Investigation of Thrust Deduction on Azimuth Thrusters in Bollard Pull Condition Using CFD

CHALMERS
UNIVERSITY OF TECHNOLOGY
Investigation of Thrust Deduction on

Azimuth Thrusters in Bollard Pull
Condition using CFD
Master of science thesis in Naval Architecture and Ocean Engineering
FOIVOS LEMONAKIS
Division of Marine Technology

Department of Mechanics and Maritime Sciences
C HALMERS U NIVERSITY OF T ECHNOLOGY
Gothenburg, Sweden 2019
MASTER OF SCIENCE THESIS IN
NAVAL ARCHITECTURE AND OCEAN ENGINEERING
Investigation of Thrust Deduction on Azimuth

Thrusters in Bollard Pull Condition using CFD
FOIVOS LEMONAKIS

Chalmers University of Technology
Investigation of Thrust Deduction on Azimuth Thrusters in Bollard Pull Condition
using CFD
© FOIVOS LEMONAKIS, 2019.
Examiner: Rickard Bensow, Chalmers University of Technology

Supervisors: Tobias Huuva, Caterpillar Propulsion AB
Simon Törnros, Caterpillar Propulsion AB
Report No.: 2019:28

SE-412 96 Gothenburg
Telephone +46 31 772 1000
Cover: Visualization constructed in ParaView v.5.6.0 of the pressure distribution

acting on a ducted azimuth thruster and a harbour tug boat, along with the stream-
lines of the magnitude of the velocity vector field.
©
Typeset in LATEX
Printed by Chalmers Reproservice
iv
Investigation of Thrust Deduction on Azimuth Thrusters in Bollard Pull Condition
using CFD
FOIVOS LEMONAKIS
Abstract
This report is result of a master’s thesis project performed together with Caterpillar
Propulsion AB, which is a revisit of an older master’s thesis with title "Hydrodynam-
ics of Conventional Propeller and Azimuth Thruster in Behind Condition" (Matin,
2011). In the previous master’s thesis, the thrust deduction factors of open shaft
propellers and ducted azimuth thrusters, installed on an offshore vessel, are studied.
However, during the recent years the ducted azimuth thrusters have become more
and more common and especially for vessels such as tug boats, drilling vessels etc.
For such vessels, the bollard pull condition is really important, which indicates high
power operation with high loading (Funeno, 2009). Thus, the purpose of the current
master’s thesis is the investigation of thrust deduction on ducted azimuth thrusters
in bollard pull condition using Computational Fluid Dynamics (CFD).
In order achieve this type of study, a harbour tug boat with twin ducted azimuth
thrusters and a combination of steady state Multiple Reference Frame (MRF) and
transient Sliding Mesh Interface (SMI) modelling methodologies have been used.
MRF has been proven to be inaccurate in behind condition and at low or high ad-
vance coefficients, as it usually results in relatively strange flow fields in the vicinity
of the propeller while SMI has shown higher accuracy. However, SMI is a much more
computationally demanding methodology (Gullberg & Sengupta, 2011). Thus, SMI
has been used for critical regions and MRF for regions of lower interest.
Outcome of this study is that during the bollard pull condition the contribution of
the nozzle is higher than the propeller in terms of thrust and the contribution of
the gear case housing is higher than the hull in terms of resistance. The highest
thrust deduction appears on the gear case housing with 8.92% while the highest
merit deduction, which shows a lower performance, appears when placing the whole
propulsion unit behind the hull with 9.09%. Furthermore, the measured bollard pull
force of the open water case is about 8.23% higher than the one of the Wageningen
CD Series. However, this high difference is not realistic and the reason has to be
identified. Finally, by comparing the MRF and SMI methodologies, it is concluded
that the MRF approach gives appropriate results for initialization and quite rea-
sonable results for medium advance coefficients at low computational cost while the
SMI approach is needed for the critical regions of low and high advance coefficients.
Keywords: Propulsion, Hydrodynamics, Thruster, Thrust deduction, Bollard pull,

CFD, Multiple Reference Frame (MRF), Sliding Mesh Interface (SMI).
v
Acknowledgements
This master’s thesis is part of the Naval Architecture and Ocean Engineering mas-
ter’s programme, which is part of the Department of Mechanics and Maritime Sci-
ences, Chalmers University of Technology.
The author of this thesis would like to thank and acknowledge the help and contri-
bution of the following:
• Professor Rickard E. Bensow, examiner of this master’s thesis and Head of

Research in Hydrodynamics at the Department of Mechanics and Maritime
Sciences at Chalmers University of Technology.
• Caterpillar Propulsion AB for providing the resources of this study project.
• Tobias Huuva, supervisor of this master’s thesis and Engineering Manager at

the Department of Hydrodynamics at Caterpillar Propulsion AB.
• Simon Törnros, supervisor of this master’s thesis and Hydrodynamics Special-

ist at the Department of Hydrodynamics at Caterpillar Propulsion AB.
• Family and friends for their support during the two intensive years of this
master’s programme.
Foivos Lemonakis, Gothenburg, June 2019
vii
Contents
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
List of Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii
List of Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvii
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxi
1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Importance of this topic . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Purpose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.4 Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.4.1 Investigation cases . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4.2 Software packages used . . . . . . . . . . . . . . . . . . . . . . 3
2 Theory 5
2.1 Introduction to Computational Fluid
Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.1 Definition of the computational domain . . . . . . . . . . . . . 6
2.1.2 Grid generation . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.3 Identification of the physical and chemical phenomena that
need to be modelled . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.4 Definition of the fluid properties . . . . . . . . . . . . . . . . . 7
2.1.5 Definition of the boundary conditions . . . . . . . . . . . . . . 7
2.1.6 Integration of the governing equations of the flow over the
finite control volumes of the domain . . . . . . . . . . . . . . . 7
2.1.7 Discretisation of the integrated equations into algebraic equa-
tions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.8 Solving of the algebraic equations by the use of iterative method 10
2.1.9 Visualization of the simulation results . . . . . . . . . . . . . . 10
2.1.10 Process of the numerical results . . . . . . . . . . . . . . . . . 10
2.2 Governing equations . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.1 The continuity equation . . . . . . . . . . . . . . . . . . . . . 11
2.2.2 The Navier-Stokes equations . . . . . . . . . . . . . . . . . . . 12
2.3 Similarity between model and full-scale . . . . . . . . . . . . . . . . . 13
2.3.1 The continuity equation . . . . . . . . . . . . . . . . . . . . . 13
ix
Contents
2.3.2 The Navier-Stokes equations . . . . . . . . . . . . . . . . . . . 14

2.3.3 Other conditions . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4 Propeller scale effects . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.5 Types of fluid flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.6 Introduction to turbulence modelling . . . . . . . . . . . . . . . . . . 17
2.6.1 Boundary layer . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.6.2 Flat plate theory . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.6.3 Numerical methods of turbulence modelling . . . . . . . . . . 22
2.6.4 The Reynolds-Averaged Navier-Stokes equations . . . . . . . . 23
2.6.5 Turbulence models . . . . . . . . . . . . . . . . . . . . . . . . 25
2.6.6 Menter SST κ − ω turbulence model . . . . . . . . . . . . . . 25
2.7 Open water characteristics . . . . . . . . . . . . . . . . . . . . . . . . 27
2.8 Propulsive factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.9 Bollard pull condition . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.10 Modelling methodologies . . . . . . . . . . . . . . . . . . . . . . . . . 30
3 Geometry 33
3.1 Main particulars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.2 Azimuth thruster . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.3 Hull . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4 Methodology 37
4.1 Pre-processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.1.1 Definition of the computational domain . . . . . . . . . . . . . 37
4.1.2 Grid generation . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.1.3 Identification of the physical and chemical phenomena that
need to be modelled . . . . . . . . . . . . . . . . . . . . . . . 43
4.1.4 Definition of the fluid properties . . . . . . . . . . . . . . . . . 44
4.1.5 Definition of the boundary conditions . . . . . . . . . . . . . . 44
4.2 Solving . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.2.1 Integration of the governing equations of the flow over the
finite control volumes of the domain . . . . . . . . . . . . . . . 49
4.2.2 Discretisation of the integrated equations into algebraic equa-
tions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.2.3 Solving of the algebraic equations by the use of iterative method 50
4.3 Post-processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.3.1 Visualization of the simulation results . . . . . . . . . . . . . . 50
4.3.2 Process of the numerical results . . . . . . . . . . . . . . . . . 51
4.4 System used to run OpenFOAM . . . . . . . . . . . . . . . . . . . . . 53
5 Results and data analysis 55

5.1 Data analysis in free sailing and bollard pull conditions . . . . . . . . 55
5.1.1 Comparison of MRF/SMI . . . . . . . . . . . . . . . . . . . . 55
5.1.2 Comparison of Case 1/Case 2 . . . . . . . . . . . . . . . . . . 62
5.1.3 Comparison of Case 1/Case 3 . . . . . . . . . . . . . . . . . . 66
5.1.4 Comparison of Case 1/Wageningen CD Series . . . . . . . . . 69
5.1.5 Flow separation . . . . . . . . . . . . . . . . . . . . . . . . . . 74
x
Contents
5.1.6 Cavitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.2 Hydrodynamic interactions in bollard pull condition . . . . . . . . . . 77
5.3 Accuracy and reliability . . . . . . . . . . . . . . . . . . . . . . . . . 83
6 Conclusions and future work 85
References 87
Appendix A Convergence of the simulated results 93

A.1 Convergence plots of Case 1 . . . . . . . . . . . . . . . . . . . . . . . 93
Appendix B Illustration of the different surface mesh grids 123

B.1 General mesh grids used in all cases . . . . . . . . . . . . . . . . . . . 124
B.2 Mesh grids used in Case 1 . . . . . . . . . . . . . . . . . . . . . . . . 129
xi
List of Abbreviations
Abbreviation Description
2D Two-Dimensional space
3D Three-Dimensional space
AHTS Anchor Handling/Tug/Supply
AMI Arbitrary Mesh Interpolation
CAD Computer Aided Design
CAE Computer Aided Engineering
CFD Computational Fluid Dynamics
DNS Direct Numerical Simulation
EAR Expanded Area Ratio
FPP Fixed Pitch Propeller
HPC High-Performance Computing
LES Large Eddy Simulation
MRF Multiple Reference Frame
PIMPLE Pressure-Implicit Method for Pressure-Linked Equation
PISO Pressure-Implicit with Splitting of Operators
QUICK Quadratic Upstream Interpolation for Convective Kinematics
RANS Reynolds-Averaged Navier Stokes
RPM Revolutions Per Minute
RSM Reynolds Stress equation Model
SIMPLE Semi-Implicit Method for Pressure-Linked Equations
SIMPLEC Semi-Implicit Method for Pressure Linked Equations - Consistent
SIMPLER Semi-Implicit Method for Pressure-Linked Equations - Revised
SMI Sliding Mesh Interface
SST Shear Stress Transport
TDMA Tri-Diagonal Matrix Algorithm
xii
List of Variables
Variable [Units] Description
D m diameter
De , Dw , Dn , Ds , Dt , Db diffusion terms of the east, west, north, south,
top, bottom faces in finite volume theory
dm kg mass of the element
dV m3 infinitesimal fluid element
e, w, n, s, t, b east, west, north, south, top, bottom faces of
the control volume in finite volume theory
E(κ) m3 /s2 spectral energy content of the large eddies in
turbulence modelling
En − Euler number
F N force
→
−
F N force vector derived from the total force acting
on the volume element in finite volume theory
F N mean force
Fe , Fw , Fn , Fs , Ft , Fb convection terms of the east, west, north, south,
top, bottom faces in finite volume theory
Fn − Froude number
g m/s2 acceleration of gravity
J − advance coefficient
k m2 /s2 turbulent kinetic energy in turbulence modelling
KQ − non-dimensional torque coefficient, torque
coefficient in behind condition
KQ − mean torque coefficient
KQo − torque coefficient in open water condition
ks µm surface roughness
KT − non-dimensional thrust coefficient
KT − mean thrust coefficient
KT,net − net sum of all the thrust coefficients
KT n − nozzle thrust coefficient
KT p − propeller thrust coefficient
KT,R − sum of the thrust coefficients that are derived
from the resistance components
KT,T − sum of the thrust coefficients that are derived
from the thrust components
L m reference length (i.e., characteristic linear
dimension) that describes the travelled length
of the fluid, obstacle width
` m characteristic length of the larger scales in
M N·m moment
M N·m mean moment
xiii
Contents
mc − merit coefficient
mc,deduct % merit deduction
n rpm rotational speed
p Pa pressure, instantaneous pressure in turbulence
modelling
P Pa steady mean value of the pressure component
P, E, W, N, S, T, B central nodal point of the control volume and
the east, west, north, south, top, bottom central
nodal points of the neighboring volumes
in finite volume theory
p0 Pa fluctuating pressure component in turbulence
modelling
PB W brake power of main engine
pd Pa downstream pressure
PD W delivered power to the propeller via the shaft
phd Pa hydrodynamic pressure
pu Pa upstream pressure
Q N·m torque
Q N·m mean torque
R N resistance
R N mean resistance
Rn − Reynolds number
Rncrit − critical Reynolds number
SM N/m3 momentum source term
◦
T C, N temperature, thrust
t s, s, % time, characteristic time of the larger scales in
turbulence modelling, thrust deduction fraction
T N mean thrust
TBP ton bollard pull force
TBP,1995kw,CD ton bollard pull force at a fixed delivered power
obtained from the Wageningen CD Series
TBP,1995kw,CF D ton bollard pull force at a fixed delivered power
obtained from CFD results
Tnet N net sum of the total thrust
Tnet,deduct N deduction of the net sum of the total thrust
Tp N propeller thrust
Ttot N total thrust of a vessel that operates in full power
U∞ m/s reference velocity, steady mean value of the
velocity component in turbulence modelling
u m/s characteristic velocity of the Kolmogorov
microscales in turbulence modelling
u, v, w m/s scalars representing the velocity in x, y and
z-directions, instantaneous velocities in x, y and
z-directions in turbulence modelling
xiv
Contents
→
−
u ,→−v ,→
−
w m/s velocity vectors in x, y and z-directions
0
u m/s fluctuating velocity component in turbulence modelling
u+ − non-dimensional velocity
uτ m/s friction velocity
VA m/s advance velocity
VS m/s ship velocity
Wn − Weber number
wT − effective Taylor’s wake fraction
x, y, z m, − components of a vector in the x, y or z-directions,
coordinates of global system
x̄, ȳ, z̄, ζ̄, ū, v̄, w̄, p̄, t̄ − non-dimensional quantities in similarity equations
y+ − non-dimensional distance from the wall
z m distance of the upper part of the nozzle and the
water surface
Z − number of blades
→
−
α m/s2 acceleration vector
Γ generalized diffusion coefficient
γ N/m surface tension or over-speed ratio in channel
δ m boundary layer thickness
δx W P m distance between the nodes W and P in
finite volume theory
∆R N increase of the resistance
ε m2 /s3 dissipation rate of turbulent kinetic energy
ζ m wave elevation
η m characteristic length of the Kolmogorov
η0 − open water efficiency
η0 − mean open water efficiency
ηD − total propulsive efficiency
ηH − hull efficiency
ηR − relative rotative efficiency
ηS − shaft efficiency
ϑ m/s characteristic velocity of the larger scales in
κ 1/m wave number
µ Pa · s dynamic viscosity of the fluid
ν m2 /s kinematic viscosity of the fluid
νt m2 /s turbulent viscosity field
π − mathematical constant
ρ kg/m3 density
τ Pa, s shear stress, characteristic time of the Kolmogorov
τw Pa wall shear stress
τxx , τyy , τzz Pa Reynolds normal stresses in RANS equations
xv
Contents
τxy , τyx , τxz , τzx , τyz , τzy Pa Reynolds shear stresses in RANS equations
φ − flow property in finite volume theory
ω 1/s specific dissipation rate of turbulent energy in
xvi
List of Figures
2.1 A finite control volume (cell) and its neighbouring nodes (Versteeg &
Malalasekera, 2007). . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Part of the grid presented in 2D (Versteeg & Malalasekera, 2007). . . 9
2.3 Mass flow in x-direction through the fluid element (Larsson & Raven,
2010). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.4 Scale effects (Carlton, 2007). . . . . . . . . . . . . . . . . . . . . . . . 16
2.5 Regions of flow (Larsson & Raven, 2010). . . . . . . . . . . . . . . . . 19
2.6 Regions of development of the velocity boundary layer on a flat plate
(Bergman, Lavine, Incropera, & Dewitt, 2011). . . . . . . . . . . . . . 20
2.7 Comparison of laminar and turbulent velocity profiles of the boundary
layer on a flat plate (Bergman, Lavine, Incropera, & Dewitt, 2011). . 21
2.8 Development of the velocity boundary layer on a flat plate (Bergman,
Lavine, Incropera, & Dewitt, 2011). . . . . . . . . . . . . . . . . . . . 21
2.9 Regions of development of the velocity boundary layer on a flat plate
(Larsson & Raven, 2010). . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.10 Arrangement of an open water test in a towing tank (Dyne & Bark,
2005). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.11 Arrangement of an open water test in a cavitation tunnel. . . . . . . 27
2.12 Open water characteristics diagram. . . . . . . . . . . . . . . . . . . . 28
3.1 Starboard ducted azimuth thruster. . . . . . . . . . . . . . . . . . . . 34

3.2 Ducted azimuth thrusters and hull. . . . . . . . . . . . . . . . . . . . 35
3.3 Ducted azimuth thrusters and hull. . . . . . . . . . . . . . . . . . . . 35
4.1 Geometry of Case 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

4.2 Domain of Case 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.3 Geometry of Case 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.4 Geometry of Case 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.5 Domain of Case 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.6 Initial mesh of Case 1. . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.7 Final mesh of Case 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5.1 Open water characteristics diagram: MRF/SMI comparison of Case 1. 56

5.2 Open water characteristics diagram: Hybrid MRF/SMI diagram of
Case 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.3 MRF simulation of Case 1 at J = 0.00. . . . . . . . . . . . . . . . . . 58
5.4 SMI simulation of Case 1 at J = 0.00. . . . . . . . . . . . . . . . . . . 58
xvii
List of Figures

5.9 Open water characteristics diagram: Case 1/Case 2 comparison. . . . 62
5.10 Pressure distributions at J = 0.00: Case 1. . . . . . . . . . . . . . . . 64
5.11 Velocity field at J = 0.00: Case 1. . . . . . . . . . . . . . . . . . . . . 64
5.17 Open water characteristics diagram: Case 1/Wageningen CD Series
comparison with constant PB . . . . . . . . . . . . . . . . . . . . . . . 70
5.18 Open water characteristics diagram: Case1/Wageningen CD Series
comparison with constant PT . . . . . . . . . . . . . . . . . . . . . . . 71
5.19 Open water characteristics: Case1/CD Series with constant pitch ra-
tio at 0.7r comparison. . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.20 Velocity field in free sailing condition: Case 1 MRF simulation at
J = 0.70. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
J = 0.75. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
J = 0.80. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
J = 0.90. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.24 Velocity field in free sailing condition: Case 1 SMI simulation at
J = 0.80. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.25 Pressure distributions in bollard pull condition: Case 1. . . . . . . . . 76
A.1 Simulations at J = 0.00. . . . . . . . . . . . . . . . . . . . . . . . . . 94

