a mechanism of varying compressive axial
force on the top of the load plate. Strain smart 7000 software was used to calibrate
the instruments, process, and acquisition of the data.
2.1 Test Specimen
The test was conducted using a square-shaped laminated rubber bearing with the
dimensions shown in Fig. 1. The specimen (50 mm × 50 mm × 50 mm) consists
of steel and rubber plates bonded together in alternate layers. The 12 mm thick steel
plates were used as top and bottom plates. The bearing consisted of five numbers of
6 mm rubber layer and four numbers of 1.5 mm steel shims with dimension 40 mm
× 40 mm × 1.5 mm. The rubber used in the bearing is neoprene with IRHD value
of 47.
2.2 Loading History
The specimen was tested under quasi-static horizontal and vertical loading. Labo-
ratory setup for shear and vertical stiffness is shown in Fig. 2. Both the tests were
conducted with displacement control method. Sets of six cycles were repeated. The
quasi-static cyclic loading was applied to get the displacement of + 3 mm in the
specimen. These specimens were applied at certain values of a vertical preloading or
an axial compressive loading of 0 kg, 20 kg, 40 kg, and 60 kg, respectively. However,
preloading was not applied during the vertical stiffness test.
@seismicisolation
@seismicisolation
Experimental and Numerical Study on Hysteretic Behavior … 417
Fig. 1 a Plan and b section of the sample laminated rubber bearing for investigation
Fig. 2 Typical laboratory test setup for (a,c) shear test, (b,d) vertical stiffness test
@seismicisolation
@seismicisolation
418 B. Chalise
200
100 Loading Loading

Force (N)
0
Unloading Unloading
-100
0 kg Compressive axial load 20 kg Compressive axial load
-200
200
100 Loading Loading

Force (N)
0
Unloading Unloading
-100
40 kg Compressive axial load 60 kg Compressive axial load

-200
-4 -3 -2 -1 0 1 2 3 4 -4 -3 -2 -1 0 1 2 3 4
Displacement (mm) Displacement (mm)
Fig. 3 Force–displacement curve (hysteresis loop) of type 1A bearing at different axial compressive
loading conditions
2.3 Shear Test and Vertical Stiffness Test
The results of the shear test of the elastomeric bearing are presented in Fig. 3. In the
plot, the vertical axis represents the imposed force on the specimen, and horizontal
axis is the displacement induced from the applied load.
The conclusion of the vertical test is presented in Fig. 4. In the plot, the vertical
axis represents the imposed vertical load, and the horizontal axis represents a relative
displacement between the base plates of a bearing. The relative displacement was
measured from the displacement transducer mounted on the center of the bearing.
The maximum vertical relative displacement varied from 1 to 3 mm. Five cycles of
loading and unloading were done during the test. The load was completely released
up to level zero once completed the setup.
2.4 Characterization of Properties of the Laminated Rubber

Bearing
The basic components of the laminated rubber bearings are steel shims and rubber
built into alternate layers (Fig. 1b). Generally, elastomeric bearing system exhibits
high damping capacity, horizontal flexibility, and high vertical stiffness. The rubber
in the isolator acts as a spring. It is very soft laterally but very stiff vertically. The high
vertical stiffness is achieved by having thin layers of rubber reinforced by steel shims.
@seismicisolation
@seismicisolation
30
25
Vertical load (kN)
20
15
10
0
0 1 2 3
Vertical displacement (mm)
Fig. 4 Force–displacement curve (hysteresis loop) for vertical test
These two characteristics allow the isolator to move laterally with relatively low
stiffness yet carry significant axial load due to their high vertical stiffness. The lead
core provides damping by deforming plastically when the isolator moves laterally
in an earthquake. For low damping rubber bearings, a lateral force–deformation
relationship can be modeled as linear (Fig. 5b). A mathematical model is shown in
Fig. 5(a) representing the linear behavior. However, large displacements can exhibit
hysteretic behavior. The restoring force in the bearing is given by.
Fb = cb ẋb + kb xb (1)
Where cb and kb are the damping and stiffness of the laminated rubber bearing,
respectively. xb , ẋb , and ẍ g are the bearing displacement, velocity, and ground motion,
respectively.
For low damping bearings. properties of the isolators can be calculated by the
equivalent linearization method [10]. The method is also known as the secant method.
Fig. 5 a Mathematical model of laminated rubber bearing, b idealized force–deformation behavior

of a low damping laminated rubber bearing
@seismicisolation
@seismicisolation
420 B. Chalise
Effective stiffness keff is given by

( )
F+ − F−
keff = (2)
(Δ+ − Δ− )
where hysteresis damping ζb defined as
ΔUh
ζ = (3)
2π × Δ20 × keff
where F + , F − are values of shear force at maximum and minimum displacement

and Δ+ , Δ− are the maximum and minimum shear displacement. ΔUh is the area
of the hysteretic loop, and Δo is the hysteretic amplitude or maximum deflection of
the curve. Once damping ratio calculated effective damping cb can be calculated by
cb = 2mζ wn (4)
where m is the total sum of the mass of the structure (m s ) and isolators (m b ) which
is m = m s + m b and wn is the angular frequency calculate as
/
keff
ωn = (5)
m
Wen [12] mentioned in his research that laminated rubber bearings hysteretic
curve can be assumed as a linear one. However, the exact shape of the hysteretic
curve is not always linear as shown in Fig. 3. The model for nonlinear hysteretic
loop is described by considering single degree of freedom oscillator with mass m
and viscous damping coefficient c subjected to base acceleration a equation of motion
is given by
m ẍ + c ẋ + q(x, t) = −ma (6)
where restoring force is given by
q = αkx + (1 − α)kz (7)
The nonlinear differential equation governing z is given by
ż = A ẍ − β|ẋ||z|n−1 z − γ ẋ|z|n (8)
Dividing Eq. (8) by ẋ yield is given by
dz |ẋ|
= A − β |z|n−1 z − γ |z|n (9)
dx ẋ
@seismicisolation
@seismicisolation
where α, β and A are the constants governing scale and general shape of the hysteretic
loop while the smoothness of the force displacement curve is governed by n [11]. Wen
[12] studied that the hysteretic relationship exists between x and z, and by adjusting
these constant, various shape and size of hysteretic loop can be produced. Equations 2
and 3 were used to determine the equivalent linear parameter. Nonlinear parameters
were obtained by solving Eqs. 7 and 8 numerically. The backward difference method
was used to solve the differential Eq. 9. The hysteresis curve obtained from the analyt-
ical method was tuned with experimental curve to identify the nonlinear parameter
of the bearing.
Figure 6 shows the force–displacement curves used for the linear and nonlinear
parameters, respectively. By taking advantage of hysteretic loop which is linear
in nature with narrow bandwidth, vertical stiffness was calculated by the linear
method. The least square method was adopted to calculate the vertical stiffness of
the specimen.
All the linear and nonlinear properties are summarized in Table 1. Effective stiff-
ness (ΔU h), yield strength ( f y ), damping ratio (ζ ), post yield stiffness ratio (α),
vertical stiffness (kv ) and yielding exponent (n) are the major parameter obtained
through the laboratory test.
200 200
150 + 150
(F ,
100 100
Loading
50 50
Force (N)
Force (N)
0 0
keff
-50 -50
Unloading -100
-100
Experimental
-150 - -150 Analytical
(F ,
-200 -200
-4 -3 -2 -1 0 1 2 3 4 -4 -3 -2 -1 0 1 2 3 4
Displacement (mm) Displacement (mm)
(a) (b)
Fig. 6 Determination of a linear parameter and b nonlinear parameter
Table 1 Linear and nonlinear parameter of test specimen

Linear parameters Nonlinear parameters (Dimensionless)
keff (N/ f y (N) Δo (mm) ζ (%) kv (N/ n α β A γ
mm) mm)
52 80 3 6.0 7990 1 0.6 0.6 1 1
@seismicisolation
@seismicisolation
422 B. Chalise
3 Numerical Investigation
The fixed based five-storied primary structure (PS) is modeled with finite element
software. Similarly, single degree of freedom system secondary structure (SS) is
modeled by mounting in five different floor levels. The geometry of primary and
secondary structure modeled for present study is shown in Fig. 7.
The single degree of freedom (SDOF) structure as shown in Fig. 7 was used as a
secondary structure. The joints of the structure were assumed as the moment carrying
fixed joints, and the damping ratio (ζ ) was taken as 2% which is assumed as inherent
damping of steel structures. Material used is Fe 250 of density 7850 kg/m3 . The
modulus of elasticity E is 2 × 1011 N/m2 . The Poisson’s ratio has taken to be 0.3.
Linear and nonlinear analysis was carried out by modeling the system as a rubber
bearing and Wen link, respectively. The effective stiffness (keff ) is used 52 N/mm,
and effective damping was calculated with 6.0% of damping ratio (ζ ). A uniform
mass of 100 kg on top and bottom of the secondary system was provided to make
effective isolation. Stiffness (k) of the SS was 107.17 N/mm, and the mass (m) of
the system was 54.41 kg (top and bottom mass not considered). All other nonlinear
data were taken from the result of the laboratory test as presented on Table 1.
Fig. 7 a Five-story primary structure, b secondary structure, and c mathematical model of five-
storied primary structure with base-isolated secondary structure
@seismicisolation
@seismicisolation
4 Result and Analysis
Newmark’s step-by-step integration method was used to analyze the system. The
2% of inherent damping of the structure was used for the analysis. The earth-
quake motions selected for the study are N00E component of 1989 Loma Prieta
earthquake recorded at Los Gatos Presentation Center, N90S component of 1994
Northridge earthquake recorded at Sylmar Station, and N00S component of 1995
Kobe earthquake recorded at JMA. The peak ground acceleration (PGA) of Loma
Prieta, Northridge, and Kobe earthquake motions are 0.57, 0.60, and 0.86 g, respec-
tively. The dominant frequency of these earthquakes ranges to maximum of less than
2 Hz.
Shake table experiments were carried out by exciting base-isolated structure as
shown in Fig. 7(c) on 3 Hz sinusoidal frequency with two different amplitudes of 4 and
6 mm. Chosen frequency of harmonic excitation is slightly higher than devastating
earthquakes. This gives the opportunity to evaluate bearing performance on higher
frequency excitations. The same setup is used to evaluate the performance of bearings
under three different earthquakes as discussed in the previous section. Remarkably,
acceleration response of the top floor under harmonic excitation of frequency of 3Hz
is amplified and unable to reduce the response. However, the top floor’s response
reveals a response reduction of up to 51%. The result implies that the base isolation
under higher frequency is not effective.
To determine the impact of laminated rubber bearing’s nonlinear properties,
another shake table experiment was conducted. Mechanical properties of bearings
are identified by the experiment as discussed in sect. 3. According to Fig. 7, the
secondary structure was placed on a different floor of the primary structure and its
base was isolated with the rubber bearings having properties as listed in Table 1 and
illustrated in Fig. 2. Setup excited with three different earthquake time histories and
response was recorded. The response quantities of interest are top floor displace-
ment of secondary structure and bearing displacement. Acceleration is the foremost
cause of the shear force induced in the structure while bearing displacements are
important for serviceability and isolation system design. Top floor acceleration of
the structure is important to predict the shear force on the secondary structure. Linear
and nonlinear responses on the isolated structure were compared and studied in this
study. Since this study is not interested in fixed base responses, they are not reported
here. Figure 8 demonstrates the acceleration response of secondary structure when
mounted in first and fifth floor.
Similarly, Fig. 10 displays the peak acceleration response of secondary structure
when mounted on different floor of primary structure. Result shows the marginal
difference in the linear and nonlinear acceleration response. The linear method
is predicting more compared to nonlinear method except in Kobe earthquake
ground motion. Displacement responses are shown in Figs. 9 and 11, where the
linear hysteretic model forecasts increased acceleration on the top floor. Nonlinear
responses outperformed linear responses when it comes to displacement response.
@seismicisolation
@seismicisolation
424 B. Chalise
2 2.05 Linear
2 Linear Non-linear
Non-linear
0
0
-2 2.05 -2 2.01 Loma Prieta, 1989

2.0 Loma Prieta, 1989
Top floor acceleration(g)
Top floor acceleration(g)

2 1.66 2
1.42 1.42 1.66
0
0
-2
Northridge, 1994
Northridge, 1994
-2
2.53 2.39
2 2 2.53
0 0
-2 -2
2.77
Kobe, 1995 Kobe, 1995
-4 -4
5 10 15 20 25 0 5 10 15 20 25
Time (sec) Time (sec)
(a) (b)
Fig. 8 (a) SS top floor acceleration when SS mounted on first floor of primary structure and (b)
SS top floor acceleration when SS mounted on fifth floor of primary structure
13.61 13.61
Linear Linear
10 9.20 10 9.21 Non-linear
Non-linear
0 0
-10 -10 Loma Prieta, 1989

Loma Prieta, 1989
10 10
Bearing displacement (cm)
Bearing displacement (cm)
0 0
-5
6.43 6.43
8.47 8.47 Northridge, 1994
-10 Northridge, 1994 -10
20 15.91 20
15.91
12.51 12.51
10 10
0 0
-10 -10
Kobe, 1995 Kobe, 1995

-20 -20
0 5 10 15 20 25 0 5 10 15 20 25
Time (sec) Time (sec)
(a) (b)
Fig. 9 (a) Bearing displacement when SS mounted on first floor of primary structure and (b)
bearing displacement when SS mounted on fifth floor of primary structure
Shear force induced in bearing at different floor levels shows that there is marginal
difference in the response as shown in Fig. 12. Shear force is the result of the acceler-
ation which is discussed here and shows almost no differences in linear and nonlinear
analysis.
5 Conclusions
Behavior of high damping rubber bearing exhibits complex and highly nonlin-
earity behavior of restoring force. Hysteretic behavior of elastomeric bearing is
important to predict the response of the isolated structure accurately. In this study,
hysteretic behavior of the laminated rubber bearing is determined by cyclic shear and
vertical stiffness test. Cyclic shear tests under varying vertical load were performed
@seismicisolation
@seismicisolation
Linear
Non-linear
4
Top floor peak acceleration (g)
Loma Prieta, 1989 Northridge, 1994 Kobe, 1995

0
0 1 2 3 4 5 6 1 2 3 4 5 0 1 2 3 4 5 6
Floor Level Floor Level Floor Level
Fig. 10 SS top floor acceleration when SS mounted on different floor of primary structure
30
Linear
25 Non-linear
Bearing peak Displacement (cm)
20
15
10
Loma Prieta, 1989 Northridge, 1994 Kobe, 1995

0
0 1 2 3 4 5 6 1 2 3 4 5 0 1 2 3 4 5 6
Fig. 11 Bearing displacement when SS mounted on different floor of primary structure
to assess the influence of varying vertical load on bearing hysteretic behavior.

