Full scale validation of CFD model of

self-propelled ship
Master Thesis

Henrik Mikkelsen
Mads Lund Steffensen
June 2016
Technical University of Denmark
Department of Mechanical Engineering
Nils Koppels Allé, building 404,
2800 Kongens Lyngby, Denmark
Phone +45 4525 1960
info@mek.dtu.dk
www.mek.dtu.dk
 Abstract
An important part of designing a ship is to estimate the performance of the ship and the necessary
propulsive power at the design stage.
Traditionally, the performance of a ship design is estimated by conducting towing tank tests on
a downscaled version of the ship (model scale). The results from these towing tank tests are then
extrapolated to the scale of the actual ship (full scale). An alternative to testing in a towing tank is
to simulate the flow around the ship using Computational Fluid Dynamics (CFD). CFD simulations
can be performed in full scale, which eliminates the need for extrapolation. A disadvantage of using
CFD is the modelling errors which occur because the flow physics are simplified (e.g. turbulence
in the boundary layer).
This thesis is a study of the accuracy of full scale CFD simulations in predicting the performance
of a ship design. The authors have been granted unique access to data from a series of sister ships.
The data includes ship geometry, towing tank tests, sea trial data and in-service performance data.
With access to this data, the CFD model can be validated as it is developed.
The final CFD setup simulates the self-propelled ship in full scale. The results of these full
scale self-propulsion results show an average overestimation of delivered power of 2% compared
to the sea trial results. The predicted delivered power using the towing tank approach shows an
average overestimation of delivered power of 6% compared to the sea trial results. In this case, the
CFD approach resulted in an estimation of delivered power which was closer to the sea trial results.
However, due to uncertainties, more comparison studies between full scale CFD and extrapolated
towing tank tests are required. More comparison studies could provide more knowledge about
CFD simulations of self-propelled ships and how CFD can be used instead of towing tank tests to
predict a ships performance at design stage in the future.
ii
 Preface
This thesis is submitted as partial fulfilment of the requirement for the degree of MSc in Mechani-
cal Engineering at the Technical University of Denmark (DTU), Lyngby, Denmark.

The work has been performed over the time period January 2016 to June 2016. The project
was carried out at the Section of Fluid Mechanics, Coastal and Maritime Engineering (FVM) at
the Department of Mechanical Engineering (MEK) in cooperation with DNV GL. The authors
have been stationed at the headquarters of DNV GL in Oslo in the project period. The thesis has
been supervised by Jens Honoré Walther (MEK) and Poul Andersen (MEK).

Mads Lund Steffensen and Henrik Mikkelsen

Oslo, Norway.
May, 2016.
iv
 Acknowledgements
Of the many people who helped us during this project, we would first like to thank our two
supervisors, Jens Honoré Walther and Poul Andersen for their guidance and support as well as
for some good discussions throughout the project. We would like to thank Olav Rognebakke for
inviting us to Oslo and his colleagues Jaeouk Sun, Cosmin Ciortan, Tormod Gjestland for sharing
state-of-the-art knowledge of numerical modelling and the maritime world.
Secondly, we would like to express our sincere gratitude to the Japanese shipyard for inviting
us to Japan and providing unique and rarely shared data.
Thirdly, we would like to thank Stig Staghøj Knudsen from OSK-ShipTech A/S for allowing
us to use confidential setup files for numerical simulations.
Furthermore, we would like to thank the shipowner for inviting us to the sea trial of their ship,
providing transportation to Japan and providing in-service performance data.
We are grateful to the software company CD-Adapco for providing their high-end software code
STAR-CCM+ for unlimited use for this project.
Finally, we would to thank Dansk Maritime Fond, Reinholdt W Jorch og Hustru Fond, William
Demant og Hustru Ida E. Fond and the Department of Mechanical Engineering at DTU for the
financial support that made the project possible.
vi
 Contents
Abstract i

Preface iii

Acknowledgements v

Nomenclature ix

1 Introduction 1

2 Theory 3
2.1 Ship Resistance Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Model-Ship Extrapolation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2.1 Model-Ship Scaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2.2 Form Factor Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2.3 Frictional Resistance Coefficient . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.4 Form Factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3 CFD Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3.1 RANS Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3.2 Two-Equation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3.3 Free Surface Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3.4 Fluid-Body Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3.5 Spatial Discretization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3.6 Temporal Discretization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

3 Method 11
3.1 Ship . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.1.1 Hull . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.1.2 Propeller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.1.3 Appendages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.2 Towing Tank Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.2.1 Resistance Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2.2 Open Water Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2.3 Self-Propulsion Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2.4 Resistance Extrapolation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.2.5 Open Water Extrapolation . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2.6 Self-Propulsion Extrapolation . . . . . . . . . . . . . . . . . . . . . . . . . . 22
viii Contents

3.3 Sea Trial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3.3.1 Speed Trial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.3.2 Speed Measurement and Corrections . . . . . . . . . . . . . . . . . . . . . . 23
3.3.3 Power Measurement and Correction . . . . . . . . . . . . . . . . . . . . . . 24
3.4 CFD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.4.1 Validation and Convergence Test . . . . . . . . . . . . . . . . . . . . . . . . 27
3.4.2 CFD Resistance Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.4.3 CFD Open Water Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.4.4 CFD Self-Propulsion Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

4 Results 41
4.1 Towing Tank Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.1.1 Results of Resistance Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.1.2 Results of Open Water Test . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.1.3 Results of Self-Propulsion Test . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.1.4 Extrapolation of Resistance Test . . . . . . . . . . . . . . . . . . . . . . . . 43
4.1.5 Extrapolation of Open Water Test . . . . . . . . . . . . . . . . . . . . . . . 43
4.1.6 Extrapolation of Self-Propulsion Test . . . . . . . . . . . . . . . . . . . . . . 44
4.2 Sea Trial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.3 CFD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.3.1 CFD Resistance Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.3.2 CFD Open Water Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.3.3 CFD Self-Propulsion Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

5 Discussion 71
5.1 Comparison of Sea Trial and In-Service Performance Data . . . . . . . . . . . . . . 71
5.2 Comparison of Sea Trial and Towing Tank Test . . . . . . . . . . . . . . . . . . . . 72
5.3 Comparison of Sea Trial and Full Scale CFD . . . . . . . . . . . . . . . . . . . . . 73

6 Conclusions 77

7 Future Work 79

Bibliography 80

A Required Data 85
 Nomenclature
Latin letters

AV S Projected frontal area above water of the ship [m2 ]

B Beam of ship [m]
C Resistance coefficient [-]
c0.7 Chord length at r/R = 0.7 [m]
CA Correlation coefficient [-]
CAAS Air resistance coefficient [-]
CB Block coefficient [-]
CDA Air drag coefficient [-]
CD Propeller drag coefficient [-]
Ccyl Engine cylinder coefficient [m3 ]
Ceng Engine coefficient [-]
CF Frictional resistance coefficient [-]
CF,Hu Frictional resistance coefficient from the Hughes line [-]
CF,Sh Frictional resistance coefficient from the Schoenherr line [-]
CR Residuary resistance coefficient [-]
CT Total resistance coefficient [-]
CT,smooth Total resistance coefficient, smooth hull, without bilge keel [-]
CV Viscous resistance coefficient [-]
CW Wave resistance coefficient [-]
Cµ Eddy viscosity coefficient [-]
D Propeller diameter [m]
DWDesign Design dead weight [m]
FD Skin friction correction force or Towing force [N]
FI Fuel injection index [-]
Fn Froude number [-]
g Gravitational acceleration [ sm2 ]
Hs Significant wave height [m]
J Propeller speed of advance coefficient (Advance ratio) [-]
JT Advance ratio based on thrust identity method [-]
k Form factor [-]
2
k Turbulence kinetic energy [ ms2 ]
kp Blade roughness [m]
KQ Propeller torque coefficient [-]
ks Nikuradse’s equivalent sand roughness [m]
x Nomenclature

KT Propeller thrust coefficient [-]

l∗ Non-dimensional cell size [−]
L Characteristic length [m]
LBW L Distance of the bow to 95% of the maximum beam [m]
LOA Overall length of ship [m]
LP P Length between perpendiculars [m]
LW L Waterline length of ship [m]
n Rate of rotation of propeller [s−1 ]
p Pressure [Pa]
pme Mean effective pressure [Pa]
P Engine power [W]
PD Delivered power [W]
P /D Pitch ratio [-]
PW Power correction due to waves [W]
Q Propeller torque [Nm]
r Radius or radial distance [m]
rT Total resistance coefficient [-]
R Resistance [N]
RAW L Added resistance due to waves [N]
RF Frictional resistance [N]
RP Pressure resistance [N]
RT Total resistance [N]
RV Viscous resistance [N]
RV,SH Viscous resistance with raised surface height [N]
RW Wave making resistance [N]
Re Reynolds number [-]
Rec0.7 Propeller Reynolds number at r/R = 0.7 [-]
Rep Propeller Reynolds number [-]
S Wetted surface area [m2 ]
SBK Wetted surface area of bilge keel [m2 ]
Sij Mean strain rate tensor [s−1 ]
SS Wetted surface of the hull and rudder [m2 ]
t Time [s]
t Trust deduction [-]
t∗ Non-dimensional time step [-]
t0.7 Maximum blade thickness length at r/R = 0.7 [m]
T Propeller thrust [N]
TAP Draft at aft perpendicular [m]
TDesign Design draft [m]
TF P Draft at fore perpendicular [m]
U Velocity vector [ ms , ms , ms ]
u∗ Frictional velocity [ ms ]
u′ Fluctuating part of velocity [ ms ]
V Speed [ ms ]
VDesign Design speed [ ms ]
Vθ Tangential rotational velocity [ ms ]
wT Wake fraction coefficient from trust identity method [-]
y Distance to wall [m]
y+ Non-dimensional wall distance [-]
x Spatial direction [m]
Z Number of blades [-]
Nomenclature xi

Greek letters

δij Kroneckers delta [-]

∆CF Roughness allowance coefficient [-]
∆l Cell size [m]
∆RF Roughness resistance [N]
∆RSH Added surface height resistance [N]
∆t Time step [s]
◦
∆θ Angle of rotation per time step [s]
2
ϵ Turbulence dissipation rate [ ms3 ]
ηb Propeller efficiency behind ship [-]
ηo Propeller efficiency in open water [-]
ηR Relative rotative efficiency [-]
λ Model-Ship scaling factor [-]
Ns
µ Dynamic viscosity [m 2]
Ns
µt Eddy viscosity [m 2]
m2
ν Kinematic viscosity [ s ]
ρ Fluid density [ mkg3 ]
ρair Air density [ mkg3 ]
τω Wall shear stress [Pa]
ϕ Coefficient used in the MARINTEK form factor method [-]
ω Engine rotational speed [ rad
s ]
ω Specific turbulence dissipation rate −1
[s ]
∇ Volume displacement [m3 ]

Subscripts

m Index for model scale ship

s Index for full scale ship
o Index for open water or open water equivalent
i,j Index for the three spatial directions
xii Nomenclature

Abbreviations

CFD Computational fluid dynamics

DFBI Dynamical fluid-body interaction
DGPS Differential GPS
DNS Direct numerical simulation
DOF Degrees of freedom
DTU Technical University of Denmark
DWT Dead weight tons
FVM Section of Fluid Mechanics, Coastal and Maritime Engineering at DTU
GPS Global Positioning System
IMO International Maritime Organization
ITTC International Towing Tank Committee
MCR Maximum continuous rating
MEK Department of Mechanical Engineering at DTU
MRF Moving reference frame method
NACA National Advisory Committee for Aeronautics
RANS Reynolds-average Navier-Stokes
RPM Revolutions per minute
RSB Rotating solid body method
RZ Refinement zone
SST Shear stress transport
VOF Volume-of-Fluid
CHAPTER 1
Introduction
At the design stage, it is essential to have an accurate and trustworthy estimate of the performance
of the ship. An accurate estimation is important in order to ensure fulfilment of contractual
requirements and minimize the costs of building and operating the ship and to design an eco-
efficient ship. Eco-efficiency is getting increasingly important these days because of the increasing
legislative requirements for a cleaner and greener shipping industry.
Traditionally, the performance of a ship design is estimated by conducting hydrodynamic tests
on a model scale version of the ship. These tests are conducted in a towing tank, which is a
large basin of water in which hydrodynamic tests can be conducted under controlled conditions
(see Figure 1.1). Towing tank tests have been performed for more than 100 years, which has
generated vast experience in conducting these tests and a large amount of knowledge about the
hydrodynamics of ships. The organisation of standard procedures for carrying out towing tank
tests is managed by an association named the International Towing Tank Conference (ITTC).

Figure 1.1: Experimental setup in towing tank. Courtesy of FORCE Technology.

One of the most significant weaknesses of predicting ship resistances by conducting towing tank
tests is that the viscous effects in model scale are not the same as in full scale. In order to cope
with this fact, extrapolations procedures have been developed. The procedure artificially splits
the resistance components up and scales them separately. The procedures for extrapolating the
towing tank measurements from model scale to full scale have been developed over the years to
more accurately predict the performance of the full scale ship.
An alternative to testing in a towing tank is to use Computational Fluid Dynamics (CFD) to
predict the ship resistance (see Figure 1.2). In CFD the flow is simulated by solving a large set
of equations numerically using a computer. CFD has become an increasingly viable alternative
with the rapid increase in available computational power in the last decades. One advantage of
2 1 Introduction

CFD is that the simulations can be performed in full scale, which eliminates the challenges of
extrapolation from model scale to full scale. CFD simulations also have the advantage of offering
insights into the flow, which is very difficult or impossible to obtain from experiments.

Figure 1.2: CFD self-propulsion calculation. The colors on the hull and propeller indicates the
pressure. The colors around the ship indicates the velocity magnitude of the water. Blue represents
low values and red represents high values.

One of the disadvantages of using CFD is the modelling errors which occur because of the
simplified flow physics (e.g. turbulence in the boundary layer). Extensive studies [1, 2, 3] have
shown that model scale CFD simulations can predict the resistance on the ship and the thrust and
torque of the propeller within a few percentages of towing tank measurements.
This thesis includes validation of CFD simulations in predicting the performance of a ship
both in model scale and full scale. In order to quantify the accuracy of the CFD simulations, some
reference data is needed. Such data about the performance of a ship is almost always considered
confidential information, and obtaining access to the data may be very difficult. The authors of
this thesis have been provided with unique access to data from a series of sister ships. The data
includes ship geometry, towing tank test data, sea trial data, and in-service performance data.
Towing tank test data is used as reference for the model scale CFD simulations. Sea trial data and
in-service performance data are used as reference for the full scale CFD simulations. A sea trial is
the final test of a ship before it is delivered by the shipyard to the ship owner.
The key to a useful comparison is to have highly accurate sea trial data. Sea trial data gener-
ally has significant uncertainties because of the lack of control in the experimental environment.
There is a focus in the industry on obtaining increasingly accurate sea trial results, as seen with
introduction of the new sea trial procedure ISO 15016:2015 [4]. Besides the difficulties in con-
ducting experiments at such a large scale, it is also important to remember that there are many
stakeholders in a sea trial, each with different interests. The authors have attended a sea trial in
order to gain knowledge about the sea trial procedure and the uncertainties of the data.
If full scale CFD simulations prove to be more accurate or as accurate as towing tank tests, it
could potentially change the method with which the maritime industry predicts the performance
of a ship before is built.
CHAPTER 2
Theory
This chapter presents in brief the theory which serves as the foundation for this thesis. Section 2.1
is devoted ship resistance and how the resistance can be split into relevant components. The
following Section 2.2 presents the theory and overall ideas behind extrapolating the ship resistance
from model scale to full scale. Finally Section 2.3 states the theoretical background for the CFD
simulations.

2.1 Ship Resistance Components

This section introduces some of the theory about resistance on ships travelling in calm water.
There are several ways of dividing the total calm water resistance RT of a ship into components.
One way is to divide the forces acting on the ship into two parts; frictionally resistance RF , which
is the forces acting tangentially to the ship surface, and pressure resistance RP , which is the shear
forces acting tangential to the ship surface. This gives rise to the following expression:

RT = RF + RP (2.1)

At lower speeds the frictional resistance is dominant and at higher speeds the pressure forces
becomes dominant due to the increased wave making. Alternatively, the total resistance can be
divided into two other parts; viscous resistance RV (energy lost in wake), and wave resistance RW
(energy lost in wave generation). This gives an alternative expression:

RT = RV + RW (2.2)

When a ship travels though water, waves are generated on the water surface. The generated
waves forms a wave system. A Kelvin wave is the mathematical form of the wave system made
by a point source travelling along a free surface[5, Sec. 3.1.5.2]. The details on the Kelvin wave
are explained by [5, Sec. 3.1.5.2, Sec. 7.3.4.3] and [6]. Figure 2.1 depicts an example of the wave
system. The wave system created by the translation of the ship consists of divergent and transverse
waves which is similar to a Kelvin wave.

2.2 Model-Ship Extrapolation

When designing a ship, it is important to predict the ship resistance, in order to estimate the
necessary amount of propulsion power. These predictions can be done by testing a model of the
ship in a towing tank. The results of these towing tank tests can then be extrapolated from model
4 2 Theory

Figure 2.1: Wave system created by a travelling ship. [7, p. 117]

scale to full scale. The idea behind measuring on ships in model scale and then extrapolating the
results to full scale is that the total resistance can be split into components that can be scaled
individually. The methods for extrapolating resistances from model scale to full scale were first
introduced by Froude in 1872 [8]. This method was then further developed by Hughes in the 1950s
[9]. Hughes’ method, known as the form factor method, has been adopted as the standard method
for extrapolating by ITTC [10].