A.2 MRF simulation of Case 1 at J = 0.10. . . . . . . . . . . . . . . . . . 94
xviii
List of Figures

A.43 SMI simulation of Case 2 at J = 0.00. . . . . . . . . . . . . . . . . . . 108
xix
List of Figures

B.1 Blade geometry and its mesh. . . . . . . . . . . . . . . . . . . . . . . 125

B.2 Hub geometry and its mesh: Isometric view of the hub. . . . . . . . . 126
B.3 Rotor geometry and its mesh. . . . . . . . . . . . . . . . . . . . . . . 128
B.4 Nozzle geometry and its mesh. . . . . . . . . . . . . . . . . . . . . . . 131
B.5 Shaft-hub geometries and their mesh: Isometric view of the shaft
including the hub. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
B.6 Domain geometry and its mesh. . . . . . . . . . . . . . . . . . . . . . 133
B.7 Azimuth unit geometry and its mesh. . . . . . . . . . . . . . . . . . . 135
B.8 Domain geometry and its mesh: Isometric view of the domain. . . . . 136
B.9 Hull-azimuth geometries and their mesh: Isometric view of the hull
including the azimuth unit. . . . . . . . . . . . . . . . . . . . . . . . 137
B.10 Domain geometry and its mesh. . . . . . . . . . . . . . . . . . . . . . 140
xx
List of Tables
3.1 Main particulars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.1 Number of cells: Case 1. . . . . . . . . . . . . . . . . . . . . . . . . . 42

4.4 Presentation of the boundary conditions: Case 1. . . . . . . . . . . . 46
5.1 Hybrid MRF/SMI results: Case 1. . . . . . . . . . . . . . . . . . . . . 57

5.2 SMI results: Case 1 (see Figure 5.2). . . . . . . . . . . . . . . . . . . 61
5.3 MRF results: Case 1 (see Figure 5.2). . . . . . . . . . . . . . . . . . . 61
5.4 SMI results: Case 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.5 SMI results (propeller-axis): Case 3. . . . . . . . . . . . . . . . . . . . 67
5.6 SMI results (x-axis): Case 3. . . . . . . . . . . . . . . . . . . . . . . . 67
5.7 Calculation of the brake power PB by using the SMI results of Case 1. 69
5.8 propCalcCD2.3.1 results: Constant PB . . . . . . . . . . . . . . . . . . 70
5.9 Calculation of the thrust power PT by using the SMI results of Case 1. 72
5.10 propCalcCD2.3.1 results: Constant PT . . . . . . . . . . . . . . . . . . 72
5.11 propCalcCD2.3.1 results: Constant pitch ratio at 0.7r. . . . . . . . . . 73
5.12 Decomposition of the hydrodynamic coefficients in bollard pull con-
dition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.13 Decomposition of the hydrodynamic coefficients in bollard pull con-
dition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.14 Case comparison: Presentation of the thrust, resistance, thrust de-
duction, net forces, deduction of the net forces and bollard pull forces. 80
5.15 Case comparison: Presentation of thrust, torque, net thrust coeffi-
cients, merit coefficient and the deduction of the merit coefficient. . . 80
5.16 Case comparison: Presentation of the rotational velocity, thrust and
their deduction in constant delivered power. . . . . . . . . . . . . . . 81
5.17 Case comparison: Presentation of the delivered power, thrust and
their deduction in constant rotational velocity. . . . . . . . . . . . . . 81
5.18 Case comparison: Presentation of the hydrodynamic coefficients in
bollard pull condition and comparison with the Wageningen CD Series. 82
B.1 Minimum and maximum cell size at the blades. . . . . . . . . . . . . 125

B.2 Minimum and maximum cell size at the hub. . . . . . . . . . . . . . . 126
xxi
List of Tables
B.3 Minimum and maximum cell size at the rotor. . . . . . . . . . . . . . 128

B.4 Minimum and maximum cell size at the nozzle. . . . . . . . . . . . . 131
B.5 Minimum and maximum cell size at the shaft. . . . . . . . . . . . . . 132
B.6 Minimum and maximum cell size at the domain. . . . . . . . . . . . . 133
B.7 Minimum and maximum cell size at the gear case housing. . . . . . . 135
B.9 Minimum and maximum cell size at the hull. . . . . . . . . . . . . . . 137
xxii
List of Tables
xxiii
List of Tables
xxiv
1
Introduction
In this chapter, a background of the propulsion systems; the propellers; the driving
reasons of the continuous development; and finally the computational fluid dynamics
is presented. In the rest of the chapter, it is described how this thesis contributes to
the field by explaining the purpose, the importance and the scope of this topic.
1.1 Background
Transportation through the water is one of the oldest means of transport during
humankind. Ships and boats are used for many different purposes such as trade,
fishing, defense and leisure. Therefore, the need of some kind of propulsion sys-
tem was already created since the ancient years of the stone age. For thousands of
years, the oars were the main and probably the only type of propulsion while during
the 7th century the sails were introduced. During the 16th century the number of
propulsion systems were limited and the use of sails was at its peak. At the end
of the 18th century the Industrial Revolution arrived, which brought a new era for
the shipping industry as well as for many other industries. In the beginning of the
19th century, paddle wheels were the prevailing propulsion system for steam vessels
while at the same time the screw propeller got invented, which was based on the
Archimedes’ screw.
Since then, propellers are the predominant marine propulsion system, which convert
rotational motion into thrust power. A pressure difference is generated between the
pressure side and the suction side of the propeller blades and as a result the fluid,
which in this specific case is water, is accelerated and the vessel moves. The propul-
sion of a vessel though is not any more the main goal as this is already solved. How-
ever, there is a continuous development of the marine propulsion systems, which is
mainly driven by the need of higher efficiency; both for environmental and econom-
ical reasons, noise reduction due to noise pollution and comfort reasons, operating
security, safe performance and maneuverability. In order to achieve those, the un-
derstanding of the physical phenomena around the propulsion unit and the hull of
the ship is vital. Hydrodynamics is the subject that is studying the flow of water
and the forces applied on structures immersed in water.
An important tool for the study of hydrodynamic phenomena and designs is Com-
putational Fluid Dynamics (CFD), which is the numerical analysis of fluid flow, heat
transfer and associated phenomena such as chemical reactions. The methodology of
1
1. Introduction
this kind of analysis is mostly used to run computer-based simulations by the use
of computer software packages in order to aid various engineering tasks, something
that makes CFD a part of the Computer Aided Engineering (CAE). CFD starts from
the 1960s where the aerospace industry developed CFD techniques and integrated
them into the design while from the 1990s a wider industrial usage starts (Versteeg
& Malalasekera, 2007).
1.2 Importance of this topic

This report is result of a master’s thesis project performed together with Caterpillar
Propulsion AB, which is a revisit of an older master’s thesis at Chalmers with title
"Hydrodynamics of Conventional Propeller and Azimuth Thruster in Behind Condi-
tion" (Matin, 2011). In the previous master’s thesis the thrust deduction factors of
open shaft propellers and ducted azimuth thrusters, installed on an offshore vessel,
are studied. However, during recent years the ducted azimuth thrusters have become
a more and more common propulsion system on vessels such as harbour tug boats,
Anchor Handling/Tug/Supply (AHTS) vessels, cable laying vessels, oceanographic
research vessels and drilling vessels due to their good manoeuvrability. For such ves-
sels, the bollard pull operating condition is really important, which indicates high
power operation with high loading (Funeno, 2009). However, the performance of
the azimuth thrusters is still questionable since they haven’t been studied as much
as the open shaft propulsion units. Moreover, model testing of azimuth thrusters is
not that efficient due to the high viscous effects at model scale while the conven-
tional open shaft thrusters have been validated with a huge amount of model tests
by several research institutes.
1.3 Purpose
The purpose of the current thesis is the investigation of thrust deduction on ducted
azimuth thrusters in bollard pull condition and in full-scale using CFD, in order
to develop a better understanding of the physical phenomena around this type of
geometry, understand where losses occur as well as understand which parts of those
geometries are beneficial. In order to fulfil this purpose, there is the need to investi-
gate the hydrodynamic interaction between the propeller, the nozzle, the gear case
housing and the hull.
1.4 Scope
In order perform this type of study, a harbour tug boat with twin ducted azimuth
thrusters and a combination of steady state Multiple Reference Frame (MRF) and
transient Sliding Mesh Interface (SMI) modelling methodologies have been used.
Steady state MRF has been proven to be inaccurate in behind condition and at
2
1. Introduction
low or high advance coefficients, as it usually results in relatively strange flow fields
in the vicinity of the propeller while transient SMI has shown higher accuracy.
However, transient SMI is a much more expensive methodology in terms of com-
putational cost (Gullberg & Sengupta, 2011). Thus, transient SMI has been used
for critical regions such as for advance coefficients J = 0.00, that indicate bollard
pull condition, and steady state MRF for regions of lower interest. Finally, a com-
parison with the Wageningen CD Series is done in order to develop an evaluation
tool which can be used in the future in order to evaluate the early stages of a project.
1.4.1 Investigation cases

As already mentioned, the hydrodynamic interactions between the propeller, the
nozzle, the gear case housing and the hull need to be investigated. Thus, CFD
simulations of the following cases are carried out and investigated separately:
• The propeller and the nozzle in open water condition,
• The propeller, the nozzle and the gear case housing in open water condition,
• The propeller, the nozzle, the gear case housing and the hull in behind condi-
tion.
Note: The gear case housing is considered to be part of the propulsion unit and
not part of the hull.
1.4.2 Software packages used

In order to solve the fluid flow problems with CFD, there is need of three main
steps: pre-processing, solving and post-processing. The software packages that are
used within this project in order to achieve these three steps are the following:
Pre-processor: Definition of the geometry and mesh generation

ANSA Pre-processor v.19.0.1
Developed by ’BETA CAE Systems S.A.’
Solver: Solving of the governing equations

OpenFOAM v1806
Open source software developed by ’OpenCFD Ltd’
Post-processors: Visualization of the simulation results

ParaView v.5.6.0
Open source software developed mainly by ’Kitware Inc.’ and
’Los Alamos National Laboratory’
3
1. Introduction
Process of the numerical results

MATLAB R2018b
Developed by ’The MathWorks, Inc.’
Comparison with the Wageningen CD Series

Microsoft Excel propCalcCD2.3.1 tool
Developed by ’Caterpillar Propulsion AB’
Rendering: 3D visualization of the geometry

Blender 2.79
Open source software supported by ’Blender Foundation’
4
2
Theory
In this chapter, a thorough explanation of the background theory of this thesis is pre-
sented. The first sections of this chapter demonstrate a fundamental theory of CFD,
governing equations, similarity between model and full-scale, propeller scaling, types
of fluid flow and turbulence modelling. Thus, these sections could be skipped from
a reader with basic knowledge in these topics. On the contrary, the last sections of
this chapter indicate a more specific theory of open water characteristics, propulsive
factors, bollard pull condition and modelling methodologies, which is directly related
to the current study.
2.1 Introduction to Computational Fluid

Dynamics
As mentioned before, CFD is the numerical analysis of fluid flow, heat transfer
and associated phenomena such as chemical reactions around a system. In order
to perform such an analysis, computer-based simulations by the use of a computer
software are mainly used.
CFD codes and software packages are using numerical algorithms in order to solve
fluid flow problems. All commercial CFD packages include input problem parame-
ters and most of them also include tools for analyzing the results. Thus, all CFD
packages include three main steps (Versteeg & Malalasekera, 2007):
• Pre-processing, which includes the input problem parameters and the con-
version of the inputs into a form that the solver can use. Pre-processing can
be divided into:
– Definition of the computational domain
– Grid generation
– Identification of the physical and chemical phenomena that need to be
modelled
– Definition of the fluid properties
– Definition of the boundary conditions
• Solving can be divided into various categories according to the numerical
techniques that are being used. However, the finite volume method is the
most widely used method among most of the CFD codes. Solving can be
divided into:
– Integration of the governing equations of the flow over the finite control
volumes of the domain
5
2. Theory
– Discretisation of the integrated equations into algebraic equations

– Solving of the algebraic equations by the use of an iterative method
• Post-processing, which includes the visualization of the domain geometry,
the mesh that is used in order to solve the algebraic equations and the simu-
lation results as well as the process of the numerical results. Post-processing
can be divided into:
– Visualization of the simulation results
– Process of the numerical results
In the following sections, the already mentioned main steps of CFD and their sub-
categories are presented in the same order as above.
2.1.1 Definition of the computational domain

A domain can be divided into an inner domain (i.e. geometries of interest) and
outer domain (i.e. boundaries). The definition of the outer computational domain
is dependent on each case and it has to be of a size to be able to provide results with
precision while it should not be much bigger than needed so the simulation becomes
too expensive.
2.1.2 Grid generation

The solving of the algebraic equations in order to determine flow variables such
as velocity, pressure and temperature is defined at nodes inside each cell of the
grid. Hence, the accuracy of a CFD problem is highly dependent on the number
of cells with a higher number of cells giving higher accuracy. However, the higher
the number of cells of a grid, the more expensive a simulation becomes in terms of
computer hardware and calculation time. Thus, an optimal number of cells should
be chosen without compromising the accuracy of the result. In order to achieve
that, non-uniform meshes are widely used which imply finer mesh in areas where
large geometrical or flow variations occur and coarser mesh in areas with smaller
geometrical or flow variations. However, the quality of the mesh is really important,
which implies elements with as simple geometries as possible. Thus, most of the
commercial mesh generation software packages include mesh quality checks as well
as automatic and manual correction mesh quality tools.
2.1.3 Identification of the physical and chemical phenomena

that need to be modelled
Prior to a CFD simulation, the definition of the flow problem in terms of the physical
and chemical phenomena is needed. The most usual decisions in order to define a
flow problem follow:
• Model a problem in two-dimensional space (2D) or three-dimensional space
(3D).
• Include or exclude the density variations due to ambient temperature or pres-
sure variations.
6
2. Theory
• Treatment of the small-scale turbulence.

Always within CFD, many assumptions are made in order to simplify complex flow
problems. However, it is really important to be always aware of all the assumptions
that are made.
2.1.4 Definition of the fluid properties

All the fluid properties such as the density ρ, pressure p, velocity u and temperature
T need to be defined. It should be mentioned that all the fluid properties are
functions of space and time. Thus, those properties should be defined respectively
as ρ(x, y, z, t), p(x, y, z, t), u(x, y, z, t) and T (x, y, z, t).
2.1.5 Definition of the boundary conditions

CFD problems are characterized by boundary conditions. The boundary conditions
represent a condition that needs to be satisfied within the boundary region. Those
boundary conditions enter a discretised system, either implicitly in the coefficient
matrix or explicitly in the source term. The most common boundary conditions,
within which a set of differential conditions are solved, are the following:
• Inlet
• Outlet
• Wall
• Symmetry
• Periodic/cyclic
• Prescribed pressure
2.1.6 Integration of the governing equations of the flow over

the finite control volumes of the domain
Initial step of the solving part is the integration of the governing equations of the
flow from face to face and over the finite control volumes of the domain in order
to obtain a flow property φ at its nodal point P. During this step the conservation
of various flow properties within each control volume is achieved by relating the
numerical algorithm with the physical conservation principles. Within the finite
control volume (see Figure 2.1) the conservation of a flow variable φ, such as velocity,
pressure, temperature and density, needs to be fulfilled. Thus, this conservation can
be expressed as a balance of a system (Versteeg & Malalasekera, 2007):
Rate of change Net rate of Net rate of Net rate of

of φ in the increase of φ increase of φ creation of φ
control volume = due to convection + due to diffusion + inside the control
with respect into the control into the control volume.
to time. volume. volume.
7
2. Theory
The integration of the governing equations of the flow over the finite control volumes
is more thoroughly presented in Sections 2.2 and 2.3.
Figure 2.1: A finite control volume (cell) and its neighbouring nodes
(Versteeg & Malalasekera, 2007).
2.1.7 Discretisation of the integrated equations into alge-

braic equations
The integrated governing equations need to be discretised (i.e. to go from a con-
tinuous differential equation to an algebraic discrete equation) in order to capture
the transport phenomena, the source terms and the rate of change with respect to
time. In order to do that, coefficients and gradients at the faces need to be cal-
culated (Versteeg & Malalasekera, 2007). Gradients at the faces are approximated
by calculating the distribution of the properties between two nodal points by the
use of discretisation schemes such as Gaussian integration and interpolation
schemes such as Central differencing, Upwind, Hybrid, Quadratic Upstream In-
terpolation for Convective Kinematics (QUICK) . In a similar way, divergence and
time derivative terms at the faces are approximated by the use of discretisation and
interpolation schemes (OpenCFD Ltd, 2018).
The transport phenomena include the convection part, which is the transport due
to fluid flow, and the diffusion part, which is the transport due to variations of
the flow variable φ from node to node (i.e. net movement from regions with high-
concentration to regions with low-concentration). An example of a grid in 2D and
its nodes (i.e. E, W, N, S) are presented in the following figure.
8
2. Theory
Figure 2.2: Part of the grid presented in 2D (Versteeg & Malalasekera, 2007).
Schemes can generally be split into different orders of accuracy. Schemes with
lower order of accuracy have higher stability of the results with less accuracy (due
to the high level of diffusion error). On the other hand, schemes with higher order
of accuracy have lower stability (due to false oscillations when the Peclet number
is high) with more accuracy. Peclet number is evaluated at the face of the control
volume and is defined as follows (i.e. assuming a west face):
Fw (ρU )w
P ew = = (2.1)
Dw Γ/δxW P
where:
Fw is the convection term of the west face,
Dw is the diffusion term of the west face,
Γ is the generalized diffusion coefficient,
δxW P is the distance between the nodes W and P.
Thus, sometimes hybrid schemes can be ideal since they combine a wider range of
problems (i.e. diffusion or convection dominated problems). Such an example is the
hybrid differencing scheme which employs:
• Central differencing scheme (2nd order accurate) for Peclet numbers P e < 2
• Upwind scheme (1st order accurate) by taking into account only the trans-
portiveness (i.e. upwind for convection and setting the diffusion to zero) for
Peclet numbers P e ≥ 2
Finally, the evaluation of the order of accuracy of a discretisation is done by the use
of Taylor expansion.
9
2. Theory
2.1.8 Solving of the algebraic equations by the use of itera-

tive method
Final step of the solving part is the iterative solving of the algebraic equations by the
use of mathematical algorithms in order to establish a correct relationship between
velocity and pressure. The governing equations are often solved separately and the
continuity equation is rearranged in order to achieve a better solution, hence the
use of some mathematical algorithms is needed. Examples of such mathematical
algorithms are (Versteeg & Malalasekera, 2007):
• Semi-Implicit Method for Pressure-Linked Equations (SIMPLE)
• Semi-Implicit Method for Pressure Linked Equations - Consistent (SIMPLEC)
• Semi-Implicit Method for Pressure-Linked Equations - Revised (SIMPLER)
• Pressure-Implicit with Splitting of Operators (PISO)
• Pressure-Implicit Method for Pressure-Linked Equations (PIMPLE)
The mathematical algorithms are using point-iterative methods for computing points
of iterated functions where a linear system of algebraic equations is solved (i.e.
Ax = b with unknown the x). Examples of such iterative linear solvers are:
• Gauss-Seidel
• Tri-Diagonal Matrix Algorithm (TDMA)
2.1.9 Visualization of the simulation results