The nonlinear numerical results show very good agreement with actual bearing
test results. The result of laminated rubber bearing shear test with varied axial
loads demonstrated marginal reduction up to 7% in shear stiffness with increase in
axial load. The bearing is modeled with both equivalent linearization and nonlinear
methods and applied on secondary structure with the properties obtained from the
lab test. The result shows that the response of base-isolated secondary structure is
significantly affected by hysteretic properties of the bearing. Equivalent lineariza-
tion can be done for narrowly damped laminated rubber bearings for conservative
results; however, for actual prediction of the response, size and shape of the hysteretic
@seismicisolation
@seismicisolation
426 B. Chalise
12 8 8
Linear Linear Linear
Non-linear Non-linear Non-linear
10 Loma Prieta, 1989 Northridge, 1994 Kobe, 1995
6 6
8
Shear Force (kN)
6 4 4
2 2
0 0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6 0 1 2 3 4 5 6
Fig. 12 Shear force induced in the bearings when SS placed at different floor levels
loop should be considered. No significant differences in responses indicate that

linearization is an ideal method to characterize laminated rubber bearings.
Response of structure for different earthquake excitations for base-isolated
secondary structures is analyzed. Laminated bearing well intercepted the earthquake
excitations having predominant frequencies less than 2 Hz from its base. For earth-
quake excitations, the response has been reduced up to 51%. However, the input of
a sinusoidal excitation at a frequency higher than 3 Hz was unable to function as
intended, increasing the acceleration response rather than decreasing it. Based on
these findings, conclusion can be made on laminated bearings which are effective
for lower frequency excitations. It is recommended to conduct additional studies on
the characterization of lead core rubber bearings using the same test setup to identify
potential differences in the characteristics and responses of structures.
References
1. Barbat AH, Bozzo LM (1997) Seismic analysis of base-isolated buildings. Arch Comput
Methods Eng 4(2):153–192
2. Datta TK (2010) Seismic analysis of the structures. John Wiley and Sons (Asia) Pte Ltd
3. Kelly JM, Hsiang CT (1985) Seismic response of light internal equipment in base isolation
structures. Earthquake Eng Struct Dyn 13(6):711–732
4. Khechfe H, Mohammad N, Zhikun H, James MK, Ahmadi G (2002) An experimental study
on the seismic response of base-isolated secondary systems. J Pressure Vessel Technol
5. Kalfas KN, Mitoulis SA (2017) Performance of steel-laminated rubber bearings subjected to
combinations of axial loads and shear strains. Procedia Eng 199:2979–2984. ISSN 1877-7058.
https://doi.org/10.1016/j.proeng.2017.09.533
6. Mallikarjun PV, Jagtap P, Kumar P, Matsagar VA (2013) Performance of seismic base-isolated
building for secondary system protection under real earthquakes. In: Conference on structural
mechanics in reactor technology
@seismicisolation
@seismicisolation
7. Matsagar VA, Jangid RS (2004) Influence of isolator characteristics on the response of base-
isolated structures. Eng Struct 26:1735–1749
8. Naeim F, Kelly JM (1999) Design of seismic isolated structures. Wiley, New York
9. Ma SJ, Shin T-M, Ryu J-S, Lee J-H, Koo G-H (2021) Experimental approach for the failure
mode of small laminated rubber bearings for seismic isolation of nuclear components. Appl
Sci 12(1):125
10. Solva CW (2007) Vibrations fundamentals and practice. CRC Press, Florida
11. Wen YK (1980) Equivalent linearization for hysteretic systems under random excitation. J
Appl Mech 47(1):150–154
12. Wen YK (1976) Method for random vibration of hysteretic system. J Eng Mech Div
13. Zhang B, Wang K, Lu G, Qiu W, Yin W (2022) Experimental and seismic response study
of laminated rubber bearings considering different friction interfaces. Buildings 12(10):1526.
https://doi.org/10.3390/buildings12101526
14. Zhou Z, Li Y, Hu X (2022) Analysis method of isolation layer composed of high damping
rubber bearings based on deformation history integral type model. Eng Struct 25. https://doi.
org/10.1016/j.engstruct.2021.113553
@seismicisolation
@seismicisolation
A Study on Self-Compacting Concrete
at High Elevated Temperatures
M. Maniknata, Shaik Subhan Alisha, Durga Vara Prasad Bokka,

Gottumukkala Sravya, and V. Siva Rama Raju
1 Introduction
Super-plasticizers and viscosity modifying chemicals have advanced to the point that
it is now possible to create concrete with improved fluidity and cohesiveness that does
not require external energy to compact. To maintain fire safety in structures made
with self-compacting concrete (SCC), it is essential to comprehend how elevated
temperatures affect SCC’s properties [1–5]. Studies show that SCC’s hot compressive
strength declines with rising temperature, while when subjected to high temperatures,
high-strength SCC displays a higher compressive strength than normal strength SCC
[6]. In comparison with conventional concrete (CC) mixtures of the same strength
class, SCC mixtures demonstrate greater residual compressive strength. The strength
category is the only factor that seems to affect the spalling behavior of CC and
SCC [7]. Additionally, the propensity for explosive spalling was investigated, which
revealed that test conditions and the water-to-cement ratio (w/cm) had an impact on
relative strength losses owing to high temperature exposure, resulting in complicated
behaviors [8]. According to experimental findings, self-compacting high-strength
concretes’ residual mechanical properties are comparable to those of conventional
high-strength concrete, but the risk of spalling is higher [9].
The study investigates the performance of lightweight concrete using LECA mate-
rial as a partial replacement for conventional aggregates at different amounts of
replacement (0%, 5%, 10%, 15%, 20%, and 30%). Water absorption tests and testing
for mechanical properties (strength in compression, flexure, and split tension) were
conducted, revealing 10% LECA material replacement as the optimal percentage
across all aspects [10–13]. Self-compacting concrete (SCC), developed to increase
M. Maniknata (B) · S. S. Alisha · D. V. P. Bokka

Vishnu Institute of Technology, Bhimavaram, Andhra Pradesh, India
e-mail: manikanta.manne206@gmail.com
G. Sravya · V. S. R. Raju
S.R.K.R Engineering College, Bhimavaram, Andhra Pradesh, India
@seismicisolation
@seismicisolation
430 M. Maniknata et al.
Table 1 Physical properties of cement

S. No. Property OPC result Permissible limits
1 Standard consistency 31% 25–35%
2 Initial setting time 45 min ≥ 30 min
3 Final setting time 450 min ≤ 600 min
4 Specific gravity 3.10 –
the durability of concrete structures, has since been extensively researched and widely
adopted in Japan, particularly by large construction companies [14]. In this study,
fly ash (FA) is used as an additive and as a partial replacement for cement (together
with ultra-fine natural steatite powder, or UFNSP). In SCC, FA is used to replace
cement to some extent at replacement levels of 10 and 20% by weight. UFNSP
is added to both replacement levels at varying percentages (0–25% by weight of
powder content at 5% increments). Recycling these materials is essential for envi-
ronmental conservation, as improper disposal of concrete and prefabricated elements
pollutes the environment. In Moscow, millions of cubic meters of concrete and rein-
forced concrete waste are generated, leading to significant loss of valuable materials
like concrete fillers, cement, and steel reinforcement [15]. Self-compacting concrete
(SCC) can be tailored to meet EN 206 specifications for density, strength develop-
ment, ultimate strength, and durability, which serves as the reference standard for
SCC guidelines in Europe. Notably, SCC may display more plastic shrinkage or
creep compared to ordinary concrete mixes due to its higher powder content [16].
2 Material and Methods
2.1 Cement
In India, the cement of higher grades was introduced. After 1987, 53 grade OPC
cement was used in the present investigation, conforming to IS12269:1987. The
physical properties of cement are listed in Table 1.
2.2 Fine Aggregate
The sieve sizes used for the sieve analysis of fine aggregates are in accordance with
IS 2386-Part 1. Fine aggregates comply with IS:383-1970, and details are listed in
Table 2.
@seismicisolation
@seismicisolation
A Study on Self-Compacting Concrete at High Elevated Temperatures 431
Table 2 Physical properties

S. No Property Test result
of fine aggregate
1 Specific gravity 2.64
2 Bulk density (KN/m3 ) 1635
3 Fineness modulus 2.64
4 Zone II
2.3 Coarse Aggregate
The sieve analysis of course aggregate is carried out by using different sizes of IS
sieves; those are 20, 12.5, 10, 4.75 mm, and 2.36. The cumulative percentage passing
is found in each sieve by using these IS sieves, and those values satisfied the grading
limits (IS 383-1970) for 12.5 mm size aggregate. In compliance with IS: 383-1970,
coarse aggregate of pulverized granite stone was used.
2.4 Properties of Coarse Aggregate
The specific gravity and bulk density of coarse aggregate can be calculated to be 2.67
and 1465 KN/m3 , respectively.
2.5 Water
Concrete can usually be mixed with water that is safe for drinking. Concrete may
be impacted by water impurities. Thus, in the current investigation, clean drinking
water that was readily available locally was used.
2.6 Fly Ash
Fly ash is a substantial thermal power plant by-product. There is an effort to use
a considerable proportion of fly ash as a cement addition. From the Vijayawada
Thermal Power Station (VTPS) in Vijayawada, the fly ash used in the current exper-
iment was procured. The grade of the fly ash was F, and its composition details are
listed in Table 3. The physical properties of fly ash are listed in Table 4.
@seismicisolation
@seismicisolation
Table 3 Composition of fly

S. No. Characteristic properties by mass Test results%
ash
1 SiO2 66.80
2 AL2 O3 24.5
3 Fe2 O3 4.00
4 CaO 1.50
5 MgO 0.45
6 Na2 O 0.4
7 K2 O 0.22
Table 4 Physical properties

S. No. Physical properties Value
of fly ash
1 Specific gravity 2.67
2 Fineness cm2 /gm 4069
2.7 Sugarcane Bagasse Ash (SCBA)
More than 110 countries cultivate sugarcane extensively, and the crop produces more
than 1500 million tons worldwide. After all of the economically viable sugar has been
extracted from the sugarcane, 40–45 percent of the fibrous residue is left over. This
residue is utilized as fuel for boilers to generate heat, resulting in approximately 10
million tons of SCBA being produced as waste. The chemical composition of the
residue is predominantly silicon dioxide (SiO2 ) after the burning process. Although
the ash is challenging to decompose and provides minimal nutrients, it is employed
as a fertilizer in sugarcane harvests on farms.
2.8 Super Plasticizer
Auramix 300 plus is a special blend of superplasticizers from the most recent gener-
ation, built on a polycarboxylic ether polymers with extended lateral chains. VMA
is inbuilt to control the segregation and mainly to produce free flow of concrete.
@seismicisolation
@seismicisolation
Table 5 Final mix proportions of SCC

Cement (kg/m3 ) F.A (kg/m3 ) C.A (kg/m3 ) W/C Ratio SP % OPC
480 990 790 0.38 0.6
3 Mix Detail
3.1 Mix Design of Self-Compacting Concrete (as Per

EFNARC Guidelines)
It is more practicable to take the essential components’ volume into account when
designing the mix than to take their mass into account. A thorough revision of the
mix should be taken into consideration if satisfactory performance is not achieved.
3.2 Final Mix Proportions of SCC
In the current study, the mix proportioning is carried out using the IS 10262:2009
code. After conducting laboratory tests, the obtained mixes are adjusted following
the EFNARC recommendations to attain the desired self-compacting concrete mix
proportion. After numerous attempts, the mixtures are obtained. Table 5 displays the
final mix ratios.
3.3 Casting of Concrete Specimens
In the case of conventional self-compacting concrete mixtures, nine cubes (size: 150
× 150 × 150 mm) were cast, and 7, 14, and 28 days later tests were performed.
For SCC10, SCC15, and SCC20, fly ash and SCBA were substituted in the ordinary
Portland cement mixtures. Nine cubes were cast for each mix for testing over 7, 14,
and 28 days. For testing purposes, castings were made for 7, 14, and 28 days for each
proportion of fly ash and SCBA in the self-compacting concrete mix. The 48 cubes
are casted for each temperature.
3.4 Test Methods of Fresh SCC
A concrete mix can only be categorized as self-compacting concrete, according

EFFNARC criteria, provided that the conditions for the three qualities listed below
are met.
@seismicisolation
@seismicisolation
• Passing ability;
• Filling ability;
• Resistance to stability or segregation.
4 Methodology
4.1 Slump Flow Test
Utilizing the slump flow test in the absence of any impediments, the horizontal free
flow of SCC is evaluated. The test is conducted similarly to that of conventional
concrete; however, in this case, instead of measuring the vertical slump distance, the
mean spread of the concrete is calculated.
4.2 L-Box Test
To assess if SCC can pass the test, the L-box test analyzes the vertical height of
concrete and the end of the apparatus’ horizontal component.
4.3 V-Funnel Test
The V-funnel test measures a mixture’s viscosity by using a V-shaped device with
an aperture at the bottom. The length of time it takes the mixture to empty the funnel
gives information about both its viscosity and flowability via the bottom aperture.
4.4 Casting and Curing of Specimens
The cubical molds measuring 150 × 150 × 150 mm are thoroughly cleaned to remove
dust particles and coated with mineral oil on all sides before pouring the concrete.
The well-mixed concrete is carefully filled into the molds, and any excess is removed
using a trowel to achieve a level and smooth top surface. Once cast, the samples are
kept in the molds at room temperature for 24 h without being touched. After this
initial period, they are taken out of the molds and placed in a curing pond filled with
clean, fresh water, where they are cured for the required duration.
@seismicisolation
@seismicisolation
4.5 Muffle Furnace
At 28 days, the specimens are heated in electrical furnace at 6000, 7000, 8000, 9000,
and 10000 C. Each temperature is maintained for 30, 60, 90, and 120 min to achieve
the thermal study state. After attaining the necessary temperature, specimens are
removed from the furnace, allowed to naturally cool to room temperature, and then
tested. The heated concrete specimens were handled with extreme care.
4.6 Compressive Strength Testing Machine
In order to conduct compressive testing, a standard test cube of concrete specimen

(150 × 150 × 150 mm) is positioned between the machine’s loading surfaces in a
horizontal direction. The testing device has a 2000 KN capacity. The IS 516:1959
standard would be followed in performing this test.
5.1 Workability Test Results of Self-Compacting Concrete
The workability of fresh concrete from various mixes used in the investigation was
tested, and the results are listed in Table 6.
Table 6 Workability test results of self-compacting concrete

Test Property Unit SCC 0 SCC 10 SCC 15 SCC 20 Recommended values as
methods value value value value per EFNARC guidelines
Slump Filling MM 670 680 695 710 650-800 mm
flow ability
T50 Filling SEC 3 3.2 4 3.5 2–5 s
slump ability
flow
V-funnel Filling SEC 10 9.3 9.5 8.9 8–12 s
ability
L-box Passing H2/ 0.8 0,83 0.85 0.84 0.8–1.0
ability H1
@seismicisolation
@seismicisolation
Fig. 1 Compressive strengths of cubes exposed to 600 °C
5.2 Compressive Strength Results for Cubes When Exposed

to 600 °C
Figure 1 displays the results of exposing concrete cubes to 600 °C. From Fig. 1, it is
observed that.
• At 600 °C, the specimens’ strength reduced to 11.86%, 3.39%, and 28.8% after
30 min of exposure to various percentages (10, 15, and 20).
• At 600 °C, the specimens’ strength dropped to 14.47%, 0.52%, and 34.10% after
• At 600 °C, the specimen’s strength reduced to 17.05%, 0.90%, and 29.55% after
90 min of exposure to varied percentages (10, 15, and 20).
• At 600 °C, the specimens’ strength dropped to 23.53%, 1.97%, and 43.79% after

to 700 °C
observed that
• At 700 °C, when the specimens are exposed to a duration of 30 min of different
percentages (10 and 20), strength decreased to 19.15% and 34.04%
• At 700 °C, when the specimens are exposed to a duration of 30 min at SCC 15,
percentage strength increased to 4.26%
@seismicisolation
@seismicisolation
percentage strength increased to 11.56%.