2.2.1 Model-Ship Scaling

The scaling factor λ between model scale and full scale is defined using a characteristic length of
the full scale ship Ls and characteristic length of the model scale ship Lm as following:

Ls
λ= (2.3)
Lm

The method of dimensional analysis can be used to create non-dimensional quantities which
are useful when scaling from model scale to full scale. The quantities are made non-dimensional
using the speed V , gravitational acceleration g, kinematic viscosity of the fluid ν and waterline
length of the ship LW L . Two of these non-dimensional quantities are: the Froude number F n and
the Reynolds number Re:
2.2 Model-Ship Extrapolation 5

V
Fn = √ (2.4)
gLW L
V LW L
Re = (2.5)
ν
Dimensional analysis can also be used to create a non-dimensional resistance coefficient C from
a resistance R using the fluid density ρ and the wetted surface area S:

R
C= 1 2
(2.6)
2 ρSV

The Froude number, the Reynolds number and resistance coefficients are essential non-dimensional
parameters in ship hydrodynamics and are used in the form factor method.

2.2.2 Form Factor Method

The idea of the form factor method is that the total resistance can be split into viscous and wave
resistances, as shown in eq. (2.2). The form factor method assumes that the viscous resistance
coefficient CV and wave resistances coefficient CW are dependent on the Froude number and the
Reynolds number respectively, as seen in the expression for the total resistance coefficient CT :

CT (Re,F n) = CV (Re) + CW (F n) (2.7)

The viscous resistance coefficient can be substituted by a friction resistance coefficient CF and
a form factor k (hence the name form factor method) while the wave resistance coefficient is
commonly referred to as the residuary resistance coefficient CR :

CV = (1 + k)CF (2.8)
CW = CR (2.9)

This gives an alternative expression for the total resistance coefficient:

CT = (1 + k)CF + CR (2.10)

The total resistance coefficient can be determined from the total resistance measured in the
towing tank, and the frictional resistance coefficient is calculated using an empirical formula called
a friction line. The residuary resistance coefficient can then be calculated using eq. (2.10).
By testing the model ship at the same Froude number as the actual ship, the residuary resistance
coefficient in model scale is equal to the residuary resistance coefficient in full scale, because the
residuary resistance coefficient is assumed to be only Froude number dependent. As in model scale,
the full scale frictional resistance coefficient can be found using an empirical friction line. With
the full scale frictional and residuary resistance coefficient known, the full scale total resistance
can be calculated using eq. (2.10).
6 2 Theory

2.2.3 Frictional Resistance Coefficient

In the form factor method described in the previous section, the frictional resistance coefficient
is estimated using a flat plate friction line since it cannot be measured directly in a towing tank
test. Many friction lines were proposed as candidates for a standard friction line during the initial
development of the first ITTC procedure in the 1950s. Ultimately, the Hughes 2D flat plate line
[11, Eq.(12.16)] was chosen to estimate the frictional resistance coefficient CF,Hu :

0.066
CF,Hu = (2.11)
(log10 (Re) − 2.03)2

The Hughes 2D flat plate line was modified by rounding off the values and by adding 11.94%
in order to account for 3D effects before being adopted as the ITTC-57 correlation line [11, 12, 10]:

0.067
CF = (1 + 0.1194)
(log10 (Re) − 2)2
0.075
= (2.12)
(log10 (Re) − 2)2

Some researchers [13] note that the ITTC-57 correlation line is a model ship correlation line,
rather than a flat plate friction line because of the corrections to the original Hughes friction line
described above. Some alternative friction lines proposed in the literature are: The Schoenherr
line [14], the Grigson line [15] and the Katsui line [16].
The Schoenherr Line estimate of the frictional resistance coefficient CF,Sh is given by [14]:

0.242
√ = log10 (ReCF,Sh ) (2.13)
CF,Sh

The alternative friction lines are compared in [13]. The existence of many different empirical
friction lines could indicate that it is difficult to develop a robust and versatile friction line. In this
thesis, the ITTC correlation line eq. (2.12) and the Schoenherr friction line eq. (2.13) will be used.

2.2.4 Form Factor

The form factor in the form factor method is a constant dependent on the shape of the hull. The
form factor can be estimated either by performing a form factor test in a towing tank or using an
empirical formula. In the towing tank test, the ship model is dragged through the water at a very
low speed. At very low speeds, the viscous resistance becomes dominant and the residual resistance
coefficient becomes negligible. When the residual resistance coefficient becomes negligible, the form
factor can be estimated by measuring the total resistance. The ITTC recommended procedure for
estimating the form factor is the Prohaska method [17, Sec. 3.6.2].
Originally the form factor was assumed to be independent of the Reynolds number. Recent
studies [18, 13, 19, 20, 21] show that this assumption is not correct. The conclusion of the research
has been confirmed by the ITTC Resistance Committee. In 2011 the Committee states that [12,
Sec. 5.3]: ”Modern bulbous bows in “offdesign” conditions (e.g. smaller draughts than the design
2.3 CFD Modelling 7

draught) make it nearly impossible to determine a reliable form factor from model tests.” and
”Although the original intent is that the form factor is independent of Reynolds number this does
not appear to be the case.”
This clearly indicates that the form factor is dependent on the Reynolds number, which contra-
dicts the original assumption of the form factor. Furthermore, ITTC concludes that it is difficult
to determine a form factor at off-design conditions and for ships with a bulbous bow. Today many
ships have bulbous bows and/or are tested in ballast condition, which is an off-design condition at
the sea trial. The sea trial is performed at ballast condition because it is practically impossible to
load the ship to design draft at the sea trial.
Another method to determine the form factor is to use an empirical method. An example of
an empirical method is the MARINTEK form factor method [22, p. 2]. The method uses main
dimensions of the ship as block coefficient CB , draft at aft TAP , draft at fore TF P and beam B to
estimate the form factor:

CB √
k = 0.6ϕ + 75ϕ3 , ϕ= (TAP + TF P ) B (2.14)
LW L

The uncertainties in the estimation of the form factor is a direct contribution to the uncertain-
ties of using towing tank tests to predict the performance of a ship design.

2.3 CFD Modelling

One of the main goals of this thesis is to model the fluid flow around a ship using CFD. This
section introduces some of the background behind the used CFD method. The flow around the
ship is modelled using either a single phase flow of water or a two phase flow of water and air.
The modelling of turbulence in the flow is an important aspect of CFD simulations. The most
ideal approach would be to simulate the turbulence all the way down to the smallest scales in
the flow, which are the Kolmogorov scales. This approach is called direct numerical simulation
(DNS). Using DNS requires an enormous computational effort, and is practically impossible for
most applications with the computational power available today. Another approach is to solve the
Reynolds-averaged Navier-Stokes (RANS) equations using a closure model. The RANS approach
is the method most commonly used in the ship building industry.

2.3.1 RANS Equations

The RANS equations are a time-averaged solution to the Navier-Stokes equations. Detailed de-
viations of the RANS equations are presented in a number of studies e.g. [23, Sec. 2.3]. The
flow is assumed incompressible, because the water phase is a liquid which is generally assumed
incompressible, and because the flow speeds are much lower than the speed of sound. For an
incompressible Newtonian fluid, the RANS equations are [23, Eq. (2.24)]:

∂Ui ∂Ui ∂p ∂ ( )
ρ + ρUj =− + 2µSji − ρu′j u′i (2.15)
∂t ∂xj ∂xi ∂xj
8 2 Theory

where Ui is the velocity vector, t is time, p is pressure, µ is dynamic viscosity, Sij is the mean
strain rate and u′ is the fluctuating part of the velocity.
The RANS equations in three spatial dimensions and a mass continuity equation give four in-
dependent equations with ten unknown variables. The imbalance between equations and unknown
variables makes the problem impossible to solve analytically. Normally, this is called the closure
problem of turbulence. The presence of the Reynolds stresses (ρu′j u′i ) make the turbulence problem
difficult.
In order to close the problem the Reynolds stresses are modelled using a linear eddy viscosity
model. The linear constitutive relationship is [23, Eq. (4.6)]:

2
−ρu′i u′j = 2µt Sij − ρkδij (2.16)
3
where µt is called turbulent viscosity or eddy viscosity and k is the turbulent kinetic energy and δij
is the Kronecker’s delta. The turbulent viscosity can be estimated using one of multiple methods.
One of these methods is to use a two-equation model.

2.3.2 Two-Equation Model

In this thesis, the turbulent viscosity is calculated using the realisable k − ϵ turbulence model.
The k − ϵ model is a two equations model using the turbulent kinetic energy k and the turbulent
dissipation rate ϵ [23, Eq. (4.40)]:

k2
µt = ρCµ (2.17)
ϵ
Cµ is a combination of constants and available flow parameters. More details on the k − ϵ model
can be found in [23, Sec. 4.3.2].
In the derivation of the k − ϵ model, it is assumed that the flow is fully turbulent [24]. The
model is known for its accuracy within flow types as modelled in this thesis. An alternative two-
equation model which is also used in the industry is the k − ω model where the second transported
value is the specific dissipation ω formulated by Wilcox[25, 26]. A revised version of the k − ω
model is the k − ω Shear Stress Transport (SST) [27], which combines the k − ϵ and the k − ω
model.

2.3.3 Free Surface Model

The calm water surface between the water and the air is called the free surface. It is important to
model the free surface around the ship, as the generated waves on the free surface contributes to
the total resistance on the ship, especially for high Froude numbers.
The free surface is modelled using the volume of fluid (VOF) method [28, p. 5337-5357]. The
VOF model works by calculating the fraction of water and air in each cell. In cells where with
both water and air, a weighted average of the density and viscosity is used, based on the volume
fraction. The position of the free surface is the interface between cells with volume fractions of air
larger than 50% and cells with volume fractions of water larger than 50%.
2.3 CFD Modelling 9

Heave
Pitch

Figure 2.2: Heave and pitch motions of a ship. The black dot represents the longitudinal center
of floatation. The direction of travel is left to right.

2.3.4 Fluid-Body Interaction

The interaction between the fluid and the ship is very important for the flow and the resistance on
the hull. The two primary degrees of freedom (DOF) for ships in calm water are heave and pitch.
The pitch and heave motions are illustrated in Figure 2.2.
The pitch and heave of the ship is dependent on the speed of the ship. At a certain speed the
ship obtains a pitch and a heave through equilibrium of static and hydrodynamic forces acting on
the ship. The equilibrium pitch and heave are called the dynamic pitch and heave.
The fluid-body interaction is modelling using a dynamic fluid-body interaction (DFBI) model
in STAR-CCM+ [28, p. 5358-5433]. The DFBI model calculates the forces on the ship at certain
intervals and translates and rotates a number of times the ship until the ship reaches an equilibrium.
An adjustment i

2.3.5 Spatial Discretization

The domain is the spatial region in which the CFD simulation takes place. The domain is dis-
cretization by dividing it into small cells. The collection of all cells besides the boundary layer
mesh in the domain is referred to as the volume mesh.
Geometry such as hull and propeller are inserted into the domain, in order to simulate the flow
around the ship. The surfaces of these geometries are also discretized by dividing the surface into
small triangular surfaces. The collection of the triangular surfaces is called the surface mesh.
It is important to model the boundary layer flow as it is used to calculate the frictional resistance
on the ship. For the cases studied in this thesis, the Froude number are relatively low. The low
Froude number causes the majority of the total resistance to be frictional resistance. A non-
dimensional distance from a cell to the nearest wall is called the y + value. The y + value for the
cells near the wall is called the wall y + value, and it is an indication of how well the boundary
layer is discretized. The y + value is defined as:

√
+ u∗ y τw
y = , u∗ = (2.18)
ν ρ
where u∗ is the frictional velocity, y is the distance to the nearest wall and τw is the wall shear
stress. The boundary layer consists of three parts: the viscous sub layer (y + < 5), the buffer layer
(5 < y + < 30) and the log-law layer (30 < y + < 200 − 300) [29, Chapter 2].
The boundary layer region is discretized using a boundary layer mesh. The boundary layer
mesh consists of several layers of thin rectangular cells, these are also known as prism layers, on all
surfaces of the ship. It is important to have a smooth transition in cell size from the boundary layer
10 2 Theory

Surface

Boundary
Layer
Mesh

Volume
Mesh

Figure 2.3: A boundary layer mesh near a surface.

cells to the volume mesh cells. Therefore, the nearer to the volume mesh the thicker he boundary
layer cells are, and they increase in thickness with each layer. An example of the boundary layer
mesh can be seen in Figure 2.3.
There are two approaches to model the boundary layer flow: fully resolving the boundary layer
flow with sufficient cells or using a wall function to approximate the shear stress. If the wall y +
value is sufficiently low (below 1), it means that there are sufficient cells to resolve the boundary
layer flow all the way down to the viscous sublayer. This method requires more cells and therefore
more computational effort. If the wall y + value indicates that the first cell is in the log layer (above
30), it means that there are no cells in the viscous sublayer alone. When the first cell includes part
of the log layer, wall functions can be used to approximate the wall shear stress. Both methods
will be used in this thesis.
When using wall functions, the CFD code STAR-CCM+ user guide advises using values larger
than 30 [28, p. 3403]. The recommendation of [30] is wall y + value of 30 to 60 and [31, p. 35-36]
recommends a wall y + value of 50. For all subsequent simulations which uses the method of wall
functions, the aim is to use an average wall y + value in the range of 35 to 50.

2.3.6 Temporal Discretization

Some of the CFD simulations requires a transient solver. These include the VOF model, the DFBI
model and the modelling of a rotating propeller. For all the simulations which requires a transient
solver, a first order implicit unsteady solver is used. The size of the time step used for the transient
solver is important. The time step size is a compromise between a small discretization error and
computational resources.
CHAPTER 3
Method
This chapter describes the methods used in this thesis. The work of the thesis is based on a
bulk carrier. The thesis consists of three elements: towing tank tests, sea trial tests and CFD
simulations, all conducted on the bulk carrier. The towing tank tests and the CFD simulations
each consists of resistance tests, open water tests and self-propulsion tests.
Firstly, the studied ship is presented in Section 3.1 followed by Section 3.2 which describes the
method used for the towing tank tests and the extrapolation of the towing tank results. Secondly,
the background of the sea trial is presented in Section 3.3 including the methods used to do the
measurements and correct the results. Finally, Section 3.4 is devoted how the CFD simulations
are set up for the resistance, open water and self-propulsion tests.

3.1 Ship
This section describes the ship studied in this thesis. The ship consists of a hull, a rudder, a
propeller, a set of pre-swirl stator fins and bilge keels. The section has three subsections, each
describing parts of the ship. The studied ship is a bulk carrier built by a Japanese shipyard. The
main dimensions of the ship can be seen in Table 3.1.

3.1.1 Hull
The hull has been provided as a 3D geometry by the Japanese shipyard. The hull and water lines
of the design and sea trial conditions is shown in Figure 3.2. The displacement of the ship in
sea trial condition is 40.2 % of the displacement in the design condition. The sea trial condition
is significantly lighter because the ship is completely empty, except for the ballast tanks. The
coordinate system used throughout the thesis is shown in Figure 3.2. The z-direction is called the
vertical direction, the y-direction is called the transverse direction and the x-direction is called the
longitudinal direction.

Table 3.1: Main dimensions of the studied ship.

Length overall LOA 200 m

Beam B 32.26 m
Design draft TDesign 11.00 m
Design dead weight DWDesign 62000 DWT
Design speed VDesign 14.55 knots
12 3 Method

Figure 3.2: Hull geometry. The black line is the water line of the design condition, and the grey
line is the water line of the sea trial condition. Seen from starboard side.

Table 3.3: Data of the actual propeller.

Diameter D 6.00 m
Boss ratio 0.1417
Pitch ratio (0.7r) 0.7284
Expanded area ratio 0.48
Number of blades 4
Direction of turning Right handed

Table 3.4: Data of the stock propeller (No. 2141R).

Boss ratio 0.18

Pitch ratio (constant) 0.76
Expanded area ratio 0.55
Number of blades 4
Direction of turning Right handed

3.1.2 Propeller
Two different propeller designs are associated with the ship: a stock propeller and an actual
propeller. The stock propeller has a standard design and is only used for testing a model of the
ship in a towing tank. The actual propeller is the propeller mounted on the actual ship. The data
of the actual propeller can be seen in Table 3.3.
Due to confidentiality it has not been possible for the propeller manufacturer to provide the
3D geometry of the actual propeller. However, it has been possible for them to provide the 3D
geometry of the stock propeller used in the towing tank test. Details of the stock propeller can be
seen in Table 3.4, and the geometry can be seen in Figure 3.5.
The provided stock propeller geometry is not of sufficient quality to be used in CFD simu-
lations. The leading and trailing edges of the propeller blades are flat, as opposed to rounded.
The flat edges, especially the leading edge, significantly disturb the flows and cause separation. A
propeller specialist from DNV GL has recreated the propeller blade based on the blade sections
characteristics such as thickness, chamber, chord length etc. from the stock propeller. The recre-
ated propeller geometry is based on the modified NACA 66 profile. An illustration of the original
and recreated propeller profile can be seen in Figure 3.6. A modified hub and end cap have also
been created with the same boss ratio as the original stock propeller. The modified stock propeller
with hub and end cap can be seen in Figure 3.7.
3.1 Ship 13

Figure 3.5: Original stock propeller geometry.

Figure 3.6: Illustration of the propeller profile. The provided stock propeller in light gray, and
the recreated leading and trailing edges in dark grey. The recreated areas are exaggerated for the
purpose of illustration.

Figure 3.7: Modified stock propeller geometry.

14 3 Method

Figure 3.8: Definition of propeller parts. Light grey is the leading and trailing edges, dark grey
is the suction and pressure sides, and black is the hub.