There are many ways to process and present the simulation results. Usually, the
domain geometry and its mesh is displayed through commercial software packages
while various flow variables φ such as velocity u, pressure p, temperature T and
density ρ, as well as turbulent quantities are also displayed around this domain
geometry and within this mesh. Moreover, vector plots and 2D or 3D surface plots
are used to present the convergence and residuals as well as a flow variable φ that
may be of interest. All these tools are helping to get a better understanding of the
simulation results in order to be able to yield to a conclusion.
2.1.10 Process of the numerical results

Even though the visualization of the results gives a really good understanding of
the quality of the results and helps to draw conclusions, it is still really important
to process the specific quantities that are important for the specific study. Thus,
a series of calculations is needed in order to complete the whole picture with the
numerical conclusions.
2.2 Governing equations

Changes in the mass, momentum and energy of a fluid due to the appearance of
a source or multiple sources within the fluid element and during a flow along the
boundaries lead to the fluid flow and heat transfer equations that govern the whole
system. The governing equations of the conservative or divergence form of the
10
2. Theory
time-dependent 3D viscous flow of a Newtonian fluid are presented in the following

sections. However, since the fluid is considered incompressible, which indicates that
the density ρ is constant within the volume and through time, some rearrangements
and simplifications are made.
2.2.1 The continuity equation

The continuity equation, also known as mass conservation, is derived from the vol-
ume, which is the infinitesimal fluid element dV , as shown in Figure 2.3. The net
transport of mass out of the element should be zero in the case where there is not
any mass source. Thus, the continuity equation is formed as:
∂ρ
+ div(ρu) = 0 (2.2)
∂t
which can also be written as:
∂ρ ∂ ∂ ∂
+ ρu · dV + ρv · dV + ρw · dV = 0 (2.3)
∂t ∂x ∂y ∂z
where:
dV = dx · dy · dz,
u is a scalar representing the velocity in x-direction,
v is a scalar representing the velocity in y-direction,
w is a scalar representing the velocity in z-direction.
Figure 2.3: Mass flow in x-direction through the fluid element

(Larsson & Raven, 2010).
However, since the fluid is considered as incompressible, the continuity equation can
be re-written as:
∂u ∂v ∂w
+ + =0 (2.4)
∂x ∂y ∂z
11
2. Theory
2.2.2 The Navier-Stokes equations

The Navier-Stokes equations are derived through a long derivation by applying New-
ton’s second law to the infinitesimal fluid element dV (Larsson & Raven, 2010):
→
−
d F = dm · →
−
α (2.5)
where:
→
−
F is the force vector derived from the total force acting on the element (i.e. arrow
represents vector),
dm is the mass of the element,
→
−
α is the acceleration vector.
After the long derivation, the Navier-Stokes equations for incompressible flow are
obtained, assuming that the gravity g is the only body force along the z-axis (Ver-
steeg & Malalasekera, 2007):
x-momentum:
∂(ρu) ∂p
+ div(ρuu) = − + div(µ grad u) + SM x (2.6)
∂t ∂x
! !
∂u ∂u ∂u ∂u ∂p ∂ 2u ∂ 2u ∂ 2u
ρ +u +v +w =− +µ + + + SM x
∂t ∂x ∂y ∂z ∂x ∂x2 ∂y 2 ∂z 2
Dividing by ρ, the following equation is obtained (Larsson & Raven, 2010):

!
∂u ∂u ∂u ∂u 1 ∂p ∂ 2u ∂ 2u ∂ 2u 1
+u +v +w =− +ν 2
+ 2 + 2 + SM x (2.7)
∂t ∂x ∂y ∂z ρ ∂x ∂x ∂y ∂z ρ
In a similar way, the equations for the other two directions are obtained as follows:
y-momentum:
!
∂v ∂v ∂v ∂v 1 ∂p ∂ 2v ∂ 2v ∂ 2v 1
+u +v +w =− +ν + + + SM y (2.8)
∂t ∂x ∂y ∂z ρ ∂y ∂x2 ∂y 2 ∂z 2 ρ
z-momentum:
!
∂w ∂w ∂w ∂w 1 ∂p ∂ 2w ∂ 2w ∂ 2w 1
+u +v +w =− +ν + + + SM z − g (2.9)
∂t ∂x ∂y ∂z ρ ∂z ∂x2 ∂y 2 ∂z 2 ρ
where:
g is the gravity in z-direction,
SM is the momentum source term.
By looking at Eq.(2.7), (2.8) and (2.9), the left-hand side of these equations repre-
sents the acceleration of the fluid element in the x, y and z-directions respectively,
thus the convection terms. The right-hand side represents the forces due to pressure
(i.e. pressure gradient), viscous (i.e. diffusion terms) and body (i.e. gravity).
12
2. Theory
2.3 Similarity between model and full-scale

When carrying out model tests, it should be kept in mind to keep some similarities
to the full-scale. Those similarities can be divided into: geometric, which include
the shape and roughness (i.e. roughness is included in the viscous resistance); kine-
matic, which means that all the velocities are scaled; and dynamic, which means
that all the forces of the flow are scaled (Larsson & Raven, 2010).
In order to apply the similarity laws between the model and full-scale, the governing
equations and the boundary conditions need to be independent of dimension. Thus,
the following dimensionless quantities are introduced:
x y z ζ
x̄ = , ȳ = , z̄ = , ζ̄ =
L L L L
u v w phd tU∞
ū = , v̄ = , w̄ = , p̄ = 2
, t̄ =
U∞ U∞ U∞ ρU∞ L
where:
x, y, z are components of a vector in the x, y or z-directions,
ζ is the wave elevation,
L is the reference length (i.e. characteristic linear dimension) that describes the
travelled length of the fluid,
U∞ is the reference velocity,
phd is the hydrodynamic pressure,
t is the time.
By introducing these dimensionless quantities into the governing equations, which

are the continuity equation (see Eq.(2.4)) and the Navier-Stokes equations (see
Eq.(2.7), (2.8) & (2.9)), the equations below are obtained.
2.3.1 The continuity equation
∂(ūU∞ ) ∂(v̄U∞ ) ∂(w̄U∞ )

+ + =0 (2.10)
∂(x̄L) ∂(ȳL) ∂(z̄L)
U∞
By dividing this equation by L
, the following equation is obtained:
∂ ū ∂v̄ ∂ w̄
+ + =0 (2.11)
∂ x̄ ∂ ȳ ∂ z̄
13
2. Theory
2.3.2 The Navier-Stokes equations

x-momentum:
∂(ūU∞ ) ∂(ūU∞ ) ∂(ūU∞ ) ∂(ūU∞ )
+ ūU∞ + v̄U∞ + w̄U∞ = (2.12)
∂ t̄L ∂(x̄L) ∂(ȳL) ∂(z̄L)
U∞
!
2
1 ∂(p̄ρU∞ ) ∂ 2 (ūU∞ ) ∂ 2 (ūU∞ ) ∂ 2 (ūU∞ )
− +ν + + + SM x
ρ ∂(x̄L) ∂(x̄L)2 ∂(ȳL)2 ∂(z̄L)2
2
U∞
By dividing this equation by L
, the following equation is obtained:
∂ ū ∂ ū ∂ ū ∂ ū
+ ū + v̄ + w̄ = (2.13)
∂ t̄ ∂ x̄ ∂ ȳ ∂ z̄
!
∂ p̄ ν ∂ 2 ū ∂ 2 ū ∂ 2 ū
− + + + + SM x
∂ x̄ U∞ L ∂ x̄2 ∂ ȳ 2 ∂ z̄ 2
y-momentum:
∂v̄ ∂v̄ ∂v̄ ∂v̄
+ ū + v̄ + w̄ = (2.14)
∂ t̄ ∂ x̄ ∂ ȳ ∂ z̄
!
∂ p̄ ν ∂ 2 v̄ ∂ 2 v̄ ∂ 2 v̄
− + + + + SM y
∂ ȳ U∞ L ∂ x̄2 ∂ ȳ 2 ∂ z̄ 2
z-momentum:
∂ w̄ ∂ w̄ ∂ w̄ ∂ w̄
+ ū + v̄ + w̄ = (2.15)
∂ t̄ ∂ x̄ ∂ ȳ ∂ z̄
!
∂ p̄ ν ∂ 2 w̄ ∂ 2 w̄ ∂ 2 w̄
− + + + − g + SM z
∂ z̄ U∞ L ∂ x̄2 ∂ ȳ 2 ∂ z̄ 2
The governing equations are taken into account when similarity is about to be
achieved in order to see which equations are dependent on the scale. The only
parameter that appears in these equations is the Reynolds number Rn (which
appears as 1/Rn and is circled in the above equations) in the dimensionless Navier-
Stokes equations (see Eq.(2.13), (2.14) & (2.15)), which is defined as:
ρuL uL
Rn = = (2.16)
µ ν
where:
u is a scalar representing the velocity in x-direction in respect to the object,
µ is the dynamic viscosity of the fluid,
ν is the kinematic viscosity of the fluid.
14
2. Theory
2.3.3 Other conditions

Other conditions are also taken into account when similarity is to be achieved. Such
conditions are the boundary conditions for solid surfaces, the free-surface boundary
conditions and the infinity condition (Larsson & Raven, 2010). However, the only
parameters that appear in those equations are in the free-surface boundary condi-
tions. Those parameters are:
• the Euler number En, which is used to determine the cavitation,
pu − pd
En = 2
(2.17)
ρU∞
where:
pu is the upstream pressure,
pd is the downstream pressure.
• the Froude number F n, which is used to determine the resistance,
U∞
Fn = √ (2.18)
gL
• the Weber number W n, which is used to determine the spray, the wave break-
ing and the surface waves,
2
ρU∞ L
Wn = (2.19)
γ
where:
γ is the surface tension or over-speed ratio in channel.
2.4 Propeller scale effects

Looking at the performance characteristics of a propeller in both model and full-
scale, and taking into account that full-scale indicates higher Reynolds number,
larger diameter and lower rotational speed, the thrust coefficient KT is almost iden-
tical while the torque coefficient KQ in full scale is lower according to Carlton (2007)
as shown in Figure 2.4.
It should be mentioned that the above statement applies for open shaft arrangement
where a huge number of model testing has been done. On the other hand, this state-
ment is not true for azimuth thrusters due to the high model testing uncertainties
that are derived from the high viscous effects that appear at model scale. For this
reason, full-scale CFD might be a better alternative for azimuth thrusters compared
to model testing.
15
2. Theory
Figure 2.4: Scale effects (Carlton, 2007).
The viscous effects are the main reason for the differences between the model and
full-scale due to the phenomena that are dependent on the Reynolds number. The
dependency of these phenomena to the Reynolds number and consequently to the
viscosity is explained thoroughly in Sections 2.2 and 2.3. The Reynolds number
should be the same at both scales in order for the viscosity effects to be correct.
However, due to the methods of testing model propellers, the Reynolds number ends
up being different between model and full-scale and as a result a different boundary
layer structure to the flow over the blades and nozzle arises. A consequence of that
is the too high viscous resistance in model scale.
2.5 Types of fluid flow

• Steady - Unsteady
A flow is described as steady (state) when the flow variables φ, such as velocity,
pressure, temperature, and density, do not change through time. On the other
hand, a flow is described as unsteady (also known as transient flow) when the
flow variables φ change through time.
• Uniform - Non-uniform
A flow is described as uniform when the velocity does not change through space
at a specific time. On the other hand, a flow is described as non-uniform when
the velocity changes through space at a specific time.
• Laminar - Turbulent
A flow is described as laminar when the fluid particles follow smooth, well-
defined paths. On the other hand, a flow is described as turbulent when the
fluid particles follow random, chaotic paths.
16
2. Theory
• Compressible - Incompressible
A flow is described as compressible when the density of the fluid ρ is dependent
on the pressure, thus the density is changing within the control volume that
moves with the flow velocity at regions with different pressure. On the other
hand, a flow is described as incompressible when the density of the fluid ρ
is independent on the pressure and thus, constant within the control volume
that moves with the flow velocity at regions with different pressure.
• Viscous - Inviscid
A flow is described as viscous when the fluid of the flow is viscous. On the
other hand, a flow is described as inviscid when the fluid of the flow has no
viscosity (i.e. viscosity is equal to zero).
• Rotational - Irrotational
A flow is described as rotational when vorticity, which is the curl of the velocity
vector, appears all over the flow field. On the other hand, a flow is described
as irrotational when vorticity is zero all over the flow field.
• 1D - 2D - 3D
A flow is described as 1D when the flow variables φ, such as velocity, pressure,
temperature and density are defined within one dimensional coordinate system
in space. A flow is described as 2D and 3D when the flow variables φ are defined
within two or three dimensional coordinate systems respectively in space.
• Potential flow
A potential flow is a flow that is described by the velocity potential φ(x, y, z)
for some scalar field φ, being a function of space and time and assuming
incompressible, inviscid and irrotational flow. The flow velocity → −v is a vector
field that is equal to the gradient of the velocity potential as shown below
(Larsson & Raven, 2010):
→
−v (x, y, z) = 5φ (2.20)
2.6 Introduction to turbulence modelling

All types of fluid flow become unstable when a specific Reynolds number is exceeded.
Flows are considered laminar at low Reynolds numbers while at high Reynolds num-
bers flows become turbulent (Versteeg & Malalasekera, 2007).
The Reynolds number associates the inertia forces that are related to the convec-
tive effects and viscous forces with the flow. A critical Reynolds number, Rncrit is
introduced, which determines the type of the flow. In cases where Rn < Rncrit the
flow is considered as laminar while when Rn ≥ Rncrit the flow is considered as tur-
bulent. At turbulent flows the instantaneous flow variables φ are decomposed into
a mean value and a fluctuating value. The instantaneous velocity u and pressure p
are presented below:
u = U + u0 and p = P + p0 (2.21)
where:
U is the steady mean value of the velocity component,
17
2. Theory
P is the steady mean value of the pressure component,

u0 is the fluctuating velocity component,
p0 is the fluctuating pressure component.
Turbulent flows include rotational flows that are called turbulent eddies. These ed-
dies bring the fluid particles closer to each other, which results in a high heat, mass
and momentum exchange. This yields to an effective mixing, which rises the diffu-
sion coefficients. Turbulent eddies that are large enough to interact with the mean
flow and extract energy from it are called vortices while vortex stretching shows how
much a fluid particle has been stretched.
The characteristic velocity ϑ and the characteristic length ` of the larger eddies (e.g.
vortices) is observed to be of the same order as the mean velocity U and length L
of the mean flow. Thus, by looking at Reynolds equation (see Eq.2.16), it can be
considered that vortices are highly influenced by inertia effects while viscous effects
can be neglected due to the same order of the velocity and length. Since vortices
are considered inviscid and the angular momentum is preserved, the rotation rate
gets increased while the rotation radius gets decreased. This yields to creation of
motion at the smaller length scales and within smaller time steps. As a result, vor-
tex stretching provides the required energy and the turbulence is maintained.
Smaller eddies are mainly streched by larger eddies while mean flow has less im-
pact on them. The so called energy cascade explains the transfer of kinetic energy
from the larger eddies to the smaller ones. The smallest eddies that are dominated
equally by inertia effects and viscous effects are called Kolmogorov microscales. The
energy within these microscales is dissipated and converted into thermal energy,
thus resulting in high energy losses. Kolmogorov microscales are expressed by the
dissipation rate of turbulent kinetic energy ε and the fluid’s kinematic viscosity ν.
Magnitude ratios of the larger scales and the Kolmogorov microscales are used in
order to proceed to a dimensional analysis. These magnitude ratios, that are ex-
pressed by the characteristic velocity ϑ, length ` and time t of the larger scales and
the characteristic velocity u, length η and time τ of the Kolmogorov microscales,
are shown below:
• Velocity scale ratio:
u −1/4
≈ Rn` (2.22)
ϑ
• Length scale ratio:
η −3/4
≈ Rn` (2.23)
`
• Time scale ratio:
τ −1/2
≈ Rn` (2.24)
t
As it is already mentioned, the large eddies are considered as inviscid while they
are highly dependent on the characteristic velocity ϑ and the characteristic length
`. As a result, the spectral energy content E(κ) of the large eddies should be
directly proportional to the characteristic velocity ϑ and the characteristic length
18
2. Theory
` as shown in Eq.(2.25) and (2.26). The large eddies are considered anisotropic as
the characteristic length ` is linked with components such as the boundary layer
thickness δ, the obstacle width L and the surface roughness ks .
E(κ) ∝ ϑ2 ` (2.25)
where the wave number κ is defined as:

1
κ= (2.26)
`
On the other hand, Kolmogorov microscale eddies are dependent on the dissipation
rate of turbulent kinetic energy ε and the fluid’s kinematic viscosity ν. As a result,
the energy content of the Kolmogorov microscale eddies should be directly propor-
tional to the fluid’s kinematic viscosity ν and the dissipation rate ε as shown in
Eq.(2.27) and (2.28). The Kolmogorov microscale eddies are considered isotropic as
the spectral energy content E(κ) is linked with the dissipation rate ε.
E(κ) ∝ ν 5/4 ε1/4 (2.27)
where the wave number κ is defined as:

1
κ= (2.28)
η
2.6.1 Boundary layer

An interaction between the molecules of a solid surface, which is also known as wall,
and a fluid element arises due to their collision. Hence, a tiny layer between the solid
surface and the fluid elements is created where two different velocity fields are mixed.
Nevertheless, these two different velocity fields get smoothed and finally become
identical (i.e. the fluid element sticks to the solid surface and the difference of their
velocities becomes zero). This interaction is described by the "no-slip" condition
where the viscous effects (i.e. friction is included) are taken into consideration.
Moving away from the solid surface, the velocity increases within the so called
boundary layer. In Figure 2.5, the different regions of the flow are presented where
the viscous sublayer and the turbulent boundary layer are part of the boundary
layer while the inviscid flow is part of the free stream (Larsson & Raven, 2010).
Figure 2.5: Regions of flow (Larsson & Raven, 2010).
19
2. Theory
2.6.2 Flat plate theory
The boundary layer is the most important region of flow where the transition from
laminar to turbulent flow occurs. The simplest boundary layer is considered the one
that is developed on a flat plate, a plate with really small thickness and parallel
to the flow on which the pressure is considered constant. The flat plate consists
of three regions: the laminar boundary layer, the transition region from laminar
to turbulent flow and the turbulent boundary layer. The development of velocity
boundary layer on a flat plate within these regions is shown in Figure 2.6 (Larsson
& Raven, 2010).
Figure 2.6: Regions of development of the velocity boundary layer on a flat plate
(Bergman, Lavine, Incropera, & Dewitt, 2011).
Laminar flow is the only flow within the laminar boundary layer, which is char-
acterized by ordered streamlines, until the transition region where an intermediate
flow dominates (i.e. laminar flow within the viscous sublayer and intermediate flow
within the buffer layer). Turbulent flow follows within the turbulent boundary layer
where a turbulence flow dominates (i.e. laminar flow within the viscous sublayer,
intermediate flow within the buffer layer and turbulent flow within the turbulent
region) (Bergman, Lavine, Incropera, & Dewitt, 2011).
Moving streamwise from the laminar to the turbulent region (i.e. increasing x-
direction), the boundary layer thickness δ grows (see Figure 2.6) and the velocity
gradients on the flat plate decrease (see Figure 2.7).
20
2. Theory
Figure 2.7: Comparison of laminar and turbulent velocity profiles of the

boundary layer on a flat plate (Bergman, Lavine, Incropera, & Dewitt, 2011).
The reduction of the velocity is associated with the shear stresses τ acting in parallel
to the velocity (see Figure 2.8) in order to conserve the particle motion at a distance
δ from the solid surface.
Figure 2.8: Development of the velocity boundary layer on a flat plate