to 800 °C
observed that
• At 800 °C, when the specimens are exposed to a duration of 60 min SCC 15,
@seismicisolation
@seismicisolation

to 900 °C
observed that
percentages (10, 15, and 20), strength decreased to 13.51%, 4.32%, and 32.43%
percentages (10, 15, and 20), strength decreased to 16.06% and 30.05%
@seismicisolation
@seismicisolation

to 1000 °C
Figure 5 displays the results of exposing concrete cubes to 1000 °C. From Fig. 5, it
is observed that
@seismicisolation
@seismicisolation
percentages (10, 15, and 20), strength decreased to 19.13% and 29.89%
percentage SCC 15, strength increased to 13.89%
percentages (10, 15, and 20), strength decreased to 54.78%, 17.7%, and 57.39%.
6 Conclusions
1. The strength of concrete weakens with rising temperatures, regardless of the

proportions of fly ash and SCBA used as cement replacements.
2. The strength of concrete diminishes with rising temperatures, regardless of the
ratios of fly ash and SCBA employed as cement replacements.
3. When fly ash and SCBA are used in place of cement at different exposures (10,
15, and 20), the strength of the SCC diminishes.
4. Irrespective of the varying amounts of cement replaced with fly ash and SCBA,
the decrease in strength increases as the specimen is exposed to elevated
temperatures for longer durations.
5. At 6000 C, the compressive strength of the SCC 15 shown equal performance
to SCC 0
6. At 7000 C, at 30, 60, 90, and 120 min exposure, the compressive strength of
the SCC 15 has performed better than normal concrete
7. At 8000 C, at 30, 60, 90, and 120 min exposure, the compressive strength of
the SCC 15 has performed better than normal concrete
8. At 9000 C, at 60 and 120 min exposure, the compressive strength of the SCC
15 has performed better than normal concrete
9. In 10000 C, at 60 min exposure, the compressive strength of the SCC 15 has
performed better than normal concrete
10. Compared to all percentages self-compacting concrete SCC 15 performed well
when compared to normal concrete (SCC0).
Citations using labels or the author/year convention are also acceptable. The
following bibliography provides a sample reference list with entries for journal arti-
cles [1], an LNCS chapter [2], a book [3], proceedings without editors [4], as well
as a URL [5].
@seismicisolation
@seismicisolation
References
1. Anand N, Arulraj P (2014) Stress-strain behaviour of normal compacting and self-compacting

concrete under elevated temperatures. J Struct Fire Eng 5(1):63–75, March 2014
2. Anand N, Arulraj P (2013) Effect of grade of concrete on the performance ofself-compacting
concrete beams subjected to elevated temperatures. Fire Technol
3. Benzaid M, Benmarce A (2017) Behaviour of self-compacting concrete mixed with different
additions at high-temperature. J Mater Environ Sci JMES 8(9):3081–3092
4. EFNARC (2005) (European Federation of national trade associations representing producers
and [2] Applicators of specialist building products), Specification andGuidelinesfor Self-
Compacting Concrete. May 2005, Hampshire, UK
5. Yang F (2004) A report on self-consolidatingconcrete 3–21
6. Tao J, Yuan Y, Taerwev L (2010) Compressive strength of self-compacting concrete during
high temperature exposure. J Mater Civ Eng 1005–1011, Oct 2010
7. Sideris KK (2007) Mechanical characteristics of self consolidating concretes exposed to
elevated temperatures. J Mater Civ Eng 19:648–654
8. Phan LT, Carino NJ (2002) Effects of test conditions and mixture proportions on behavior of
high strength concrete exposed to high temperatures. ACI Mater J Jan-February 54–62
9. Noumowe A, Carre H, Daoud A, Toutanji H (2006) High-strength self-compacting concrete
exposed to fire test. J Mater Civ Eng 754758, Dec 2006
10. Rukavina MJ, Bjegovic D, Gabrijel I (2013) Mechanical properties of self-compacting concrete
with different mineral additive after high temperature exposure. Appl Struct Fire Eng Prague,
Czech Republic, 19–20 Apr 2013
11. Srinivasa Rao K, Potha Raju M, Raju PSN (2006) Effect of Elevated temperature on
compressive strength on HSC made with OPC & PPC. Indian Concr J Aug 2006, 43–48
12. Vasusmitha R, Rao PS (2012) Effect of elevated temperature on mechanical properties of high
strength self-compacting concrete. Int J Eng Res Technol (IJERT) 1(8), Oct 2012
13. Murugan K, Palaniappan M, Kalappan KK (2022) Experimental studies on light weight
concrete using LEC a material. Mater Today Proc 74:10351041. https://doi.org/10.1016/j.
matpr.2022.11.467
14. Anjaneyulu B, Alisha SS, Neeraja S, Divya K (2017) Experimental study on self-compacting
concrete with various percentage of steel fibres. IJIRST Int J Innovative Res Sci Technol
3(10), Mar 2017 ISSN (online): 2349-601, performances elf-consolidating concrete at elevated
temperatures
15. Gnanaraj SC, Chokkalingam RB, Lizia Thankam G, Pothinathan SKM (2021) Durability Prop-
erties of self-compacting concrete developed with fly ash and ultra fine natural steatite powder.
J Mark Res 13:431–439
16. Premkumar R, Chokkalingam RB, Kumar VJ (2021) Self-compacting concrete properties of
recycled coarse aggregate. Mater Res Proc 115–122
@seismicisolation
@seismicisolation
Optimization Studies on Bracing Systems
and Its Effective Placement
to Counteract Earthquakes in Very High
Damage Risk Zone
S. Usha , H. A. Ajay , D. T. Abhilash , and B. A. Brunda
1 Introduction
India has a devastating history of earthquakes. Prime reason for high frequency and
intensity of earthquakes is that the Indian plate is driving into Asia approximately at
a rate of 47 mm every year. The recent version of earthquake zoning map of India
as given in the IS 1893 (Part 1): 2016 assigns four zone factors depending on the
level of seismicity. In other words, Indian seismic zoning map has 4 zones (Zone
II, III, IV, and V) unlike the previous version, which consisted of 5 zones. Zone 5
expects the highest level of seismicity whereas Zone II is associated with the lowest
level of seismic activity. The ground movement because of seismic waves imparts
vibrations to the structures and hence they undergo displacement from their original
position. Vibrations are considered as lateral loads being acted upon the structures.
These loads have to be resisted by the building frames. However, when the loads
are prominent then OMRF is not sufficient to counteract the lateral loads, in which
case, we will adopt additional lateral stiffness systems to minimize the deformation
of structures.
S. Usha · H. A. Ajay (B)

Presidency University, Bangalore, India
e-mail: ajay@presidencyuniversity.in
D. T. Abhilash
Adichunchanagiri Institute of Technology, Chikmagalur, India
B. A. Brunda
Vijaya Vittala Institute of Technology, Bangalore, India
@seismicisolation
@seismicisolation
444 S. Usha et al.
1.1 Lateral Load Resisting Systems (LLRS)
A type of load resisting system in which the lateral load is resisted and sway or
deflection of the structure is reduced. Lateral loads can either be wind loads or
seismic loads. Mainly we have Moment resisting frames, Braced frames, and Shear
wall as LLRS. Moment resisting frames could be either Ordinary or Special Moment
resisting frames. There are different types of bracings as shown in Fig. 1 viz. concen-
tric bracings, eccentric bracings, and buckling restrained braces. The mechanism of
resisting loads is different for each case. Concentric bracings are the bracings which
have a common center point. These bracings have different shapes like X-shape and
V-shape. In this system, the sway caused by lateral loads can be resisted from any
direction. Eccentric Bracings are the bracings which have a different Center point.
When load is acted upon these frames, then they behave like a multiple framed
structure. High-rise structures can also be made earthquake resistant using outrigger
system. These systems reduce response of structures to lateral forces.
Fig. 1 Frames with

X-bracing and Eccentric
bracing
@seismicisolation
@seismicisolation
Optimization Studies on Bracing Systems and Its Effective Placement … 445
2 Literature Review
Bagheri et al. [1] have analyzed a twenty-storey high-rise building with plan irreg-
ularity using ETABS and SAP 2000 for earthquake Zone V. Analysis is done by
considering earthquake data from El-Centro (1949) and Chi-Chi Taiwan (1999).
They have concluded that the displacement values obtained from static analysis are
higher than dynamic analysis and make the design uneconomical. Patil et al. [2]
have performed seismic analysis considering tall structures by Response spectrum
method. In the study, they considered a G + 14 storey structure for the analysis. They
have concluded that the effect of bracings and shear walls is relatively more when
placed along the longer span than along shorter spans and corners. The storey drift
and storey displacement are reduced significantly.
Abhijeet and Kanchan [3] have conducted a behavioral study of LLRS of various
heights in the structure in a variety of soils like Hard, Medium, and Soft in Zone
V. They have concluded that when the soil type changes from hard to soft, the
displacement, drift, base shear, and time period also increase and special care should
be taken in case of structures built on soft soils. Shear walls in the middle periphery of
the building yielded better results than along the corners. Thejaswini and Rashmi [4]
have performed analysis of irregular shaped buildings with varied structural forms.
They have concluded that higher number of exterior columns increases the stiffness of
structure. Outrigger systems help reduce the drift. Torsional irregularity also reduces
with proper shape and symmetry in buildings.
Piyush Gupta and Neeraja [5] have performed seismic analysis of 30 storey
building models with various RC lateral force resisting systems using ETABS. They
have concluded that outrigger is the most effective type of LLRS. The outrigger
system provided with belts and bands placed at the top and the middle portion yielded
better results. Muralidhar and Swathi [6] have studied the performance of LLRS in
the buildings with Flat slabs. They had concluded that performance of the building
is increased with the use of lateral systems. Shear wall models gave less time period
than bracings. Increase in base shear is observed due to the additional weight from
the shear walls. Vaseghi [7] recorded response of the structure for three earthquakes
compared with Friction Damper (FD) and EBF. The result showed FD in comparison
with EBF, significantly reduced the storey displacement, support reaction, and base
shear. Zashiah [8] has done the investigation on different types of bracings such as
concentric and eccentric bracings using HSS section. The lateral displacement has
reduced with the use of different types of bracings as compared with bare frames.
Bhojkar [9] found out from their studies that X-bracings play a major role in the
structure stiffness and reduce the inter-storey drift.
Khan [10] has done the investigation on eccentric bracings for a 7-storey building
using nonlinear static pushover analysis. The results of pushover analysis show that
buildings with eccentric bracings have significantly lower drift demand and proba-
bility of collapse. Ghanem [11] has done a parametric study to understand behavior of
steel eccentric braced frames. The study aims to determine the ability and efficiency
of building structure to dissipate seismic energy through an inelastic deformation in
@seismicisolation
@seismicisolation
446 S. Usha et al.
plastic hinges under different peak ground acceleration level. The outcome shows
that the height of the structure has an inverse relationship with the force reduction
factor.
David [12] evaluated the seismic performance of EBF in various configurations.
EBFs are subjected to both linear and nonlinear analysis using SAP 2000. The
linear analysis sheds light on mode forms and percentages of mass participation.
The pushover analysis, which offers details on the collapse mechanisms and perfor-
mance points, is a component of the nonlinear analysis. The result obtained is that
all the selected configuration of eccentrically braced frames undergo small displace-
ment at the top. Ajay et al. [13] have carried out seismic analysis of a mid-rise and
high-rise building structure by response spectrum method. In their study, they have
modeled a 6 storey and 15 storey building to compare the behavior and effect of
X-bracings and shear walls. They have concluded that both mid-rise and high-rise
structures behave identically when subjected to response spectrum loads. Bracings
placed at middle periphery and shear walls like a core yielded better performance
of the structure. High-rise structures can also be made earthquake resistant using
outrigger system. These systems reduce response of structures to lateral forces [14].
3 Methodology
Dynamic analysis of G+14 building models is performed using the FEA software
tool ETABS v.20. The typical plan of the model is shown in Fig. 2. X—Bracings,
Inverted V—Bracings and Eccentric Bracings are placed at Locations A (Middle
periphery along Y-direction), B (Middle periphery along X-direction), C (Corner
periphery along X and Y-direction), D (Center of the building). Bay length is 5 m
and 4 m along X and Y-direction respectively. Building model with X-bracings at
middle periphery is shown in Fig. 3. The dimensions of the columns, beams, slab,
and bracings have been indicated in Table 1.
The models are subjected to dynamic loads by response spectrum method as
per standard plot of spectral acceleration and time. The maximum displacement,
Fig. 2 Placement location

of braces
@seismicisolation
@seismicisolation
Fig. 3 X-bracings
positioned at A and B
locations
Table 1 Dimensions of building components

Members Lower 6 stories, Middle 4 stories, Top 5 stories,
sizes (mm) sizes (mm) sizes (mm)
Columns 300 × 600 230 × 450 230 × 300
Beams 300 × 600 230 × 450 230 × 300
Slabs 150 150 150
X-Bracings 300 × 600 230 × 450 230 × 300
Inverted V-Bracings 300 × 600 230 × 450 230 × 300
Eccentric Bracings 300 × 600 230 × 450 230 × 300
overturning moment, stiffness, time period, and base shear values are recorded from
the analyzed models. The best LLRS and its position are concluded based on the
above-mentioned seismic parameters.
The forces are applied as area loads on the slabs as per IS875 (Part 1 and 2): 1987.
DL: Program calculated, LL: 3.5 kN/m2 , FF: 1.5 kN/m2 .
@seismicisolation
@seismicisolation
448 S. Usha et al.
The models without braces and with LLRS such as X-braced frame, Inverted V-
braced frame, and Eccentric braced frames are analyzed using Response spectrum
method. The parameters viz. maximum storey displacement, base shear, overturning
moment, storey stiffness, and time periods are tabulated for spectrum in X-direction
and Y-direction represented as SPECx and SPECy respectively. These are plotted
graphically, and the inference is made based on the results obtained.
4.1 Maximum Storey Displacement
The results of maximum storey displacement are shown in Figs. 4 and 5 respectively.
For the bare frame model, the displacement is 466 mm and 549 mm respectively
for SPECx and SPECy loading. From the analysis results, it is clearly evident that
the storey displacements will significantly reduce when braced frames are used. It
can also be noted that when braces are provided along loading direction (position A
for SPECy and position B for SPECx), the displacement reduces by almost 45–55%
due to increased stiffness along those directions. When braces are placed along the
middle periphery i.e., both positions A and B, then the displacement reduces by
50% for any direction of loading. A similar performance can be expected from the
structure when braces are provided at the center of the building (position D) like a
braced core system. The performance of all the types of bracings in every position
is identical.
MAXIMUM STOREY DISPLACEMENT - SPEC X