Table 3.9: Surface size settings of the surface wrapper for each of the three propeller parts.

Part of propeller Target surface size Minimum surface size

Leading and trailing edge 5.32 × 10−4 D 2.66 × 10−4 D
Suction and pressure side 5.32 × 10−3 D 5.32 × 10−4 D
Hub 1.06 × 10−2 D 1.06 × 10−3 D

The surface of the propeller geometry is wrapped in STAR-CCM+ in order to obtain a good
surface mesh for for the CFD simulations. The propeller is divided into three parts; hub, blade
surface and blade edges. The three parts can be seen in Figure 3.8. In ensure that the geometry
is closed and has no sharp edges a function in STAR-CCM+ called surface wrapper is used. The
settings of the surface wrapping are different for each of the three propeller parts. In order to
correctly simulate the flow around the blade edges, it very important to resolve the important
parts of the propeller geometry. The propeller is wrapped with the finest surface mesh around the
blade edges, a coarser mesh around the blade surface, and the coarsest mesh around the hub. The
surface wrapper uses a target surface size to control the intended size of the surface mesh, and a
minimum surface size to prevent an excessive number of surface cells. The surface size settings
can be seen in Table 3.9.
Two different versions of the propeller geometry are used; one version for the open water test
and another for the self-propulsion test. The open water version can be seen in Figure 3.10, and
the self-propulsion version can be seen in Figure 3.7. The difference between the two versions is
the end cap. In the self-propulsion version, the end cap is cut off in order to have the correct gap
between the end cap and the rudder bulb. In the open water version, the end cap is fully rounded.
3.2 Towing Tank Tests 15

Figure 3.10: Propeller configuration of the open water test. The flow direction is right to left.

It is assumed that the cut-off of the end cap in the self-propulsion version has very little influence
on the propeller performance since the end cap is downstream of the propeller blades. In the open
water test, the end cap is upstream of the propeller blades and thereby has a significant influence
on the flow field to the propeller. Therefore, the end cap is fully rounded in the open water version.

3.1.3 Appendages
The actual ship is mounted with a rudder, a number of upstream stator fins and bilge keels. The
”to-be-built” 2D-drawings of the stator fins and rudder including rudder bulb have been provided
by the Japanese shipyard. Based on these drawings, the stator fins and rudder geometry have been
created in 3D, by the authors.
The stator fins are mounted upstream of the propeller. The idea behind the stator fins are to
produce a so-called pre-swirl in the flow in the direction opposite of the propeller rotation direction.
The stator fins increase the performance of the propeller since the propeller produces more thrust
for the same power.
Bilge keels are not included in any CFD simulations since they are not mounted on the model
ship in the towing tank test. They are instead accounted for in the towing tank and CFD simula-
tions by other means.

3.2 Towing Tank Tests

The design of the ship has been tested in a Japanese towing tank prior to this project, using the
ITTC procedure. The towing tank tests include a resistance test, an open water test and a self-
propulsion test. The results of the towing tank test have been provided by the Japanese shipyard.
16 3 Method

Table 3.11: List of the resistance test parameters provided by the towing tank.

Resistance Test
Ship speed Vm
Frictional resistance coefficient CF m,Sh
Total resistance coefficient CT m

All of the towing tank tests are conducted with rudder and pre-swirl fins mounted and in the sea
trial condition.
In this section, the parameters of the towing tank tests are presented and discussed, and along
with the procedure used to extrapolate the towing tank results from model scale to full scale.

3.2.1 Resistance Test

The resistance test is a test of the drag resistance of the ship at different speeds. The resistance
test is conducted by dragging the test model through water while measuring the speed and the
total resistance force on the model ship. In this case, the rudder and stator fins were mounted on
the hull during the resistance test. The provided parameters of the resistance test can be seen in
Table 3.11. The resistance test has been conducted at twelve different speeds.
The Froude number is calculated from the model scale ship speed Vm , and the waterline length
of the model ship using eq. (2.4).
The frictional resistance coefficient is calculated using the Schoenherr line eq. (2.13). The
Schoenherr line is not the recommended friction line in the ITTC procedure [17, Sec. 2.1], but
has been used by this particular towing tank. From the results of the resistance test, the residual
resistance coefficient is calculated using the form factor method [17, Sec. 3.6.1]:

CR = CT m − (1 + k)CF m,Sh (3.1)

The results of the resistance test are presented in Section 4.1.1.

3.2.2 Open Water Test

The open water test is a test of the propeller without the presence of the ship in a towing tank or
cavitation tunnel where the performance of the propeller, also called the propeller characteristics,
are measured. The test is usually conducted by measuring water inflow velocity V0 , thrust T0 ,
torque Q0 and the rotation rate of the propeller n0 . It is conducted on the propeller alone, without
the presence of the ship, which eliminates the interaction between the hull and the propeller.
Results of the open water test for both the actual propeller and the stock propeller have been
provided and can be seen in Table 3.12. The open water test of the actual propeller has been
conducted at four different advance ratios while the open water test of the stock propeller has
been conducted at eight different advance ratios.
3.2 Towing Tank Tests 17

Table 3.12: List of the open water test parameters provided by the towing tank.

Open Water Test

Propeller Reynolds number Rep
Advance ratio Jo
Thrust coefficient KT o
Torque coefficient KQo
Propeller open water efficiency ηo

The open water test of the stock propeller has been carried out at a specific propeller Reynolds
number Rep . The rate of rotation of the propeller no is determined by the propeller Reynolds
number using the following formula:

Rep ν
no = = 17.59 rps (3.2)
D2
The parameters of the open water test are defined in the ITTC Open Water procedure [32, Sec.
2.1]. The advance ratio of the propeller:

Vo
Jo = (3.3)
no D
The thrust coefficient:

To
KT o = (3.4)
ρn2o D4
The torque coefficient:

Qo
KQo = (3.5)
ρn2o D5
The open water propeller efficiency:

Jo K T o
ηo = (3.6)
2πKQo

The results of the open water test are presented in Section 4.1.2.

3.2.3 Self-Propulsion Test

The self-propulsion test is conducted similarly to the resistance test, but with the stock propeller
mounted on the ship. The ship moves forward by rotating the propeller, which creates thrust.
A towing force is also applied to the ship. During the self-propulsion test, the propeller thrust
T , torque Q, and rotation rate along with the towing force are measured at a given speed. The
18 3 Method

Table 3.13: List of the self-propulsion test parameters provided by the towing tank.

Self-Propulsion Test
Ship speed Vm
Total resistance coefficient rT s
Thrust deduction t
Advance ratio (thrust identity) JT
Wake fraction (thrust identity) wT
Relative rotative efficiency ηR
Propeller open water efficiency ηo

provided parameters of the self-propulsion test can be seen in Table 3.13. The self-propulsion test
has been conducted at the same twelve speeds as the resistance test.
The resistance coefficients provided by the towing tank are normalized using two different
methods. One normalization method is used for coefficients denoted with r, another for coefficients
denoted with C. The two normalization methods are defined as:

R
C= 1 2
(3.7)
2 ρSV
R
r= (3.8)
ρ∇2/3 V 2
where ∇ is the volume displacement. Coefficients normalized with one of the methods can be
transformed to the other method„ as both the displacement and the wetted surface are known.
A towing force FD is applied to the ship in the self-propulsion test. The towing force or skin
correction force, as it is also called, is applied in order to compensate for the difference between
the model scale and full scale frictional resistances. There are three different methods for applying
the skin correction force [33, Sec. 3.4.1]: Load Varying (or Constant Speed) Method, Constant
Loading Method, and Mixed Loading Method. In this case, the towing tank uses the Constant
Loading Method. The skin correction force is given by [33, Sec. 2.1]:

1
FD = ρm Sm Vm2 ((1 + k)(CF m − CF s ) − ∆CF ) (3.9)
2
where ∆CF is the roughness allowance coefficient. This expression can be rewritten using
eq. (3.7) and eq. (3.8).

FD = ρ∇2/3 V 2 (rT m − rT s ) (3.10)

The model scale total resistance coefficient CT m is provided in the resistance test. This co-
efficient is renormalized into rT m . The full scale resistance coefficient rT s is provided in the
self-propulsion test. Thereby, the skin correction force can be calculated using eq. (3.10).
The thrust of the self-propulsion test is calculated from the thrust deduction, the towing force,
and the total resistance of the resistance test [33, Sec. 3.5]:
3.2 Towing Tank Tests 19

RT m − FD
Tm = (3.11)
(1 − t)

The towing tank uses the thrust identity method to derive additional parameters. The thrust
identity method assumes that the thrust coefficient of the propeller behind the ship is equal to
that of the open water test for a specific advance ratio [5, Sec. 8.7.4.3]. Using the thrust identity
method and the parameters given by the towing tank, other parameters are calculated.
The wake fraction and the advance ratio, both derived using thrust identity, are given by the
towing tank. Using these parameters, the rate of rotation of the propeller (nm ) is calculated [33,
Sec. 3.5]:

Vm (1 − wT )
nm = (3.12)
JT D

The thrust coefficient is calculated using the thrust and the rotational rate of the propeller [33,
Sec. 2.1]:

Tm
KT m = (3.13)
ρn2m D4

The propeller open water efficiency and the relative rotative efficiency have been provided by
the Japanese shipyard. The efficiency of the propeller in open water is defined in eq. (3.6), and
the relative rotative efficiency is defined as [33, Sec. 3.5]:

KQom
ηR = (3.14)
KQm

The efficiency of the propeller behind the ship (ηb ) is calculated using the propeller open water
efficiency and the relative rotative efficiency:

ηb = ηo ηR (3.15)

The propeller torque coefficient behind the ship (KQm ) is calculated using:

JT K T m
KQm = (3.16)
2πηb

The results of the self-propulsion test are presented in Section 4.1.3.

20 3 Method

3.2.4 Resistance Extrapolation

A variant of the form factor method has been adopted by the ITTC as the recommended procedure
for extrapolating resistance tests from model scale to full scale. The current version of the extrap-
olation procedure is described in [10]. The form factor method assumes that the form factor is
independent of the Reynolds number. The validity of this assumption is discussed in Section 2.2.4.
A resistance test is performed. The model scale total resistance coefficient and the form factor
are obtained from the resistance test. The model scale frictional resistance coefficient is derived us-
ing a friction line. Based on the form factor, total and frictional coefficients, the residual resistance
coefficient is calculated using the form factor method eq. (3.1).
The full scale ship speed is calculated using the Froude scaling law and the scaling factor:

√
Vs = Vm λ (3.17)

The full scale Reynolds number is calculated:

Vs LW Ls
Res = (3.18)
νs
The total resistance coefficient in full scale without bilge keel and with smooth hull CT s,smooth
is given by this expression:

CT s,smooth = (1 + k)CF s + CR (3.19)

The full scale total resistance coefficient, including hull roughness, air resistance, general cor-
relation and added resistance of the bilge keel is given by [10, Sec. 2.4.1]:

S + SBK
CT s = ((1 + k)CF s + ∆CF + CA ) + CR + CAAS (3.20)
S
where SBK is the wetted surface of the bilge keels, CA is the correlation coefficient, and CAAS is
the air resistance coefficient.
The full scale residuary resistance coefficient is assumed to be equal to the model scale residuary
resistance coefficient for any specific Froude number as described in Section 2.2.
The full scale frictional coefficient is calculated using the ITTC-57 model-ship correlation line
eq. (2.12).
The roughness allowance coefficient takes into account the additional roughness of the actual
hull due to welding seams, fouling, anodes etc. The roughness allowance coefficient is given by [10,
Sec. 2.4.1]:

(( )1 )
ks 3 −1
∆CF = 0.044 − 10Res 3 + 0.000125 (3.21)
LW Ls
ks = 150 · 10−6 m
3.2 Towing Tank Tests 21

where ks is the Nikuradse’s equivalent sand roughness. The value of ks indicates the roughness
of the hull surface. The ITTC procedure recommends a value of 150 µm if no data of the hull
roughness exists.
The correlation allowance corrects for systematic errors in the extrapolation procedure. It is
expressed as [10, Sec. 2.4.1]:

CA = (5.68 − 0.6 log Res ) · 10−3 (3.22)

The air resistance coefficient takes into account the added air resistance of the full scale ship.
It is given by [10, Sec. 2.4.1]:

ρA A V S
CAAS = CDA (3.23)
ρs S s

where CDA is the air drag coefficient, ρA is the air density, AV S is the projected frontal area above
water, and Ss is the wetted surface of the hull and the rudder. The ITTC procedure recommends
an air drag coefficient value of 0.8 as a default.
The extrapolated resistance results can be seen in Section 4.1.4.

3.2.5 Open Water Extrapolation

ITTC states that the self-propulsion extrapolation has to use the full scale open water characteristics[32].
The open water characteristics for the stock propeller have only been given in model scale and
therefore have to be extrapolated to full scale. Several methods are avalible and these are discussed
in [34]. The ITTC procedure for extrapolating the propeller characteristics of the open water test
from model scale to full scale is described in [10]. The extrapolation is based on results from the
open water test.
The local Reynolds number on the propeller blade at r/R equal to 0.7 Rec0.7 is given by [32,
Sec. 2.1]:

c0.7 (Vo2 + (0.7πnD)2 )1/2

Rec0.7 = (3.24)
νm

where c0.7 is the chord length at r/R equal to 0.7. The propeller drag coefficient in model scale
CDm and full scale CDs are given by [10, Sec. 2.4.2]:

( )( )
t0.7 0.044 5
CDm = 2 1 + 2 − (3.25)
c0.7 (Rec0.7 )1/6 (Rec0.7 )2/3
( )( )
t0.7 c0.7 −2.5
CDs = 2 1 + 2 1.89 + 1.62 ln (3.26)
c0.7 kp

where t0.7 is the maximum thickness of the propeller blade at r/R equal to 0.7, and kp is
the blade roughness. The ITTC procedure recommends a blade roughness value of 30 · 10−6 m as
22 3 Method

default. The default value is used in this extrapolation. The full scale thrust and torque coefficients
are then calculated [10, Sec. 2.4.2]:

cZ
KT os = KT om + 0.3(CDm − CDs )P /D (3.27)
D
cZ
KQos = KQom − 0.25(CDm − CDs ) (3.28)
D

where P /D is the pitch ratio and Z is the number of blades. The extrapolated open water
results can be seen in Section 4.1.5.

3.2.6 Self-Propulsion Extrapolation

The procedure for extrapolating self-propulsion results from model scale to full scale are described
in [10]. The extrapolation is based on results of the resistance test, the extrapolated resistance
test and the self-propulsion test.
Firstly, the full scale wake fraction is calculated using [10, Sec. 2.4.3]:

(1 + K)CF s + ∆CF
wT s = (t + 0.04) + (wT m − t − 0.04) (3.29)
(1 + K)CF m

Then the full scale propeller load is calculated [10, Sec. 2.4.3]:

KT s Ss CT s
= (3.30)
2
Js 2Ds (1 − t)(1 − wT s )2
2

From the propeller load, the advance ratio and torque coefficient are interpolated from the full
scale propeller characteristics of the actual propeller and the stock propeller. The interpolation
is a linear interpolation between the two nearest data points of the propeller characteristics. The
rate of rotation is given by [10, Sec. 2.4.3]:

(1 − wT s )Vs
ns = (3.31)
JT s Ds

The delivered power of the ship PDs is given by [10, Sec. 2.4.3]:

KQs
PDs = 2πρDs5 n3s (3.32)
ηR
(3.33)

The extrapolated self-propulsion results can be seen in Section 4.1.6.

3.3 Sea Trial 23

3.3 Sea Trial

A sea trial is a series of tests conducted in order to verify that the newly build ship meets the
requirements of the contract, the classification society and the flag state. A sea trial consists of
many tests. One of these tests is the speed trial, which is important for this thesis. The speed
trial is a test of the speed at which the ship is able to sail with a certain output of the main engine.
In this thesis, the results of the speed trial provide an opportunity to compare the performance of
the actual full scale ship with results from the towing tank and the CFD simulations.
This section will firstly describe the sea trial method briefly. Furthermore, the measurements
and corrections subsequently will be described in more detail with the measurements and correc-
tions.

3.3.1 Speed Trial

Four sister ships of the same design have been built in total. Speed trials have been conducted
for each of the four sister ships. In 2015, the International Maritime Organisation (IMO) and
ITTC released ISO15016:2015 [4], an updated version of the ISO 15016:2002 standard procedure
for carrying out and correcting speed trials. The speed trial of the fourth ship has been conducted
mostly as described in [4]. The speed trials of the three other ships have been conducted prior to
the release of that procedure, but is assumed to have been conducted using similar procedures.
The speed trial is conducted at a number of different speeds and using double runs for each
speed. Double runs refer to the method of conducting one speed run, turning 180 the ship degrees
and conducting a second speed run. The data gathered from a speed trial are the measured power
from the main engine and the speed of the ship.
The Japanese shipyard has provided the results of these speed trials for all four ships. Fur-
thermore, the authors were attending the sea trial of the last sister ship. This gave he authors
the opportunity to experience how a speed trial is conducted, the measurements performed, and
factors influencing the speed trial causing uncertainties.
The main measurements in the speed trial are speed, propeller rate of rotation, and the engine
power. After the speed trial, the measurements are corrected for the environment forces to the
final results.