(Bergman, Lavine, Incropera, & Dewitt, 2011).
The velocity within the boundary layer needs to be expressed in terms of the fluid
density ρ, the dynamic viscosity µ of the fluid and the shear stresses at the wall
τw . It should be mentioned that the velocity should be kept independent of the
boundary layer thickness δ. Thus, the following equations are introduced (Larsson
& Raven, 2010):
• The non-dimensional velocity u+ :
u
u+ = (2.29)
uτ
where the friction velocity uτ is defined as
s
τw
uτ = (2.30)
ρ
• The non-dimensional distance from the wall y + :
yuτ
y+ = (2.31)
ν
Finally, the boundary layer can be split into four regions:
• The viscous sublayer: 0 ≤ y + ≤ 5
This region (i.e. region I) is part of the inner layer where a laminar flow
dominates.
21
2. Theory
• The buffer layer: 5 ≤ y + ≤ 30

This region (i.e. region II) is part of the inner layer where an intermediate
flow dominates.
• The turbulent (logarithmic) region: 30 ≤ y + ≤ 500
This region (i.e. region III) is part of the inner layer where a turbulent flow
dominates.
• The wake region: 500 ≤ y + ≤ 10000
This region (i.e. region IV) is part of the outer layer. Therefore, it is not
further explained.
Figure 2.9: Regions of development of the velocity boundary layer on a flat plate
(Larsson & Raven, 2010).
2.6.3 Numerical methods of turbulence modelling

The instantaneous flow variables are divided into a mean component and a turbulent
fluctuating component. The fluctuating component consists of all scales of eddies.
Based on this statement, the numerical methods that are used in order to treat the
effects of turbulence are the following:
• Reynolds-Averaged Navier-Stokes (RANS)
RANS uses time averaged Navier-Stokes equations to compute the steady mean
flow, which is the same order as the large scale eddies (Versteeg & Malalasek-
era, 2007). Then, the effects of all the scales of eddies to the mean flow are
modelled (Larsson & Raven, 2010).
• Large Eddy Simulation (LES)
LES uses space filtering of the unsteady Navier-Stokes equations to compute
the flow including large scale eddies (Larsson & Raven, 2010). Then, the
effects of the small scale eddies are modelled.
• Direct Numerical Simulation (DNS)
DNS uses the unsteady Navier-Stokes equations to compute the flow including
all the scales of eddies, even the smallest ones which are the same scale as the
Kolmogorov microscale eddies.
LES and DNS methods provide higher resolution and a more detailed description
22
2. Theory
of turbulence compared to RANS. However, both LES and DNS are much more
expensive methods and the level of resolution and details that both provide is usually
not needed. On the other hand, RANS yields a decent level of resolution and details,
which are the result of the averaged flow properties such as the mean values of the
velocities, pressures and stresses. Due to this fact, RANS method has been used the
most among the three.
2.6.4 The Reynolds-Averaged Navier-Stokes equations

Taking the average of each term, one gets u = u + u0 = U + u0 . The overbar (e.g.
u) or capital letters (e.g. U ) indicate averaging and u0 = 0. Hence, the continuity
equation is formed as (Fletcher, 1991):
∂ρ
+ div(ρu) = 0 (2.32)
∂t
which can also be written as:
∂ρ ∂ ∂ ∂
+ ρu · dV + ρv · dV + ρw · dV = 0 (2.33)
∂t ∂x ∂y ∂z
However, since the fluid is considered incompressible (i.e. ρ = ρ) and by taking into
account that div(u) = div(U ), the continuity equation can be re-written as:
div(U ) = 0 (2.34)
Then, by taking the average of each term of the Navier-Stokes equations (see
Eq.(2.7), (2.8) & (2.9)) for incompressible flow and assuming that the gravity g is
the only body force along the z-axis, the RANS equation are obtained:
x-momentum:
∂(ρu) ∂p
+ div(ρuu) = − + div(µ grad u) + SM x (2.35)
∂t ∂x
Then, the mean and fluctuating components (see Eq.(2.21)) are introduced into the
time-average of each term:
∂(ρU ) ∂P
+ div(ρU U ) + div(ρu0 u0 ) = − + div(µ grad U ) + SM x (2.36)
∂t ∂x
where:
∂(ρu) ∂(ρU )
= , div(ρuu) = div(ρU U ) + div(ρu0 u0 ) (2.37)
∂t ∂t
∂p ∂P
− =− , div(µ grad u) = div(µ grad U )
∂x ∂x
23
2. Theory
Rearranging Eq.(2.36) and moving the fluctuating components on the right hand
side, the following equation is obtained:
! !
∂U ∂U ∂U ∂U ∂P ∂ 2U ∂ 2U ∂ 2U
ρ +U +V +W =− +µ + + + (2.38)
∂t ∂x ∂y ∂z ∂x ∂x2 ∂y 2 ∂z 2
" #
∂(−ρu02 ) ∂(−ρu0 v 0 ) ∂(−ρu0 w0 )
+ + + SM x
∂x ∂y ∂z
Dividing by ρ, the following equation is obtained:
!
∂U ∂U ∂U ∂U 1 ∂P ∂ 2U ∂ 2U ∂ 2U
+U +V +W =− +ν + + + (2.39)
∂t ∂x ∂y ∂z ρ ∂x ∂x2 ∂y 2 ∂z 2
" #
1 ∂(−ρu02 ) ∂(−ρu0 v 0 ) ∂(−ρu0 w0 ) 1
+ + + SM x
ρ ∂x ∂y ∂z ρ
y-momentum:
!
∂V ∂V ∂V ∂V 1 ∂P ∂ 2V ∂ 2V ∂ 2V
+U +V +W =− +ν + + + (2.40)
∂t ∂x ∂y ∂z ρ ∂x ∂x2 ∂y 2 ∂z 2
" #
1 ∂(−ρu0 v 0 ) ∂(−ρv 02 ) ∂(−ρv 0 w0 ) 1
+ + + SM y
z-momentum:
!
∂W ∂W ∂W ∂W 1 ∂P ∂ 2W ∂ 2W ∂ 2W
+U +V +W =− +ν + + + (2.41)
∂t ∂x ∂y ∂z ρ ∂x ∂x2 ∂y 2 ∂z 2
" #
1 ∂(−ρu0 w0 ) ∂(−ρv 0 w0 ) ∂(−ρw02 ) 1
+ + − g + SM z
By looking at the Navier-Stokes equations (see Eq.(2.7), (2.8) & (2.9)) and the
RANS equations (see Eq.(2.39), (2.40) & (2.41)), nine extra terms can be observed,
which represent three normal stresses:
τxx = −ρu02 , τyy = −ρv 02 , τzz = −ρw02 (2.42)
and three double shear stresses:
τxy = τyx = −ρu0 v 0 , τxz = τzx = −ρu0 w0 , τyz = τzy = −ρv 0 w0 (2.43)
These stresses are knows at the Reynolds stresses. The Reynolds stresses are addi-
tional turbulent shear stresses within the fluid layers, which are result of momentum
exchange derived from the convective transport by the eddies. This momentum ex-
change results to a deceleration of the faster moving fluid layers and an acceleration
of the slower moving fluid layers. The symmetry of the double shear stresses is
explained by the identical net flux of the x, y and z-momentum accordingly and the
identical mean component of this net flux of the x, y and z-momentum out of the
control volume (Versteeg & Malalasekera, 2007).
24
2. Theory
2.6.5 Turbulence models

For most CFD purposes, there is no need to resolve the details of the turbulent
fluctuations. Thus, most of the turbulent flow computations are based on RANS
equations. However, the effects of turbulence on the mean flow are important. In
order to be able to compute turbulent flows with RANS that include the turbulence
effects, turbulence models to predict the already mentioned Reynolds stresses as
well as scalar transport terms are needed (Versteeg & Malalasekera, 2007).
Some of the most common turbulence models that are usually used in CFD are
presented below:
• κ − ε model
• Reynolds stress equation model (RSM)
• Wilcox κ − ω model
• Menter Shear Stress Transport (SST) κ − ω model
2.6.6 Menter SST κ − ω turbulence model

Menter SST κ−ω model is a hybrid model that uses the original Wilcox κ−ω model
to solve the boundary layer due to its high sensitivity to the arbitrary assumed val-
ues and the κ − ε model to solve the regions located far from the wall, which are
fully turbulent.
The turbulent kinetic energy k, which is derived from the Navier-Stokes equations,
is defined as the half of the sum of the fluctuating velocity components (Versteeg &
Malalasekera, 2007):
1
k = (u02 + v 02 + w02 ) (2.44)
2
The dissipation rate of turbulent kinetic energy ε is defined as:
ε = 2νs0ij · s0ij (2.45)
Finally, the specific dissipation rate of turbulent energy ω is defined as:

ε
ω= (2.46)
k
For the k − ε model, the velocity scale ϑ and the length scale ` are defined as:
ϑ = k 1/2 (2.47)
k 3/2
`= (2.48)
ε
25
2. Theory
For the k − ω model, the velocity scale ϑ is the same as above while the the length
scale ` is defined as:
k 1/2
`= (2.49)
ω
The Reynolds stress and the k-equation are identical to the original Wilcox k − ω
model:
!
2 ∂Ui ∂Uj 2
τij = −ρu0 v 0 = −ρu0i u0j = 2µt Sij − ρkδij = µt + − ρkδij (2.50)
3 ∂xj ∂xi 3
∂(ρk) µt

+ div(ρkU ) = div µ+ grad(k) + Pk − β ∗ ρkω (2.51)
∂t σk
while the ε-equation is transformed into an ω-equation by substituting ε = kω:
" ! # !
∂(ρω) µt 2 ∂Ui
+ div(ρωU ) = div µ+ grad(ω) + γ2 2ρSij · Sij − ρω δij −
∂t σω,1 3 ∂xj
ρ ∂k ∂ω
β2 ρω 2 + 2 (2.52)
σω,2 ω ∂xk ∂xk
The turbulent eddy viscosity is defined as:
α1 ρk
µt = (2.53)
max(α1 ω, SF2 )
where:
q
S= 2Sij Sij (2.54)
the blending function is defined as (Menter, Kuntz, & Langtry, 2003):

"√ !#2 
2 k 500ν
F2 = tanh max ∗ , 2  (2.55)
β ωy y ω
and Pk is defined as:

!
∗ 2 ∂Ui
Pk = min 10β ρkω, 2µt Sij − ρk δij (2.56)
3 ∂xj
while the revised model constants are the following (Versteeg & Malalasekera, 2007):
σk = 1.000, σω,1 = 2.000, γ2 = 0.440 β2 = 0.083 β ∗ = 0.090
and α1 = 5/9 according to Menter et al. (2003).
26
2. Theory
2.7 Open water characteristics

The open water propeller characteristics are performance characteristics of a pro-
peller operating in a uniform flow, that are usually obtained by model testing in a
towing tank (see Figure 2.10) or in a cavitation tunnel (see Figure 2.11). In cases
where the propeller is placed behind the hull, the inflow is not uniform. The thrust
T and the torque Q of the propeller are measured while the rotational speed n is
most of the times kept constant and the advance velocity VA varies (Dyne & Bark,
2005).
Figure 2.10: Arrangement of an open water test in a towing tank

(Dyne & Bark, 2005).
Having everything known, the advance coefficient J is then obtained:

VA
J= (2.57)
nD
which is a non-dimensional value that describes the operating condition of the pro-
peller.
Figure 2.11: Arrangement of an open water test in a cavitation tunnel.
Then, the non-dimensional thrust coefficient KT and torque coefficient KQ are ob-
tained:
T
KT = 2 4 (2.58)
ρn D
27
2. Theory
Q
KQ = (2.59)
ρn2 D5
The delivered power to the propeller via the shaft PD is defined as:
PD = 2πnQ (2.60)
and the thrust power that is created from the propeller PT is:
PT = T VA (2.61)
The open water efficiency η0 is finally defined as:

PT J KT
η0 = = · (2.62)
PD 2π KQ
The results of an open water test are presented through an open water characteristics
diagram as shown in Figure 2.12.
Figure 2.12: Open water characteristics diagram.
2.8 Propulsive factors

The propulsive factors are factors that express specific quantities of the flow that
are relevant to the operation of a ship’s propeller. Moreover, these factors also
express the hydrodynamic interaction between the propeller and the hull, which
is really important during the design of a propeller. The propulsive factors are
determined based on values measured during open water tests, resistance tests and
self propulsion tests. These propulsive factors are:
28
2. Theory
• The thrust deduction fraction t, which is defined as:

∆R
t= (2.63)
T
where:
∆R is the increase of the resistance due to the propulsor.
• The effective mean wake wT M , which is defined as:

VS − VAT
wT = (2.64)
VS
where:
VAT is the advance mean velocity where T stands from the thrust,
VS is the ship velocity.
• The relative rotative efficiency ηR , which is defined as:

KQo
ηR = (2.65)
KQ
where:
KQo is the torque coefficient in open water condition,
KQ is the torque coefficient in behind condition.
• The total propulsive efficiency ηD , which is defined as:
ηD = ηo · ηR · ηH (2.66)
where the hull efficiency ηH is defined as:
R/T 1−t
ηH = = (2.67)
VA /VS 1 − wT
and R is the resistance of the hull.
Another useful factor often connected to the hydrodynamic interactions, even if itself
is a mechanical interaction, is the shaft efficiency ηS , which is defined as:
PD
ηS = (2.68)
PB
where:
PB is the brake power.
Empirically, ηS is set to 0.97 for open shaft configurations and to 0.95 for azimuth
thrusters.
29
2. Theory
2.9 Bollard pull condition

In Section 2.8, the conventional ways to express the hydrodynamic interactions
that are commonly used in free sailing condition are presented. However, during
bollard pull condition the movement of a vessel is restricted. The real applica-
tion that the bollard pull condition tries to capture is the pushing or pulling of a
really heavy structure (e.g. a vessel) or equipment (e.g. anchors for drilling rigs),
which results to a high power operation and a vessel speed really close to zero. Dur-
ing bollard pull testing, a vessel operates in full power while it is connected on a
bollard on shore through a tow-line. As a result, the ship’s velocity VS is zero and
the advance velocity VA looses its meaning since the ship is not moving. This yields
an advance coefficient J equal to zero.
In bollard pull condition, the thrust deduction fraction t is mainly used to describe
the increase of the resistance due to the suction of the propeller and to determine
the generated thrust. On the other hand, some of the other conventional hydro-
dynamic interactions such as the effective Taylor’s wake fraction wT , the relative
rotative efficiency ηR and the propulsive efficiency ηD cannot be applied and some
alternative ways have to be used instead. The merit coefficient mc is introduced
as a replacement of the total propulsive efficiency ηD in order to be able to express
the performance of the thruster. The merit coefficient is defined as:
(KT /π)3/2
mc = (2.69)
KQ
As it has already been mentioned, a vessel operates in full power while it is connected
on a bollard on shore through a tow-line while bollard pull testing. This force applied
on the line is represented in tons and is defined as:
Ttot
TBP = (2.70)
g
Bollard pull condition is really important for tug boats among other types of vessels.
However, it should be acknowledged that tug boats usually operate in really low
power while high power is used only for a short period of time, which is usually
less than the full power that bollard pull condition indicates. Though, full power is
mainly used only during bollard pull testing during the vessel’s sea trials.
2.10 Modelling methodologies

The CFD modelling of a propeller indicates a relative motion between the stationary
parts (i.e. static parts such as the nozzle, gear case housing, hull) and the rotor (i.e.
rotating parts such as the propeller blades, hub) of a propulsion unit and the vehicle
(e.g. ship, airplane) that this propulsion unit is mounted on. Two methodologies
have been widely used within the industry through the years among many other
methodologies that are trying to capture the propeller motion. Those two method-
ologies are the Multiple Reference Frame (MRF) and the Sliding Mesh Interface
30
2. Theory
(SMI) (Tabib, Siddiqui, Rasheed, & Kvamsdal, 2017).
MRF is a steady state approach, which indicates a frozen rotor hypothesis. The
RANS equations are solved for both the stationary parts and the rotor while the
rotation effect of the rotor is taken into account by including the coriolis and cen-
trifugal forces in additional source terms in the momentum equations.
SMI is a transient approach, which indicates a rotation of the rotor relative to the
stationary parts. In a similar way as before, the RANS equations are solved for both
the stationary parts and the rotor while the rotation effect of the rotor and the un-
steady interactions between the stationary parts and the rotor are taken into account
by solving the conservation equations on a moving (i.e. sliding) mesh. Moreover,
the rotation effect of the moving volume mesh is taken into account by including
the mesh motion flux in the face mass flux computation in the convective terms of
the governing equations. Finally, the position of the mesh of the rotor is updated
after every time step.
Here it should be mentioned that for both the MRF and SMI approaches, an Ar-
bitrary Mesh Interpolation (AMI) takes place between the stationary parts and the
rotor.
Comparing the two numerical modelling methodologies, it is concluded that the

SMI approach reproduces representative hydrodynamic interactions such as the tip
vortices, steep velocity gradients etc. and the transient behavior of the wake region
is more realistic. On the other hand, MRF approach is weaker on calculating these
significant details. Moreover, SMI shows higher accuracy on predicting the power
and thrust. Nevertheless, it should be considered that SMI is much more expensive
in terms of computational time compared to the MRF methodology and as a result,
the use of MRF is often more realistic (J. Liu, Lin, & Purimitla, 2016).
31
2. Theory
32
3
Geometry
In this chapter, the main particulars of the vessel and the machinery that are used
within this project are presented. Moreover, an explanation and representation of
the geometries follows.
3.1 Main particulars

The classification and the dimensions of the vessel that is used within this study
as well as the machinery, which includes the main engines, the propulsion and the
propeller specifications, are presented in Table 3.1.
Classification Harbour tug boat
Dimensions Length overall, LOA 26.50 m

Rule length, LP P 22.60 m
Max. breadth, B 11.40 m
Draft (design), T 5.50 m
Machinery Main Engines

Make 2x Caterpillar
Type B-rated 3516E
Effect, each engine 2100 kW
Rpm 1600
Propulsion
Make 2x Caterpillar
Type Simplified MTA627
Propellers
Make 2x Caterpillar
Type FPP
Diameter, DP 2.70 m
Expanded blade area ration, EAR 0.67
Design pitch ratio at 0.7r 1.136
No. blades, Z 4
Table 3.1: Main particulars
33
3. Geometry
3.2 Azimuth thruster