X-Bracings Inverted V-Bracings Eccentric Bracings
Maximum Displacement, mm
472
471
471
353
343
338
329
319
310
275
263
209
207
204
203
A B AB C D
Bracings Position
Fig. 4 Graph showing maximum displacement at the top most storey when response spectrum load
is acted upon X-direction
@seismicisolation
@seismicisolation
MAXIMUM STOREY DISPLACEMENT - SPEC Y

561
558
557
Maximum Displacement, mm
396
387
373
350
347
346
328
327
320
319
313
312
A B AB C D
Bracings Position
Fig. 5 Graph showing maximum displacement at the top most storey when response spectrum load
is acted upon Y-direction
4.2 Maximum Storey Shear
The results of maximum storey shear are shown in Figs. 6 and 7 respectively. The
maximum base shear is observed at the bottom most storey and for the bare frame
model, the base shear is 8244 kN and 6866 kN respectively for SPECx and SPECy
loading. From the analysis results, it is clear that base shear is going to increase with
the addition of braces as the dead would automatically increase. It is also observed
that, placing the bracings at middle periphery of the structure would increase base
shear by almost 150% and hence is not recommended. Design would make the
sections heavier and more prone to earthquakes. Hence, placement of bracings at D
would be ideal as it would only increase the shear up to 80%. Specially, the eccentric
bracings at D would only increase base shear by only 40% and are an ideal choice.
4.3 Maximum Overturning Moment
The results of maximum overturning moment are shown in Figs. 8 and 9 respectively.
The maximum overturning moment observed in the bare frame model at the bottom
most storey is 208 MNm and 172 MNm for SPECx and SPECy loading respec-
tively. When the braces are added and response spectrum analysis is performed,
the overturning moment has almost increased by 100% for braces placed in middle
periphery of the building, which is significantly high. But braces at position D have
only increased the moment by 50% for both directions of loading. Eccentric bracing
performed well by only allowing the moment to increase by 40% when placed as
braced core.
@seismicisolation
@seismicisolation
450 S. Usha et al.
MAXIMUM BASE SHEAR - SPEC X

20007
19771
19134
19066
15310
15253
15229
14181
13980
Base Shear, kN
13860
12235
11560
8565
8487
8424
A B AB C D
Bracings Position
Fig. 6 Graph showing maximum shear at the bottom most storey when response spectrum load is
acted upon X-direction
MAXIMUM BASE SHEAR - SPEC Y

15530
15275
14856
14793
14730
14635
13218
13193
12107
11973
11851
11203
Base Shear, kN
7278
7194
7147
A B AB C D
Bracings Position
Fig. 7 Graph showing maximum shear at the bottom most storey when response spectrum load is
acted upon Y-direction
4.4 Maximum Storey Stiffness
The results of storey stiffness are shown in Figs. 10 and 11 respectively. The storey
stiffness of the bare frame model is 502 MN/m and 241 MN/m for SPECx and SPECy
loading respectively. From the analysis results it is evident that the storey stiffness
could be increased to a very large extent if braces are placed in both the directions
along middle periphery of the structure by 6 times. But stiffer structures are more
prone to earthquake effects and hence structure has to be ductile. The bracings in
position D especially the eccentric bracings would only increase the stiffness by
@seismicisolation
@seismicisolation
MAXIMUM OVERTURNING MOMENT - SPEC X

493.8
Overturning Moment, MNm
495
489
488
384
325.6
363
322.4
304.7
304.3
281.7
292
211.7
210.6
210.4
A B AB C D
Bracings Position
Fig. 8 Graph showing maximum overturning moment at the bottom most storey when response
spectrum load is acted upon X-direction
MAXIMUM OVERTURNING MOMENT - SPEC Y

319.8
319.3
317.1
315.2
313.2
311.4
286.2
285.4
282.5
Overturning Moment, MNm
266.9
249.6
269
175.1
174.8
177
A B AB C D
Bracings Position
Fig. 9 Graph showing maximum overturning moment at the bottom most storey when response
spectrum load is acted upon Y-direction
about 3 times keeping a perfect balance. Structure is neither too stiff nor too flexible
to cause huge displacement and drift.
4.5 Maximum Time Period
The results of maximum time period is shown in Fig. 12. Time period for the bare
frame model is 2.26 s. Lesser the time period higher the performance of the structure
exhibiting less displacements. The time period for structure with braces at middle
periphery and center is almost the same as per the analysis results. Time period
@seismicisolation
@seismicisolation
452 S. Usha et al.
STOREY STIFFNESS - SPEC X
3194
3167
3110
3106
Storey Stiﬀness, MN/m
2188
1967
1774
1619
1252
1202
1023
907
503.7
506
505
A B AB C D
Bracings Position
Fig. 10 Graph showing maximum storey stiffness at the bottom most storey when response
spectrum load is acted upon X-direction
STOREY STIFFNESS - SPEC Y
1786.8
1774
1759
1577.6
1558.5
1511.8
1577
1273
Storey Stiﬀness, MN/m
707.6
700.3
700
700
242.5
241.9
242
A B AB C D
Bracings Position
Fig. 11 Graph showing maximum storey stiffness at the bottom most storey when response
spectrum load is acted upon Y-direction
reduces by 80% compared to structure without braces. Hence brace core system is
more efficient in absorbing the energy from vibrations.
@seismicisolation
@seismicisolation
TIME PERIOD, SECONDS

3.00 2.56
2.57 2.55
2.50 2.08
2.08 2.07
Time Period, Seconds
1.66 1.45
2.00 1.41 1.71 1.78 1.77
1.58 1.48
1.44
1.50
1.00
0.50
0.00
A B AB C D
Bracings Position
Fig. 12 Graph showing maximum time period of building models for various bracing locations
5 Conclusion
From a detailed investigation on the use of bracings as a lateral load resisting system
for a very high damage zone, the following conclusions have been made.
1. Bracings would help decrease the storey displacements and drifts by as much as
50% making the sway less.
2. Bracings will be effective when placed at the middle periphery of the building in
both directions rather than at corners to get efficient performance.
3. Positioning the bracings like a core system at the center of the building is the
most efficient placement of braces as the displacements, overturning moments,
base shear, and time period would reduce significantly compared to any other
position.
4. Eccentric bracing is an ideal choice as it optimizes performance of the building
making it flexible without any sacrifice on displacement, moments, base shear,
and time period aspects as can be seen in many literatures.
References
1. Bagheri B (2012) Comparative study of the static and dynamic analysis of multi-storey irregular
building. Int J Civ Environ Struct Constr Architectural Eng 6(11):1045–1049
2. Patil SS (2013) Seismic analysis of high-rise building by response spectrum method. Int J
Comput Eng Res 3(3):272–279
3. Baikerikar A (2014) Study of lateral load resisting systems of variable heights in all soil types
of high seismic zone. Int J Res Eng Technol 3(10):109–119
@seismicisolation
@seismicisolation
454 S. Usha et al.
4. Thejaswini RM (2015) Analysis and comparison of different lateral load resisting structural
forms. Int J Eng Res Technol 4(7):827–833
5. Gupta P (2016) Analysis of various RCC lateral force resisting systems and their comparison
using ETABS. Int J Latest Trends Eng Technol 6(4):175–182
6. Muralidhar GB (2016) Comparison of performance of lateral load resisting systems in multi
storey flat slab building. Int J Res Eng Technol 5(3):15–19
7. Vaseghi J (2009) A parametric assessment of friction damper in eccentric braced frame. Int J
Civ Environ Eng 3(10):361–365
8. Tafheem Z (2019) Structural behavior of steel building with concentric and eccentric bracing:
a comparative study. Int J Civ Struct Eng 5(2):56–63
9. Bhojkar N (2015) Seismic evaluation of high-rise structure by using steel bracing system. Int
J Innovative Sci Eng Technol 2(3):264–269
10. Khana D (2016) Nonlinear seismic analysis of masonry infill RC buildings with eccentric
bracings at soft storey level. In: World multidisciplinary civil engineering-architecture-urban
planning symposium 2016. Procedia Engineering, Prague, Czech Republic, pp 9–17
11. Ghanem A (2019) Effect of the height of structures on the force reduction factor of eccentric
braced frames. Thesis submitted to Cairo University
12. Rinu G (2020) Seismic performance of eccentrically braced frames. In: International virtual
conference on emerging research trends in structural engineering. IOP Conference Series:
Materials Science and Engineering, VIT Chennai, pp 16–17
13. Ajay HA (2016) Seismic analysis of a multistorey RC building for various lateral load resisting
systems by response spectrum method. Int J Res Appl Sci Eng Technol 4(7):657–665
14. Priyanka NM, Abhilash DT, Ajay HA, Mohan HS, Apoorva S (2022) Comprehensive analysis
of outrigger system for high rise structures subjected to wind and earthquake loadings. In:
Nandagiri L, Narasimhan MC, Marathe S, Dinesh S (eds) Sustainability trends and challenges
in civil engineering, vol 162. Lecture Notes in Civil Engineering. Springer, Singapore, pp
601–618
@seismicisolation
@seismicisolation
Performance Evaluation of Fire Exposed
RC Structure Using Pushover Analysis
K. Mahammad Jafar and V. Sachin
1 Introduction
During a fire event, a structure is subjected to both thermal and structural loads.
The thermal loads cause serious damage to the material and owing to this structural
element undergoes large deformations [1, 2]. The heat and flames from a fire can
weaken and damage the structural integrity of a building. This can result in the
failure of structural elements or collapse of the whole structure [3]. It is important
to note that the accuracy and reliability of the analysis results depend on the quality
of the input data and the assumptions made during the analysis. By performing
a pushover analysis for fire, structural engineers can evaluate the performance of
a structure under the combined effects of thermal and structural loads and make
informed decisions about the safety and viability of the structure after a fire event.
After conducting a thorough review of literature on the study of pushover analysis
(non-linear static analysis), it is observed that most of the study has been done on
pushover analysis (performance evaluation) considering gravity load and effect of
lateral loads i.e.; (seismic and wind loads) [4]. However, performance of RC framed
structures during post-fire events has a limited amount of study. The standard fire
effect on RC Buildings is not considered to perform pushover analysis. It is observed
from the literature summary that performance of RC framed structures exposed to
fire scenarios has limited studies [5, 6]. Hence, these objectives were formulated
to determine performance of RC structure exposed to standard fire of 2 h duration
adopting pushover analysis.
• To formulate fire scenarios and generate the temperature profile of RC beams and
columns exposed to standard fire.
• To determine performance of reinforced concrete structure before and after being
exposed to standard fire by adopting pushover Analysis.
K. M. Jafar (B) · V. Sachin

Department of Civil Engineering, The National Institute of Engineering, Mysuru 570 008, India
e-mail: jappi098@gmail.com
@seismicisolation
@seismicisolation
456 K. M. Jafar and V. Sachin
2 Description of Building
The building selected to carry out the performance evaluation (pushover analysis) is
G+9 RC commercial building. The RC structure is subjected to both future seismic
loads and fire effects to check the building performance pushover analysis is carried
out. The building is assumed to be located in seismic zone III on type II soil (Medium
soil). A square plan configuration is considered with 4 bays in both X and Y directions,
and each bay of length 6m is adopted in the present study. The story height of 3m
is considered for all the 10 stories of the structure and a height of 2m is considered
from plinth level to foundation. The total height of the structure from base of the
footing to the Terrace level is found to be 32 m. To carry out pushover analysis
ETABS Ultimate 19.0.0 software is used. The plan, elevation, and isometric view of
the Building is shown in Figs. 1, 2, and 3. The technique adopted for non-linear static
analysis without the Fire effect is explained in Sect. 3.4. The analysis was carried
out in reference with Indian standard codal provisions IS-875 [7–10].
2.1 Frame Material and Section Properties are Defined

in Table 1
2.2 Defined Load Combinations are Discussed in Table 2
2.3 Fire Calculation of 500 °C Isotherm Method
The concrete with temperature above 500 °C has no compressive strength, and
concrete below 500 °C has full compressive strength. For this, the temperature profiles
Fig. 1 Plan
@seismicisolation
@seismicisolation
Performance Evaluation of Fire Exposed RC Structure Using Pushover … 457
Fig. 2 Elevation
Fig. 3 3D isometric
EN 1992-1-2 (2004) were used. Determine a new width and new effective depth of
cross section by excluding the concrete outside the 500 °C isotherm.
The fire spread scenario is considered at five different compartments i.e., case 1
with 25% floor area at corner, case 2–25% floor area at interior, case 3 with 25%
floor area at side, case 4 with 50% floor area, and case 5 with 100% floor area at
1st story, 5th story, and 10th story exposed to standard fire effect of 2h duration as
shown in Fig. 4.
@seismicisolation
@seismicisolation
Table 1 Frame material and section properties defined

Element name Floor levels Section (mm) Grade Clear cover (mm)
Plinth beam GF (Plinth level) B 300 × 450 M35 25
Floor beam 1st FL to terrace FL B 300 × 600 M35 25
Floor slab 1ST GL to terrace FL S 150 M25 20
Exterior column GF to 6th C 300 × 600 M50 40
7th FL to terrace FL C 450 × 600 M50 40
Interior column GF to 6th C 450 × 750 M50 40
7th FL to terrace FL C 300 × 450 M50 40
Table 2 Load combinations

0.9 (DL + SDL)-1.5 (EQ X) 1.2 (DL + LL + SDL + EQ Y)
0.9 (DL + SDL)-1.5 (EQ Y) 1.2 (DL + LL + SDL + WIND X)
0.9 (DL + SDL)-1.5 (WIND X) 1.2 (DL + LL + SDL + WIND Y)
0.9 (DL + SDL)-1.5 (WIND Y) 1.5 (DL + LL + SDL)
0.9 (DL + SDL) + 1.5(EQ X) 1.5 (DL + SDL)
0.9 (DL + SDL) + 1.5(EQ Y) 1.5 (DL + SDL)-1.5 (EQ X)
0.9 (DL + SDL) + 1.5(WIND X) 1.5 (DL + SDL)-1.5 (EQ Y)
0.9 (DL + SDL) + 1.5(WIND Y) 1.5 (DL + SDL)-1.5 (WIND X)
1.2 (DL + LL + SDL)-1.2 (EQ X) 1.5 (DL + SDL)-1.5 (WIND Y)
1.2 (DL + LL + SDL)-1.2 (WIND X) 1.5 (DL + SDL + EQ X)
1.2 (DL + LL + SDL)-1.2 (EQY) 1.5 (DL + SDL + EQ Y)
1.2 (DL + LL + SDL)-1.5 (WIND Y) 1.5 (DL + SDL + WIND X)
1.2 (DL + LL + SDL + EQ X) 1.5 (DL + SDL + WIND Y)
2.4 Wickström’s Method
Wickström’s method of calculating the temperatures in a concrete (internal tempera-

ture) due to standard fire effect is based on the fire-exposed surface temperature then
Tx, y is calculated as follows.
For 2h of standard fire exposure, the temperature in the fire compartment is.
Fire temperature, Tf = 20 + 345log(8t + 1) = 1049 °C where, t = time (min).
Tf = 1049 °C for 2h of standard fire duration.
Surface temperature, ηs = 1 – 0.06th – 0.90.
Concrete temperature, ηx = ηy = 0.172ln( t/x2 ) – 0.74.
Internal temperature, Ts = Tc = Tx,y = ηs(ηx + ηy -2ηxηy) + ηx ηw)Tf.
Yield strength of bars, fy (N/mm2).
Reduced yield strength, fy’ = (1.53 – Ts/470)fy.
@seismicisolation
@seismicisolation
Fig. 4 Fire spread scenarios at five different compartment cases
2.5 Temperature Profile
Sections of Beams and Columns subjected to standard fire. Internal temperature

curve generated inside the sections in the excel spreadsheet figure is shown in Figs. 5
and 6.
3.1 Comparison of Capacity Curve Along Push-X Before

and After Fire Effect
The comparison of the capacity curve along Push-X without fire and with fire effect.
The plot for comparison of the capacity curve along Push-X is presented in Fig. 7. The
unaffected building shows higher base shear value of 6353.0021 KN for displacement
of 116.101 mm along X compared to 1st story case 5 for the same displacement of
109.944 mm obtained low base shear value of 4751.6576 KN. Because at 1st story
case 5, the entire beams and columns sections are exposed to fire which is the major
critical case among all. Hence the unaffected building shows high stiffness compared
@seismicisolation
@seismicisolation
Fig. 5 500 °C isotherm profile of beam
Fig. 6 500 °C isotherm profile of column
to fire cases. Because there is no reduction of sections (Beam and columns). Even
at the 10th story case 5, entire story exposed does not show much variation even for
displacement of 113.57 mm obtained base shear value of 5594.9656 KN.
@seismicisolation
@seismicisolation
Fig. 7 Comparison of
capacity curve along Push-X
before and after fire effect
3.2 Comparison of Capacity Curve Along Push-Y Before