3.3.2 Speed Measurement and Corrections

The speed of the ship during the speed trial is measured using a Differential-GPS (DGPS). There
is already a DGPS installed on board, but a temporary and more precise DGPS is installed on top
of the bridge during the sea trial. The temporary DGPS is connected to a laptop on the bridge of
the ship. From the DGPS data, the ship speed over ground is calculated by assuming no lateral
drift. The ship speed is recorded during each speed run, with a sampling frequency of about 1 Hz.
From the recorded data, a mean ship speed is calculated.
The speed over ground signal is visually inspected, showing only minor fluctuation. Based
on observations, it is concluded that the speed signal is reliable and that the mean is a good
measurement of the actual speed over ground of the ship. To obtain the speed through water, the
speed measurement has to be corrected for current.
24 3 Method

Tidal Current Correction

The speed measured by the DGPS is the speed over ground. The measured speed is corrected for
the tidal current in order to obtain the speed through water. The location of the speed trial has
significant tidal current. A tidal current curve is used to calculate the speed of the tidal current
in the east-west direction as a function of time. The origin of the tidal current curve is unknown
to the authors. The tidal current is subtracted from the measured speed, which gives the speed
though water. The tidal current curve shows that the tidal current varies by more than two knots
in the time period of the speed trial. A speed correction of this magnitude is significant. To
minimize the influence of the tidal current and possible discrepancies in the tidal current speed
curve, the speed trial is conducted using two runs: one in the direction against the current and
one with the current.

3.3.3 Power Measurement and Correction

The power of the engine is calculated from the engine rate of rotation and the amount of fuel
injected (fuel index) using a formula provided by the engine manufacturer Mitsui. Both the rate
of rotation and the fuel index are sampled directly from the engine control system. The RPM of
the engine is measured directly on the shaft, and there is no gear box. The engine manufacturers
formula for calculating the engine power is:

P = Ccyl · Ceng · ω · pme (3.34)

3
Ccyl = 5.339 m
60
Ceng = · 10 · 0.7355 · 10−3
2π
Where P is the engine power, Ccyl is a cylinder coefficient provided by the engine manufacturer,
Ceng is an engine constant also provided by the engine manufacturer, ω is the rate of rotation, pme
is the mean effective pressure, and F I is the fuel index. The mean effective pressure is calculated
from the fuel index F I, which is indicating the amount of fuel injected, using a linear relation.
The linear relation is also given by the engine manufacturer:

pme = 19375F I − 105938 (3.35)

The values of the constants in eq. (3.34) and eq. (3.35) are given by the engine manufacturer.
It can be difficult to create a good model of the power output of an engine. The uncertainties of
the engine formula are not easily quantified so the accuracy of the engine formula is not known,
and neither can it be compared to other methods of measuring the engine power.
ISO 15016:2015 states that the power has to be measured using a torque meter mounted directly
on a shaft. Such a device is also installed, but the measurements are not used in the speed trial
results.

Displacement correction
The measured power is corrected for the difference between the planned and the actual displacement
of the ship [4, Annex H]. The planned displacement of the ship in sea trial condition is calculated
3.3 Sea Trial 25

before conducting the sea trial, and the actual displacement is measured at the beginning of the
sea trial. The correction is conducted by the shipyard. The difference in displacement for the sea
trial used in this thesis is very small. Therefore the correction of the power is also very small.

Wind Correction
The performance of the ship is corrected for the added wind resistance [4, Annex C]. The wind
correction is based on an estimated drag coefficient, an estimated projected front area and a
relative wind speed. The relative wind speed cannot be measured directly and is therefore found
by subtracting the vessel velocity from the absolute wind velocity. The absolute wind velocity is
measured by a wind anemometer.
The wind anemometer is placed above the bridge, just behind a rotating rectangular radar. The
authors estimate that the wind anemometer is not significantly influenced by the super structure
if the absolute wind direction is from the front. It is estimated by the authors that the wind
measurements might be influenced by the rotating radar, especially when the wind direction is
from the starboard side.

Wave Correction
The results of the speed trials are not corrected for waves. The speed trial procedure ISO15016:2015
specifies that the speed trial data should be corrected for waves [4, Sec. 7.3]. In this case, the
Japanese shipyard and the ship owner agreed not to correct for waves. Omitting the wave correction
gives a conservative estimate on the performance.
A simplified method for power correction due to waves is described in [4, D.1]. This method
is used to correct the speed trial results for weather waves. The method requires significant wave
height and direction of the waves. These two parameters were not recorded during any of the
sea trials of the four sister ships. Instead, the parameters have been provided by Eirik Eisinger
from DNV-GL. Eirik Eisinger is writing his master’s thesis [35] at DNV-GL in Høvik in the spring
of 2016. In Eisinger’s thesis, ERA-Interim data [36] is used. The ERA-Interim data is based on
satellite weather data and a meteorological model. The provided significant wave heights Hs and
wave directions is a mean value of a 6 hour window. It was shown that significant wave height
data has a relatively low uncertainty, but the wave direction has a relatively high uncertainty. The
uncertainties are acceptable since the magnitude of the wave corrections are relatively small.
The added resistance due to waves RAW L is estimated using following formula [4, D.12]:

√
1 B
RAW L = ρs g(Hs)2 B (3.36)
16 LBW L

LBW L is the distance from the ship bow to 95% of the maximum beam on the water line, as
illustrated in Figure 3.14.
The power will be corrected with the power correction due to waves PW :

PW = RAW L Vs
26 3 Method

Figure 3.14: Definition of LBW L , the distance from the ship bow to 95% of the maximum beam
on the water line. [4, Figure D.2]

Resistance Test Open Water Test

Model Scale Model Scale

Resistance Test Self-Propulsion Test

Full Scale Model Scale

Self-Propulsion Test
Full Scale

Figure 3.15: The five CFD setups. The arrows indicate the direction of the work flow.

This method has several restrictions. First of all, it is only applicable if the significant wave height
satisfies the following criterion:

√
Hs ≤ 2.25 LP P /100 (3.37)

Furthermore, the pitch and heave of the ship must be small (vertical acceleration at the bow
must to be less than 0.05g). The correction is only applied if the wave direction is from ahead
(from 0◦ to ±45◦ ). The fact that waves within 45◦ degrees result in full correction and waves not
within 45◦ result in no correction shows that the correction method is very coarse. Other wave
correction methods are described in [4].

3.4 CFD
This section describes the CFD setups developed and used in this project. All CFD simulations
in this project are performed in the commercial CFD-code STAR-CCM+ v.10.04.011 from CD-
Adapco. The code is widely used in the marine industry and is well-known for its capabilities
within marine applications.
In total five CFD setups are developed: model scale resistance setup, full scale resistance setup,
open water setup, model scale self-propulsion setup, and full scale self-propulsion setup. The five
CFD setups can be seen in Figure 3.15.
3.4 CFD 27

In the beginning of the project period, it was not known if the rudder and stator fins were
included in the towing tank tests. Furthermore there has been some doubt about the provided
drawings of the stator fins as the angles of attack of the fins are much higher than what is normally
encountered by the authors and experts at DNV-GL. Therefore, the rudder and stator fins are not
included in some of the CFD simulations. Later, it was confirmed from the shipyard that the
rudder and stator fins were mounted in all towing tank tests and that the specifications of the
stator fins were correct. Therefore, there are differences between the appendages used in some of
the CFD simulations and the towing tank tests to which they are compared. These differences are
mentioned when they occur.

3.4.1 Validation and Convergence Test

It is essential to trust a CFD setup before conclusions are based on the results of the setup. Several
methods are used in this thesis to gain trust in the CFD setup. Both convergence studies and
comparison studies are performed, where possible. This section explains the details of the methods
used.
A convergence study is conducted on a CFD setup in order to determine the spatial (See Sec-
tion 2.3.5) and temporal (See Section 2.3.6) discretization settings of the setup. The discretization
settings are a balance between minimizing the discretization error and using the least amount of
computational effort.
A convergence study is conducted by creating a series of simulations from the same CFD
setup. The simulations are identical in all respects, except for either the spatial or the temporal
discretization. The discretization is changed for each simulation, creating a series of simulations
with a range in the discretization from coarse to fine. The solution of the simulations are expected
to converge towards a certain value as the discretization becomes finer because the discretization
error is minimized. From the results of the series of simulation the balance between the accepted
discretization error and the computational effort is determined.
In total, three different types of convergence studies are conducted in this thesis: boundary
layer mesh convergence study, volume mesh convergence study, and time step convergence study.
These studies determine the discretization settings of the boundary layer mesh, the volume mesh
and the time step respectively.
After the convergence study, the CFD results are compared to experimental data, if available.
If the CFD results show convergence and are similar to experimental measurements, the CFD
setup is considered validated. A validation does not mean that the all results from the setup can
be trusted, but it is a good indication of the trustworthiness. The more cases the setup is validated
for, the more the CFD setup can be trusted.

3.4.2 CFD Resistance Test

This section documents the CFD setup of the calm water resistance test, both in model scale and
in full scale. First, the model scale CFD setup is created and validated by comparing results with
the towing tank results. The validated model scale setup is then modified to a full scale setup.
Convergence studies are conducted in both the model and full scale setup. Results of the full scale
CFD setup are compared with the extrapolated resistance data from the towing tank data.
28 3 Method

Table 3.17: Domain size of the CFD resistance setup.

Longitudinal length 5 LP P
Vertical length 3 LP P
Transverse length 1.5 LP P

All resistance simulations are performed in the sea trial condition (see Section 3.1.1) similarly
to the towing tank test.
The initial CFD setup used to estimate the calm water resistance is an automated CFD setup
developed by OSK-ShipTech A/S. CFD Specialist Stig Staghøj Knudsen from OSK-ShipTech A/S
has accepted the use of the setup in this thesis. The author Henrik Mikkelsen has worked exten-
sively with the setup both in the development and validation phases during his employment at
OSK-ShipTech A/S. The automated setup is a set of Java-scripts which can be played as a macro
in STAR-CCM+.

Model Scale Resistance

The CFD setup presented in this section simulates the resistance test in model scale.

Physics The CFD resistance setup uses a number of different physics models to simulate the
flow around the ship. The physics models used for the model scale resistance setup are briefly
discussed here. For more details on the models, see Section 2.3.
The setup uses the k - ϵ two-equation model to model the turbulence of the flow. Free surface
waves are modelled using volume of fluid. The hull is allowed dynamic pitch and heave motions,
by use of the DFBI model, as described in Section 2.3.4. In order to model the free surface waves
and the dynamic pitch and heave, a transient solver is required. The simulations are solved using
a 1st order implicit unsteady solver.

Domain The domain of the simulation is the spatial region in which the simulation takes place.
The shape of the domain is a rectangular box. Since the propeller is not present, it is reasonable
to assume that the flow around the hull is symmetrical, with a vertical plane dividing the port and
starboard sides. Due to this assumption, only the starboard side or port side of the ship needs
to be simulated. This halves the number of cells in the mesh, which reduces the computing time
significantly. The domain can be seen in Figure 3.16. The size of the domain as function of length
between perpendiculars LP P can be seen in Table 3.17, and the boundary conditions can be seen
in Table 3.18.

Spatial Discretization The volume mesh consists of hexahedrons in a structured grid. The
mesh is generated using the trimmer-mesh function in STAR-CCM+. Boundary layer mesh is used
on the hull surface in order to make a good estimation of the shear stresses on the hull.
3.4 CFD 29

Top Top

Hull Inlet Hull Side

Free Surface
Outlet Symmetry

Bottom Bottom
Figure 3.16: Domain of the resistance setup, with boundary names. Seen from starboard side
(left) and from the front (right).

Table 3.18: Boundaries and boundary conditions of the CFD resistance setup.

Name of Boundary Boundary Condition

Inlet Velocity inlet, prescribed with Vm
Top Velocity inlet, prescribed with Vm
Bottom Velocity inlet, prescribed with Vm
Outlet Pressure outlet
Symmetry Symmetry plane
Side Symmetry plane
Hull No-slip wall

Table 3.19: Refinement zones of the CFD resistance setup.

Name of RZ Refinement Direction

Far Free Surface RZ Longitudinal and transverse
Middle Free Surface RZ Longitudinal and transverse
Near Free Surface RZ Longitudinal and transverse
Bulbous Bow RZ All
Transom RZ All
Vertical Free Surface RZ Vertical

Six refinement zones (RZ) are used to refine the mesh in important regions. These six refinement
zones can be seen in Table 3.19, Figures 3.20 and 3.21. The purpose of the far, middle, near and
vertical free surface refinement zones is to resolve Kelvin waves. The theory behind the Kelvin
waves is briefly presented in Section 2.1. The bulbous bow and transom refinement zones ensures
sufficient resolution of the flow at the bow and stern. The refinement direction and cell size of the
refinement zones can be seen in Table 3.19.
30 3 Method

Transom RZ Bulbous Bow RZ

Near Free Surface RZ

Middle Free Surface RZ

Far Free Surface RZ

Figure 3.20: The far, middle, near, bulbous bow and transom refinement zones of the CFD
resistance setup. Seen from the top.

Transom RZ Bulbous Bow RZ

Vertical Free Surface RZ

Figure 3.21: The vertical free surface, bulbous bow and transom refinement zones of the CFD
resistance setup. Seen from starboard side.
3.4 CFD 31

Table 3.22: Appendages used in the model scale CFD resistance studies.

Study With Rudder With Stator Fins

Boundary Layer Mesh Convergence No No
Volume Mesh Convergence No No
Time Step Convergence No No
Comparison Yes No

Table 3.23: Appendages used in the full scale CFD resistance studies.

Study With Rudder With Stator Fins

Volume Mesh Convergence No No
Time Step Convergence No No
Comparison Yes No

Studies Four simulation studies have been conducted using the model scale resistance setup: a
boundary layer mesh convergence study, a volume mesh convergence study, a time step convergence
study and a comparison with the results of the towing tank resistance test. The appendages
mounted on the hull is different in the four setups, and can be seen in Table 3.22.
The results of the model scale resistance studies can be seen in Section 4.3.1.

Full Scale Resistance

The full scale resistance setup is created by scaling up the model scale resistance setup with the
scaling factor (2.3). Three studies have been conducted on the full scale resistance setup: volume
mesh convergence study, time step convergence study and a comparison with the extrapolated
resistance test from the towing tank. The appendages mounted on the hull is different in the three
setups are shown in Table 3.23.

3.4.3 CFD Open Water Test

This section documents the CFD setup of the model scale open water test of the stock propeller.
The stock propeller is described in Section 3.1.2. Modelling the open water test using CFD is less
complex than modelling the self-propulsion test using CFD because the hull-propeller interaction
is not included in the open water setup. The idea is that a separately validated resistance and
open water setup can be combined to a self-propulsion setup. This self-propulsion setup can then
be validated in model scale by comparing it with self-propulsion results from towing tank tests.

Physics

The setup uses the k - ϵ turbulence model. A single phase with the properties of water is used in
the open water setup. For details on the physics models used in this thesis, see Section 2.3.
32 3 Method

Top Top
Interface Interface
Inlet
Outlet Right
Left

Bottom Bottom

Figure 3.24: Domain of the CFD open water setup, with boundary names. Seen from the side
(left) and from the front (right).

Table 3.25: Domain size of the CFD open water setup.

Stationary Domain
Upstream length 9D
Downstream length 20D
Vertical length 8D
Transverse length 8D
Rotating Domain
Upstream length 2.5D
Downstream length 1.5D
Diameter 2D

Table 3.26: Boundary conditions of the CFD open water setup.

Name of Boundary Boundary Condition

Inlet Velocity inlet equal to Vo
Outlet Pressure outlet
Top Symmetry plane
Bottom Symmetry plane
Left Symmetry plane
Right Symmetry plane
Interface Wall and internal interface
Shaft, Blades, Hub No-slip wall

Domain

The open water setup uses two domains: A stationary domain and a rotating domain. The two
domains are connected by an interface. The shape of the stationary domain is a rectangular
cubeoid. The shape of the rotating domain is a cylinder which is located inside the static domain
and around the propeller. A illustration of the domains can be seen in Figure 3.24. The size of
the domains can be seen in Table 3.25, and the boundary conditions of the domains can be seen in
Table 3.26. The size of the domains are determined, based on experience from experts at DNV-GL.
3.4 CFD 33

Table 3.27: Refinement zones of the CFD open water setup.

Name of RZ Refinement Direction

Far Propeller RZ All
Near Propeller RZ All

Far Propeller RZ

Interface

Near Propeller RZ

Figure 3.28: The far and near propeller refinement zones of the open water setup. Seen from
the side.

Spatial Discretization

The volume mesh consists of hexahedrons in a structured grid. Boundary layer mesh is used on
the shaft, blades and hub.
In order to better resolve the flow around the propeller, two refinement zones are used to
refine the mesh in the volume around the propeller. The two refinement zones are illustrated in
Table 3.27 and Figure 3.28. The near propeller refinement zone generates a finer mesh than the
far propeller refinement zone.

Propeller Rotation

The rotation rate of the propeller in the open water setup is equal to the setup used in the open
water test of the towing tank. The rotation rate of the open water test is defined in eq. (3.2).
The propeller movement can be modelled using different two methods: the moving reference
frame method (MRF) or the rotating solid body method (RSB).
MRF works by adding a rotational velocity component to the flow in the rotating domain
while the propeller geometry is held stationary. The added rotational velocity component Vθ can
be expressed as function of the rotational rate and the radial distance from the propeller center r:
34 3 Method

Vθ = 2πnr

MRF can be simulated using a steady-state solver and it has a faster convergence compared to the
rotating solid body method, which requires a transient solver.
RSB works by rotating the propeller geometry in increments at each new time step. This
method requires more computational time, but is a better model of the actual propeller physics
compared to the MRF method. The RSB method models the hull-propeller interaction while the
MRF model does not.
The two methods can be combined for computational efficiency. The first step in the combi-
nation approach is to firstly accelerate the flow using MRF. When the simulation has converged,
it is changed to the more precise and computational heavy RSB and continued until it converges
again. The combination of methods saves a significant amount of computational time.
Both simulations using MRF and the combination approach have been performed. It is con-
cluded from the results of these simulations that the increased accuracy of using the combination
of methods is negligible for the open water simulations, compared to using only MRF. Therefore,
it is decided only to use MRF for all open water simulations in order to save computational time.