In Figure 3.1, the starboard ducted azimuth thruster (i.e. azimuth thruster that
includes a nozzle) that is studied within this project is presented. During the CFD
simulations the thruster is split into four different components: the propeller, which
is also split into the four blades and the hub, the nozzle, the gear case housing and
the lid. It should be mentioned that the lid is a geometry that does not exist in
reality since the azimuth thruster is connected directly to the hull. However, its
presence is important in order to have a closed geometry at the top of the gear case
housing so that no strange flow is created around this region.
Figure 3.1: Starboard ducted azimuth thruster.
3.3 Hull
In Figure 3.2, the azimuth thrusters and the hull that are studied within this project
are presented. The lid that is mentioned above is removed and the gear case hous-
ing is connected directly to the hull. Moreover, the hull is split into two different
components: the main hull and the skeg, which is the "fin" under the hull that is
used for stability during sailing.
Moreover, the azimuth thruster is rotated 6◦ around x-axis and 6◦ around y-axis as
shown in Figure 3.3.
Finally, it is important to keep in mind that during the CFD simulations a symmetry
plane is used on top of the domain while the water surface in Figure 3.2 is used only
for visualization and understanding of the geometries.
34
3. Geometry
Figure 3.2: Ducted azimuth thrusters and hull.
(a) Side view. (b) Rear view.
Figure 3.3: Ducted azimuth thrusters and hull.
35
3. Geometry
36
4
Methodology
In this chapter, the methodology used in order to accomplish the CFD simulations
is presented. In the first sections the three main steps of CFD (i.e. pre-processing,
solving and post-processing) are presented while in the end the system used to run
OpenFOAM is demonstrated.
As already mentioned in Section 1.4.1, the hydrodynamic interactions between the

propeller, the nozzle, the gear case housing and the hull need to be investigated.
Thus, the following cases are defined and then investigated separately:
• Case 1:
The propeller and the nozzle in open water condition,
• Case 2:
The propeller, the nozzle and the gear case housing in open water condition,
• Case 3:
The propeller, the nozzle, the gear case housing and the hull in behind condi-
tion.
In order to achieve such an investigation, as also mentioned in Section 2.1, the input
problem parameters need to be defined and specific tools need to be developed for
analyzing the results. Thus, the following sections present the mandatory steps for
the setup of a series of investigations within this study project.
4.1 Pre-processing
4.1.1 Definition of the computational domain
The main geometries used within this study project are modelled prior to this study
by Caterpillar Propulsion AB, using Computer Aided Design (CAD) software pack-
ages. These geometries are later provided to the author and researcher of this
project. However, small geometrical changes are made by the author using the
Computer Aided Engineering (CAE) software ANSA pre-processor v.19.0.1. The
flow domain is then defined using ANSA pre-processor, which is built around the
main geometries. Here it should be noted that all the simulation models are defined
in full-scale.
In Figure 4.1, the geometry of Case 1 is presented. A shaft that is connected to the
hub and to the inlet of the domain is added in order to get rid of the stagnation
point that would be created at the center of the hub in case the shaft was not added.
37
4. Methodology
The velocity of the fluid would end up being zero at the stagnation point and thus,
strange flows at this region would be created.
Figure 4.1: Geometry of Case 1.
In Figure 4.2, the domain of Case 1 is presented. The dimensions of the domain are:
• Diameter d = 20DP = 54m
• Length L = 30DP = 81m
Figure 4.2: Domain of Case 1.
38
4. Methodology
In Figure 4.3, the geometry of Case 2 is presented while the domain of Case 2 is
identical to Case 1. As already mentioned in Chapter 3, the lid is an additional
geometry that does not exist in reality since the azimuth thruster is connected
directly to the hull. However, its presence is important in order to have a closed
geometry so that no strange flow is created around this region.
(a) Azimuth thruster pressure side. (b) Azimuth thruster suction side.
In Figure 4.4, the geometry of Case 3 is presented,
39
4. Methodology
while in Figure 4.5, the domain of Case 3 is presented. The dimensions of the
domain are:
• Width W ≈ 20DP = 2.3LP P = 52m
• Height H ≈ 20DP = 2.3LP P = 52m
• Length L ≈ 58DP = 6.9LP P = 156m
Figure 4.5: Domain of Case 3.
Looking at both the geometry and the domain of Case 3, it is clear that the the
hull geometry is split into half and the reason is due to the geometrical symmetry.
Thus, there is no need of simulating the whole geometry since this would result
into a much more expensive computational simulation. A symmetry boundary layer
during the CFD simulations is set on the top and side boundaries instead.
4.1.2 Grid generation

The grid that is used within this study is generated with ANSA pre-processor, using
semi-automated functions and then manual corrections. The grid is a non-uniform
mesh where finer mesh size is used in regions with large geometrical variations and
coarser mesh size in regions with smaller geometrical variations.
Initial step of the meshing is the creation of a surface mesh with triangle shapes.
Then, layers are created around the boundaries, which are a type of volume mesh
that are created by the extrusion of the triangle surface mesh. In the next step, a
volume mesh is created within the whole domain, which consists of elements with
pyramid, tetrahedral and hexahedral shapes.
40
4. Methodology
In Figure 4.6, the different types of mesh used in Case 1 are presented as an example.
More precisely, the surface mesh appearing on the blades, the hub and the inner part
of the nozzle is highlighted with magenta color, the layers around the solid geometries
are highlighted with green and the volume mesh is highlighted with brown. Finally,
the surface mesh on the nozzle is highlighted with lilac color.
(a) Mesh of the (b) Mesh of the

rotor. rotor and the nozzle.
(c) Demonstration of the different types of mesh.
Figure 4.6: Initial mesh of Case 1.
Finally, the triangle (i.e. trias) and the square (i.e. quads) elements of the surface
mesh are converted into polygons, the extruded triangle shapes of the layers are
converted into extruded polygon shapes and the pyramid, tetrahedral (i.e. tetras),
pentahedral (i.e. pentas) and hexahedral (i.e. hexas) elements of the volume mesh
are converted into polyhedral elements.
41
4. Methodology
In Figure 4.7, the converted polyhedral mesh of Case 1 is presented.
Figure 4.7: Final mesh of Case 1.
The cell sizes vary from 0.027% · DP ≈ 0.75mm at regions that require really high
accuracy (i.e. leading and trailing edge of the propeller blades) to 124.000% · DP ≈
3348.00mm at regions that require lower accuracy (i.e. domain). Moreover, seven
layers are created around the boundaries with a growth factor of 1.5 and the first
layer having a height with an aspect ratio of 0.05 (i.e. first height = base length ·
0.05). In the following tables the total numbers of cells of all the three cases used
within this study are demonstrated. The naming of the following elements and
categories is presented as in ANSA Pre-processor v.19.0.1. The shell elements are
surface elements that are referring to the trias, quads and polygons. On the other
hand, the volume elements are referring to the the pyramids, tetras, pentas, hexas
and polyhedrals. The layers that have already been discussed are part of the volume
elements.
Element Surface mesh Volume mesh Volume mesh

type (Polyhedrals)
No. trias 790,424 1,488,950 0
Shell quads 0 9,590 0
elements polygons 0 0 1,208,839
Total 790,424 1,498,540 1,208,839
pyramids - 1,009,107 0
No. tetras - 6,139,306 0
Volume pentas - 5,514,141 0
elements hexas - 3,324,486 0
polyhedrals - 0 9,813,408
Total - 15,987,040 9,813,408
Table 4.1: Number of cells: Case 1.
42
4. Methodology

type (Polyhedrals)
No. trias 785,710 1,483,579 0
Total 785,710 1,491,216 1,217,205
pyramids - 1,051,531 0
No. tetras - 6,378,624 0
Total - 16,314,565 9,970,230

type (Polyhedrals)
No. trias 970,058 1,786,267 0
Total 970,058 1,803,788 1,415,604
pyramids - 1,187,549 0
No. tetras - 7,371,853 0
Total - 18,741,337 11,161,882
The quality of the final mesh is highly connected to the quality of the surface mesh,
thus the surface mesh is the one with the highest importance and one should put a
lot of effort into it. A detailed presentation of the different surface mesh types and
cell sizes follows in Appendix B.
4.1.3 Identification of the physical and chemical phenomena

that need to be modelled
All the cases are modeled in 3D. In the mean time, the flow is considered incom-
pressible with a constant density ρ = 1025kg/m3 . Finally, the turbulent flow is
modelled since turbulence has a large contribution to the flow.
43
4. Methodology
4.1.4 Definition of the fluid properties

All the fluid properties such as density, pressure and velocity are defined in space
(i.e. x, y, z-directions) and in time during the transient simulations. Steady state
MRF simulations are used for the initialisation of the flow properties and transient
SMI simulations follow in order to have higher accuracy. The results that are finally
used are derived from the transient SMI simulations.
4.1.5 Definition of the boundary conditions

Initial step before applying the boundary conditions in OpenFOAM v1806 is to
break the boundaries into patches, where one patch might include several bound-
ary surfaces. A geometric patch type is then selected for each patch. The patches
are located in the <case>/constant/polyMesh/boundary directory (OpenCFD Ltd,
2018). It should be mentioned that the polyMesh folder is created during the grid
generation step by ANSA pre-processor.
Second step is to define the boundary conditions for each field that needs to be
solved (i.e. velocity U , pressure p). The boundary conditions are located in the
<case>/0/* directory (e.g. <case>/0/U ).
Final step is to define the wall functions for the turbulence models (i.e. turbulent
kinetic energy k, dissipation rate of turbulent kinetic energy ε, specific dissipation
rate of turbulent energy ω, turbulent viscosity field νt ). The wall functions are
treated in the same way as the boundary conditions, which means that the wall
functions are also applied on individual patches. The wall functions are located in
the <case>/0/* directory (e.g. <case>/0/nut).
A detailed presentation of the different patches, geometric patch types, boundary

conditions and wall functions that are used within this project study follows. The
patches that are demonstrated in Table 4.4, Table 4.5 and Table 4.6 that include
an asterisk (*) are split into more components, thus more patches in the actual
OpenFOAM case setup files. More specifically, the following patches are split into:
• Blades
– Blade 000
– Blade 090
– Blade 180
– Blade 270
• Nozzle
– Nozzle inner
– Nozzle outer
• Shaft
– Shaft inner
– Shaft outer
44
4. Methodology
• Azimuth
– Azimuth body (MTA) inner
– Azimuth body (MTA) outer
– Stay right
– Stay left
• Domain (in Case 3)
– Domain inner
– Domain outer
– Domain top
– Domain bottom
It should be mentioned that the naming of the patches is free of choice. On the other
hand, the geometric patch types, boundary conditions and wall functions should be
set in openFOAM as they appear in the following tables. A thorough explanation
of the patch types and the wall functions is given by F. Liu (2017) for a full under-
standing of the wall functions.
The boundary conditions used within this project are explained in the following list
(OpenCFD Ltd, 2018):
• calculated: This boundary condition is not evaluated. It is rather assumed
that the value is assigned vie field assignment.
• fixedValue: This boundary condition sets a fixed value constraint.
• zeroGradient: This boundary condition sets a zero gradient condition from
the patch internal field to the patch faces.
• cyclicAMI: This boundary condition determines a cyclic condition between
a pair of boundaries where communication between the patches is achieved
using an Arbitrary Mesh Interpolation (AMI).
• movingWallVelocity: This boundary condition sets a velocity condition for
cases with moving walls.
• slip: This boundary condition sets a slip constraint.
Finally, the patches, patch types, boundary conditions and wall functions are pre-
sented in Table 4.4, Table 4.5 and Table 4.6. Here it should be mentioned that the
slip boundary condition is the same as symmetry boundary condition for the fields
over which the boundaries are imposed. However, symmetry boundary condition
alters also the fields internal to the solvers. Thus, a slip condition is used.
45
46
Patch Geometric patch type Boundary conditions
boundary U p p_rgh
Inlet patch fixedValue zeroGradient zeroGradient
Outlet patch zeroGradient fixedValue fixedValue
4. Methodology
Domain patch slip slip slip

Blades∗ wall movingWallVelocity zeroGradient zeroGradient
Hub wall movingWallVelocity zeroGradient zeroGradient
Nozzle∗ wall fixedValue zeroGradient zeroGradient
Shaft∗ wall fixedValue zeroGradient zeroGradient
cyclic_half0 cyclicAMI cyclicAMI cyclicAMI cyclicAMI
Patch Wall functions

nut (νt ) epsilon (ε) k (κ) omega (ω) alpha.water
Inlet calculated fixedValue fixedValue fixedValue fixedValue
Outlet calculated zeroGradient zeroGradient zeroGradient zeroGradient
Domain slip slip slip slip slip
Blades∗ nutUSpaldingWallFunction epsilonWallFunction kqRWallFunction omegaWallFunction zeroGradient
Hub nutUSpaldingWallFunction epsilonWallFunction kqRWallFunction omegaWallFunction zeroGradient
Nozzle∗ nutUSpaldingWallFunction epsilonWallFunction kqRWallFunction omegaWallFunction zeroGradient
Shaft∗ nutUSpaldingWallFunction epsilonWallFunction kqRWallFunction omegaWallFunction zeroGradient
cyclic_half0 cyclicAMI cyclicAMI cyclicAMI cyclicAMI cyclicAMI
Table 4.4: Presentation of the boundary conditions: Case 1.

boundary U p p_rgh
Domain patch slip slip slip
Azimuth∗ wall fixedValue zeroGradient zeroGradient
Lid wall fixedValue zeroGradient zeroGradient

Domain slip slip slip slip slip
Azimuth∗ nutUSpaldingWallFunction epsilonWallFunction kqRWallFunction omegaWallFunction zeroGradient
Lid nutUSpaldingWallFunction epsilonWallFunction kqRWallFunction omegaWallFunction zeroGradient
47
4. Methodology
boundary U p p_rgh
48
Domain∗ patch slip slip slip
4. Methodology

Azimuth∗ wall fixedValue zeroGradient zeroGradient
Cone wall fixedValue zeroGradient zeroGradient
Hull wall fixedValue zeroGradient zeroGradient
Skeg wall fixedValue zeroGradient zeroGradient

Domain∗ slip slip slip slip slip
Azimuth∗ nutUSpaldingWallFunction epsilonWallFunction kqRWallFunction omegaWallFunction zeroGradient
Cone nutUSpaldingWallFunction epsilonWallFunction kqRWallFunction omegaWallFunction zeroGradient
Hull nutUSpaldingWallFunction epsilonWallFunction kqRWallFunction omegaWallFunction zeroGradient
Skeg nutUSpaldingWallFunction epsilonWallFunction kqRWallFunction omegaWallFunction zeroGradient

4. Methodology
4.2 Solving
4.2.1 Integration of the governing equations of the flow over

the finite control volumes of the domain
The numerical method used to model the flow and capture the effects of turbu-
lence is the RANS model. Moreover, the Menter SST κ − ω turbulent model is
used to model the boundary layer and the effect of the unresolved fully turbulent
regions far from the wall. Both RANS and Menter SST κ − ω model are specified
in <case>/constant/RASProperties directory.
4.2.2 Discretisation of the integrated equations into alge-

braic equations
The numerical schemes used in this project are:
• Interpolation schemes: Central differencing (i.e. linear in OpenFOAM) is
used, which is a second order, unbounded scheme. Interpolation schemes are
specified in <case>/system/fvSchemes/interpolationSchemes.
• Surface normal gradient schemes: Limited explicit non-orthogonal cor-
rection (i.e. limited corrected in OpenFOAM) is used. Surface normal gradient
schemes are specified in <case>/system/fvSchemes/snGradSchemes.
• Gradient schemes: Gaussian integration with central differencing (i.e. Gauss
linear in OpenFOAM) is used, which is a second order, unbounded scheme.
Gradient schemes are specified in <case>/system/fvSchemes/gradSchemes.
• Laplacian schemes: Gaussian integration with central differencing and a
blend of corrected and uncorrected numerical behaviour (i.e. Gauss linear
limited corrected in OpenFOAM) is used. Laplacian schemes are specified in
<case>/system/fvSchemes/laplacianSchemes.
• Divergence schemes:
– Gaussian integration with upwind differencing
(i.e. Gauss upwind in OpenFOAM), which is a first order, bounded
scheme. This scheme is used for the ∇ · (ρU κ) term, represented by the
div(phi,k) and for the ∇·(ρU ω) term, represented by the div(phi,omega).
– Gaussian integration with linear upwind differencing
(i.e. Gauss linearUpwind in OpenFOAM), which is a first/second order,
bounded scheme. This scheme is used for the ∇·(ρU U ) term, represented
by the div(phi,U).
Divergence schemes are specified in <case>/system/fvSchemes/divSchemes.
• Time schemes:
Euler (i.e. Euler in OpenFOAM) is used in transient simulations, which is
a first order, bounded and implicit scheme. Time schemes are specified in
<case>/system/fvSchemes/ddtSchemes.
49
4. Methodology
4.2.3 Solving of the algebraic equations by the use of itera-

tive method
The algorithms used in this project are:
• SIMPLE for the steady state MRF simulations,
• PIMPLE for the transient SMI simulations.
The following iterative linear solvers are used for both the steady state MRF and
transient SMI simulations:
• Generalised geometric-algebraic multi-grid (GAMG) (i.e. GAMG in
OpenFOAM) with a Gauss-Seidel smoother (i.e. GaussSeidel in Open-
FOAM) for the pressure p, which is specified in <case>/system/fvSolution/
solvers/p.
• Smoother solver (i.e. smoothSolver in OpenFOAM) with a symmetric
Gauss-Seidel smoother (i.e. symGaussSeidel in OpenFOAM) for the ve-
locity U , the turbulent kinetic energy k, the dissipation rate of turbulent
kinetic energy ε and the specific dissipation rate ω, which is specified in
<case>/system/fvSolution/solvers/
<U|k|...>.
while the transient simulations also need:
• Generalised geometric-algebraic multi-grid (GAMG) with a Diago-
nal incomplete-Cholesky with Gauss-Seidel (symmetric) smoother (i.e.
DICGaussSeidel in OpenFOAM) for the pressure correction, which is specified
in <case>/system/fvSolution/solvers/"(pcorr.*)".
• Generalised geometric-algebraic multi-grid with a Diagonal incomplete-
Cholesky with Gauss-Seidel (symmetric) smoother for the pressure range
that matters, thus the pressure without the hydrostatic pressure which is for-
mulated as p_rgh = p−ρgh. GAMG is specified in <case>/system/fvSolution/
solvers/p_rgh.
Moreover, two types of modelling regarding the mesh take place within the project:
• MRF, which is specified in <case>/system/fvOptions.
• SMI, which is specified in <case>/constant/dynamicMeshDict.
4.3 Post-processing
4.3.1 Visualization of the simulation results
Initial step of the post-processing is to look at flow variable fields, which are a
results of the CFD simulations. The following flow fields are observed thoroughly
in ParaView v.5.6.0:
• Velocity field U
• Pressure field p
• The turbulent kinetic energy field k
• The specific dissipation rate of turbulent energy field ω
Next step is to plot the convergence of the forces F and moments M of all the cases
50
4. Methodology
and for each advance coefficient J and part (e.g. blades, nozzle, hull) that are lo-
cated in <case>/postProcessing/<part>/0/forces.dat. This is done in MATLAB
R2018b.
4.3.2 Process of the numerical results