The comparison of the capacity curve along Push-Y before and after fire effect.
The plot for comparison of capacity curve along Push-X is presented in Fig. 8. The
unaffected building obtained base shear value of 5793.5236 KN for displacement
of 128.81 mm along X compared to 1st story case5 around same displacement of
122.315 mm obtained low base shear value of 3778.9464 KN. Because at 1st story
case5, the entire beams and columns sections are exposed to fire which is the major
critical case among all. Hence the unaffected building shows stiffness compared to
fire cases because there is no reduction of sections (Beam and columns). Even at
the 10th story case 5, entire story exposed does not show much variation even to for
displacement of 113.57 mm obtained base shear value of 5594.9656 KN.
3.3 Comparison of Response Spectrum Along Push-X Before

The comparison of Response spectrum along Push-X before and after fire effect.
The plot for comparison of Response spectrum along Push-X is presented in Fig. 9.
Among which the unaffected case obtained lower spectral displacement (sd) of
84.283 mm for the applied spectral acceleration (sa) 00.072547 g because the unaf-
fected building to fire there is no reduction in cross section of structural elements
at any storys so it has high stiffness to resist these lateral acceleration applied to it.
Among fire affected storys for 1st story case 5 for lower spectral acceleration (sa)
0.054754 g got the higher spectral displacement (sd) of 87.901mm with a constant
time period of 2.542 s compared to other story 5th and 10th story Fire exposed. But the
@seismicisolation
@seismicisolation
capacity curve along Push-Y
response spectrum curve
along push-X before and
after fire effect
10th story obtained a spectral displacement of 88.299 mm for spectral acceleration

of 0.065225.
3.4 Comparison of Response Spectrum Along Push-Y Before

The comparison of Response spectrum along Push-Y before and after fire effect is
presented. The plot for comparison of Response spectrum along Push-Y is presented
in Fig. 10. Among which the unaffected case shows lower spectral displacement
(sd) of 104.86 mm for the applied spectral acceleration (sa) 0.070192 g for the time
@seismicisolation
@seismicisolation
response spectrum curve
along Push-Y before and
after fire effect
period of 2.452 s because the unaffected building to fire there is no reduction in

cross section of structural elements at any story so it has high stiffness to resist these
lateral acceleration applied to it. Among fire affected storys for 1st case 5 story for
lower spectral acceleration (sa) 0.041871g got the higher spectral displacement (sd)
of 100.674 mm with a constant time period of 3.111s compared to other story 5th and
10th Fire exposed. But the 10th story case 5 got lower value of spectral displacement
102.574 mm for spectral acceleration of 0.061905 g because in this case, the top
story affected by fire does not depend on other story. The Push-Y response is more in
terms of displacements compared to Push-X because along push-Y direction all the
exterior columns are oriented in Push-X direction to resist the applied forces against
displacements the depth orientation is more and has sufficient.
3.5 Comparison of Maximum Story Roof Displacement

Along Push-X Before and After Fire Effect
The comparison of maximum story displacement along Push-X before and after
fire effect is shown, and the plot for all cases of fire scenario is shown in Fig. 11.
The comparison of maximum story displacement along Push-X unaffected building
shows lower value of displacement 157.571 mm along X compare to all fire scenario
cases because there is no reduction of sections (Beam and columns) hence the unaf-
fected building shows stiffness compare to fire cases. But among these fire scenario
cases i.e., 1st story, 5th story, and 10th story, the 5th story shows maximum story
displacement of 246.383 mm along X which is the major critical case among all
cases. Up to 4th story, the unaffected portion shows displacement of 146.839 mm
but as the 5th story reaches fire affected area shows a sudden increase of displacement
from 146.839 to 191.602 mm due to its reduction cross section.
@seismicisolation
@seismicisolation
maximum story roof
displacement along Push-X
3.6 Comparison of Maximum Story Roof Displacement

Along Push-Y Before and After Fire Effect
The comparison of maximum story displacement along Push-Y is shown in Fig. 12.
The unaffected building shows lower value of displacement 161.405 mm along X
compared to fire cases of 1st story, 5th story, and 10th story exposed because there is
no reduction of sections (Beam and columns). But among these fire scenario cases, the
affected story i.e., 1st story, 5th story, and 10th story, the 5th story shows maximum
story displacement of 226.73 mm along X which is the major critical case among
all. Up to 4th story, the unaffected portion shows displacement of 124.291 mm but
as the 5th story reaches fire affected area shows sudden a increase of displacement
from 124.291 mm to 178.062 mm due to its reduced cross section.
3.7 Comparison of Maximum Story Drift Ratio Along

Push-X Before and After Fire Scenario
The comparison of maximum story drift ratio along Push-X before and after fire
scenario. The plot for all cases of fire scenario is presented in Fig. 13. The unaffected
building shows lower value of drift ratio of 0.008704 along Push-X compared to all
fire scenario cases because there is no reduction of sections (Beam and columns)
hence the unaffected building shows stiffness compared to fire cases. But among
these fire scenario cases 1st story, 5th story, and 10th story, the 1st story case 4
shows maximum story drift ratio of 0.017758 along X. Up to ground floor story, the
unaffected portion shows story drift ratio of 0.009428 but as the 5th story reaches
fire affected area shows sudden increase of displacement from 0.00942 to 0.017758
@seismicisolation
@seismicisolation
maximum story roof
displacement along Push-Y
maximum story drift ratio
along Push-X before and
after fire scenario
due to its reduced cross section by the effected fire case. The drift ratio at 5th story
case 5 also changes its drift ratio from 0.00671 to 0.014939 but lower compare 1st
story case 4.
3.8 Comparison of Maximum Story Drift Ratio Along

Push-Y Before and After Fire Scenario
The comparison of maximum story drift ratio along Push-Y is shown in Fig. 14. The
unaffected building shows lower value of drift ratio of 0.010813 along Y compared to
@seismicisolation
@seismicisolation
maximum story drift ratio
along Push-Y before and
after fire scenario
fire cases of 1st story, 5th story and 10th story exposed because there is no reduction
of sections (Beam and columns). But among these fire scenarios, the affected story
cases i.e.; 1st story, 5th story, and 10th story, the 1st story case 3 shows maximum
story drift ratio of 0.020894 along Y which is the major critical case among all. Up to
ground floor story, the unaffected portion shows story drift ratio of 0.016819 but as
the 5th story reaches fire affected area shows sudden increase of displacement from
0.016819 to 0.020894 due to its reduced cross section by the effected fire case. The
drift ratio at 5th story case 5 also changes its drift ratio from 0.005303 to 0.017926
but shows lower compared to 1st story case 4.
4 Conclusion
The performance of G+9 RC Commercial building frames was investigated using the
pushover Analysis. These are the conclusions drawn from the analyses and results
carried out.
• Compared to unaffected building the pushover analysis of fire scenario building
gave lower base shear value of 25.21%, 12.4%, and 11.94% for 1st story, 5th
story, and 10th story along Push-X direction and 34.78%, 14.8%, and 9.6% for
1st story, 5th story and 10th story exposed to fire scenario along Push-Y direction.
• The maximum story roof displacement before and after fire scenario from the
obtained results and plot shows that the 5th story case 5 fire scenario shows
maximum roof displacement. Because at the 5th story case 5 fire scenario all
members (Beams and columns) size is reduced due to which members loses their
stiffness. From the plot obtained observed that up to 4th story, displacement was
@seismicisolation
@seismicisolation
gradually increasing but as the 5th story case5 region reaches it shows sudden
increase in displacement due to fire scenario case at this stage.
• The maximum drift ratio before and after fire scenario effect at 1st story 104%, 5th
story 67%, and 10th story 74.3%. The maximum drift ratio increased by 104%.
• The Response spectrum method concludes that no performance point is generated
from demand vs capacity curve even for this 1st story case 5 critical case because
the overall capacity of the structure is sufficient to carry the external loads acts
on the structure.
• It is observed that the results of maximum roof displacements, maximum story
drift ratio, capacity, and base shear value remain unaffected when 10th story of
the structure is subjected to all five fire scenarios.
• The last step of hinge status of performance level shows that maximum failure of
RC members (Beams and columns) in the range of IO to LS (Immediate occupancy
to Life safety level) and greater than CP (Collapse level) is at ground floor (Plinth
level) and first floor level of building.
• Hinge status and performance level before and after fire effect compared to all
fire scenario cases the 1st story case 5 is critical and shows higher Hinge status
at ground floor (Plinth level) 36 exterior and interior beams are in the range of
(life safety level to collapse prevention level). At 1st floor 9 are greater than CP
(Collapse level) are collapsed completely.
References
1. Sunagar P, Bhashyam A, Shashikant M, Sreekeshava KS, Chaurasiya AK (2021) Effect of