Studies
Two simulation studies have been conducted with the open water setup: a volume mesh convergence
study and a comparison with the open water test from the towing tank. The results of the open
water simulations can be seen in Section 4.3.2.

3.4.4 CFD Self-Propulsion Test

The following sections present the background for the self-propulsion simulations. The first section
describes the simulations in model scale and it is followed by a section describing the changes made
from the model scale setup to the full scale setup.

Model Scale Self-Propulsion

This section describes the CFD setup that simulates the calm water self-propulsion test in model
scale. The setup is created by combining the model scale resistance CFD setup (see Section 3.4.2)
with the propeller and the open water CFD setup (see Section 3.4.3).

Physics The physics of the self-propulsion setup is similar to the physics of the resistance setup,
but some changes are made. For details on the physics models used in this project see Section 2.3.
The setup uses the k - ϵ turbulence model. Based on the experience from DNV-GL, it has been
decided not to model the free surface waves using the VOF model. Experience from DNV-GL
shows that including the VOF model in the self-propulsion setup causes large fluctuations in the
resistance, thrust and torque signals. The magnitude of these fluctuations is so large that the
fluctuations make it difficult to obtain reliable signals.
Without the free surface waves, it is not possible to model the dynamic pitch and heave of the
ship using the DFBI model. This is because the DFBI model works by balancing the forces acting
3.4 CFD 35

Top Hull Top Hull

Inlet
Outlet Starboard Side Port Side

Bottom Bottom

Figure 3.29: Static domain of the self-propulsion setup, with boundary names. Seen from
starboard side (left) and from the front (right).

Rotating Domain

Figure 3.30: Rotating domain of the self-propulsion setup. Seen from starboard side.

on the ship. Without the free surface waves, the wave forces on the ship are not modelled, and
the DFBI can not correctly balance the forces. Therefore the DFBI model is not included in the
self-propulsion setup.
Instead of modelling the dynamic pitch and heave using the DFBI model, the dynamic pitch
and heave of the resistance simulations are prescribed to the self-propulsion simulations before
they are simulated. The influence of the propeller dynamics on the dynamic pitch and heave of
the ship is neglected by using the dynamic pitch and heave of the resistance simulations instead
of using the DFBI model in the self-propulsion simulations. Neglecting this influence introduces a
modelling error which has not been quantified, but is assumed to be smaller than other modelling
errors.

Domain As in the open water setup, the self-propulsion setup uses two domains; a large static
domain, with a small rotating domain inside. The shape of the static domain is a rectangular
cubeoid, and the shape of the rotating domain is a cylinder.
The static domain can be seen in Figure 3.29. The size of the domains can be seen in Table 3.31,
and the boundary conditions of the domains can be seen in Table 3.32.
The static domain of the self-propulsion setup includes both the port and starboard sides of
the domain. This is different from the resistance setup, where only one side is simulated while the
opposite side is accounted for using a symmetry plane. The reason for simulating both sides in
36 3 Method

Table 3.31: Domain size of the CFD self-propulsion setup.

Stationary Domain
Longitudinal length 5 LP P
Vertical length 3 LP P
Transverse length 3 LP P
Rotating Domain
Upstream length 0.11D
Downstream length 0.23D
Diameter 1.20D

Table 3.32: Boundary conditions of the CFD self-propulsion setup.

Name of Boundary Boundary Condition

Inlet Velocity inlet, prescribed with Vm
Bottom Velocity inlet, prescribed with Vm
Outlet Pressure outlet
Top Symmetry plane
Starboard Side Symmetry plane
Port Side Symmetry plane
Interface Wall and internal interface
Hull, Shaft, Blades, Hub No-slip wall

the self-propulsion setup is because propeller flow is unsymmetrical. Therefore, the self-propulsion
calculations are not modelled using a symmetry plane in the center line.
Without the use of the VOF model, the self-propulsion setup only needs to model the water
phase of the flow. Therefore the top of the static domain of the self-propulsion setup is at the free
surface, and is prescribed with a symmetry plane boundary condition.
Removing the free surface waves from the simulations gives rise to another problem. In reality,
a stern wave is created by the ship. The stern wave increases the water height above the propeller,
fully submerging the propeller even at low draft conditions such as the sea trial condition. Without
the free surface waves, the propeller is not fully submerged. In order to fully submerge the propeller
in the self-propulsion setup, the free surface is raised approximately 3% of the draft in the sea trial
condition.

Free Surface Correction The decision not to model the free surface waves and to raise the
calm water surface decreases the measured resistance on the ship. In order to compensate for this
change, a free surface correction is performed on the results of the self-propulsion simulations. The
correction is performed in two steps. The free surface correction procedure is shown in Figure 3.33.
The first step is to correct for not modelling the free surface waves, which reduces the hull
resistance by removing the wave making resistance. The wave making resistance is quantified
by simulating the resistance setup both with and without the free surface waves. The difference
between those two results is the wave making resistance:
3.4 CFD 37

Free surface waves

RT Initial surface height

No free surface waves

RV Initial surface height

No free surface waves

RV,SH Raised surface height

Figure 3.33: The three simulation setups of the free surface correction procedure.

RW m = RT m − RV m (3.38)

where RW m is the wave making resistance and RV m is the total resistance without free surface
waves, also referred to as the viscous resistance.
The second step is to correct for the raised free surface which increases the resistance on the hull,
because a larger surface area is submerged. The increase in resistance is quantified by simulating
the resistance CFD setup both with the initial and the raised surface heights. The difference
between those two results is the added resistance due to the raised surface ∆RSHm given by:

∆RSHm = RV m,SH − RV m (3.39)

The resistance of the self-propulsion simulations can then be corrected for wave making resis-
tance and added resistance due to the raised surface height:

RT m = RV m,SH + RW m − ∆RSHm (3.40)

Spatial Discretization The self-propulsion setup uses two separate meshes; one for the station-
ary domain and one for the rotating domain around the propeller. Using two separate meshes
allows mesh options from the resistance setup to be used for the static domain and mesh options
from the open water setup to be used for the rotating domain.
Four mesh refinement zones are used in the self-propulsion setup. The four refinement zones
can be seen in Table 3.34 and Figure 3.35.

Propeller Rotation Two different methods or a combination of them can be used to model the
rotation of the propeller. The two methods are the moving reference frame method (MRF) and
the rotating solid body method (RSB). The methods are described in Section 3.4.3.
38 3 Method

Table 3.34: Refinement zones of the CFD self-propulsion setup.

Name of RZ Refinement Direction

Bulbous Bow RZ All
Transom RZ All
Far Propeller RZ All
Near Propeller RZ All

Transom RZ Bulbous Bow RZ

Near Propeller RZ
Far Propeller RZ

Figure 3.35: The four refinement zones of the self-propulsion setup. Seen from starboard side.

For the self-propulsion simulations a combination of the two methods is used. The modelling
of the propeller rotation is initialized by using the MRF method, followed by the RSB method.
In the self-propulsion tests, a towing force is applied to the model ship, as described in Sec-
tion 3.2.3. The CFD self-propulsion setup is carried out using the same principle. The forces on
the model ship must be balanced, which leads to this expression:

RT m = Tm + FD (3.41)

The towing force has been provided by the towing tank and is defined in eq. (3.10). The
corrected total resistance is calculated from eq. (3.40). The thrust is adjusted by changing the
propeller rotation rate during the simulation until the forces are in balance.
In total, the self-propulsion simulation uses three stages. First, the simulation is initialized
using MRF. Next, the RSB which changes the propeller rotation rate in order to balance the
forces is applied. Lastly, the propeller rotation rate is frozen at the force equilibrium and the
simulation is continued using RSB for one whole rotation. The three stages of the self-propulsion
simulation are illustrated in Figure 3.36.

Temporal Discretization The self-propulsion setup requires a transient solver in order to sim-
ulate the rotation of the propeller. An unsteady implicit solver with a first-order scheme is used.
3.4 CFD 39

MRF RSB RSB

const rotation rate variable rotation rate const rotation rate

Figure 3.36: The three stages of the CFD self-propulsion simulations.

Table 3.37: Appendages used in the model scale CFD self-propulsion studies.

Study With Rudder With Stator Fins

Time Step Convergence Yes No
Volume Mesh Convergence Yes No
Comparison Yes Yes

The choice of time step size is based on the physics of the propeller rotation. The time step
can therefore be expressed in terms of the number of degrees the propeller rotates between two
time steps:

∆θ = n · ∆t · 360◦ (3.42)

where ∆θ is the angle per time step and ∆t is the time step size.

Studies Three simulation studies have been conducted on the model scale self-propulsion setup.
The three studies are: a time step convergence study, a volume mesh convergence study and a
comparison with the self-propulsion results from the towing tank. The appendages mounted in
each of the three studies can be seen in Table 3.37.
The results of the model scale self-propulsion studies can be seen in Section 4.3.3.

Full Scale Self-Propulsion

The full scale self-propulsion setup is created by scaling up the model scale self-propulsion setup
according to the scaling factor (2.3).

Free Surface Correction The free surface correction method is changed slightly from model
scale to full scale in order to account for the roughness resistance, air resistance and bilge keel
resistance which are only relevant for the full scale simulations. The method is still conducted by
using the three simulations setups seen in Figure 3.33. The wave making resistance and the added
resistance of the increased surface height are calculated similarly to eq. (3.38) and eq. (3.39):

RW s = RT s − RV s (3.43)

∆RSHs = RV s,SH − RV s (3.44)

40 3 Method

Table 3.38: Appendages used in the full scale CFD self-propulsion studies.

Study With Rudder With Stator Fins

Time Step Convergence Yes Yes
Volume Mesh Convergence Yes Yes
Comparison Yes Yes

The full scale resistance corresponding to eq. (3.20) is then calculated using:

S + SBK
RT s = (RV s,SH + ∆RF ) + RW s + RAAS − ∆RSHs (3.45)
S
where ∆RF is the roughness resistance of eq. (3.21), and RAAS is the air resistance of eq. (3.23).
The correlation allowance of eq. (3.22) is not included in eq. (3.45) because its purpose is to correct
for systematic errors in the towing tank extrapolation procedure.
An alternative to using the ITTC procedure to estimate the roughness, air and bilge keel
resistances is to model them in the CFD simulations. The roughness resistance can be modelled
using a wall roughness model [28, p. 3405-3406]. The air resistance requires the superstructure of
the ship as a 3D geometry. If the 3D geometry of the superstructure is available, the air resistance
can be estimated either by including the superstructure directly in the CFD simulations, or by
running a separate set of simulations with the superstructure alone. The added resistance from
the bilge keels can quite easily be estimated by adding the bilge keels to the hull geometry.

Propeller Rotation The full scale self-propulsion setup uses the same method as the model
scale setup for modelling the propeller rotation (see Figure 3.36). The force balance used in the
full scale setup is changed slightly from eq. (3.41) because there is no towing force in full scale:

RT s = Ts (3.46)

The thrust is adjusted by changing the propeller rotation rate until the thrust is equal to
the resistance. The adjustments are performed by a proportional controller programmed by the
authors.

Studies The studies conducted on the full scale self-propulsion setup are a time step convergence
study, a volume mesh convergence study, and a comparison with the extrapolated self-propulsion
simulation from the towing tank. Both the rudder and stator fins are mounted in all full scale
self-propulsion simulations, as seen in Table 3.38.
The results of the full scale self-propulsion simulations can be seen in Section 4.3.3.
CHAPTER 4
Results
This chapter presents the results from towing tank tests in Section 3.4 followed by the speed trial
results in Section 3.4. Finally the results from the CFD simulations are presented in Section 4.3.

4.1 Towing Tank Test

The design of the ship has been tested in a towing tank prior to the project period of this thesis. The
conducted tests include resistance test, open water tests and self-propulsion tests. All of the towing
tank tests are conducted with rudder and pre-swirl fins mounted and at the sea trial condition. The
sea trial condition has 40.2% of the displacement of the design condition as explained Section 3.1.1.
The method behind the towing tank tests and the procedure for extrapolating from model scale
to full scale are presented in Sections 3.2.4 to 3.2.6. The following sections present the provided
towing tank test results and the extrapolated full scale results.

4.1.1 Results of Resistance Test

A resistance test has been carried out at the towing tank, as described in Section 3.2.1. The
model scale total resistance coefficient and the model scale frictional resistance coefficient have
been provided by the towing tank. The frictional resistance coefficient is based on the Schoenherr
frictional line in eq. (2.13). The residual resistance coefficient is calculated using eq. (3.1). The
model scale total, frictional and residual resistance coefficients are presented in Figure 4.1.
From Figure 4.1 it is seen that the frictional resistance is the largest part of the total resistance
in the studied Froude number range. It is also seen as expected, that the frictional resistance
coefficient becomes smaller as the Froude number and the Reynolds number increases, and that
the residual resistance coefficient becomes larger.

4.1.2 Results of Open Water Test

An open water test of the stock propeller has been carried out in the towing tank, as described
in Section 3.2.2. Data of the stock propeller can be seen in Table 3.4. The model scale propeller
characteristics of the stock propeller provided by the towing tank are presented in Figure 4.2.
As seen in Figure 4.2 the open water test is only conducted for advance ratios in the range
relevant for this ship design. Therefore propeller characteristics are not provided in the the bollard
pull range or in the range where the open water efficiency decreases.

4.1.3 Results of Self-Propulsion Test

A self-propulsion test has been carried out at the towing tank, using the stock propeller. The self-
propulsion test is described in Section 3.2.3. Data of the stock propeller is presented in Table 3.4.
42 4 Results

×10-3
6
Coefficients [-]

0
0.14 0.15 0.16 0.17 0.18 0.19 0.20 0.21
Fn [-]
Figure 4.1: Results of the resistance test from the towing tank. Model scale total resistance
coefficients CT m (circle), frictional resistance coefficients CF m,Sh (square) and residual resistance
coefficients CR (triangle) as a function of Froude number F n.

0.6
Coefficients [-]

0.5

0.4

0.3

0.2

0.1
0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60
J [-]
Figure 4.2: Results of the open water test from the towing tank. Propeller thrust coefficient
KT om,stock (circle), 10 times torque coefficient 10KQom,stock (diamond) and open water efficiency
ηom,stock (square) of the stock propeller in model scale as a function of advance ratio J.
4.1 Towing Tank Test 43

0.26
Coefficients [-]

0.24

0.22

0.14 0.15 0.16 0.17 0.18 0.19 0.20 0.21

Fn [-]
Figure 4.3: Results of the self-propulsion test from the towing tank in model scale. Propeller
thrust coefficient KT m (circle) and propeller torque coefficient KQm (square) as a function of
Froude number F n.

The thrust identity method has been used to derive parts of the self-propulsion results and is
explained briefly in Section 3.2.6 and [5, Sec 8.7.4.3]. The thrust coefficient eq. (3.13) and the
torque coefficient eq. (3.16) is shown in Figure 4.3. It is observed that the thrust and torque
coefficients have a limited dependence on the Froude number.

4.1.4 Extrapolation of Resistance Test

The results of the resistance test shown in Section 4.1.1 are extrapolated to full scale using the
ITTC extrapolation method [10, Sec. 2.4.1] as explained in Section 3.2.4. The extrapolation
is based on the full scale propeller characteristics of the actual propeller. The total resistance
coefficient eq. (3.20), the smooth total resistance coefficient eq. (3.19), the frictional resistance
coefficient eq. (2.12) and the residuary resistance coefficient eq. (3.1) are presented in Figure 4.4.
From Figures 4.1 and 4.4 it is seen that the frictional resistance is still the largest part of the
total resistance in full scale, but it is a smaller part than in model scale. This means that in
full scale, the frictional resistance is less of the total resistance and the wave making resistance
(residuary resistance) is more of the total resistance.

4.1.5 Extrapolation of Open Water Test

The open water characteristics of the actual propeller in full scale have been given by the shipyard.
Data of the actual propeller can be seen in Table 3.3. However, the open water characteristics of
the stock propeller provided by the shipyard is in model scale. Data of the stock propeller can
be seen in Table 3.4. In order to compare the actual propeller with the stock propeller, the open
water characteristics of the stock propeller presented in Section 4.1.2 need to be extrapolated. The
44 4 Results

×10-3
4

3
Coefficients [-]

0
0.14 0.15 0.16 0.17 0.18 0.19 0.20 0.21
Fn [-]
Figure 4.4: Results of the extrapolated resistance test from the towing tank. Full scale total
resistance coefficient CT s (circle), total resistance coefficient with smooth hull and without bilge
keels CT s,smooth (diamond), frictional resistance coefficient CF s (square) and residuary resistance
coefficient CR (triangle) as a function of Froude number F n. The extrapolation is conducted using
the standard ITTC extrapolation method.

ITTC procedure for extrapolating open water characteristics is described in Section 3.2.5. The
extrapolated open water characteristics of the stock propeller are compared to the provided open
water characteristics of the actual propeller in Figures 4.5 and 4.6.
A difference between the performance characteristics of the actual and stock propellers can be
seen in Figures 4.5 and 4.6. The actual propeller delivers about 1% to 3% more thrust and about
4% to 9% more torque than the stock propeller. This results in a lower open water efficiency of the
actual propeller of 2% to 6% compared to the stock propeller, as seen in Figure 4.6. It is important
to remember that the percentages are within the expected magnitude of the measurement and
extrapolation uncertainties.