A mean value of the forces and moments of the last 720 iterations, which is equivalent
to the last two full revolutions of the propeller of the total number of ten revolutions,
of each case, advance coefficient J and part that are mentioned above is calculated.
For a better understanding, these mean values are demonstrated below:
• Mean force F , which is either expresses as mean thrust T or
mean resistance R,
• Mean moment M , which is expressed as mean torque Q.
Later on, by the use of these mean values the following variables are calculated:
• Mean thrust coefficient KT ,
• Mean torque coefficient 10 · KQ ,
• Mean open water efficiency η0 .
Note: Due to simplification purposes, from now on all the forces, moments, co-
efficients and open water efficiencies will not be referred as mean values and the
overline of these symbols will be neglected, even though they are all derived from a
mean value.
The thrust deduction factors t are then calculated in order to describe the in-
crease of the resistance due to the suction of the propeller, determine the generated
thrust and obtain the different contributions of the gear case housing and the hull
from a thrust perspective:
• Case 1
0
t= =0 (4.1)
Tblades + Tnozzle
• Case 2
−Rgear_case_housing
t= (4.2)
Tblades + Tnozzle
• Case 3
−(Rgear_case_housing + Rhull )
t= (4.3)
Tblades + Tnozzle
Next step is to calculate the merit coefficients mc , which are a replacement of the
total propulsive efficiency ηD :
• Case 1
[(KT blades + KT nozzle )/π]3/2

mc = (4.4)
KQblades
51
4. Methodology
• Case 2
[(KT blades + KT nozzle + KT gear_case_housing )/π]3/2
mc = (4.5)
KQblades
• Case 3 (excluding hull)
[(KT blades + KT nozzle + KT gear_case_housing )/π]3/2
mc = (4.6)
KQblades
• Case 3 (including hull)
[(KT blades + KT nozzle + KT gear_case_housing + KT hull )/π]3/2
mc = (4.7)
KQblades
Here it should be mentioned that the merit coefficient of Case 3 is calculated by
both including and excluding the contribution of the hull. This is done in order to
obtain the flow field that is derived from the case simulation with the hull (i.e. Case
3) but also at the same time to include or exclude the contribution of the hull itself.
Then, the open water characteristic diagrams for all the cases are generated.
Both the MRF and SMI simulation results that have been performed in OpenFOAM
are plotted for comparison and evaluation purposes.
A complete open water diagram is plotted for Case 1 with both MRF and SMI
results in order to get an overall idea of the behaviour of this specific arrangement
in bollard pull and free sailing conditions and in both steady state and transient
approaches. The values that are plotted are the following:
• MRF results for J = 0.00, 0.10, 0.20, 0.30, 0.40, 0.50, 0.60, 0.65, 0.70, 0.75, 0.80
0.85, 0.90, 1.00,
• SMI results for J = 0.00, 0.20, 0.40, 0.60, 0.80.
The SMI results are chosen for J = 0.00 and the MRF results for all the other J val-
ues and a combined MRF-SMI open water characteristics diagram is created. This
diagram, that from now on is referred to as "hybrid diagram", is later used as Case 1.
A complete open water diagram is also plotted for Case 2 but this time with only SMI
results in order to get an overall idea of the behaviour of this specific arrangement
in bollard pull and free sailing conditions and in transient approach. The values
that are plotted are the following:
• SMI results for J = 0.00, 0.20, 0.40, 0.60, 0.70, 0.80, 0.90, 1.00.
In the end, only the values for J = 0.00 are plotted for Case 3 with SMI results but
this time with the only focus in bollard pull condition and in transient approach.
Finally, the open water case (i.e. Case 1) is compared with the Wageningen CD
Series in three different ways (i.e. brake power similarity, thrust power similarity
and geometrical similarity) in order to develop an evaluation tool which can be used
in the future in order to evaluate the early stages of a project.
It should be noted that all the above calculations and plots are done in MATLAB.
52
4. Methodology
4.4 System used to run OpenFOAM

A High-Performance Computing (HPC) system is used since parallel processing is
needed. The CFD simulations within this project are run on a cluster owned by
Caterpillar Inc. The simulations are decomposed into 144 cores and computed in
parallel, something which is known as domain decomposition in OpenFOAM.
Through the domain decomposition, the geometry and the flow fields are split and
then distributed into separate processors, which are then separately solved. After
the completion of the simulation the decomposed cases are reconstructed into one
case.
The steady state simulations take about two hours to be completed while the tran-
sient simulations are run overnight.
53
4. Methodology
54
5
Results and data analysis
In this chapter, the results of all the CFD simulations are presented. In addition,
discussion of the results and data analysis is achieved. In the first section of this
chapter a data analysis in both free sailing and bollard pull conditions is achieved
in order to get an overall idea of the behaviour of all the cases. The hydrodynamic
interactions are presented in the second section only in bollard pull condition since
this is the main focus of this study project.
5.1 Data analysis in free sailing and bollard pull

conditions
The performance characteristics of a propeller or generally a propulsion unit are
usually presented by open water characteristic diagrams. Thus, this type of dia-
grams are used in order to conduct various comparisons.
5.1.1 Comparison of MRF/SMI

In this section, a comparison of the two different modelling methodologies used
within this project is achieved. For this type of comparison, results from both MRF
and SMI of Case 1 have been used.
In Figure 5.1, two open water characteristic diagrams derived from the SMI (see
Table 5.2) and MRF (see Table 5.3) results are presented. Since this comparison is
done only for Case 1, the hydrodynamic coefficients discussed are the thrust coeffi-
cients of the blades KT blades and nozzle KT nozzle , the torque coefficient of the blades
10 · KQblades and the open water efficiency η0 . Moreover, identical rotational speed
n = 191rpm for all the advance coefficients J has been used. In this way, the differ-
ence of the resulting thrust and torque can be observed.
In Figure 5.2, the already mentioned hybrid open water characteristics diagram that
combines the SMI results for J = 0.00 and the MRF results for J = 0.10 − 1.00 is
presented.
55
5. Results and data analysis
Figure 5.1: Open water characteristics diagram: MRF/SMI comparison

of Case 1.
Figure 5.2: Open water characteristics diagram: Hybrid MRF/SMI

diagram of Case 1.
The values of the hybrid diagram are presented in Table 5.1. The complete results
are fully presented in Table 5.2 for the SMI and in Table 5.3 for the MRF.
56
Blades Nozzle Open water
efficiency
J T KT Q 10KQ T KT Q 10KQ η0
[−] [kN] [-] [kNm] [-] [kN] [-] [kNm] [-] [-]
0.00 190.98 0.3460 94.15 0.6317 208.20 0.3772 -0.09 -0.0006 0.0000
0.10 190.97 0.3460 93.22 0.6255 158.92 0.2879 -0.14 -0.0009 0.1613
0.20 188.07 0.3407 92.10 0.6179 126.03 0.2283 -0.13 -0.0009 0.2931
0.30 183.15 0.3318 90.20 0.6052 97.40 0.1764 -0.12 -0.0008 0.4010
0.40 176.08 0.3190 87.44 0.5867 72.84 0.1320 -0.11 -0.0008 0.4893
0.50 166.82 0.3022 83.78 0.5621 52.16 0.0945 -0.10 -0.0007 0.5616
0.60 155.19 0.2811 79.09 0.5306 35.00 0.0634 -0.08 -0.0005 0.6201
0.65 148.43 0.2689 76.31 0.5120 27.61 0.0500 -0.07 -0.0005 0.6444
0.70 139.82 0.2533 72.72 0.4879 17.96 0.0325 -0.06 -0.0004 0.6526
0.75 119.69 0.2168 64.10 0.4301 -0.37 -0.0007 -0.05 -0.0004 0.5999
0.80 105.18 0.1905 57.75 0.3875 -11.90 -0.0216 -0.05 -0.0003 0.5552
0.85 89.97 0.1630 51.14 0.3431 -23.47 -0.0425 -0.06 -0.0004 0.4749
0.90 73.63 0.1334 44.09 0.2958 -35.27 -0.0639 -0.06 -0.0004 0.3365
1.00 37.47 0.0679 28.35 0.1902 -59.44 -0.1077 -0.06 -0.0004 0.3330
Table 5.1: Hybrid MRF/SMI results: Case 1.
57
The total forces and moments extracted from the OpenFOAM post-processing sim-
ulation files for both MRF and SMI, in bollard pull condition (i.e. J = 0.00) and
for all the three directions are presented in the following figures.
x-direction
(a) Propeller (b) Inner part of (c) Outer part of

blades. the nozzle. the nozzle.
Figure 5.3: MRF simulation of Case 1 at J = 0.00.

Figure 5.4: SMI simulation of Case 1 at J = 0.00.
y-direction

58

z-direction


59
Looking at Figure 5.1 and by comparing the complete results for SMI in Table
5.2 and the complete results for MRF in Table 5.3, it is obvious that most of the
MRF results are following the SMI results. However, a big difference is observed at
J = 0.00 that bollard pull condition occurs. Observing the convergence plots above,
it is obvious that the x-direction results seem to be converged for both MRF and
SMI. However, the MRF results for y-direction and z-direction are not converged
while the SMI results are. Moreover, it should be mentioned that an oscillation of
the results appears for the SMI. Nevertheless, this oscillation is constant and thus
an average value of the last 720 iterations is taken.
Considering that the results for bollard pull condition seem more realistic by taking
into account the CFD results of other similar studies and by observing the conver-
gence of SMI and the convergence issues of MRF, the already mentioned statement
that the transient SMI methodology gives much more accurate results
compared to the steady state MRF for low J values can be supported. Mean-
while, the results for medium J values computed with the MRF approach are quite
reasonable.
Since the transient SMI methodology gives much more accurate results, the ideal
would be to run every simulation with SMI. On the other hand, SMI is much more
computationally expensive compared to MRF.
60
efficiency
[−] [kN] [-] [kNm] [-] [kN] [-] [kNm] [-] [-]
0.00 190.98 0.3460 94.15 0.6317 208.20 0.3772 -0.09 -0.0006 0.0000
0.20 188.78 0.3420 93.29 0.6259 132.26 0.2396 -0.12 -0.0008 0.2958
0.40 176.08 0.3190 87.44 0.5867 72.84 0.1320 -0.11 -0.0008 0.4893
0.60 157.22 0.2848 80.61 0.5408 36.29 0.0657 -0.09 -0.0006 0.6190
0.80 105.18 0.1905 57.75 0.3875 -11.90 -0.0216 -0.05 -0.0003 0.5552
Table 5.2: SMI results: Case 1 (see Figure 5.2).

efficiency
[−] [kN] [-] [kNm] [-] [kN] [-] [kNm] [-] [-]
0.00 210.64 0.3816 102.36 0.6868 200.99 0.3641 -0.12 -0.0008 0.0000
0.10 190.97 0.3460 93.22 0.6255 158.92 0.2879 -0.14 -0.0009 0.1613
0.20 188.07 0.3407 92.10 0.6179 126.03 0.2283 -0.13 -0.0009 0.2931
0.30 183.15 0.3318 90.20 0.6052 97.40 0.1764 -0.12 -0.0008 0.4010
0.40 176.08 0.3190 87.44 0.5867 72.84 0.1320 -0.11 -0.0008 0.4893
0.50 166.82 0.3022 83.78 0.5621 52.16 0.0945 -0.10 -0.0007 0.5616
0.60 155.19 0.2811 79.09 0.5306 35.00 0.0634 -0.08 -0.0005 0.6201
0.65 148.43 0.2689 76.31 0.5120 27.61 0.0500 -0.07 -0.0005 0.6444
0.70 139.82 0.2533 72.72 0.4879 17.96 0.0325 -0.06 -0.0004 0.6526
0.75 119.69 0.2168 64.10 0.4301 -0.37 -0.0007 -0.05 -0.0004 0.5999
0.80 105.18 0.1905 57.75 0.3875 -11.90 -0.0216 -0.05 -0.0003 0.5552
0.85 89.97 0.1630 51.14 0.3431 -23.47 -0.0425 -0.06 -0.0004 0.4749
0.90 73.63 0.1334 44.09 0.2958 -35.27 -0.0639 -0.06 -0.0004 0.3365
1.00 37.47 0.0679 28.35 0.1902 -59.44 -0.1077 -0.06 -0.0004 -0.3330
61
Table 5.3: MRF results: Case 1 (see Figure 5.2).

5.1.2 Comparison of Case 1/Case 2

In Figure 5.9, two open water characteristic diagrams derived from Case 1 and Case
2 are presented. The thrust coefficient of the gear case housing KT gear_case_housing
is also included among the other coefficients. Moreover, as in the previous section
identical rotational speed n = 191rpm for all the advance coefficients J has been
used.
Figure 5.9: Open water characteristics diagram: Case 1/Case 2 comparison.
The values of Case 2 are presented in Table 5.4. It should be mentioned that while
the hydrodynamic coefficients of Case 1 are derived from the hybrid MRF/SMI re-
sults, the coefficients of Case 2 are obtained only from the SMI results since only
SMI was used in this case in order to have slightly better accuracy for the gear case
housing within all the open water diagram.
Looking at Figure 5.9, it is noticed that the propeller thrust is really similar in
both cases for J = 0.00 − 0.40 while the difference starts to increase after this point.
For J = 0.70 − 1.00 the thrust is much higher for Case 2. The difference of the pro-
peller torque appears to be quite large between J = 0.00 − 0.40 while after this
point the difference gets even larger with the largest difference occurring between
J = 0.70 − 1.00. Looking at the nozzle thrust, it is clear that in Case 2 the thrust
is much higher than in Case 1 within the whole diagram. Moreover, it can be seen
that the gear case housing thrust remains constant within the whole diagram.
Since the thrust is having a negative value, it is considered as resistance instead.
Finally, it is observed that the open water efficiency is lower for Case 2 between
J = 0.00 − 0.70. Looking specifically at J = 0.00, the open water efficiency looses
its meaning (i.e. open water efficiency is always zero since the velocity of the ship is
zero). In this moment, the merit coefficient is introduced, however this is further
discussed in Section 5.2.
62
Blades Nozzle Gear Case Housing Open water
efficiency
J T KT Q 10KQ T KT Q 10KQ T KT Q 10KQ η0
[−] [kN] [-] [kNm] [-] [kN] [-] [kNm] [-] [kN] [-] [kNm] [-] [-]
0.00 193.00 0.3496 94.99 0.6373 242.24 0.4388 -0.10 -0.0007 -38.81 -0.0703 -1.10 -0.0074 0.0000
0.20 191.09 0.3462 94.14 0.6316 160.72 0.2912 -0.10 -0.0007 -37.56 -0.0680 1.70 0.0114 0.2869
0.40 180.09 0.3262 89.84 0.6028 101.65 0.1841 -0.08 -0.0005 -33.40 -0.0605 0.49 0.0033 0.4751
0.60 162.42 0.2942 82.57 0.5540 59.15 0.1072 -0.06 -0.0004 -30.25 -0.0548 -2.72 -0.0183 0.5974
0.70 149.66 0.2711 77.14 0.5176 42.56 0.0771 -0.05 -0.0003 -28.49 -0.0516 -4.50 -0.0302 0.6384
0.80 121.62 0.2203 65.00 0.4361 14.84 0.0269 -0.07 -0.0005 -28.56 -0.0517 -3.09 -0.0207 0.5707
0.90 88.37 0.1601 50.03 0.3357 -9.49 -0.0172 -0.07 -0.0005 -29.70 -0.0538 -2.64 -0.0177 0.3801
1.00 55.02 0.0997 35.85 0.2405 -33.30 -0.0603 -0.13 -0.0009 -30.21 -0.0547 -3.53 -0.0237 -0.1018
Table 5.4: SMI results: Case 2.
63
In Figure 5.10, the pressure distribution in bollard pull condition of Case 1 is pre-
sented. The pressure distribution seems normal with higher pressure at the pressure
side and lower pressure at the suction side. Looking at Figure 5.11 that presents
the velocity field of Case 1, it is clearly demonstrated by the velocity magnitude
field that the flow is straight. A small decrease of the velocities in the centerline is
observed, which is due to the hub that is blocking the flow.
(a) Pressure side. (b) Suction side.
Figure 5.10: Pressure distributions at J = 0.00: Case 1.
Figure 5.11: Velocity field at J = 0.00: Case 1.
In Figure 5.12, the pressure distribution of Case 2 is presented. The pressure dis-
tribution seems normal. Furthermore, it is observed that Case 2 has an inflow with
lower pressure compared to Case 1 due to the gear case housing that is blocking
the flow. Thus, the pressure difference in Case 2 is higher, which indicates higher
acceleration of water from the pressure side to the suction side (i.e higher advance
velocity). As a result, the higher thrust power that is created from the propeller
towards the outflow is explained. Looking at Figure 5.13 that presents the velocity
field of Case 2, it is observed that some fluctuations have been introduced to the
flow. Most likely this result occurs due to the change of the quality of the mesh
within this region. However, further investigation should be done in order to answer
whether this result is dependent on the mesh or whether it occurs due to physical
phenomena.
64
In addition, looking at Figure 5.14 that presents a closer view of the velocity field of
Case 2, it can be observed that the inlet velocity at the nozzle leading edge is higher
than the inlet velocity at the blades, hence the higher increase of the nozzle thrust
compared to the blade thrust in Case 2. However, due to the increase of resistance
at the same time, the net thrust generated in Case 2 is 0.68% lower than Case 1
(see Table 5.12).
65
(a) Side view. (b) Rear view.
5.1.3 Comparison of Case 1/Case 3

In Table 5.5, the hydrodynamic coefficients derived from Case 3 for the propeller-
axis and in Table 5.6 for the x-axis are presented. It should be mentioned that the
coefficients of Case 3 are obtained only from the SMI results in bollard pull condition
(i.e. J = 0.00). The thrust coefficient of the hull KT hull is also included among the
other coefficients that have already been discussed in the previous sections. More-
over, as in the previous section identical rotational speed n = 191rpm for all the
advance coefficients J has been used.
It is obvious that the results are presented in two different perspectives. By exam-
ining the first perspective (see Table 5.5), the values are derived from the actual
thrust and presented in respect to the propeller-axis that the thrust is created. By
examining the second perspective (see Table 5.6), the values are derived from the
x-component of the actual thrust and presented in respect to the x-axis. Looking at
the same tables, it could be noted that the hull values are calculated only for x-axis
since the resistance of the hull is created only in this direction. Thus, the hull values
are identical in both tables.
Moreover, it is noticed that the propeller thrust is increased significantly compared

to Case 1 as well as Case 2. For a better understanding, it is highly recommended
to look at the diagrams in parallel with Table 5.12. In a similar way, the propeller
torque appears to be as well significantly larger compared to both previous cases.
The nozzle thrust is also increased significantly. Furthermore, the gear case
housing resistance is represented, which is almost identical to Case 2. Then, the
hull resistance is introduced, which is also considered as resistance since it’s nega-
tive. However, the hull resistance is really close to zero since the ship’s speed is zero.
66
Blades Nozzle
(propeller-axis) (propeller-axis)
J T KT Q 10KQ T KT Q 10KQ
[−] [kN] [-] [kNm] [-] [kN] [-] [kNm] [-]
0.00 199.64 0.3617 97.81 0.6563 229.82 0.4163 -38.63 -0.2592
Gear case housing Hull Open water