different base isolation techniques in multistoried RC regular and irregular building. In:
Narasimhan MC, George V, Udayakumar G, Kumar A (eds) Trends in civil engineering and
challenges for sustainability. Lecture notes in civil engineering, vol 99. Springer, Singapore.
https://doi.org/10.1007/978-981-15-6828-2_29
2. Sreekeshava KS, Rodrigues H, Arunkumar AS (2023) Response of masonry-infilled reinforced
concrete frames strengthened at interfaces with geo-fabric under in-plane loads. Buildings
13(6):1495
3. Gowda RS, Sunagar P, Nruthya K, Manish SD, Sreekeshava KS, Chaurasiya AK, Priyanka
(2020) Analytical and finite element buckling and post buckling analysis of laminated plates.
Int J Civ Eng Technol 11(5):84–92
4. Sreekeshava KS, Arunkumar AS, Ravishankar BV (2020) Experimental studies on polyester
geo-fabric strengthened masonry elements. In: Prashant A, Sachan A, Desai C (eds) Advances
in computer methods and geomechanics. Lecture notes in civil engineering, vol 55. Springer,
Singapore. https://doi.org/10.1007/978-981-15-0886-8_58
5. Sreekeshava KS, Arunkumar AS, Ravishankar BV (2021) Strengthening of brick masonry
wallette by polyester geo-fabric reinforced cementitious matrix. Mater Today Proc 46:2392–
2399
6. Endo Y, Pelà L, Roca P (2017) Review of different pushover analysis methods applied to
masonry buildings and comparison with nonlinear dynamic analysis. J Earthq Eng 21(8):1234–
1255
7. IS 875-1 (1987) Code of practice for design loads (Other than earthquake) for buildings and
structures, Part 1: Dead Loads
@seismicisolation
@seismicisolation
8. IS 875-2 (1987) Code of practice for design loads (Other than earthquake) for buildings and
structures, Part 2: Imposed Loads
9. IS 875-3 (2016) Code of practice for design loads (other than earthquake) for buildings and
structures - Part 3: Wind loads
10. IS 1893-1 (2016) Criteria for Earthquake resistant design of structures
@seismicisolation
@seismicisolation
Internet of Things Enabled Structural
Health Monitoring Using Fiber Bragg
Grating Sensors
K. Chethana, Somesh Nandi, A. P. Guruprasad, and S. Ashokan
1 Introduction
In the Internet of Things, Cloud computing and related technologies are used for
information and communication exchange. The IoT has become so important to
modern IT because it can manage large amounts of data. Because of its benefits,
optical-fiber-based sensing technology has a big impact on the Internet of Things.
The ongoing advancement of communication technology has helped to make optical-
fiber-based sensing technology possible in recent years [1]. Fiber-optic sensors, such
as functional fiber and non-functional fiber sensors, have therefore been quickly
developed because of their benefits, which include small size, high sensitivity, strong
anti-interference ability, and more. The sensors are widely used in a different industry,
like shipping, civil engineering, and the production of pharmaceuticals. Fiber-optic
sensor advancements have sped up the Internet of Things’ development and increased
the value of sensors in various applications. Through the employment of advanced and
accurate monitoring instruments, modern technology enables continuous structure
monitoring. This is why fiber Bragg grating (FBG) sensing has demonstrated a wide
variety of applications for monitoring vital structures such as buildings, bridges,
tunnels, and mines [2]. FBG sensing has been employed extensively due to its high
level of sustainability, endurance, long-term stability, spatial resolution, and strong
tolerance to electromagnetic fields.
The Internet of Things (IoT) has shown its potential in recent years with its rapid
growth in a number of industries, including smart grids, emergency response systems,
K. Chethana
Department of ECE, Jyothy Institute of Technology Bangalore, Bangalore, India
S. Nandi (B)
Department of AIML, R.V. College of Engineering Bangalore, Bangalore, India
e-mail: someshnandi@rvce.edu.in
A. P. Guruprasad · S. Ashokan
Department of IAP, IISC, Bangalore, India
@seismicisolation
@seismicisolation
470 K. Chethana et al.
home automation, mine tailing monitoring, and mine environment monitoring [3–
5]. Increasing the structural safety of structural structures like building bridges via
means of FBG monitoring, automatic data processing for damage identification, and
remote transmission might be a promising course of action [6].
Similar to this, a distributed sensing technique based on the Internet of Things
(IoT) will offer greater scalability than an offline monitoring system, by doing
so, FBG sensing systems can be included into an IoT framework for improved
signal processing, historical data storage, scalable solutions, and intelligent deci-
sion support additional to the notification creation [7]. Additionally, these systems
provide a complete monitoring, processing, and notification (MPN) strategy where
cross-platform service is not an issue [8]. The biggest benefit is that it connects
a lot of people to the suggested sensor system via smart devices like tablets and
Android/IOS phones [9]. The IoT enabled fiber Bragg grating sensor based real-time
strain monitoring setup is described in this paper, and the electronic sensor setup is
used to validate the sensor data obtained from the FBG sensor. We have also used
the Microsoft Azure cloud gateway for real monitoring, and it allows users to get
real-time data, considerably aiding in the monitoring of civil constructions.
2 Principle of FBG for Strain Measurement
To find the Bragg wavelength’s sensitivity to strain, we start with the basic FBG
equation and Bragg’s law states that the reflected wavelength is according to what
follows in Eq. (1) [10].
λB = 2n eff ⌃ (1)
where neff is the core refractive index and ⌃ is the spatial period of the index modula-
tion, the reflected wavelength shifts from the sensor since there is change in strain in
the grating region and helps to calculate variations in the strain. The basic equation
of the FBG’s sensitivity to strain is the partial derivative with regard to displacement,
and given in Eq. (2) [10].
ΔλB ∂⌃ ∂n eff
= 2n eff + 2⌃ (2)
ΔT ∂T ∂T
ΔλB 1 ∂⌃ 1 ∂n eff
= λB + λB (3)
ΔT ⌃ ∂T n eff ∂ T
The strain on the grating period caused by the fiber’s expansion is characterized
as the first term in Eq. (2). Suppose FBG is inscribed in the fiber of the length L, if
we exert a tension on the fiber of L, we will consequently experience a relative strain
of ΔL/L. The FBG will experience a strain of ΔLFBG/LFBG if it has a length of
LFBG, however, since the FBG is in the fiber, ΔLFBG/LFBG = ΔL/L. Since the
@seismicisolation
@seismicisolation
Internet of Things Enabled Structural Health Monitoring Using Fiber … 471
displacement of the grating period with the same extension equals the displacement
of the Bragg displacement, the first term in the equation is the unit. The photo-elastic
coefficient (ρe), or modification of the index of refraction with light, is the second
term in Equation 2. The Bragg wavelength displacement with strain is a classic
phenomenon that results from the interaction of both occurrences mentioned and the
equation can be deduced to
ΔλB
= (1 − ρe)εz (4)
λB
n ef
ρe = [ρ12 − v(ρ11 + ρ12)] (5)
2
where the strain optic sensor ρ11 and ρ12 has the value of 0.113 and 0.232, the
Poisson ratio (v) has the value of 0.17, neff has value of 1.4545, and εz is the
longitudinal strain of the grating, by further deducing, we can get the equation below.
ΔλB
= (1 − ρe)εz + (α + η)ΔT (6)
λB
3 Modeling of the Sensor
To be able to ensure that the intended structure for strain measurement will work as
expected, the cantilever structure was analyzed and designed in the Ansys software.
There are two stages to the sensor modeling process. Phase one involves designing
the cantilever with the proper dimensions and specifications. To extract the strain
and deformation data from the cantilever beam’s surface, finite element analysis
(FEM) calculations are performed on the planned cantilever. In phase 2, Comsol
multiphysics software (FEM) is utilized to get the mechanical deformation data of
the cantilever. The design parameters of the sensor are length = 50 μm, width =
15 μm, thickness (t) = 220 nm, material = Silicon, Young’s modulus = 130 Gpa,
and Poisson’s ratio = 0.3. The proposed design as shown in Fig. 1.
SOI platform
Fig. 1 Proposed design of the Cantilever Bea
@seismicisolation
@seismicisolation
Fig. 2 Simulation of Cantilever beam
3.1 Stress Distribution of the Surface of the Cantilever Beam
Analysis of the cantilever beam’s stress distribution revealed that the vertical
displacement of the beam changes linearly with applied force, peaking at 23 μm
for an applied force of 1 N. It can be seen in Fig. 2, the cantilever base junction
will experience the greatest amount of stress, as an outcome and the place where the
deformation is greatest also has the greatest change in refractive index.
4 IoT System Architecture
IoT architecture is made up of the hardware, networks, and cloud technologies that
enable interoperability between IoT devices. A basic Internet of Things architecture
comprises three layers namely sensors, gadgets, and other technologies that aid with
perception, Network layer (the interconnection of devices), and the application layer
(The layer with which the user interacts). The fundamental components of data
gathering and monitoring are an FBG interrogator and strain sensors. The perceptive
layer was created with statistical models and analytical tools to discover any anomaly
in the datasets of the strain measured and it represents the convergence of theory and
technology. The application layer is always given top priority in the architecture
@seismicisolation
@seismicisolation
Fig. 3 IoT system architecture
of IoT services since it enables remote information sharing. Microsoft Azure IoT
platform is used for real-time data processing and distant information exchange. The
IoT system architecture is portrayed in Fig. 3.
5 Experimental Setup
The FBG sensor was installed on the cantilever beam and implanted in the stainless-
steel tube with reference to the sensor’s modeling in the Ansys software as depicted
in Fig. 4.
@seismicisolation
@seismicisolation
Fig. 4 Proposed design of the Cantilever Beam
The information provided by the fiber Bragg grating sensor positioned on the
cantilever beam needs to be evaluated, in order to achieve this an electronic or analog
sensor setup was developed utilizing analog strain gauge sensor, coupled to ESP8266.
6 Results
The host structure was put under various thermal loadings. The mechanical strain
from the host structure was switched to the bonded FBG sensor and with the aid of the
FBG interrogator the sensor readings were logged at the interval of 60 s as shown in
Fig. 5, and calibration factor of 0.8168 pm/μm was achieved for strain measurement.
To be able to validate the FBG sensor readings, the strain sensor additionally results
from the analog sensor setup were acquired. Once the sensor data was logged into
the FBG interrogator, it was transferred to the local gateway and from local gateway.
The data was sent to the Microsoft Azure IOT Operating system, which enables a
cloud-hosted solution back end for connecting almost any device and it is extremely
secure and dependable connectivity between sensor and the devices it oversees.
Using this platform, the real-time strain sensor data was obtained and the same is
depicted in Fig. 6.
@seismicisolation
@seismicisolation
Fig. 5 Strain measurement of strain Gauge Sensor and FBG strain sensor
Fig. 6 Real-time sensor data
@seismicisolation
@seismicisolation
7 Conclusion
This paper describes a real-time strain monitoring configuration based on a IoT

enabled fiber Bragg grating sensor. The electronic sensor setup is utilized to evaluate
the sensor data from the FBG sensor and the calibration factor of 0.8168 pm/μm is
achieved for strain measurement and using the Microsoft Azure cloud gateway, the
real-time strain data is measured and the same is described in the paper.
References
1. Her SC, Lin WN (2020) Simultaneous measurement of temperature and mechanical strain
using a fiber bragg grating sensor. Sens (Basel) 20(15):4223. https://doi.org/10.3390/s20154
223.PMID:32751291;PMCID:PMC7435463
2. Mansoursamaei M, Malakzadeh A (2021) Simultaneous measurement of temperature and strain
using a single fiber bragg grating on a tilted cantilever beam. Opt Rev 28:289–294. https://doi.
org/10.1007/s10043-021-00660-w
3. Malakzadeh A, Mansoursamaei M, Pashaie R et al (2021) Simultaneous measurement of
temperature and strain based on peak power changes and wavelength shift using only one
uniform fiber bragg grating, 15 Mar 2021, PREPRINT (Version 1) available at Research Square
[https://doi.org/10.21203/rs.3.rs-287636/v1]
4. Majumder M, Gangopadhyay TK, Chakraborty AK, Dasgupta K, Bhattacharya DK (2008)
Fibre Bragg gratings in structural health monitoring—Present status and applications, sensors
and actuators a: physical. 147(12008):150–164. ISSN 0924-4247, https://doi.org/10.1016/j.
sna.2008.04.008
5. Hill KO, Meltz G (1997) Fiber Bragg grating technology fundamentals and overview. Light
Wave Technol 15(8)
6. Nishihara H, Haruna M, Suhara T (1985) Optical integrated circuits. McGraw-Hill, New York
7. Grattan KTV, Sun T (2000) Fiber optic sensor technology and overview. Sens Actuators A
Phys 82:40–61
8. Othonos A, Kalli K (1999) Fiber bragg gratings: fundamentals and applications in telecommu-
nications and sensing. Artech House
9. Nandi S, Indumathi TS, Sai Priya RV (2019) Aparna kori: analysis of fiber bragg grating
spectral characteristics using couple mode theory for sensor applications. IMIS 2019: 868-879
10. Miller SE (1969) Integrated optics: an introduction. Bell Syst Technical J 48(7):1538–7305
@seismicisolation
@seismicisolation
Seismic Analysis of Five-Storied Building
Using U-Shaped Hybrid Isolator
and Lead Rubber Bearing Isolator
Srushti Gaikwad, Priti R. Satarkar, and A. Manchalwar
1 Introduction
One of the most common techniques for shielding a building from earthquake effects
is base isolation. It is one of the most effective seismic engineering tools for passive
structural vibration reduction. A set of bearing pads that are positioned between the
buildings and the building foundation sustains an isolated base structure. With the aid
of a building resting on frictionless rollers, the idea of base isolation is shown. The
rollers freely roll when the ground trembles, while the building above remains still.
As a result, the ground shaking has no effect on the building; the structure is simply
not affected by the earthquake. Now, part of the effects of the ground trembling will
be transferred to the building if it is supported by flexible pads that provide resistance
to lateral motions. If the flexible pads are chosen effectively, the forces caused by
ground shaking may be several times lower than those felt by a fixed base building,
which is a structure that is erected directly on the ground. The flexible pads are known
as base isolators, and the buildings they protect are known as base-isolated buildings.
The introduction of flexibility into the structure is the base isolation technology’s key
feature. As a result, a sturdy medium-rise reinforced concrete or masonry building
gains a high degree of flexibility. The procedure for foundation isolation in an RCC
construction is depicted in the image below. The fundamental idea underlying base
isolation is that the structure’s or building’s response is altered so that the ground
below can move with little to no motion being transmitted to the structure above.
Only in an ideal system is a complete separation feasible. To transfer vertical loads to
the base in a real-world setting, a vertical support is required. Since the acceleration
S. Gaikwad (B)
AISSMS College of Engineering, Pune, Maharashtra, India
e-mail: srushtijgaikwad99@gmail.com
P. R. Satarkar · A. Manchalwar
Deparment of Civil Engineering, AISSMS College of Engineering, Pune, Maharashtra, India
@seismicisolation
@seismicisolation
478 S. Gaikwad et al.
introduced in the structure is the same as that of ground motion, the relative displace-
ment of the ground and the structure is zero for a fully rigid, zero-period structure.
In contrast, a perfect flexible structure would not experience any acceleration; as
a result, the relative displacement of the structure would be equal to the displace-
ment of the ground. Since no construction is perfectly rigid or flexible; its response
will fall somewhere in the middle of the two described above. For periods between
zero and infinity, the maximum acceleration and displacements are a function of the
earthquake. Although relative displacements may not be greater than peak ground
displacements, there will be a range of times during earthquakes when acceleration
in the building will be enhanced beyond maximum ground acceleration. The best
solution for this is base isolation, which reduces the transfer of motion and limits
building displacement. For fixed base structures, displacement happens around the
centre of gravity (CG), which will be around two-thirds of the height for build-
ings, and at the isolation plane for base-isolated structures, with less displacement
inside the structure. Mazza and Vulcano [1] discovered that increasing pulse period
values were related to growing isolator displacement. Controlling the displacement
of the isolator can be done by adding viscous damping at the base. Torunbalci and
Ozpalanlar [2] examined the values for storey shear, relative drift, and acceleration
and concluded that this value is sufficient for totally removing the structure from
the earthquake’s resonant range. When the storey accelerations are examined, it can
be seen that, in the fixed base scenario, moving up to higher storeys results in high
acceleration values. Up to 1.0 g level, in the base isolator scenario, the accelera-
tion values on the basis of storeys do not change that much and remain fixed at
rather low values. Balamonica et al. [3] discovered that, for the same soil conditions
and time duration, a rise in damping percentage causes a decrease in the isolator’s
thickness. If the needed time period and damping are the same, buildings located in
soft soil conditions need thicker isolators than those in hard soil conditions. When
using an isolator with rubber that has a lower shear modulus for a specific value
of axial load that needs to be transferred, the diameter of the isolator rises. If the
soil quality and the rubber being utilized are both the same, the isolator’s diameter
reduces as the intended value of the time period grows. Shirule et al. [4] demonstrated
that base isolation using lead rubber bearings and friction bearings can significantly
lessen the structure’s reaction. As a result, lead rubber bearings are more effective at
reducing base shear and displacements, making them better suited for multi-storey
buildings located in higher zones. In base-isolated buildings, displacements increase
over time and as storey height increases. Deosarkar and Gowardhan [5] designed
6-storey RCC bases that are fixed and isolated. They use nonlinear time history anal-
ysis to look at the response of the combined isolation system used to isolate the
building, which consists of a friction pendulum system at the foot of the column,
a lead rubber bearing, and high-damping rubber bearings. They discovered that all
base-separated building models exhibit an increase in lateral displacement at the base,
in contrast to fixed base buildings, which have zero displacement at the base of the
building. Murota et al. [6] researchers in Turkey used numerical and experimental
methods to study the suitability of high-damping rubber bearings for the seismic
isolation of residential buildings. The recently updated Turkish Building Seismic
@seismicisolation
@seismicisolation
Seismic Analysis of Five-Storied Building Using U-Shaped Hybrid … 479
Code-2018 (TBSC2018) guided the design of a seismic isolation system made up

of high-damping rubber bearings. An actual building database was used to choose
three model buildings of varying heights in an anticipated seismically active region.
These models served as the basis for time histories and equivalent lateral force assess-
ments. Bagheri et al. [7] examined rotational friction dampers, sometimes referred
to as Friction Damper Devices (FDD), and their performance in comparison with
U-shaped metallic-yielding dampers in steel structure frames. They found that by
including dampers in both systems, nonlinear behaviour was transferred from frame
components to the damper system, reducing or eliminating structural damage. The
numerical findings show that adding dampers reduces the structure’s maximum base
shear and the roof’s lateral displacements by 6% to 40% and 13% to 57%, respec-
tively, when using U-shaped damper and 15% to 41% and 9% to 53% when using
the FDD.
According to the findings mentioned above, lead rubber bearings are primarily
used in isolation systems. The behaviour of a 5-storey RC structure under various
earthquakes is investigated in the current study, both with and without an isolated
base. A fresh attempt has been made for this goal, using a U-shaped steel damper
for the isolation system and comparing it to a lead rubber bearing system. In the
authors’ prior study of U-shape dampers, their effectiveness was only evaluated
experimentally. Four separate earthquakes are being used for nonlinear time history
analysis.
1.1 U-Shaped Damper
The U-shaped damper is a brand-new steel damper constructed from high-quality