4.1.6 Extrapolation of Self-Propulsion Test

The results of the self-propulsion test and the open water test shown in Section 4.1.3 are extrap-
olated to full scale using the ITTC extrapolation method [10, Sec. 2.4.3] as explained in the
Section 3.2.6.
The extrapolation is done using both the actual propeller characteristics and the extrapolated
stock propeller characteristics. The delivered power of the extrapolated self-propulsion results can
be seen in Figure 4.7.
The extrapolation using the stock propeller predicts a 4 − 5% lower delivered power than the
extrapolation using the actual propeller. The difference in delivered power between the stock
propeller and the actual propeller is discussed in Section 5.3.
4.1 Towing Tank Test 45

0.30

0.25
Coefficients [-]

0.20

0.15

0.10
0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60
J [-]
Figure 4.5: Results of the extrapolated open water test from the towing tank. Thrust coeffi-
cient KT os,stock (plus sign) and 10 times the torque coefficient 10KQos,stock (triangle) of the stock
propeller as function of advance ratio J. Thrust coefficient KT os,actual (circle) and 10 times the
torque coefficient K10Qos,actual (diamond) of the actual propeller as function of advance ratio J.
The extrapolation is conducted using the standard ITTC extrapolation method.

0.80

0.70

0.60
ηos [-]

0.50

0.40

0.30
0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60
J [-]
Figure 4.6: Results of the extrapolated open water test from the towing tank. Open water
efficiency ηos,stock of the stock propeller (asterisk) and open water efficiency ηos,actual of the actual
propeller (square) as a function of advance ratio J. The extrapolation is conducted using the
standard ITTC extrapolation method.
46 4 Results

160
Delivered Power [% MCR]
140

120

100

80

60

40

0.14 0.15 0.16 0.17 0.18 0.19 0.20 0.21

Fn [-]
Figure 4.7: Results of the extrapolated self-propulsion test from the towing tank. Delivered
power PDs in full scale using stock propeller (asterisk) and actual propeller (triangle) as a function
of Froude number F n. The extrapolation is conducted using the standard ITTC extrapolation
method.

4.2 Sea Trial

This section presents the results from the speed trial. The results provided by the shipyard are the
measured values corrected for wind and current. The procedures of the sea trial and corrections
are described in Section 3.3.
The weather data shows that only one of the ships has to be corrected for waves using the ISO
15016:2015 standard. The relative difference between the uncorrected and the corrected power due
to waves is between 1% and 3.5%. The corrected speed trial data of the four sister ships can be
seen in Figure 4.8.

4.3 CFD
The following sections present the results of the CFD simulations. Details on the simulations can
be found in Section 3.4. First, the results of the resistance test are presented followed by the
results of the open water. Finally, the results of the self-propulsion results are presented. The used
computational power for calculating one speed for the different test types can be seen in Table 4.9.

4.3.1 CFD Resistance Test

This section presents results of the model scale and full scale resistance setup. First, the model
scale results are presented, and then the full scale results are presented. The model scale and full
scale resistance CFD setups are described in Section 3.4.2.
4.3 CFD 47

Delivered Power [% MCR]

100

90

80

70

60

50

0.14 0.15 0.16 0.17 0.18 0.19 0.2 0.21

Fn [-]
Figure 4.8: Results of the speed trial for all four sister ships. Delivered power as a function of
Froude number F n. Conducted and corrected using the ISO 15016:2015 procedure [4].

Table 4.9: Used computational power for a single simulation. All simulations are performed on
16-core nodes (Either Xeon E5-2650 or Xeon E5-2670 running at 2.60 GHz)

Simulation Computational Cost

Resistance 450 CPU hours
Open water 32 CPU hours
Self-propulsion 550 CPU hours

Model Scale Resistance

Before trusting the results of the CFD resistance simulations, convergence tests and validation with
model scale resistance experiments are essential as explained in Section 3.4.1. The following sec-
tions present validation and convergence studies of the boundary layer mesh, spacial discretization,
temporal discretization and finally a comparison with towing tank test results.

Boundary Layer Mesh The boundary layer mesh is the part of the mesh situated closest to
the hull (See Figure 2.3). The frictional resistance is dependent on the wall y + value of the hull,
which is controlled by the thickness of the inner most boundary layer as explained in Section 2.3.5.
A wall y + study is conducted in order to determine the boundary layer thickness to be used in the
subsequent simulations.
There are multiple ways of defining a representative global y + value for a simulation. In this
study, the global wall y + value is assumed to be the average wall y + of all cells on the wetted
surface of the hull.
It has been decided to perform the y + study at the lowest speed tested in the towing tank,
which corresponds to a Froude number equal to 0.1480. The reason for performing the study at
48 4 Results

×10-3
3.4

CFm [-] 3.2

2.8

2.6
0 20 40 60 80 100
+
Average wall y [-]
Figure 4.10: Results of the wall y + study in model scale, showing the frictional resistance
coefficient CF m from CFD (circle) as a function of average wall y + value, together with the ITTC
Mode-Ship Correlation Line eq. (2.12) (dash line). Simulations are performed at a Froude number
F n equal to 0.1480, using 2M cells.

Table 4.11: Boundary layer settings of the model scale CFD resistance setup, used to achieve
wall y + values in the range of 35 to 50.

Number of boundary layers 7

Total boundary layer thickness 8.4 × 10−3 Lpp
boundary layer stretching 1.4

the lowest speed is that the frictional resistance is the largest part of the total resistance at the
lowest speed as described in Section 2.1.
The wall y + study is conducted by running a series of simulations with different values of
boundary layer thickness while keeping all other mesh settings constant. The frictional resistance
coefficients of the series of simulations can be seen in Figure 4.10, along with the ITTC Model-Ship
Correlation Line eq. (2.12).
It is seen in Figure 4.10 that the frictional resistance coefficient converges at a wall y + value
of approximately 35 to 50, which is expected as the inner cell is in the log layer as described in
Section 2.3.5. This range of wall y + values is achieved by using the boundary layer settings shown
in Table 4.11 for all speed.

Spatial Discretization A mesh convergence study is conducted in order to determine the mesh
settings of the volume mesh and the six refinement zones. The refinement zones are described in
Section 3.4.2.
The mesh convergence study is conducted by running a series of simulations with different mesh
settings while all other settings of the simulations are kept constant. Three of these simulation
series are conducted, each at a different Froude number. The three Froude numbers are chosen
4.3 CFD 49

1
Discrepancy in C Tm [%]
0

-1

-2

-3
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
6
Number of cells [×10 ]
Figure 4.12: Results of the volume mesh convergence study for the model scale resistance setup.
Total resistance coefficient CT m of the CFD simulations relative to the results of the towing tank
resistance test at three speeds as a function of the number of cells. Froude number F n equal to
0.1634 (circle), Froude number F n equal to 0.1736 (diamond) and Froude number F n equal to
0.1940 (square).

from the range of Froude numbers used in the towing tank tests. The mesh convergence study is
performed on the hull without the rudder and without the stator fins as seen in Table 3.22.
The results of the three simulations series show that mesh convergence is obtained at the
medium and the highest speeds, but not at the lowest speed. The lack of mesh convergence at
the lowest speed is investigated by inspecting the free surface height. It is concluded that the lack
of mesh convergence at the lowest speed is caused by insufficient mesh resolution of the Kelvin
wave zone in the longitudinal and transverse directions. The Kelvin waves are briefly described in
Section 2.1. Lower ship speeds generate divergent waves with shorter wave lengths. Therefore, the
mesh of the Kelvin waves should be finer in order to properly resolve the waves at all ship speeds.
The settings of the free surface refinement zones are modified with finer mesh in the longitudinal
and transverse directions. The results of the mesh convergence study with the modified settings
can be seen in Figure 4.12.
The resistance coefficients of the mesh convergence study are slightly lower than those of the
towing tank results because the rudder and stator fins are not included in the simulations. The
conclusions of the study are still valid as the effect of the rudder and stator fins on the convergence
is assumed to be negligible. Furthermore, it is important to remember that the discrepancies are
similar to the uncertainties of the measurements from the towing tank [37].
50 4 Results

Table 4.13: Cell sizes of refinement zones in the model scale CFD resistance setup. Cell sizes are
made non-dimensional using eq. (4.1).

Name of RZ Cell Size l∗ [-]

Far Free Surface RZ 96
Middle Free Surface RZ 192
Near Free Surface RZ 384
Bulbous Bow RZ 77
Transom RZ 128
Vertical Free Surface RZ 1281

From Figure 4.12 it is seen that the results are converged at 2M cells, but with lower discrepancy
at 3.7M cells. It is decided not to optimize the volume mesh further because most of the results
are within the uncertainty of the towing tank measurements data [37]. The cell size ∆l is non-
dimensionalized using the ship length:

Lpp
l∗ = (4.1)
∆l
where l∗ is the non-dimensional cell size. The non-dimensional cell size can be interpreted as the
number of cells needed to obtain one ship length. Larger values of l∗ indicate smaller cell sizes.
The mesh with 3.7M cells is used for all subsequent resistance simulations in model scale, and
the settings of the refinement zones can be seen in Table 4.13.

Temporal Discretization The CFD setup is simulated using an implicit unsteady solver as
described in Section 3.4.2. The size of the time step is determined by conducting a time step study.
The time step is non-dimensionalized using the ship speed and ship length:

Vdesign
t∗ = ∆t (4.2)
LP P
where t∗ is the non-dimensional time step. A physical interpretation of this non-dimensional
time step is the number of ship lengths the ship translates per time step.
The time step study is conducted by running a series of simulations with different time step
sizes while all other settings are kept constant. The simulation series are conducted using the 3.7M
cell mesh at a Froude number equal to 0.1736. The study is performed on the hull, without the
rudder and without the stator fins as seen in Table 3.22. The total resistance coefficient of the
CFD simulations are compared to the resistance test data from the towing tank. The results of
the time step study can be seen in Figure 4.14.
The resistance coefficients of the time step study are slightly lower than those of the towing
tank results as the rudder and stator fins are not present in the simulation series. The conclusions
of the study are still valid as the effect of the rudder on the convergence is assumed to be negligible.
Figure 4.14 shows that a smaller time step gives a smaller discrepancy between the resistances
of the CFD and the towing tank. As in the mesh convergence study in Section 4.3.1, it important
to keep in mind that the discrepancies are in the same order of magnitude as the uncertainty of the
4.3 CFD 51

-1.0
Discrepancy in C Tm [%]

-1.2

-1.4

-1.6
0.000 0.002 0.004 0.006 0.008 0.010 0.012
*
t [-]
Figure 4.14: Results of the time step study of the model scale CFD resistance setup. Total resis-
tance coefficient CT m of the CFD simulations relative to the results of the towing tank resistance
test as a function of time step size t∗ . The simulations are conducted at a Froude number F n
equal to 0.1736 and with 3.7M cells.

towing tank test [37]. It can be seen in Figure 4.14 that a reduction in time step from the largest
to the smallest time step size only reduces the discrepancy by approximately half a percent point
while requiring over six times the computational time. In this case, it is a high computing cost
compared to the reduction of the discretization error. A time step size of 6.6 × 10−3 is determined
to be the right balance between reduced discretization error and computational cost.
Other time step study simulations have also been conducted. The results, which are not in-
cluded in the thesis, show that when the number of cells is increased, the impact on the discrepancy
of changing the time step size is reduced. This means that increasing the number of cells reduces
the requirement of the time step size in order to obtain similar discrepancies.
For the full scale resistance CFD simulations, the increased cost of a small time step can be
reduced by varying the time step size during the simulation. This method is described in detail in
Section 4.3.1.

Comparison A series of simulations are conducted using the final model scale resistance setup.
The series of simulations are conducted at the same speeds as the resistance test of the towing
tank. The setup parameters of the series of simulations can be seen in Tables 4.11 and 4.13. The
series of simulations are performed on the hull with the rudder, but without the stator fins as seen
in Table 3.22.
The discrepancy in the total resistance coefficient between CFD simulation and towing tank
test data at each speed can be seen in Figure 4.15.
The total resistance coefficient estimated using CFD is very similar to the towing tank results.
For the speed range of Fn 0.16 to 0.19, which is in the range of the sea trial data, the discrepancies
are within 1%. The discrepancy is within the uncertainty of the towing tank test. Furthermore, it
is important to state that the towing tank tests are conducted with fins and the CFD calculations
52 4 Results

0.5

Discrepancy in C Tm [%] 0.0

-0.5

-1.0

-1.5

-2.0
0.14 0.15 0.16 0.17 0.18 0.19 0.20
Fn [-]
Figure 4.15: Results of the model scale resistance setup. Total resistance coefficient CT m of the
CFD simulations relative to the results of the towing tank resistance test as a function of Froude
number F n. The simulation series are conducted using a 3.7M cell mesh with a time step size t∗
equal to 6.6 × 10−3 .

are conducted without. Including the fins in the CFD calculations would increase the resistance
slightly.
As additional validation of the resistance setup, the setup has been blind tested on an additional
ship. The data were provided by the DNV GL office in Hamburg. The only data provided for the
simulations were the ship geometry, displacement and draft. After a short spatial and temporal
convergence study, the final simulations were performed and the results were sent to Hamburg.
The results are not included in this thesis. After the results were received, the Hamburg office
provided the towing tank test results. The results showed a relative difference in total resistance
below 1% between CFD and towing tank test. This is the same discrepancy as in the case studied
in this thesis.
With the satisfying convergence studies of both the temporal and spacial discretization and
the very small discrepancies from model tank test in the studied case and in the blind testing, the
model scale resistance setup is considered trustworthy.

Full Scale Resistance

The setup validated in the previous sections is modified from model to full scale. The following
sections present validation and convergence studies of the spatial and temporal discretization,
followed by a comparison with the extrapolated resistance results from the towing tank.

Boundary Layer Mesh A wall y + study in model scale has been carried out in Section 4.3.1.
It has been decided not to study the influence of the y + value on the frictional resistance in full
scale, but to use the results of the model scale study. It was concluded that wall y + values in the
range of 35 to 50 gives a reasonable estimation of the frictional resistance. Due to the increases
flow velocity in full scale, the number and thickness of the boundary layers have to be changed, in
4.3 CFD 53

Table 4.16: Boundary layer settings of the full scale CFD resistance setup, used to achieve wall
y + values in the range of 35 to 50.

Number of boundary layers 15

Total boundary layer thickness 7.9 × 10−4 LP P
boundary layer stretching 1.4

4
Discrepancy in C Ts,smooth [%]

-2

-4
0 1 2 3 4 5 6 7 8
6
Number of cells [×10 ]
Figure 4.17: Results of the volume mesh convergence study for the full scale resistance setup.
Total resistance coefficient with smooth hull and without bilge keels CT s,smooth of the CFD simula-
tions relative to the extrapolated resistance results of the towing tank as a function of the number
of cells. The simulations are conducted at a Froude number F n equal to 0.1736.

order to maintain the average wall y + value. Model scale and full scale boundary layer settings
can be seen in Table 4.11 and Table 4.16 respectively.

Spatial Discretization The volume mesh and refinement zones of the full scale CFD setup are
similar to those of the model scale setup, as seen in Section 4.3.1. A mesh convergence study of the
full scale setup is conducted by running a series of simulation with different mesh settings while
all other settings are kept constant. The simulation series is conducted on the hull without the
rudder and without the stator fins (see Table 3.23) and at a Froude number equal to 0.1726. The
results of the mesh convergence study can be seen in Figure 4.17.
The resistance of the rudder and stator fins are included in the extrapolated resistance data,
but not in the CFD setup. This could be one reason for the discrepancies. It is assumed that
the exclusion of the rudder does not influence the convergence of the study, and therefore the
conclusions are still valid.
It is found that the settings of the vertical free surface, bulbous bow and transom refinement
zones directly scaled from the model scale simulations are sufficient for properly resolving the full
scale free surface. It is found that the model scale settings of the far, middle and near refinement
zones are too coarse and therefore smaller cell sizes in these three zones are used. It has been
decided to use the 4.7M cell mesh for the final full scale simulations. The final cell sizes of the
54 4 Results

Table 4.18: Cell sizes of refinement zones in the full scale CFD resistance setup. Cell sizes are
made non-dimensional using eq. (4.1).

Name of RZ Cell size l∗ [-]

Far Free Surface RZ 137
Middle Free Surface RZ 275
Near Free Surface RZ 549
Bulbous Bow RZ 69
Transom RZ 92
Vertical Free Surface RZ 915

First Stage Second Stage Third Stage

Small t* Large t* Medium t*

Figure 4.19: The three stages of the time step study of the full scale resistance setup.

refinement zones in model and full scale can be seen in Table 4.13 and Table 4.18 respectively. The
cell size is made non-dimensional using eq. (4.1).

Temporal Discretization Similarly to the model scale setup, the full scale setup is simulated
using an implicit unsteady solver. A time step study has been conducted to determine the time
step size.
The full scale simulations have shown to be quite sensitive to the time step size during the
first couple of time steps. If the time step size is too large, the simulation experiences numerical
instability, which causes the simulation to break down. The time step size needed to successfully
simulate the first few time steps is very small. Using this time step size throughout the simulation
would require inconveniently large computational effort. Therefore, it has been decided to run
the simulation in three stages, each using a different time step size: Firstly a small time step for
initializing the flow (first stage), then a large time step for fast convergence (second stage), and
lastly a medium time step for minimizing the discretization error (third stage). This approach
balances the required computational cost, the numerical instability, and the discretization error.
The three stages are illustrated in Figure 4.19.
The time step size of the first stage is chosen purely to avoid numerical instabilities. The time
step sizes of the second and third stages are studied in the time step study. The study is conducted
on the hull without the rudder and without the stator fins (see Table 3.23). The results of the
study are compared to the results of the extrapolated towing tank test. The results of the study
can be seen in Figure 4.20. The time step size is made non-dimensional using eq. (4.2).
The resistance of the rudder and the stator fins are included in the extrapolated towing tank
results, but not in the CFD setup. This gives an offset to the discrepancies. It is assumed that the
exclusion of the rudder and stator fins does not influence the convergence behaviour of the study,
and therefore the conclusions are still valid.
Simulations with a non-dimensional time step size larger than 0.011 have a very fluctuating
signal which makes it difficult to compute a reliable time average. The same is true for time step
sizes smaller than 0.0022. This gives a range of stable time step sizes between 0.0022 and 0.011.
4.3 CFD 55

Discrepancy in C Ts,smooth [%] -2.00

-2.50

-3.00

-3.50

-4.00
0.000 0.005 0.010 0.015 0.020 0.025 0.030 0.035 0.040
*
t [-]
Figure 4.20: Results of the time step study of the full scale CFD resistance setup. Total resistance
coefficient with smooth hull and without bilge keels CT s,smooth of the CFD simulations relative to
the extrapolated resistance results of the towing tank as a function of time step size t∗ . The
simulations are conducted at Froude number F n equal to 0.1712 and with 4.7M cells.