(propeller-axis) (x-axis) efficiency
[−] [kN] [-] [kNm] [-] [kN] [-] [kNm] [-] [-]
0.00 -35.72 -0.0714 -1.31 -0.0088 -5.08 -0.0092 62.52 0.4195 0.0000
Table 5.5: SMI results (propeller-axis): Case 3.
Blades Nozzle
(x-axis) (x-axis)
J T KT Q 10KQ T KT Q 10KQ
[−] [kN] [-] [kNm] [-] [kN] [-] [kNm] [-]
0.00 198.55 0.3597 97.27 0.6527 228.56 0.4140 -38.42 -0.2578
Gear case housing Hull Open water

(x-axis) (x-axis) efficiency
[−] [kN] [-] [kNm] [-] [kN] [-] [kNm] [-] [-]
0.00 -35.52 -0.0710 -1.30 -0.0087 -5.08 -0.0092 62.52 0.4195 0.0000
Table 5.6: SMI results (x-axis): Case 3.
67
In a similar way as before, the open water efficiency looses its meaning and the total
propulsive efficiency ηD becomes zero. Hence, the merit coefficient is used as a re-
placement of the total propulsive efficiency, which is further discussed in Section 5.2.
In Figure 5.15, the pressure distribution of Case 3 is presented. The pressure dis-
tribution seems reasonable. Furthermore, it is observed that Case 3 has an inflow
similar to Case 2. The increase of the thrust in Case 3 compared to Case 1 is ex-
plained in a similar way as Case 2. Looking at Figure 5.16 that presents the velocity
field of Case 3, it is observed that a small "step" is observed at the top of the flow
and under the hull. Most likely this result occurs due to a similar reason as Case 2,
meaning the change of the quality of the mesh within this region. However, further
investigation should also be done.
Finally, in a similar way as Case 2, the inlet velocity at the nozzle leading edge is
higher than the inlet velocity at the blades, hence the higher increase of the nozzle
thrust compared to the blade thrust in Case 3. However, the net thrust generated
68
in Case 3 is 4.09% lower than Case 1 due to the increase of the resistance (see Table
5.12).
5.1.4 Comparison of Case 1/Wageningen CD Series

In the following figures, two open water characteristic diagrams derived from Case 1
and the Wageningen CD Series are presented for each of all the three different ways
of comparison. For this comparison, the Microsoft Excel propCalcCD2.3.1 tool is
used that has been developed by Caterpillar Propulsion AB.
propCalcCD2.3.1 needs some input parameters, such as brake power PB , shaft effi-
ciency ηS , expanded blade area ratio, propeller diameter DP , ship velocity VS , wake
fraction wT , rotational speed n, design pitch ratio at 0.7r and blade number. By the
use of tabulated data from the Wageningen CD Series as well as from Caterpillar
Propulsion AB test data, this tool predicts the propeller thrust, nozzle thrust, pro-
peller torque, pitch ratio at 0.7r etc. These predicted values are used for evaluation
of the early stages of a project.
Constant brake power PB

In this section, the first comparison of Case 1 and the Wageningen CD Series is
achieved.
In Figure 5.17, the hydrodynamic coefficients of Case 1 and the Wageningen CD Se-
ries are demonstrated. A brake power similarity is achieved between the compared
cases by setting up the same brake power at the Wageningen CD Series as in Case
1. All the hydrodynamic coefficients are included while the pitch ratio at 0.7r is also
plotted, which shows the geometrical similarity between the compared cases.
The brake power PB is calculated by using the torque from the blades Q (see Table
5.2).
J VS [kn] KQ [-] Q [kNm] PB [kW]

0.00 3.3415 0.0632 94.15 1883
0.20 3.3415 0.0626 93.29 1866
0.40 6.6829 0.0587 87.44 1749
0.60 10.0244 0.0541 80.61 1612
0.80 13.3659 0.0387 57.75 1155
Table 5.7: Calculation of the brake power PB by using the SMI results of Case 1.
69
Figure 5.17: Open water characteristics diagram: Case 1/Wageningen CD Series

comparison with constant PB .
Then, the brake power PB together with the ship’s velocity VS as well as a series of
constant parameters such as:
• nS = 1
• EAR = 0.67
• D = 2.7m
• w=0
• n = 191rpm
• Design pitch ratio at 0.7r = 1.136
• No. blades = 4
are set as input parameters into the propCalcCD2.3.1 tool.
Finally, the results of propCalcCD2.3.1 are presented in the following table.
J KT p [-] KT n [-] KQ [-] Pitch ratio at 0.7r [-] Tp [kN] η0 [-]

0.00 0.3276 0.3387 0.0632 1.195 180.90 0.0000
0.20 0.3216 0.2221 0.0626 1.205 177.50 0.2765
0.40 0.2972 0.1310 0.0587 1.211 164.00 0.4644
0.60 0.2623 0.0604 0.0541 1.237 144.80 0.5696
0.80 0.1804 -0.0176 0.0387 1.211 99.60 0.5356
Table 5.8: propCalcCD2.3.1 results: Constant PB .
Looking at Figure 5.17, it is noticed that the propeller thrust of Case 1 is higher
than the one that the Wageningen CD Series suggests. For a better understanding,
it is recommended to look at Table 5.8 for the Wageningen CD Series and at Table
5.12 for Case 1. On the other hand, the propeller torque appears to be identical
(i.e. brake power similarity is achieved). The nozzle thrust behaves in a similar
70
way for most of the advance coefficients while there is a quite big difference between
J = 0.00−0.10 with Case 1 having a higher thrust at bollard pull condition. Finally,
the pitch ratio at 0.7r remains almost identical within the whole diagram for the
Wageningen CD Series, which has values really close to the constant pitch ratio of
Case 1, which is 1.136.
Constant thrust power PT
In this section, the second comparison of Case 1 and the Wageningen CD Series is
achieved.
In Figure 5.18, the hydrodynamic coefficients of Case 1 and the Wageningen CD Se-
ries are demonstrated. A thrust power similarity is achieved between the compared
cases by setting up the same thrust power at the Wageningen CD Series as in Case 1.
The hydrodynamic coefficients and the pitch ratio at 0.7r are included as previously.
Figure 5.18: Open water characteristics diagram: Case1/Wageningen CD Series

comparison with constant PT .
The thrust power PT is calculated by using the thrust from the propeller blades Tp
(see Table 5.2).
71
J VS [kn] Tp [kN] KT p [-] PT [kW]

0.00 3.3415 190.98 0.3460 0
0.20 3.3415 188.78 0.3420 325
0.40 6.6829 176.08 0.3190 605
0.60 10.0244 157.22 0.2848 811
0.80 13.3659 105.18 0.1905 723
Table 5.9: Calculation of the thrust power PT by using the SMI results of
Case 1.
Then, the thrust power PT together with the ship’s velocity VS as well as the same
constant parameters as previously are set as input parameters into propCalcCD2.3.1.
J KT p [-] KT n [-] KQ [-] Pitch ratio at 0.7r [-] PB [kW] η0 [-]

0.00 0.3460 0.3572 0.0691 1.241 2061 0.0000
0.20 0.3420 0.2400 0.0692 1.256 2064 0.2677
0.40 0.3190 0.1454 0.0655 0.163 1953 0.4514
0.60 0.2850 0.0716 0.0606 1.287 1807 0.5615
0.80 0.1905 -0.0136 0.0414 1.232 1234 0.5440
Table 5.10: propCalcCD2.3.1 results: Constant PT .
Looking at Figure 5.18, it is noticed that the propeller thrust appears to be

identical (i.e. thrust power similarity is achieved). On the other hand, the propeller
torque appears to be quite higher for the Wageningen CD Series compared to Case
1. For a better understanding, it is recommended to look at Table 5.10 for the
Wageningen CD Series and at Table 5.12 for Case 1. The nozzle thrust appears to
be quite similar in both compared cases. However, Case 1 achieves a higher nozzle
thrust at bollard pull condition while in the rest of the diagram the Wageningen
CD Series is leading with higher thrust. Finally, the pitch ratio at 0.7r remains
almost identical between J = 0.00 − 0.20 and J = 0.60 − 0.80 while a difference
appears for J = 0.20 − 0.60.
Constant pitch ratio at 0.7r
In this section, the third and last comparison of Case 1 and the Wageningen CD
Series is achieved.
In Figure 5.19, the hydrodynamic coefficients of Case 1 and the Wageningen CD

Series are demonstrated. A geometrical similarity is achieved between the compared
cases by setting a constant pitch ratio at 0.7r.
72
Figure 5.19: Open water characteristics: Case1/CD Series with constant pitch
ratio at 0.7r comparison.
In this case, a geometrical comparison is achieved. Thus, the pitch ratio at 0.7r is
the only input parameter into propCalcCD2.3.1, which is same as before:
• Design pitch ratio at 0.7r = 1.136
J KT p [-] KT n [-] KQ [-] PB [kW] Tp [kN] η0 [-]

0.00 0.3048 0.3152 0.0570 1698 168.24 0.0000
0.10 0.3025 0.2528 0.0565 1685 166.99 0.1563
0.20 0.2951 0.1992 0.0553 1647 162.91 0.2847
0.30 0.2834 0.1530 0.0533 1590 156.45 0.3906
0.40 0.2670 0.1121 0.0507 1512 147.40 0.4759
0.50 0.2457 0.0748 0.0473 1411 135.63 0.5390
0.60 0.2190 0.0396 0.0431 1285 120.88 0.5728
0.70 0.1855 0.0045 0.0378 1127 102.40 0.5599
0.80 0.1440 -0.0320 0.0312 930 79.51 0.4571
0.90 0.0941 -0.0709 0.0233 693 51.97 0.1433
1.00 0.0338 -0.1135 0.0136 406 18.68 -0.9300
Table 5.11: propCalcCD2.3.1 results: Constant pitch ratio at 0.7r.
Looking at Figure 5.19, it is obviously recognised that the propeller thrust is

much lower for the Wageningen CD Series within the whole range of the diagram.
Similarly, the propeller torque is much lower for the Wageningen CD Series com-
pared to Case 1. For a better understanding, it is recommended to look at Table
5.11 for the Wageningen CD Series and Table 5.12 for Case 1. The nozzle thrust
appears to be again lower for the Wageningen CD Series with a significant difference
73
between J = 0.00 − 0.20.
Finally, it should be mentioned that there are other differences between the geo-
metrical properties of the propeller as well as differences between the geometrical
properties of the nozzle between Case 1 and the Wageningen CD Series, thus the
plotted Wageningen CD Series values might differ in reality. Thus, the
above comparisons can only provide an approximation of the behaviour.
5.1.5 Flow separation

Looking at Figure 5.2, a small change in the slope of the hydrodynamic coefficients
between J = 0.70 and J = 0.80 is observed. By visualizing the pressure distribu-
tion and the velocity fields of both cases in ParaView 5.6.0, derived from the MRF
simulations, it is observed that flow separation occurs within this region. More
specifically, looking at Figure 5.21 it is observed that separation arises on the nozzle,
starting at the advance coefficient J = 0.75 while prior to this point no separation
appears. After this point, separation increases proportionally with the advance co-
efficient.
Figure 5.20: Velocity field in free sailing condition: Case 1

MRF simulation at J = 0.70.

74


Finally, comparing the MRF (see Figure 5.22) and SMI (see Figure 5.24) results for
the advance coefficient J = 0.80 where the flow separation has already occured, it is
obvious that the separation is of similar scale. On the other hand, the flow field of
MRF is slightly different from the one of the SMI, with the SMI flow field resulting
in a more straight flow compared to the one of MRF.

SMI simulation at J = 0.80.
5.1.6 Cavitation
Looking at the three following figures, it is observed that a really low pressure in
Case 2 and Case 3 occurs at the connection point of the gear case housing and the
nozzle as well as at the tip of the blade that is behind the gear case housing. This
observation indicates that cavitation will first appear at these points.
75
In this low pressure region the pressure value derived from the CFD simulations,
which is read from ParaView, is equal to:
−200kP a
The atmospheric pressure of the air is equal to:
1atm = 101.325kP a
and the hydrostatic pressure is equal to:
ρ · g · z = 1025kg/m3 · 9.81m/s2 · 2.4m = 24.133kP a
where:
z is the distance of the upper part of the nozzle and the water surface.
By summing up the three pressure values above, the actual pressure at the connec-
tion point of the gear case housing and the nozzle is calculated, which is equal to
−74.54kP a. This is an unrealistic pressure since the vapor pressure for sea water at
20◦ C is equal to 2.29kP a. In other words, this means that the minimum value that
the pressure can get is 2.29kP a, while when this minimum value is exceeded, cav-
itation will occur. Thus, this should be further investigated by running cavitation
CFD simulations.
Figure 5.25: Pressure distributions in bollard pull condition: Case 1.
76
5.2 Hydrodynamic interactions in bollard pull con-

dition
In this section, the hydrodynamic interactions of all the cases in bollard pull con-
dition (i.e. J = 0.00) are demonstrated and a comparison among these cases is
achieved. Moreover, a short explanation of some of the variables is given since there
are variables with special subscripts that haven’t been discussed prior to this mo-
ment.
In Table 5.12, the hydrodynamic coefficients of all the cases are decomposed and a
detailed presentation of the thrust of each component is presented. Furthermore,
the torque coefficients derived from the propeller blades are also presented. It can
also be observed that Case 3 is divided into two parts. This is done in order to
express the forces in respect to the x-axis as well as the axis that the propeller is
77
rotating.
Coefficients Case 1 Case 2 Case 3 Case 3

x-axis propeller-axis
[-] [-] [-] [-]
KT blades 0.3460 0.3496 0.3597 0.3617
10KQ blades 0.6317 0.6373 0.6527 0.6563
KT nozzle 0.3772 0.4388 0.4140 0.4163
KT gear_case_housing - -0.0703 -0.0710 -0.0714
KT hull - - -0.0092 -0.0092
KT,net 0.7231 0.7182 0.6935 0.6935
Table 5.12: Decomposition of the hydrodynamic coefficients in bollard pull

condition.
The above table is then presented in percentages, which makes it easier to identify
the contribution of each component to the thrust and resistance accordingly.
Coefficients Case 1 Case 2 Case 3 Case 3

x-axis propeller-axis
[%] [%] [%] [%]
KT blades 47.85 48.68 51.87 52.16
10KQ blades - - - -
KT nozzle 52.16 61.10 59.70 60.03
KT gear_case_housing - -9.79 -10.24 -10.30
KT hull - - -1.33 -1.33
KT,net 100.00 100.00 100.00 100.00
Table 5.13: Decomposition of the hydrodynamic coefficients in bollard pull

condition.
Looking at Table 5.13, it is observed that the nozzle has the highest contribu-
tion in thrust for all cases. Besides that, the gear case housing has the highest
contribution to resistance for all cases.
In the following tables (see Table 5.14, Table 5.15, Table 5.16 & Table 5.17), a com-
parison among the cases is presented by demonstrating values derived from the CFD
simulation in OpenFOAM as well as by demonstrating the calculated hydrodynamic
interactions.
In Table 5.14, Tnet represents the net sum of the total thrust (i.e. T − R), Tnet,deduct
represents its deduction and TBP represents the bollard pull force. In Table 5.15,
KT,T represents the sum of the thrust coefficients that are creating thrust (i.e.
KT p + KT n ), KT,R represents the sum of the thrust coefficients that are derived
from the parts that are creating resistance (i.e. Case 2: KT gear_case_housing and Case
78
3: KT gear_case_housing + KT hull ), KT,net represents the net sum of all the thrust co-
efficients (i.e. KT,T − KT,R ), mc represents the merit coefficient and finally mc,deduct
represents its deduction.
In Table 5.16, n1995kw represents the rotational velocity of the propeller at a fixed
delivered power and n1995kw,deduct represents its deduction. Here it should be men-
tioned that a delivered power of PD = 1995kW is derived from the main engine’s
power (see Table 3.1) with a reduction factor of 5%. Looking again at the definition
of the variables, T1995kw represents the generated thrust at a fixed delivered power
and T1995kw,deduct represents its deduction. PD,191rpm represents the delivered power
at a fixed rotational speed and PD,191rpm,deduct represents its deduction. Finally,
T191rpm represents the generated thrust at a fixed rotational speed and T191rpm,deduct
its deduction. It should be mentioned that the 191 rpm is an estimation of another
CFD study that was conducted within Caterpillar Propulsion AB.
Note: All the calculations below are presented for half the vessel and one az-
imuth thruster unit. Moreover, the values presented in Table 5.14 and Table 5.15
are derived from the CFD results while in Table 5.16 and Table 5.17 the values are
calculated by assuming constant delivered power and rotational velocity accordingly.
79
80
Case Parts included T [kN] R [kN] t [%] Tnet [kN] Tnet,deduct [%] TBP [ton]
Case 1 Blades+Nozzle 399.17 0.00 0.00 399.17 0.00 40.69
Case 2 Blades+Nozzle+Gear Case Housing 435.24 38.81 8.92 396.44 0.69 40.41
excluding hull
including hull Hull
Table 5.14: Case comparison: Presentation of the thrust, resistance, thrust deduction,
net forces, deduction of the net forces and bollard pull forces.
Case KT,T [-] KT,R [-] 10KQ [-] KT,net [-] mc [-] mc,deduct [%]
Case 1 0.7231 0.0000 0.6317 0.7231 1.7481 0.00
Case 2 0.7885 0.0703 0.6373 0.7182 1.7150 1.90
Case 3 0.7737 0.0710 0.6527 0.7027 1.6210 7.27
excluding hull
Case 3 0.7737 0.0797 0.6527 0.6935 1.5892 9.09
including hull
Table 5.15: Case comparison: Presentation of thrust, torque, net thrust coefficients,
merit coefficient and the deduction of the merit coefficient.
Case n1995kw [rpm] n1995kw,deduct [%] T1995kw [kN] T1995kw,deduct [%]
Case 1 194.71 0.00 414.82 0.00
Case 2 194.14 0.31 409.56 1.27
Case 3 192.60 1.08 394.45 4.91
excluding hull
Case 3 192.60 1.08 389.28 6.16
including hull
Table 5.16: Case comparison: Presentation of the rotational velocity,

thrust and their deduction in constant delivered power.
Case PD,191rpm [kW] PD,191rpm,deduct [%] T191rpm [kN] T191rpm,deduct [%]