rolled steel. This minimizes movement during earthquakes by absorbing seismic
energy and utilizing balanced recovery characteristics. Nearly same hysteretic
behaviour is displayed by the base isolation U-shaped damper in all directions (360o).
In comparison to standard base isolation steel rod dampers, it is more affordable and
has a higher level of shear yield strength. After an earthquake, the extent of the damage
can be determined visually. In the odd case that it becomes essential, the damper part
can also be changed after the earthquake activity [8]. Experimental research on U-
shape yielding dampers showed that this system has a high-energy absorbability and
the capacity for substantial deformation. This system’s energy dissipation is based
on the plate’s camber section’s plastic deformation during the bending moment as a
result of deformation brought on by cyclic loading [9]. Fig. 1 depicts the arrange-
ment of a U-shaped steel damper, which includes four arms and low damping rubber
bearing. The U-shaped isolator UD40 has nonlinear characteristics including yield
force of 112 kN, initial stiffness of 400 kN/m, post stiffness of 100 kN/m, and yield
deformation of 0.0189 m.
@seismicisolation
@seismicisolation
Fig. 1 U-shaped steel damper [10]
1.2 Lead Rubber Bearing
In 1975, New Zealand became the first country to use lead rubber bearings (LRB),
which were later widely adopted by Japan and the United States [11]. A laminated
rubber and steel bearing with steel flange plates for installation to the structure makes
up lead core rubber bearings (LRB). Our isolators have a lead core that dissipates
energy in 90% of cases. The isolator’s rubber acts as a spring. Although quite rigid
vertically, it is very soft laterally. Thin rubber layers strengthened by steel shims are
used to create the high vertical stiffness. The isolator can move laterally with relatively
little stiffness thanks to these two features. When the isolator is cut to exacting
specifications by laser, the lead core provides damping by deforming plastically.
Computer-controlled milling machines that consistently and accurately deliver high
production rates are used to produce the steel mounting plates [12].
2 Numerical Study
In this work, a five-storey reinforced concrete building is taken into account. In X

and Y directions, respectively, there are two bays of 4 m and four bays of 3 m.
Consideration is given to a 3 m constant floor height. There is a structure in seismic
zone V. It is decided to use a reinforced concrete slab that is 120 mm thick with an
infill that is 230 mm thick. All of the columns and beams are 600 × 600 mm [13,
14]. The software determines the dead load, and the live load is taken into account
at 3 kN/m2. Concrete has a 30 MPa compressive strength, while the yield strength
of reinforcing steel is 415 MPa. SAP 2000 is used to model and study a three-
dimensional building with and without an isolator [15]. The study takes into account
four distinct earthquakes, as illustrated in Table 1. Calculations of the characteristics
of LRB were done using Q/W = 0.1 and T-2 s.
@seismicisolation
@seismicisolation
Table 1 Earthquake data

Earthquake ID PGA (g)
considered for study
Hector Mine, USA (1999) 120,522 0.34
Imperial Valley, Mexico (1979) 120,612 0.35
Kocaeli, Turkey (1999) 120,812 0.36
Superstition Hills, USA (1987) 121,211 0.36
2.1 Dynamic Response of Superstructure
A structure’s acceleration reveals how much force is being applied to it as a result

of the motion of an earthquake. Figure 3 illustrates the top acceleration of the super-
structure for four different earthquakes and contrasts the fixed foundation with the
solitary building. Both isolation technologies significantly lessen the superstructure’s
top-floor acceleration. The U-shaped damper can therefore also serve as an isolator.
When subjected to ground motion, the joints no longer accelerate as quickly as they
would have under fixed base conditions because of the use of the LRB isolator. Less
energy is transmitted into the superstructure because as a result of the high-energy
seismic waves, the base isolators at the base of the structure deform themselves.
Therefore, as can be seen from Figs. 2 and 3 for 5-storey buildings, the amount of
storey drift decreases. The U-shaped isolator instance exhibits a similar pattern, with
the U-shaped isolated building having less flexibility than the LRB isolated building.
120522
1.5
1
Acceleration (m/s2)
0.5
0
0 10 20 30 40 50
-0.5
-1
-1.5
Time (s)
Fixed LRB
Fig. 2 Comparison of acceleration with time period for fixed base building and LRB for different
earthquake cases
@seismicisolation
@seismicisolation
120612
1.5
Acceleration (m/s2) 1
0.5
0
-0.5 0 20 40 60 80 100 120
-1
-1.5
Time (s)
Fixed LRB
120812
1
Acceleration (m/s2)
0.5
0
0 10 20 30 40 50 60
-0.5
-1
Time (s)
Fixed LRB
121211
1
0.5
Acceleration (m/s2)
0
0 20 40 60 80 100
-0.5
-1
Time (s)
Fixed LRB
Fig. 2 (continued)
@seismicisolation
@seismicisolation
120522
1.5
Acceleration (m/s2) 1
0.5
0
0 10 20 30 40 50
-0.5
-1
-1.5
Time (s)
Fixed U-shaped
120612
1.5
1
Accelerance (m/s2)
0.5
0
0 20 40 60 80 100 120
-0.5
-1
-1.5
Time (S)
Fixed U-shaped
120812
1
0.5
Acceleration (m/s2)
0
0 10 20 30 40 50 60
-0.5
-1
Time (s)
Fixed U-shaped
Fig. 3 Comparison of acceleration with time period for fixed base building and U-shaped isolator
for different earthquake cases
@seismicisolation
@seismicisolation
121211
1
Acceleration (m/s2) 0.5
0
0 20 40 60 80 100
-0.5
-1
Time (s)
Fixed U-shaped
Fig. 3 (continued)
3 Conclusion
The findings of the current analytical work indicate a tendency between the isolated
foundation conditions and fixed base conditions. Since they lessen the overall reaction
of the structure to any seismic activity, LRB and U-shaped isolators have been used
successfully to isolate foundations. This study definitely highlights the U-shaped
damper’s effectiveness as an isolator and as a potential future tool. We get to the
conclusion that the U-shaped isolator has a higher energy dissipation capability than
the lead rubber isolator. The reduction in top-floor acceleration of the superstructure
caused by the U-shaped isolator is comparable to the LRB. There is a significant
decrease in acceleration as a result of the U-shaped isolator. Thus, one of the best
seismic base isolation techniques for minimizing the seismic impact on a structure
is this novel technology.
References
1. Mazza F, Mazza M, Vulcano A (2018) Base-isolation systems for the seismic retrofitting of
RC framed buildings with soft-storey subjected to near-fault earthquakes. Soil Dyn Earthq Eng
109:209–221
2. Torunbalci N, Ozpalanlar G (2008) Earthquake response analysis of mid-story buildings
isolated with various seismic isolation techniques. In: The 14th world conference on earthquake
engineering, Beijing, China
3. Keerthana S, Kumar KS, Balamonica K, Jagannathan DS (2014) Seismic response control
using base isolation strategy. Int J Emerg Technol Adv Eng. ISSN, 2250–2459
4. Shirule PA, Jagtap LP, Sonawane KR, Patil TD, Jadwanir N, Sonar SK (2012) Time history
analysis of base isolated multi-storyed building 5:809–816
@seismicisolation
@seismicisolation
5. Deosarkar MU, Gowardhan SD (2015) Non linear dynamic response of combined isolation
system on symmetric and asymmetric buildings. IJIFR 3(3):1021–1035, Nov 2015, p-ISSN:
2347-1697
6. Murota N, Suzuki S, Mori T, Wakishima K, Sadan B, Tuzun C, Sutcu F, Erdik M (2021)
Performance of high-damping rubber bearings for seismic isolation of residential buildings in
Turkey. Soil Dyn Earthq Eng 143:106620
7. Bagheri S, Barghian M, Saieri F, Farzinfar A (2015) U-shaped metallic-yielding damper
in building structures: seismic behavior and comparison with a friction damper. Structures
(3):163–171. Elsevier
8. Sravya GJS, Manchalwar A (2019) Seismic isolation system using U-shaped steel damper. Int
J Recent Technol Eng 8:12336–12339
9. Ebadi Jamkhaneh M, Ebrahimi AH, Shokri Amiri M (2019) Experimental and numerical
investigation of steel moment resisting frame with U-shaped metallic yielding damper. Int J
Steel Struct 19:806–818
10. Manchalwar A, Bakre SV (2020) Seismic performance of structure with isolated foundation
using u-shape steel damper as an isolator. Soil Mech Found Eng 57:170–177
11. Su L, Ahmadi G, Tadjbakhsh IG (1991) Performance of sliding resilient-friction base-isolation
system. J Struct Eng 117(1):165–181
12. Mishra P, Awchat GD (2017) Lead rubber bearings as base isolating devices for the construction
of earthquake resistant structures-a review. SSRG Int J Civ Eng (SSRG-IJCE) 4(7):18–20
13. IS 456 (2000) Plain and reinforced concrete-code of practice (fourth revision). BIS, New Delhi,
India
14. IS 13920 (2016) Ductile detailing of reinforced concrete structures subjected to seismic forces-
code of practice. BIS, New Delhi, India
15. SAP2000, Integrated software for structural analysis & design, technical reference manual.
Computers & Structures, Inc.
@seismicisolation
@seismicisolation
Structural Health Monitoring of Bridges
Using IoT
Deepak V. Ingale , K. Chethana , Gowthami P. Jain , S. Aditya ,

and V. Venkatesh
1 Introduction
Bridges are critical components of transportation infrastructure, and their structural

integrity is crucial for public safety [1]. Bridges that are not adequately maintained
or monitored may experience progressive deterioration, which can lead to sudden
failures or collapses. Such incidents can result in severe injuries, loss of life, and
property damage and compromise the integrity of the transportation infrastructure.
Also, insufficient attention to structural management can lead to higher maintenance
costs in the long run. By identifying these issues promptly, appropriate measures
can be taken to ensure the safety of bridge users and prevent catastrophic failures
[2]. Traditional bridge maintenance practices often rely on periodic inspections or
repairs based on fixed schedules, which may not effectively capture the real-time
condition of the structure [3]. SHM provides a proactive approach by offering real-
time data on the structural health of bridges [4]. This allows engineers and bridge
managers to optimize maintenance planning by addressing specific areas of concern
and allocating resources efficiently, thereby maximizing the lifespan of bridges and
minimizing maintenance costs. SHM enables engineers to assess the performance
of bridges under different loading conditions, such as heavy traffic, seismic events,
or extreme weather. By collecting data on parameters such as strain, displacement,
vibration, and environmental factors, researchers can analyze the bridge’s response
to these conditions and gain insights for design improvements, operational changes,
and structural optimization. Several methods and technologies have been employed
for structural health monitoring (SHM) of bridges [5–8]. Here are some commonly
D. V. Ingale (B)
CIIRC, Jyothy Institute of Technology, Thataguni, 82, Bengaluru, Karnataka, India
e-mail: deepak.vi@ciirc.jyothyit.ac.in
K. Chethana · G. P. Jain · S. Aditya · V. Venkatesh
Electronics and Communication Engineering, Jyothy Institute of Technology, Thataguni,
Bengaluru, Karnataka 82, India
@seismicisolation
@seismicisolation
488 D. V. Ingale et al.
used techniques: Wireless Sensor Networks (WSNs), Acoustic Emission, Vibration-

Based Monitoring, Remote Sensing Technologies, Sensors and Instrumentation,
Non-Destructive Testing (NDT), etc. Wireless Sensor Networks (WSNs) offer [9,
10], and [11] several advantages when compared to other methods and technologies
used in structural health monitoring (SHM) of bridges. Here are some key reasons
why WSNs are considered advantageous: Real-time monitoring, scalability and flex-
ibility, cost-effectiveness, data accuracy and reliability, easy installation and main-
tenance, long-term monitoring, integration with data analysis tools, remote accessi-
bility, etc. IoT (Internet of Things) technology and wireless sensor networks (WSNs)
are closely related and often used in conjunction with each other to enable various
applications. IoT relies on network connectivity to enable communication and data
exchange between devices and systems. Wireless sensor networks provide a wireless
infrastructure that allows IoT devices to connect and communicate with each other
and with the broader Internet. WSNs serve as a key component in establishing the
connectivity layer for IoT applications. Further data collected by the IoT is trans-
ferred to the cloud through the ThingSpeak platform which can be used for analysis
of the condition of the bridge, anywhere anytime. The reviewed research papers [12–
16] shed light on the advancements and applications of IoT-based data application
systems in the field of structural health monitoring, with a specific focus on bridge
monitoring. These papers explore various topics such as sensor deployment strate-
gies, data acquisition techniques, wireless communication protocols, data processing
algorithms, anomaly detection methods, and cloud-based data storage and analysis
platforms. They emphasize the importance of integrating IoT devices, such as the
NodeMCU, with sensor technologies like the strain gauge module and temperature
sensors to enable real-time monitoring, remote access, and data-driven decision-
making in bridge SHM. In this paper, we present an overview of the research papers
related to IoT-based data application systems for structural health monitoring of
bridges, with a specific emphasis on the integration of the strain gauge module and
temperature sensor with the NodeMCU. We discuss [17, 18] the methodologies, find-
ings, and contributions of these papers, highlighting their implications for enhancing
the effectiveness and efficiency of bridge SHM. Furthermore, we analyze [19, 20]
the challenges and opportunities associated with this integration, considering factors
such as sensor placement, data accuracy, wireless communication reliability, and
cloud-based data storage and analysis. By leveraging the advantages of IoT technolo-
gies, including real-time monitoring, continuous data acquisition, and cloud-based
analysis, the proposed IoT-based data application system holds great potential for
improving the safety and performance of bridges.
2 Methodology
IoT-based data application system is implemented for structural health monitoring

of bridges by integrating the BF350-3AA strain gauge module and a temperature
sensor with the NodeMCU ESP8266 development board as indicated by the block
@seismicisolation
@seismicisolation
Structural Health Monitoring of Bridges Using IoT 489
Fig. 1 Methodology used

for the structural health
monitoring of bridges
diagram shown in Fig. 1. The BF350-3AA strain gauge module provides accurate
measurements of strain or deformation, while the temperature sensor ensures precise
temperature readings. The NodeMCU ESP8266 serves as the central processing
unit, facilitating data acquisition, processing, and wireless communication. Further-
more, the collected data is uploaded to the ThingSpeak cloud platform for real-time
visualization and analysis.
2.1 Hardware
To interface the NodeMCU with the strain gauge BF350-3AA and the NodeMCU
to read data from the sensors and perform the desired actions. The BF350-3AA
is a type of strain gauge that measures strain or pressure changes. It typically has
@seismicisolation
@seismicisolation
four wires: excitation (+V), ground (GND), signal (+S), and shield (SH). You will
need to connect the excitation and ground wires to appropriate power and ground
sources on the Node MCU and connect the signal wire to an analog input pin. The
DS18B20 is a digital temperature sensor that uses the One Wire protocol. It provides
accurate temperature readings with a resolution of up to 12 bits. It has three wires:
power (+V), ground (GND), and data (DQ). You will need to connect the power and
ground wires to the Node MCU’s power and ground sources, respectively. The data
wire needs to be connected to a digital input/output pin that supports the One Wire
protocol to write the code to read data from both sensors. For the strain gauge, you
will read the analog value from the appropriate pin and convert it into meaningful
units based on the calibration data for the BF350-3AA. For the temperature sensor,
you will use the One Wire library to communicate with the DS18B20 and retrieve the
temperature readings. Once you have the sensor data, you can transmit it to a remote
server or perform any other actions based on your requirements. The NodeMCU
Wi-Fi capabilities enable you to send the data over the internet to a cloud service or
display it on a web page. Temperature sensor DS18B20, you will need to follow a
few steps, as illustrated in Fig. 2.
The NodeMCU is an open-source development board that uses the ESP8266 Wi-
Fi module. It allows you to connect to the internet and control various electronic
components. You will need to program the NodeMCU to read data from the sensors
and perform the desired actions. The BF350-3AA is a type of strain gauge that
measures strain or pressure changes. It typically has four wires: excitation (+V),
ground (GND), signal (+S), and shield (SH). You will need to connect the excitation
and ground wires to appropriate power and ground sources on the Node MCU and
connect the signal wire to an analog input pin. The DS18B20 is a digital temperature
sensor that uses the One Wire protocol. It provides accurate temperature readings
with a resolution of up to 12 bits. It has three wires: power (+V), ground (GND), and
data (DQ). You will need to connect the power and ground wires to the Node MCU’s
Fig. 2 Block diagram
@seismicisolation
@seismicisolation
power and ground sources, respectively. The data wire needs to be connected to a
digital input/output pin that supports the One Wire protocol to write the code to read
data from both sensors. For the strain gauge, you will read the analog value from
the appropriate pin and convert it into meaningful units based on the calibration data
for the BF350-3AA. For the temperature sensor, you will use the One Wire library
to communicate with the DS18B20 and retrieve the temperature readings. Once you
have the sensor data, you can transmit it to a remote server or perform any other
actions based on your requirements. The NodeMCU Wi-Fi capabilities enable you
to send the data over the internet to a cloud service or display it on a web page.
2.2 Program Description
As indicated in Fig. 3, to establish a network connection between an ESP8266 device