Table 4.21: Time step sizes of the three stages used in the full scale CFD resistance setup.

Stage Time step size t∗ [-]

First 7.6 × 10−5
Second 1.1 × 10−2
Third 2.2 × 10−3

The largest time step size in the stable range is chosen for the second stage, and the smallest time
step size in the range is chosen for the third stage. The time step size of the three stages in the
final setup can be seen in Table 4.21.

Comparison The full scale CFD setup is simulated at all speeds tested in the towing tank and
compared to the extrapolated resistance results of the towing tank. The simulations are performed
using the hull and the rudder, but not the stator fins (see Table 3.23). The setting of the final
setup can be seen in Tables 4.16, 4.18 and 4.21. The discrepancy in the total resistance coefficient
between the CFD simulations and the extrapolated resistance results of the towing tank can be
seen in Figure 4.22.
From Figure 4.22 it is seen that the total resistance coefficients of the CFD setup is very similar
to the extrapolated towing tank results. For the speeds in the range of the sea trial from a Froude
number of 0.16 to 0.19, the discrepancies are below 1.5 %, which is satisfying since it is about the
same as the uncertainty of towing tank measurements. The discrepancies are also similar to those
of the model scale setup, as seen in Figure 4.15. Again, the underestimation could be caused by
the fact that the fins are not included in the CFD simulations.
The setup has shown independence of spatial and temporal discretization and small discrepan-
cies compared to the extrapolated resistance from the towing tank.
56 4 Results

Discrepancy in C Ts,smooth [%] 0

-1

-2

-3
0.14 0.15 0.16 0.17 0.18 0.19 0.2
Fn [-]
Figure 4.22: Results of the full scale resistance setup. Total resistance coefficient with smooth
hull and without bilge keels CT s,smooth of the CFD simulations relative to the extrapolated results
of the towing tank as a function of Froude number F n. The simulation series are conducted using
a 4.7M cell mesh, with the time step sizes shown in table 4.21.

Table 4.24: Boundary layer mesh settings of the CFD open water setup, used to achieve wall y +
values below 1.

Propeller and Hub

Number of boundary layers 11
Total boundary layer thickness 2.7 × 10−3 D
Near wall boundary layer thickness 3.7 × 10−5 D

4.3.2 CFD Open Water Test

As described in Section 3.4.1, a validation procedure is required in order to trust the results of the
CFD simulations. This section presents the results of the validation for the open water simulations
in model scale.

Boundary Layer Mesh

A number of boundary layers (see Section 2.3.5) are used on the surface of the propeller hub and
blades as well as at the boundary between the rotating and stationary domains.
It is decided to fully resolve the boundary layer of the propeller hub and blades. To fully resolve
the boundary layer, the simulations are performed with wall +y values of the propeller blade and
hub below 1. A single boundary layer is used on each side of the interface between the static and
rotating domain in order to make a good alignment of the two domains. The obtained y + values
on the propeller in open water can be seen in Figure 4.23, and the boundary layer settings can be
seen in Table 4.24.
4.3 CFD 57

Figure 4.23: Wall y+ value on the pressure side (left) and suction side (right) of the propeller
in the open water setup. Advance ratio is equal to 0.6.

4
Discrepancy in K Tom [%]

-2

-4

-6
3 4 5 6 7 8 9 10 11 12
6
Number of cells [×10 ]
Figure 4.25: Results of the volume mesh convergence study for the open water setup. Thrust
coefficient KT om of the CFD simulations relative to the results of the towing tank open water test
as a function of the number of cells. Advance ratio J equal to 0.30 (circle), advance ratio J equal
to 0.45 (diamond) and advance ratio J equal to 0.60 (square).

Spacial Discretization
A mesh convergence study has been conducted in order to determine the volume mesh settings
and the settings of the refinement zones. The mesh convergence study is conducted by running a
series of simulations with different volume mesh settings while all other settings are kept constant.
Three simulation series are conducted at J equal to 0.30, 0.45 and 0.60 respectively. The propeller
characteristics of the CFD simulations are compared with the open water test data from the towing
tank. The discrepancies of the propeller characteristics are shown in Figures 4.25 to 4.27.
58 4 Results

5
[%]
Qom 4
Discrepancy in K

0
3 4 5 6 7 8 9 10 11 12
6
Number of cells [×10 ]
Figure 4.26: Results of the volume mesh convergence study for the open water setup. Torque
coefficient KQom of the CFD simulations relative to the results of the towing tank open water test
as a function of the number of cells. Advance ratio J equal to 0.30 (circle), advance ratio J equal
to 0.45 (diamond) and advance ratio J equal to 0.60 (square).

0
Discrepancy in ηom [%]

-2

-4

-6

-8
3 4 5 6 7 8 9 10 11 12
6
Number of cells [×10 ]
Figure 4.27: Results of the volume mesh convergence study for the open water setup. Open
water efficiency ηom of the CFD simulations relative to the results of the towing tank open water
test as a function of the number of cells. Advance ratio J equal to 0.30 (circle), advance ratio J
equal to 0.45 (diamond) and advance ratio J equal to 0.60 (square).
4.3 CFD 59

Table 4.28: Cell sizes of the refinement zones in the CFD open water setup.

Name of RZ Cell size

Far Propeller RZ 4.26 × 10−2 D
Near Propeller RZ 2.13 × 10−2 D

2
Discrepancy [%]

-2

-4

-6

-8
0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60
J [-]
Figure 4.29: Results of the CFD open water test. Thrust coefficient KT om (circle), torque
coefficient KQom (diamond) and efficiency ηom (square) of the stock propeller in open water relative
to the results of the towing tank open water test as a function of advance ratio J.

It can be seen in Figures 4.25 to 4.27 that the solution converges for all advance ratios. It can
therefore be concluded that the propeller characteristics are independent of the mesh size for sizes
larger than approximately 7.5M cells. In all subsequent open water simulations, the mesh with
7.5M cells will be used.
Furthermore, it can be seen in the Figures 4.25 to 4.27 that the thrust coefficient is overes-
timated by 1 − 2% for J equal to 0.3 and 0.45, and underestimated by 3% for J equal to 0.6,
compared to the towing tank test. The torque is overestimated by 3 − 5% for all advance ratios.
The open water efficiency is a combination of the thrust and torque coefficients and is therefore
underestimated by under 3% for J equal to 0.3 and 0.45 and 7% for J equal to 0.6.
The final mesh setting of the open water setup can be seen in Table 4.28.

Comparison
The propeller performance of the stock propeller is calculated with the mesh settings with ap-
proximately 7.5M cells. The CFD results compared with the towing tank results can be seen in
Figure 4.29.
It can be seen in Figure 4.29 that the discrepancies are reasonable. Generally the torque is
overestimated by 3 − 4%. The thrust is overestimated by approximately 2% at the lower advance
ratios and underestimated by approximately 2 − 4% at the higher advance ratios. This results in
an underestimation of the efficiency of up to 7%. There are several possible explanations for the
60 4 Results

Figure 4.30: Flow regimes on the suction side of a propeller blade in model scale, from [38, 39].

discrepancy between the CFD results and the towing tank results seen in Figure 4.29. The propeller
geometry used in the CFD setup is slightly different from the stock propeller geometry used in
the towing tank (see Section 3.1.2). This difference could cause discrepancies in the propeller
characteristics. Another explanation for the discrepancy is the turbulence modelling used in the
CFD setup. The near wall flow regime on the propeller in model scale is laminar at first and
changes to turbulent somewhere on the blade as illustrated in Figure 4.30.
The CFD simulation in this report enforces turbulent flow in the domain both for model scale
and full scale since the k − ϵ model assumes that the flow is fully turbulent. Therefore, the CFD
setup is not able to model the laminar flow on the propeller. The full scale flow on the propeller is
fully turbulent so this possible discrepancy will only occur in model scale. The use of a transition
model in the model scale CFD setup has been discussed, but ultimately it has been decided not
use such a model. The use of ordinary turbulence models [40] and transition-sensitive turbulence
model in analysis of marine propeller performance [41] have been studied in the literature and a
sensitive in the choice of turbulence model is found. Not implementing a transitional model saves
a significant amount of time and it only causes modelling errors in model scale.

4.3.3 CFD Self-Propulsion Test

This section presents the results of the self-propulsion simulations in model scale and full scale.
The two setups are presented in Section 3.4.4.

Model Scale Self-Propulsion

The self-propulsion setup is the combination of the full scale resistance setup and the open water
setup, as described in Section 3.4.4. Before the results from the self-propulsion simulations can be
trusted, convergence studies of the temporal and spacial discretization must performed. After the
convergence studies, the results of the self-propulsion simulation are compared to the results of
4.3 CFD 61

Figure 4.31: Wall y + value on the pressure side (left) and suction side (right) of the propeller in
self-propulsion simulation at Froude number F n equal to 0.1736.

the self-propulsion test from the towing tank in order to validate the self-propulsion setup. This
section presents the results of the convergence studies and the comparison between the CFD and
the towing tank results.
Only the mean value of the signal has been investigated in the convergence studies. The
amplitude and the nature of the signal have not been studied. The time-dependent details of the
thrust and torque signal are relevant in e.g. cavitation studies, noise and vibration studies and
fatigue fracture studies, but is out of the scope of this thesis. For this thesis, only the time averaged
thrust and torque signals are important because only the time-averaged delivered power is needed
for the comparison with the towing tank results.

Boundary Layer Mesh The self-propulsion setup uses the same boundary layer mesh setting
as the model scale resistance setup on the hull. The settings can be seen in Table 4.11.
The boundary layer on the propeller in the self-propulsions setup is fully resolved, exactly as
it is for the open water setup. Therefore, the simulations are performed with wall y + values below
1, using the same settings as the open water setup. The setting of the boundary layer mesh on
the propeller can be seen in Table 4.24. The wall y + on the propeller in a self-propulsion test in
model scale can be seen in Figure 4.31.
It is seen from Figure 4.31 that the wall y + values are lower than they are in the open water
simulations, as shown in Figure 4.23. This is expected since the velocities are smaller due to the
wake of the ship hull. Furthermore, it is seen that most of the cells have a wall y + below 1 as
required for fully resolving the boundary layer.

Temporal Discretization A time step convergence study has been conducted in order to deter-
mine the time step size of the transient solver. The study has been conducted by running a series
of simulations with different time step sizes while all other settings of the setup are kept constant.
The time step sizes are shown in non-dimensional values using eq. (3.42). The simulation series
are conducted using a mesh with 10.7M cells in the rotating region and 1.2M cells in the static
region at a Froude number equal to 0.1736.
62 4 Results

0.188

0.187
K Tm [-]

0.186

0.185

0.184
0 1 2 3 4 5 6 7
∆ θ [°]
Figure 4.32: Results of the time step study for the self-propulsion setup. Thrust coefficient KT m
of the CFD simulations relative to the results of the towing tank self-propulsion test as a function
of time step size ∆theta. The simulations are conducted at a Froude number F n equal to 0.1736.

0.029

0.028
K Qm [-]

0.027

0.026

0.025
0 1 2 3 4 5 6 7
∆ θ [°]
Figure 4.33: Results of the time step study for the self-propulsion setup. Torque coefficient KQm
of the CFD simulations relative to the results of the towing tank self-propulsion test as a function
of time step size ∆theta. The simulations are conducted at a Froude number F n equal to 0.1736.

The study has been conducted on the hull with the rudder, but without the stator fins, as
shown in Table 3.37. The simulations has been conducted with the same propeller rotation rate.
From the simulation series, the propeller thrust and torque coefficients are obtained. The results
of the time step study can be seen in Figures 4.32 and 4.33.
In can be seen in Figures 4.32 and 4.33 that the solution in not completely converging in the
temporal discretization. A possible explanation for the lack of convergence could be that the vessel
has a large wake fraction resulting in separation and/or that transient flow structures are present.
4.3 CFD 63

Figure 4.34: Nominal wake field, normalized with the free stream velocity. Result of the model
scale resistance setup including stator fins. Black ring is the propeller diameter. The simulation
is conducted at a Froude number F n equal to 0.1736.

Separation and transient flow structures could be dependent on the temporal discretization and
thereby influence the convergence. The nominal wake field can be seen in Figure 4.34.
The time step study simulations are computationally heavy especially at ∆θ equal to 0.5 and
1.0. Therefore, is was not possible, with the available computational power, to use time steps
smaller than 0.5∆θ. Even through the solution has not completely converged, the solution is not
considered to be diverging. It has been decided to perform all further self-propulsion simulations
in model scale with ∆θ equal to 2. This decision is a trade-off between temporal discretization
error and usage of computational resources.

Spatial Discretization To ensure that the solution is independent of the mesh discretization,
a mesh convergence study has been performed. The mesh around the ship has been studied in
Section 4.3.1 and the mesh around the propeller in Section 4.3.2. Only settings of the near propeller
refinement zone (see fig. 3.35) are changed in this mesh convergence study as it is assumed that the
mesh of the static region is sufficiently discretized because it has been validated in Section 4.3.1.
The meshes of the study can be seen in Table 4.35. The simulations have been conducted with
the same propeller rotation rate as the towing tank self-propulsion test, and at the same Froude
number equal to 0.1736. The simulations have been conducted on the hull with the rudder, but
without the stator fins, as shown in Table 3.37.
The results of the mesh convergence study show that the difference in thrust and torque between
all five meshes in Table 4.35 are smaller than the uncertainty of the signals. This is a strong
indication that the solution is independent of the mesh discretization. Based on the results of the
mesh convergence study, it has been decided to use Mesh 1, which is the coarsest mesh, for all
further simulations. The mesh settings of Mesh 1 can be seen in Table 4.36.
64 4 Results

Table 4.35: Studied meshes of the mesh convergence study, for the model scale CFD self-
propulsion setup.

Mesh # Number of cells

Mesh 1 6.2M
Mesh 2 6.6M
Mesh 3 7.6M
Mesh 4 9.1M
Mesh 5 21.2M

Table 4.36: Cell sizes of refinement zones in the model scale CFD self-propulsion setup. Cell
sizes are made non-dimensional using eq. (4.1).

Name of RZ Cell size l∗ [-]

Bulbous Bow RZ 69
Transom RZ 92
Name of RZ Cell size [-]
Far Propeller RZ 4.26 × 10−2 D
Near Propeller RZ 2.13 × 10−2 D

Figure 4.37: Separation downstream of the stator fins. From model scale self-propulsion simu-
lation at a Froude number equal to 0.1736. The colour indicates the velocity magnitude at r/R
equal to 0.5. Seen from starboard side (left), and from port side (right).

Stator Fins A set of self-propulsion simulations have been conducted both with and without
the stator fins in order to investigate the performance of the fins and their influence on the flow.
The results of the study show that the stator fins significantly improve the performance of the
propeller even though there is significant flow separation on the two fins on starboard side. The
flow around the stator fins can be seen in Figure 4.37.
The study has also shown that the results of the simulations with the stator fins has resulted in
smaller discrepancies relative to the towing tank self-propulsion tests, compared to the simulation
without the stator fins. This is reasonable since the stator fins have also been included in the
towing tank test. It was therefore decided to include the stator fins in all subsequent self-propulsion
4.3 CFD 65

Discrepancy in nm [%]
6

2
0.14 0.15 0.16 0.17 0.18 0.19 0.2 0.21
Fn [-]
Figure 4.38: Results of the model scale self-propulsion setup. Propeller rotation rate nm of the
CFD simulations relative to the results of the towing tank self-propulsion test as a function of
Froude number F n.

simulations. It would have been better to conduct all previous simulations, both resistance and self-
propulsion, with the stator fins, but unfortunately that has not been possible due to time limitations
of the project. It is important to remember that the resistance simulations conducted prior to the
self-propulsion test are mainly meant as validation steps to ensure trustworthy results in the self-
propulsion stimulations. When the stator fins are included in the self-propulsion simulations, the
influence of not having included the stator fins before is very limited.