Case 1 1883 0.00 399.17 0.00
Case 2 1900 -0.89 396.44 0.69
Case 3 1946 -3.31 387.91 2.82
excluding hull
Case 3 1946 -3.31 382.83 4.09
including hull
Table 5.17: Case comparison: Presentation of the delivered power,

thrust and their deduction in constant rotational velocity.
81
Looking at Table 5.14, it can be seen that 8.92% of the thrust deduction occurs
at Case 2 when the gear case housing is placed behind the propeller and the nozzle.
By placing the whole azimuth thruster unit behind the hull and by excluding the
resistance created by the hull the thrust is deducted by 8.32%. Finally, the thrust
is deducted by 9.44% by including the resistance of the hull. Thus, the highest
thrust deduction occurs at Case 2 by placing the gear case housing be-
hind the propeller and the nozzle while the hull has a smaller contribution to
the thrust deduction.
The merit coefficient is introduced as a replacement of the total propulsive effi-

ciency ηD . Looking at Table 5.15, it can be seen that the merit coefficient of Case
2 is deducted by 1.90% when the gear case housing is placed behind the propeller
and the nozzle. Meanwhile, the merit coefficient of Case 3 is deducted by 7.27%
by excluding the hull and by 9.09% by including the hull. Hence, the highest
merit deduction occurs at Case 3 by placing the whole azimuth thruster
unit behind the hull due to the flow field that is generated by the hull while the
resistance from the hull itself has a smaller contribution of only 1.82%.
By reviewing Table 5.16, it can be identified that the highest decrease of rota-
tional speed of the propeller as well as the highest thrust decrease occurs
in Case 3. However, the decrease of both values is almost unnoticeable.
In the end, by reviewing Table 5.17, similar conclusions as in Table 5.16 can be
drawn.
In Table 5.18, TBP,1995kw,CF D is representing the bollard pull force that is derived
from the generated thrust at a fixed delivered power T1995kw (see Table 5.16). This
generated thrust is calculated based on the CFD simulation results. Moreover,
TBP,1995kw,CD is the bollard pull force that the Wageningen CD Series gives, which
is calculated in propCalcCD2.3.1 by using as inputs the delivered power PD =
1995kW and the rotational velocity at a fixed delivered power n1995kw among other
parameters.
Case TBP,1995kw,CF D TBP,1995kw,CD TBP,1995kw,CD /TBP,1995kw,CF D

[ton] [ton] [%]
Case 1 42.29 38.81 8.23
Case 2 41.75 - -
Case 3 40.21 - -
excluding hull
Case 3 39.68 36.10 9.02
including hull
Table 5.18: Case comparison: Presentation of the hydrodynamic coefficients in

bollard pull condition and comparison with the Wageningen CD Series.
82
A reduction factor of 7% is added at TBP,1995kw,CD for the hull (i.e. Case 3 including
hull) while the reductions that appear on Case 2 and Case 3 by excluding the hull
are unknown.
5.3 Accuracy and reliability

The highest accuracy of the results is immensely connected with the size and quality
of the mesh, the choice of the numerical methods of turbulence modelling and the
choice of the turbulence model. A really high quality mesh has been achieved during
this project since almost half of this project in respect to time was dedicated to
the mesh generation. A lot of manual work was required to generate the mesh,
which means that every part of the mesh has thoughtfully been created by the user.
The reliability of the results can be obtained by comparing different mesh sizes,
numerical methods and turbulence models as well as by comparing the convergence
of the results. However, this is very time consuming, thus, the above choices were
made according to standard procedures that have been validated and used for a long
time by Caterpillar Propulsion AB.
83
84
6
Conclusions and future work
In this chapter, a conclusion of this study project is drawn and the degree that the
initial purposes are fulfilled within this project is presented. In the end of the chap-
ter, suggestions for further investigations and things that should be kept in mind are
presented as part of the future work.
The CFD modelling of a propeller indicates a relative motion between the stationary
parts and the rotor of a propulsion unit and the vehicle that this propulsion unit
is mounted on. Two methodologies have been used within this project in order to
capture this motion: the steady state Multiple Reference Frame (MRF) and the
transient Sliding Mesh Interface (SMI). By comparing the simulation results of SMI
and MRF and by looking at their convergence plots, it is concluded that the tran-
sient SMI methodology gives much more accurate results compared to the steady
state MRF for low advance coefficient J values. Meanwhile, the results for medium
J values computed with the MRF approach are quite reasonable. Finally, MRF gives
appropriate results at low computational cost for initializing the flow properties of
the CFD simulations in a steady state so then these flow properties can be used to
proceed to transient SMI simulations in order to increase the accuracy of the results.
Looking into the results of interest of this study project, it is concluded that the
contribution of the nozzle is higher than the propeller in terms of thrust while the
contribution of the gear case housing is higher than the hull in terms of resistance.
Thus, the highest thrust deduction occurs on the gear case housing. Bollard pull
condition indicates advance coefficient J and velocity of the ship VS both equal to
zero. As a result, the open water efficiency looses its meaning and the propulsive
efficiency ηD becomes zero. Thus, the performance of the propulsion unit in bollard
pull condition needs to be expressed with the merit coefficient mc , which is a replace-
ment of the total propulsive efficiency ηD . Moving back to the results of interest, the
highest merit deduction occurs when placing the gear case housing behind the hull.
Finally, a higher bollard pull force of the open water case (i.e. Case 1) compared to
the Wageningen CD Series is observed.
During the current study, a symmetry boundary condition has been used at the top
and side of the domain during the CFD simulations. As part of the future work
would be to include the free surface effect at the top boundary instead. It would
also be suggested to run more transient SMI simulations, whether enough resources
are available to proceed to these computationally expensive simulations, in order to
complete the open water characteristic diagrams only with transient SMI results.
85
6. Conclusions and future work
CFD simulations in order to get the towing resistance of the hull for various J values
and investigation of the behavior of the gear case housing and hull in free running
condition would also be suggested. Furthermore, cavitation should be investigated
more by running cavitation CFD simulations. Moreover, comparison of numerical
methods and turbulence models should be included in order to obtain the reliability
of the results. Last but not least, a grid refinement study should be done.
86
References
Bergman, T. L., Lavine, A. S., Incropera, F. P., & Dewitt, D. P. (2011). Fundamen-
tals of Heat and Mass Transfer (7th ed.). Chichester, United Kingdom: John
Wiley and Sons Ltd.
Carlton, J. S. (2007). Marine Propellers and Propulsion (2nd ed.). Burlington, USA:
Elsevier Ltd.
Dyne, G., & Bark, G. (2005). Ship Propulsion: Compendium for part of the MSc
course "Ship resistance and propulsion". Gothenburg, Sweden.
Fletcher, C. A. J. (1991). Computational Techniques for Fluid Dynamics: Specific
Techniques for Different Flow Categories (2nd ed.). Berlin; New York: Springer-
Verlag.
Funeno, I. (2009). Hydrodynamic Optimal Design of Ducted Azimuth Thrusters.
In Proceedings of the First International Symposium on Marine Propulsors -
smp’09, Trondheim, Norway.
Gullberg, P., & Sengupta, R. (2011). Axial Fan Performance Predictions in CFD,
Comparison of MRF and Sliding Mesh with Experiments. doi:https : / / doi -
org.proxy.lib.chalmers.se/10.4271/2011-01-0652
Larsson, L., & Raven, H. C. (2010). The Principles of Naval Architecture Series:
Ship Resistance and Flow. Jersey City, USA: The Society of Naval Architects
and Marine Engineers.
Liu, F. (2017). A Thorough Description Of How Wall Functions Are Implemented
In OpenFOAM. In Proceedings of the CFD with OpenSource Software, 2016,
Edited by Nilsson. H. Gothenburg, Sweden. Retrieved from http://www.tfd.
chalmers.se/~hani/kurser/OS_CFD_2016
Liu, J., Lin, H., & Purimitla, S. R. (2016). Wake field studies of tidal current turbines
with different numerical methods.
Matin, F. (2011). Hydrodynamics of Conventional Propeller and Azimuth Thruster
in Behind Condition. (Master’s thesis, Chalmers University of Technology,
Department of Shipping and Marine Technology).
Menter, F. R., Kuntz, M., & Langtry, R. (2003). Ten Years of Industrial Experience
with the SST Turbulence Model. In Proceedings of the Fourth International
Symposium on Turbulence, Heat and Mass Transfer, Redding, USA.
OpenCFD Ltd. (2018). OpenFOAM User Guide v1812. Retrieved from https : / /
www.openfoam.com/documentation/user-guide/
Tabib, M., Siddiqui, M. S., Rasheed, A., & Kvamsdal, T. (2017). Industrial scale
turbine and associated wake development - comparison of RANS based Ac-
tuator Line Vs Sliding Mesh Interface Vs Multiple Reference Frame method.
87
References
In Proceedings of the 14th Deep Sea Offshore Wind R&D Conferencem EERA
DeepWind’2017, Trondheim, Norway.
Versteeg, H. K., & Malalasekera, W. (2007). An Introduction to Computational Fluid
Dynamics: The Finite Volume Method (2nd ed.). Harlow, England: Pearson
Education Limited.
88
References
89
References
90
Appendices
91
A
Convergence of the simulated
results
In this chapter, the total forces and moments extracted from the OpenFOAM simu-
lations are plotted and their convergence is observed.
The total forces and moments extracted from the OpenFOAM v1806 simulations for
all the cases used within this study project are plotted with MATLAB R2018b and
their convergence is observed. For a better understanding, all the simulated cases
are presented below:
• MRF Simulations at J = 0.00, 0.10, 0.20, 0.30, 0.40, 0.50, 0.60, 0.65,
0.70, 0.75, 0.80, 0.85, 0.90, 1.00,
• SMI Simulations at J = 0.00, 0.20, 0.40, 0.60, 0.80,
• SMI Simulations at J = 0.00, 0.20, 0.40, 0.60, 0.70, 0.80, 0.90, 1.00,
• SMI Simulations at J = 0.00.
All the convergence plots bellow are represented in all the three dimensions (i.e.
x-dir, y-dir and z-dir).
A.1 Convergence plots of Case 1
x-dir
(a) MRF.
93
A. Convergence of the simulated results
(b) SMI.
Figure A.1: Simulations at J = 0.00.
Figure A.2: MRF simulation of Case 1 at J = 0.10.
(a) MRF.
(b) SMI.
94
(a) MRF.
(b) SMI.
95
(a) MRF.
(b) SMI.
96
(a) MRF.
(b) SMI.
97
y-dir
(a) MRF.
(b) SMI.
98
(a) MRF.
(b) SMI.
99
(a) MRF.
(b) SMI.
(a) MRF.
100
(b) SMI.
101
(a) MRF.
(b) SMI.
102
z-dir
(a) MRF.
(b) SMI.
103
(a) MRF.
(b) SMI.
(a) MRF.
104
(b) SMI.
(a) MRF.
(b) SMI.
105
(a) MRF.
106
(b) SMI.
107
x-dir
Figure A.43: SMI simulation of Case 2 at J = 0.00.
108
109
110
111
y-dir
112
113
114
115
z-dir
116
117
118
119

x-dir
y-dir
120
z-dir
121
122
B
Illustration of the different surface
mesh grids
In this chapter, all the initial surface meshes used within this study project are pre-
sented. First, the general mesh grids that are used in all cases are presented while
later on the specific meshes used for each of the three different cases follow.
The definition of the geometry and mesh generation is done in ANSA Pre-processor
v.19.0.1. In Section 2.1.2, the grid generation has been discussed. It is worth men-
tioning in this point that when a geometry is changed, the mesh of this geometry
ends up being different as well. Thus, it is really important to acknowledge that
the general mesh grids mentioned in Section B.1 are slight different from case to case.
A good example to support the statement above is the hub in Case 1 that has
slightly different mesh compared to the hub in Case 2 due to the small geometrical
differences that are needed in order to adapt the hub to the shaft in Case 1 (see
Figure B.5) or to the gear case housing in Case 2 (see Figure B.7a). However, it is
attempted that consistency of the meshes of the same parts between the different
cases is kept, thus the meshes are considered almost identical.
123
B. Illustration of the different surface mesh grids
B.1 General mesh grids used in all cases
(a) Profile view of the blade.
(b) Top view of the blade.
124
(c) Leading edge of the blade.
(d) Trailing edge of the blade.
Figure B.1: Blade geometry and its mesh.
Mesh quality min [mm] max [mm]

Fine regions 0.027% · DP ≈ 0.75 0.060% · DP ≈ 1.62
Coarse regions 0.060% · DP ≈ 1.62 0.920% · DP ≈ 24.84
Table B.1: Minimum and maximum cell size at the blades.
125
Figure B.2: Hub geometry and its mesh: Isometric view of the hub.

Table B.2: Minimum and maximum cell size at the hub.
126
Here it should be mentioned that the domain of the rotating geometries, also known
as rotor, is the inner volume that is rotating within the outer volume (i.e. the outer
domain). Thus, this configuration is a SMI configuration. The outer boundaries of
the rotor are the interfaces (i.e. inlet and outer interface) and a surface attached to
the inner part of the nozzle. An Arbitrary Mesh Interpolation (AMI) takes place
between the rotor and the outer volume.
(a) Isometric view of the rotor.
127
(b) Rotor including the surface attached to the inner part of

the nozzle.
(c) Rotor excluding the surface attached to the inner part of

the nozzle.
Figure B.3: Rotor geometry and its mesh.

Table B.3: Minimum and maximum cell size at the rotor.
128
B.2 Mesh grids used in Case 1
Here it should be mentioned that part of the inner part of the nozzle (see Figure
B.3a) is geometrically identical with the surface of the rotor that is attached to the
inner part of the nozzle (see Figure B.4a). Thus, almost identical mesh is attempted
to be achieved. Moreover, this part of the inner part of the nozzle as well as the
surface of the rotor that is attached to the inner part of the nozzle are including
some refined regions. These refined regions occur at the regions where the tip of the
blades is placed.
(a) Isometric view of the nozzle.
129
(b) Inner side of the nozzle, focusing on the refined mesh of the inner part
of the nozzle.
(c) Leading edge of the nozzle.
130
(d) Trailing edge of the nozzle.
Figure B.4: Nozzle geometry and its mesh.

Table B.4: Minimum and maximum cell size at the nozzle.
131
Figure B.5: Shaft-hub geometries and their mesh: Isometric view of the shaft
including the hub.

Table B.5: Minimum and maximum cell size at the shaft.
132
(a) Isometric view of the domain.
(b) Inlet of the domain including the connection

region to the shaft.
Figure B.6: Domain geometry and its mesh.

Fine regions 1.400% · DP ≈ 37.80 38.000% · DP ≈ 1000
Coarse regions 38.000% · DP ≈ 1000 38.000% · DP ≈ 1000
Table B.6: Minimum and maximum cell size at the domain.
133
The azimuth unit excluding the blades is presented below. On the top of the gear
case housing, a lid has been used during the simulations of Case 2 while in Case 3
the gear case housing is connected directly to the hull.
(a) Isometric view of the azimuth unit excluding the blades.
134
(b) Pressure side of the azimuth unit excluding the blades.
(c) Suction side of the azimuth unit excluding the blades.
Figure B.7: Azimuth unit geometry and its mesh.

Table B.7: Minimum and maximum cell size at the gear case housing.
135
Figure B.8: Domain geometry and its mesh: Isometric view of the domain.

Fine regions 38.000% · DP ≈ 1000 38.000% · DP ≈ 1000
Coarse regions 38.000% · DP ≈ 1000 38.000% · DP ≈ 1000
136
Figure B.9: Hull-azimuth geometries and their mesh: Isometric view of the hull
including the azimuth unit.

Fine regions 0.370% · DP ≈ 10 2.320% · DP ≈ 62.64
Table B.9: Minimum and maximum cell size at the hull.
137
(a) Isometric view of the domain.
(b) Part of the domain, focusing on the refined mesh of the hull and the wake field
behind it.
138
(c) Part of the domain, focusing on the refined mesh of the wake field behind the
hull.
(d) Part of the domain, focusing on the refined mesh of the skeg of the hull.
139
(e) Part of the domain, focusing on the refined mesh of the bow of the hull.
Figure B.10: Domain geometry and its mesh.

140

Investigation of Thrust Deduction on Azimuth Thrusters in Bollard Pull Condition Using CFD

Uploaded by

Copyright:

Available Formats

Investigation of Thrust Deduction on Azimuth Thrusters in Bollard Pull Condition Using CFD

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

Investigation of Thrust Deduction on Azimuth Thrusters in Bollard Pull Condition Using CFD

Uploaded by

Copyright:

Available Formats

CHALMERS

Investigation of Thrust Deduction on

Division of Marine Technology

Investigation of Thrust Deduction on Azimuth

Division of Marine Technology

© FOIVOS LEMONAKIS, 2019.

Examiner: Rickard Bensow, Chalmers University of Technology

Report No.: 2019:28

Cover: Visualization constructed in ParaView v.5.6.0 of the pressure distribution

Keywords: Propulsion, Hydrodynamics, Thruster, Thrust deduction, Bollard pull,

• Professor Rickard E. Bensow, examiner of this master’s thesis and Head of

• Caterpillar Propulsion AB for providing the resources of this study project.

• Tobias Huuva, supervisor of this master’s thesis and Engineering Manager at

• Simon Törnros, supervisor of this master’s thesis and Hydrodynamics Special-

Foivos Lemonakis, Gothenburg, June 2019

2.3.2 The Navier-Stokes equations . . . . . . . . . . . . . . . . . . . 14

5 Results and data analysis 55

6 Conclusions and future work 85

Appendix A Convergence of the simulated results 93

Appendix B Illustration of the different surface mesh grids 123

Variable [Units] Description

Variable [Units] Description

Variable [Units] Description

Variable [Units] Description

3.1 Starboard ducted azimuth thruster. . . . . . . . . . . . . . . . . . . . 34

4.1 Geometry of Case 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

5.1 Open water characteristics diagram: MRF/SMI comparison of Case 1. 56

5.5 MRF simulation of Case 1 at J = 0.00. . . . . . . . . . . . . . . . . . 58

A.1 Simulations at J = 0.00. . . . . . . . . . . . . . . . . . . . . . . . . . 94

A.15 Simulations at J = 0.00. . . . . . . . . . . . . . . . . . . . . . . . . . 98

A.61 SMI simulation of Case 2 at J = 0.40. . . . . . . . . . . . . . . . . . . 117

B.1 Blade geometry and its mesh. . . . . . . . . . . . . . . . . . . . . . . 125

3.1 Main particulars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

4.1 Number of cells: Case 1. . . . . . . . . . . . . . . . . . . . . . . . . . 42

5.1 Hybrid MRF/SMI results: Case 1. . . . . . . . . . . . . . . . . . . . . 57

B.1 Minimum and maximum cell size at the blades. . . . . . . . . . . . . 125

B.3 Minimum and maximum cell size at the rotor. . . . . . . . . . . . . . 128

1.2 Importance of this topic

1.4.1 Investigation cases

1.4.2 Software packages used

Pre-processor: Definition of the geometry and mesh generation

Solver: Solving of the governing equations

Post-processors: Visualization of the simulation results

Process of the numerical results

Comparison with the Wageningen CD Series

Rendering: 3D visualization of the geometry

2.1 Introduction to Computational Fluid

– Discretisation of the integrated equations into algebraic equations

2.1.1 Definition of the computational domain

2.1.2 Grid generation

2.1.3 Identification of the physical and chemical phenomena

• Treatment of the small-scale turbulence.

2.1.4 Definition of the fluid properties

2.1.5 Definition of the boundary conditions

2.1.6 Integration of the governing equations of the flow over

Rate of change Net rate of Net rate of Net rate of

2.1.7 Discretisation of the integrated equations into alge-

2.1.8 Solving of the algebraic equations by the use of itera-

2.1.9 Visualization of the simulation results

2.1.10 Process of the numerical results