and another device and to allow the user to view the network’s SSID and password Wi-
Fi connection established or not if the connection has already been made, the status
will read “Wi-Fi connected” otherwise it will indicate that the network connection is
pending. It will print the statistics following connection establishment. We are estab-
lishing a connection with the ThingSpeak client–server and plotting the measured
readings on ThingSpeak using Wi-Fi and reading the strain gauge sensor values and
temperature sensor values reading data on a serial monitor. The loop then repeats,
continuously measuring the temperature and strain gauge values, printing them to
the serial monitor, and sending the data to ThingSpeak for storage and visualization.
Data Acquisition: The sensor data is collected using data acquisition systems, which
convert the analog signals from the sensors into digital form for further analysis. The
acquired data is typically stored in databases or transmitted wirelessly to a central
monitoring system.
Data Analysis: Analyzing the collected data is a crucial step in SHM. Advanced
signal processing and data analysis techniques are used to extract meaningful infor-
mation and detect any deviations or anomalies from the expected behavior of the
structure. This may involve the use of statistical methods, pattern recognition algo-
rithms, machine learning, and artificial intelligence techniques. One of the primary
objectives of SHM is to detect the presence and location of damage or deterioration
in a structure. This can be achieved by comparing the measured responses of the
structure with a reference model or baseline data. Any deviations from the expected
response can indicate the presence of damage.
Diagnosis and Prognosis: Once the damage is detected, the next step is to diag-
nose its severity, extent, and possible causes. This may involve further analysis and
interpretation of the sensor data. Prognosis refers to predicting the future behavior of
the structure based on the current condition and known damage progression models.
The information obtained from SHM is used to make informed decisions regarding
maintenance, repair, and operational strategies.
@seismicisolation
@seismicisolation
Fig. 3 Flowchart
@seismicisolation
@seismicisolation
Fig. 4 Experimental setup of structural health monitoring of bridges using IoT
Application: Structural Health Monitoring (SHM) of Bridges using strain gauge

sensors and temperature sensors is a cutting-edge application in civil engineering.
This innovative system involves the integration of IoT technology to enable real-
time monitoring and analysis of the structural integrity and performance of bridges.
The SHM system utilizes strain gauge sensors to measure the structural strain expe-
rienced by the bridge under various loads. By strategically placing multiple strain
gauge sensors at critical points along the bridge’s span, the system conducts strain
determination at different points of the bridge. This data provides valuable insights
into how forces are distributed and transmitted throughout the structure, allowing
engineers to identify potential stress concentrations and areas prone to fatigue.
Additionally, the system incorporates temperature sensors to monitor tempera-
ture variations on the bridge as shown in Fig. 4. Temperature changes can influ-
ence the behavior of the bridge, affecting its expansion and contraction. By continu-
ously measuring temperature data, the system can account for thermal effects on the
bridge’s structural health, ensuring accurate and reliable assessments. Moreover, the
SHM system enables sway analysis, made possible through the strain gauge sensors.
Sway analysis involves monitoring the lateral movement and stability of the bridge
under different load conditions and environmental factors. This analysis helps engi-
neers understand the dynamic behavior of the bridge, allowing for better evaluation
of its safety and performance. The real-time data collected by the strain gauge sensors
and temperature sensors is wirelessly transmitted to a central data processing unit
through the IoT infrastructure. This data is then analyzed and presented using data
analytics and visualization tools. Engineers can access critical information remotely,
facilitating informed decision-making and proactive maintenance strategies.
This paper successfully developed a structural health monitoring system as we can

interface the strain gauge sensor and temperature sensor with the ESP8266, The
ESP8266 collects data from these sensors and performs the following actions.
@seismicisolation
@seismicisolation
Serial Monitor: The collected strain and temperature data is displayed on the Serial
Monitor in real-time. This provides a convenient way to monitor the data locally and
ensure proper sensor functioning.
ThingSpeak: The ESP8266 establishes a connection to the ThingSpeak platform
using the provided API key. The strain and temperature data is sent to ThingSpeak,
where it is stored and made accessible for further analysis. ThingSpeak provides
features like data visualization, alerts, and data export options.
ExcelSheet: The ESP8266 establishes a serial communication link with the
ExcelSheet using the Software Serial library. The strain and temperature data is
formatted and transmitted to the Excel sheet, where it can be further processed,
analyzed, and visualized using spreadsheet functions and formulas.
By utilizing these three outputs—Serial Monitor, ThingSpeak, and ExcelSheet—
we can effectively monitor, store, and analyze the strain and temperature data from
the sensors. This enables us to gain insights into the behavior of the monitored object
and make informed decisions based on the collected data as shown in Figs. 5 and 6.
Fig. 5 Result: Real-time Temperature data plotted on ThingSpeak platform
@seismicisolation
@seismicisolation
Fig. 6 Result: Real-time strain data plotted on ThingSpeak platform
4 Conclusion
The development of an IoT-based data application system for structural health moni-
toring of bridges by integrating the strain gauge module and a temperature sensor with
the NodeMCU development board holds significant promise in the field of structural
engineering. By leveraging the power of the Internet of Things (IoT), this system
enables real-time monitoring and analysis of structural health, providing valuable
insights into the structural integrity and performance of bridges.
Sway analysis, enabled by the strain gauge module, allows for monitoring the
lateral movement and stability of the bridge during different load conditions, offering
critical information for assessing its dynamic behavior. The system’s ability to
conduct sway analysis enhances the understanding of the bridge’s response to external
forces, enabling engineers to make more informed decisions regarding its safety and
performance.
Additionally, the integration of the strain gauge module with the NodeMCU
development board facilitates strain determination at different points of the bridge.
By strategically placing multiple sensors across the bridge’s span, engineers can
obtain a comprehensive understanding of how forces are distributed and trans-
mitted throughout the structure. This valuable data aids in identifying potential
stress concentrations and areas prone to fatigue, guiding targeted maintenance and
reinforcement efforts.
However, it is important to acknowledge that the successful implementation of this
system requires careful consideration of factors such as sensor placement, data trans-
mission security, and power management. Properly positioned sensors are crucial for
accurate data collection, while data transmission security ensures the protection of
sensitive information related to the bridge’s structural health.
@seismicisolation
@seismicisolation
Additionally, ongoing maintenance and calibration of the sensors are necessary

to ensure the accuracy and reliability of the collected data over the bridge’s lifespan.
Regular sensor maintenance and calibration procedures are essential to guarantee the
long-term effectiveness of the IoT-based structural health monitoring system.
In conclusion, the development of an IoT-based data application system for struc-
tural health monitoring of bridges, incorporating sway analysis and strain determina-
tion, represents a significant advancement in civil engineering practices. This system
empowers engineers and infrastructure managers to proactively monitor, analyze, and
manage the health of bridge infrastructure, contributing to the development of sustain-
able and resilient transportation networks for the future. By continuously capturing
strain data and temperature variations, conducting sway analysis, and ensuring real-
time monitoring, the IoT-based solution plays a crucial role in enhancing bridge
safety, optimizing resource allocation, and supporting informed decision-making for
maintaining critical bridge infrastructures efficiently and effectively.
References
1. Anwar N, Alnuaimi A, Alhassan M, He X (2020) A review of bridge health monitoring systems

for structural safety assessment
2. Sadek AW, Zhang J, Dong Q (2012) Effects of inadequate maintenance on bridge performance
and safety
3. Toro FX, Aly AM, Abdelkarim OI (2015) Traditional bridge maintenance: challenges,
opportunities, and best practices
4. Fujino Y, Abe M, Takeda T (1996) A review of structural health monitoring techniques for
civil structures
5. Reynolds TP, Veeraragavan PS, Fuhr PL (2016) A review of fiber optic sensing technologies
for structural health monitoring of civil infrastructure
6. Jiang Y, Wang X, Han Y (2015) Wireless sensor networks for structural health monitoring: a
comprehensive review
7. Xu B, Gu J, Chen W (2016) Structural health monitoring using smart sensors and intelligent
systems: a review
8. Zhu H, Zhang J, Lu Y (2016) Review of modal-based methods for structural health monitoring
9. Paek J, Chintalapudi K, Govindan R, Caffrey J, Masri S (2015). A wireless sensor network for
structural health monitoring: performance and experience. IEEE, Publisher
10. Abdulkarem M, Samsudin K, Rasid MFA. Wireless sensor network for structural health
monitoring: a contemporary review of technologies, challenges, and future direction 19(3)
11. Li J, Xu J, Guan X (2015) Title: wireless sensor networks for structural health monitoring: a
comparative study. Published: 2015
12. Chen J, Jiang J, Xiang Y (2017) Structural health monitoring of bridges: review and future
directions. J Sound Vib 400:556–575
13. Ehsani MR, Salamone SS, Reinhorn AM (2006) Bridge health monitoring: technologies and
applications. J Bridg Eng 11(4):423–431
14. Xu Z, Ye L, Li J (2016) Bridge health monitoring by using wireless sensor networks: a
comprehensive review. KSCE J Civ Eng 20(2):485–495
15. Samaan N, Gindy S, Elhattab A (2017) A review of non-destructive testing methods and their
applications in bridge health monitoring. Struct Infrastruct Eng 13(5):565–576
16. Xie S, Li L, Zhang L (2019) Structural health monitoring of bridges: a comprehensive review.
Struct Health Monit 18(2):318–336
@seismicisolation
@seismicisolation
17. Kang JS, Kim JH, Kim SH (2018) A wireless structural health monitoring system using a strain
gauge and temperature sensor for real-time monitoring. Adv Struct Eng 21(12):1813–1827
18. Tran HN, Li Y, Wang J (2020) Structural health monitoring system based on IoT and cloud
computing. J Perform Constr Facil 34(1):04019118
19. Li H, Zhou J, Du X, Li J (2016) Challenges and opportunities in bridge structural health
monitoring using wireless sensor networks. Sensors 16(5):662
20. Liu X, Law SS, Zhang Z (2018) Integration of structural health monitoring and maintenance
management for sustainable and resilient civil infrastructure. Struct Infrastruct Eng 14(7):904–
923
@seismicisolation
@seismicisolation

Civil_Engineering_for_Multi_Hazard_Risk_Reduction_Select_Proceedings

Uploaded by

Copyright:

Available Formats

Civil_Engineering_for_Multi_Hazard_Risk_Reduction_Select_Proceedings

Uploaded by

Document Information

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

Civil_Engineering_for_Multi_Hazard_Risk_Reduction_Select_Proceedings

Uploaded by

Copyright:

Available Formats

Lecture Notes in Civil Engineering

To submit a proposal or request further information, please contact the appropriate

N. Vinod Chandra Menon

ISSN 2366-2557 ISSN 2366-2565 (electronic)

Paper in this product is recyclable.

The field of civil engineering stands as a cornerstone of human progress, shaping

Bengaluru, India Dr. K. S. Sreekeshava

Civil Engineering for Multi-hazard Risk

Accessibility and Convenience

Human Activity Recognition in Construction Industry Using

Risk Analysis and Approaches

Machine Learning Algorithms for Classifying Land Use and Land

Seismic Study on Step Back Buildings and Step Back Setback

Dr. K. S. Sreekeshava is currently working as Associate Professor and Head of

K. S. Sreekeshava , Sreevalsa Kolathayar , N. Vinod Chandra Menon ,

K. S. Sreekeshava (B) · Bhargavi.C

2 Accessibility and Convenience

Bamboo as an alternative construction material when pre-treated with certain

Subsurface characteristics must be concentrated upon, when we are considering the

4 Risk Analysis and Approaches

A notable contribution to this volume is by Rajib et al. study titled Investigating

Greenhouse monitoring to maintain optimal conditions for growth of plants using

The aforesaid researches and applications in the domain of Civil Engineering

17. Bagheri S, Barghian M, Saieri F, Farzinfar A (2015) U-shaped metallic-yielding damper in

Sumana Jayaprakash and Vimala Swamy

A walkable locality is characterized by the availability of most amenities required for

2 The Study Area: Hassan

expressions of this transformation in Hassan. The distance of availability of ameni-

3.1 Primary Data Collection of Hassan City

3.2 Study and Analysis of the Present Conditions

Fig. 1 Methodology (source author’s work)

Documentation of the various amenities available in the neighborhood: After the

3.3 Data Analysis

Evaluation of the Neighborhood Localities on the Overlaid 15-min City Map.

3.4 Proposal of Neighborhood Structure Plan for Hassan

Table 1 Proposed parameters of city (HUDA and author work)

• Innovative financial and regulatory mechanisms to enhance the public–private

4 Discussions and Conclusions

Rajashri Khanai, Rajkumar Raikar, and Mrutyunjay Uppinmath

R. Khanai (B) · R. Raikar · M. Uppinmath

4 Results and Discussions

Table 1 Noise pollution data (values in dB)

4.1 Linear Regression

4.2 Gaussian Naïve Bayes Regression

Table 3 Coefficient of determination of different regression models

Fig. 1 Linear regression

Fig. 2 Gaussian Naïve

4.3 Multinomial Naïve Bayes Regression

Fig. 3 Multinomial Naïve

Fig. 4 Bayesian ridge

4.4 Bayesian Ridge Regression

4.5 Huber Regression

Fig. 5 Huber regression plot

4.6 Logistic Regression

Initially, the original dataset of 72 numbers given Table 1 is employed in the

Fig. 6 Logistic regression

Fig. 7 Quantile regression

4.7 Quantile Regression

B. K. Monnappa, B. M. Shiva Kumar, T. S. Pushpa , and S. Shilpa

B. K. Monnappa (B) · B. M. Shiva Kumar · T. S. Pushpa · S. Shilpa