Comparison This section presents the results from the self-propulsion simulations compared to
the results from the towing tank self-propulsion test. The self-propulsion simulations are conducted
at the same Froude numbers as the self-propulsion test from towing towing tank. The series of
simulations are conducted both with and without the stator fins, as discussed previously.
As described in Section 3.4.4, a solver has been used in order to obtain the correct tow force
by changing the rate of rotation of the propeller. As described in Section 3.4.4, the simulations
have been performed with a variable propeller rotation rate in order to balance the forces and
obtain the same towing force as used in the towing tank tests. The obtained towing forces of the
simulations are within 0.4% of the towing force of the towing tank test (calculated using eq. (3.9)).
The propeller rotation rate for the balanced solution can be seen in Figure 4.38.
It is seen from Figure 4.38 that the rate of rotation is overestimated in the simulations. Simu-
lations have also been conducted without balancing the forces, but using a fixed propeller rotation
rate instead. The results are not shown in this thesis. These simulations show an underestimation
of the thrust and an overestimation of the resistance compared to the towing tank results. The
over- and underestimation both contribute to an increase in the propeller rotation rate when bal-
ancing the forces because the balancing of the forces has been done according to eq. (3.41). That
explains why the propeller rotation rate in Figure 4.38 is consistently overestimated.
66 4 Results

-3

Discrepancy in K Tm [%]
-4

-5

-6

-7

-8
0.14 0.15 0.16 0.17 0.18 0.19 0.2 0.21
Fn [-]
Figure 4.39: Results of the model scale self-propulsion setup. Thrust coefficient KT m of the CFD
simulations relative to the results of the towing tank self-propulsion test as a function of Froude
number F n.

The thrust coefficient eq. (3.13) of the simulations can be seen in Figure 4.39. The thrust is
overestimated, but the thrust coefficient is underestimated. This is the case because the propeller
rotation rate is overestimated, as explained. It can be seen that the magnitude of the thrust
discrepancy in Figure 4.39 is related to the discrepancy of the propeller rotation rate in Figure 4.38.
For the low Froude numbers, the discrepancies are approximately 6 − 7% and for the high Froude
numbers the discrepancies are approximately 2 − 4%.
The obtained torque coefficients can be seen in Figure 4.40. It is seen that the torque coefficient
is overestimated by approximately 6 − 7% compared to the towing tank results. An overestimation
of the torque coefficient is also seen in the open water simulations in Figure 4.29. In the open
water simulations, the overestimation of the torque coefficient is 3 − 4%.
Based on the convergence studies and comparisons to the towing tank tests, the results of the
model scale self-propulsion setup are reasonable.

Full Scale Self-Propulsion

The full scale setup is based on the model scale setup validated in the previous section. The
modification from the model scale setup to the full scale setup is described in Section 3.4.4.
In this section, first the changes to the boundary layer are discussed followed by a brief conver-
gence study in temporal and spatial descretization. Then the wake field and flows around the stator
fins are presented and discussed. Finally, the results from self-propulsion results are presented.
For the full scale simulations, only calculations in the Froude number range of 0.175 to 0.195
will be performed. This is the range where the speed trial are conducted and where the model
scale setup shows the smallest discrepancies.
4.3 CFD 67

7
Discrepancy in K Qm [%]
6.5

5.5

5
0.14 0.15 0.16 0.17 0.18 0.19 0.2 0.21
Fn [-]
Figure 4.40: Results of the model scale self-propulsion setup. Torque coefficient KQm of the
CFD simulations relative to the results of the towing tank self-propulsion test as a function of
Froude number F n.

Boundary Layer Mesh The wall y + value on the propeller in the model scale self-propulsion
setup is below 1 as required in order to fully resolve the boundary layer. In the full scale self-
propulsion simulations, the Reynolds number is much larger and the boundary layer is much
thinner. It has not been possible to obtain numerically stable full scale simulations using a bound-
ary layer mesh with wall y + values below 1 in full scale. Therefore, it has been decided to use
wall y + values above 30 instead. This way the boundary layer is not resolved, but instead wall
functions are used to calculate the effects in the boundary layer.
To study the differences of using wall y + values below 1 and using wall y + values above
30, three model scale self-propulsion simulations have been conducted using wall y + value above
30. The results of these simulations are not included in this thesis, but they are compared to the
earlier model scale self-propulsion simulations conducted with wall y + values below 1. The relative
difference in thrust and torque between the three simulations with wall y + values above 30, and
the one with wall y + value below 1 is no larger than 0.5%. This shows that the use of either
values below 1 or above 30 has little influence on the thrust and torque results in model scale. It
is assumed that the same is true for the full scale simulations.

Temporal and Spatial Discretization The following convergence studies are performed to
ensure convergence in the temporal and spatial discretization as described is Section 3.4.1. The
studies will not be as comprehensive is the case of the resistance, open water and model scale
self-propulsion simulations.
The reason for this is that it is expected that modelling errors contribute more to the uncertainty
than to the discretization errors. The two most significant contributions to the modelling errors
are the use of the stock propeller instead of the actual propeller and that the roughness and air
resistance are based on the ITTC procedure and not modelled in the CFD simulations. Therefore,
it does not make sense to intensively study the convergence and ensure very small discretization
68 4 Results

Table 4.41: Studied time step sizes of the time step convergence study for the full scale self-
propulsion setup.

Time Step # Time Step Size

Time Step 1 1∆θ
Time Step 2 2∆θ
Time Step 3 3∆θ
Time Step 4 4∆θ

Table 4.42: Studied meshes of the mesh convergence study, for the full scale self-propulsion setup.

Mesh # Number of Cells

Mesh 1 9.6M
Mesh 2 9.7M
Mesh 3 9.8M
Mesh 4 12.0M

errors when the modelling errors is much larger. However, it is still important to ensure that the
setup does not have a high dependence on the spatial and temporal discretization.
For the full scale self-propulsion setup, a time step study has been conducted. The time step
study is a slightly smaller version of the study conducted in model scale. The study consists of four
time step sizes which can be seen in Table 4.41. The time step sizes are shown in non-dimensional
values using eq. (3.42). The studies have been conducted at a Froude number equal to 0.1736,
using a mesh of 9.8M cells. The studies have been conducted on the hull with both the rudder
and stator fins attached, as seen in Table 3.38.
The simulations have been conducted with a constant rate of rotation equal to the rate of
rotation from the extrapolated towing tank tests from eq. (3.31). The relative difference in thrust
and torque from these four simulations is smaller than 0.5%. This indicates that the solution is
almost independent of the temporal discretization and that the discretization error is small. It has
been decided to use a time step size of 2∆θ, which is the same as in the model scale self-propulsion
simulations.
As described in Section 3.4.4, the mesh is scaled from the model scale to full scale using the
scaling factor. A mesh convergence study has been conducted for the full scale self-propulsion
setup. The study is conducted using four different meshes. The meshes of the study can be seen
in Table 4.42. The study has been conducted at a Froude number equal to 0.1736. The study has
been conducted on the hull with both the rudder and stator fins attached, as seen in Table 3.38.
The relative difference in thrust and torque of the simulations in the study are smaller than 1.5%.
It is concluded that the meshes give consistent results and that the spacial discretization error
is smaller than the expected modelling error due to the use of the stock propeller instead of the
actual propeller. It is decided to use Mesh 3 with 9.8M cells for subsequent simulations. The mesh
settings of Mesh 3 are the same as for the model scale self-propulsion simulation, which can be
seen in Table 4.43.

Wake field and Stator Fins In the model scale self-propulsion simulations, it was found that
the stator fins caused separation due to the high angle of attack combined with the significant
4.3 CFD 69

Table 4.43: Cell sizes of refinement zones in the full scale CFD self-propulsion setup. Cell sizes
are made non-dimensional using eq. (4.1).

Name of RZ Cell size l∗ [-]

Bulbous Bow RZ 69
Transom RZ 92
Name of RZ Cell size [-]
Far Propeller RZ 4.26 × 10−2 D
Near Propeller RZ 2.13 × 10−2 D

Figure 4.44: Separation downstream of the stator fins. From full scale self-propulsion simulation
at a Froude number equal to 0.1736. The color indicates the velocity magnitude at r/R equal to
0.5. Seen from starboard side (left), and from port side (right)

wake field. The flow around the stator fins in full scale can be seen in Figure 4.44. It can be seen
that there is almost no separation compared to the flow in model scale seen in Figure 4.37. As
seen in Figure 4.45, the nominal wake field in full scale is much smaller than in model scale. The
increased water flow to the propeller prevents the stator fins from stalling.

Final Results With the convergence studies performed, the full scale self-propulsion setup is
ready for the simulations. In these simulations, three stages are used as shown in Figure 3.36.
The results of the simulations can be seen in Figure 4.46 where MCR is the maximum continuous
rating. The study has been conducted on the hull with both the rudder and stator fins attached,
as seen in Table 3.38.
The CFD results will be compared with sea trial data in the next chapter.
70 4 Results

Figure 4.45: Velocity field, normalized with the free stream velocity. Result of the full scale
resistance setup without stator fins. Black ring is the propeller diameter. The simulation is
conducted at a Froude number F n equal to 0.1736.

120
Delivered Power [% MCR]

110

100

90

80

70

60
0.17 0.175 0.18 0.185 0.19 0.195
Fn [-]
Figure 4.46: Results of the full scale self-propulsion CFD setup.
CHAPTER 5
Discussion
This chapter will discuss the results presented in the previous chapter. Firstly Section 5.1 of this
chapter assesses the accuracy of the sea trial results, and the following Section 5.2 compare and
discuss the results of the extrapolated towing tank results. Finally Section 5.3 is devoted the full
scale CFD simulations.

5.1 Comparison of Sea Trial and In-Service Performance Data

The results from the sea trials presented in Section 4.2 show high correlation between speed and
power, and low scatter of the data points. This indicates that the accuracy of the speed trial data
is high. However, there might be systematic errors in the speed trial data. The correction for wind,
waves and current is a possible source for systematic errors. As mentioned, the new ISO 15016:2015
procedure defines strict standards for conducting sea trials and ensures accurate measurements
and corrections. The power measurements are another possible source for systematic errors. The
constants of the engine manufacturer formula in eq. (3.34) are approximations. Systematic errors
can not be determined solely from the data provided for this thesis. A way to validate the results
from the sea trial is by comparing the results with performance measurements on the ship after
the delivery while it is in service.
The ship owner has provided in-service performance data corrected for wind and waves. The
data is the mean performance of all four sister ships in the first 3-9 months of operation. The
power is based on torque measurements directly on the shaft and not the engine manufacturer
formula as in case of the sea trial measurements. The ship owner has estimated the uncertainty to
be approximately ±3.5% MCR. Uncertainties in the performance data include uncertainties in the
raw torque measurement, the correction for wind and waves and the averaging of the data from
the four sister ships.
The results of the sea trial are compared to the in-service performance data in Figure 5.1. It
can be seen that the ships use more power during operation than during sea trial. On average,
the sea trial data shows 8% less power than the in-service performance data. This is expected
since the ships are in excellent condition at the sea trial with minimal fouling on the ship hull.
Even though both the sea trial data and the in-service performance data have uncertainties, the
relatively low offset in power shows that the sea trial data and the in-service performance data
measures delivered power in the same range. As mentioned earlier, there is only a small scatter of
the sea trial data, which indicates reasonable consistency in the measurements. The combination
of reasonable offset to the in-service data and consistency indicates that the sea trial results have
reasonable accuracy. Because the sea trial results are assumed to be accurate, the sea trial results
can be used as reference for the extrapolated towing tank results and the full scale CFD results.
72 5 Discussion

120
Delivered Power [% MCR]

100

80

60
0.170 0.175 0.180 0.185 0.190 0.195
Fn [-]
Figure 5.1: Comparison of the corrected speed trial results (dot) and the in-service performance
data (dash line). Delivered power PDs as a function of Froude number F n.

5.2 Comparison of Sea Trial and Towing Tank Test

The sea trial results presented in Section 4.2 are compared to the extrapolated self-propulsion
results from the towing tank test presented in Section 4.1.6. The comparison can be seen in
Figure 5.2. By comparing the sea trial and towing tank results directly, it is assumed that there
is no loss in the shaft. Machinery experts at DNV-GL assess that this is a fair assumption since
there is no gear.
As it can be seen in Figure 5.2, the extrapolated towing tank results give a reasonable good
estimation of the delivered power. The extrapolated towing tank results are overestimating the
power by approximately 3% to 9% with an average of 6% compared to the sea trial results. Con-
sidering that towing tank tests are conducted at a much lower Reynolds number and extrapolated,
the observed discrepancies are satisfactory. The power overestimation from the towing tank could
indicate that the ITTC procedure gives a conservative estimate of the ship performance.
The towing tank tests are based on measurements which have uncertainties, and the ITTC ex-
trapolation procedure also adds uncertainties to the estimation of the full scale ship performance.
The accuracy of the ITTC procedure is under constant debate. As an example, recent research
has questioned the assumption of Reynolds independence of the form factor, as described in Sec-
tion 2.2.4. The discrepancies seen in Figure 5.2 show that the ITTC procedure works well in this
case and that the procedure is a good method for estimating the performance of this ship design.
The ITTC procedure has been regularly improved and updated, and is based on over 100 years of
knowledge and experience from towing tanks around the world.
Is it important to remember that the extrapolation of the towing tank results are performed by
the authors using the standard ITTC procedure described in Section 3.2.4. Based on the authors
experiences, many towing tanks use slightly different coefficients and corrections than the standard
ITTC procedure recommends. These changes to the standard procedure are based on experience,
5.3 Comparison of Sea Trial and Full Scale CFD 73

120
Delivered Power [% MCR]

100

80

60
0.170 0.175 0.180 0.185 0.190 0.195
Fn [-]
Figure 5.2: Comparison of the corrected speed trial results (dot) and the extrapolated self-
propulsion results from the towing tank (triangle). Delivered power PDs as a function of Froude
number F n.

tradition, studies of systematic errors and validation studies of each tank. It has not been possible
to have the exact extrapolation procedure that the towing tank uses at our disposal. Therefore,
the extrapolation procedure used in this thesis is the standard ITTC procedure.

5.3 Comparison of Sea Trial and Full Scale CFD

The final results of the full scale self-propulsion CFD simulations can be seen in Figure 5.3 together
with the extrapolated towing tank results and the sea trial results.
It can be seen in Figure 5.3 that the CFD results are close to the sea trial results. Compared
to the sea trial results, the CFD results overestimates the delivered power by approximately 0% to
5% with an average of 2%. It is important to remember that the CFD simulations are performed
using the stock propeller and not the actual propeller. In Section 4.1.6 it is shown that the
delivered power from the extrapolated towing tank tests using the stock propeller characteristics
instead of the actual propeller characteristics is 4% to 5% lower. This could indicate that the stock
propeller performs slightly better in full scale than the actual propeller. This indication is based
on several assumptions. One assumption is that the ITTC extrapolation procedure of the open
water characteristics is accurate. Research as [34, 42] has questioned the accuracy of the procedure.
The comparison is based on open water and self-propulsion tests which both have measurement
uncertainties. The origin of the actual propeller characteristics and the method with which it is
calculated is not known to the authors.
It would be interesting to perform CFD simulations with the actual propeller, but it has not
been possible to get the actual propeller geometry due to confidentiality. Therefore, it is important
to remember that the difference of 4 to 5% is not an accurate value but more an indication that
74 5 Discussion

120
Delivered Power [% MCR]

100

80

60
0.170 0.175 0.180 0.185 0.190 0.195
Fn [-]
Figure 5.3: Comparison of the corrected speed trial results (dot), the extrapolated self-propulsion
results (triangle) from the towing tank and the full scale CFD self-propulsion results (circle).
Delivered power PDs as a function of Froude number F n.

the performance of the stock propeller is similar to or possibly slightly better than the actual
propeller.
Another source of uncertainty is the fact that the roughness resistance on the hull, the air
resistance on the super structure and the added bilge keel resistance are all estimated using the
ITTC procedure (see Section 3.2.4) and not modelled in the CFD simulations. The ITTC procedure
for estimating these resistance contributions has been developed as a part of extrapolating towing
tank tests to full scale and is therefore not meant to be used separately for full scale CFD results.
By using the procedure on the CFD results, it is expected that uncertainties are introduced. The
ITTC procedure estimates that out of the total resistance, the hull roughness is approximately
4% to 5%, the air resistance is approximately 3% to 4% and the bilge keels is approximately 1%
to 2%. It could have been interesting to model or simulate the hull roughness, air resistance and
bilge keels in the CFD simulations. However, due to time constraints, it has been decided not to
implement these elements in the CFD setups.
Prior to this thesis, experts from DNV GL have had discussions with a world leading paint
manufacturer. The paint manufacturer has expressed that a roughness of 150 µm is high. A
hull roughness of 90 µm to 120 µm is a more reasonable estimate. It is noted that the paint
manufacturer is not neutral because they have an interest in estimating a low roughness for their
paint. Furthermore, it is not known if the roughness given by the paint manufacturer takes welding
seems and anodes into account. However, this could indicate that the roughness resistance which
has been added to both the towing tank results and the CFD results is too high. As seen in
Figure 5.3, both the CFD and towing tank results are overestimating the power. That could
possibly be explained by an overestimation of hull roughness resistance.
As mentioned in Section 3.4, the CFD method simplifies the physics by modelling the turbulence
behaviour instead of resolving the flow. The choice of turbulence model is crucial to the accuracy
5.3 Comparison of Sea Trial and Full Scale CFD 75

of the results. However, by modelling the turbulence, uncertainties are expected. An example
of a study with different turbulence models in ship hydrodynamics is [3]. Furthermore the free
surface is removed from the self-propulsion simulation and the wave making resistance is estimated
separately. This modelling error caused by removing the free surface is small based on experience
from CFD experts at DNV GL.
Based on the discussion above, the results seen in Figure 5.3 show that the full scale CFD
simulations provides results similar to the sea trials which indicates a great potential for CFD
simulation in predicting the performance of ships. Previously it has not been possible to validate
towing tank results in the fully loaded condition because speed trial data in fully loaded condition
is very rare. The reason for the limited speed trial data is the practicalities related to fully load
the ship at the speed trial. If full scale CFD is shown trustworthy, CFD simulations can easily be
performed in the fully loaded condition, which is much more interesting than the ballast condition.
In order to gain more trust in CFD simulations of self-propelled ships in full scale, more test cases
are required. If other cases show results similar to this thesis, it could change the way the maritime
world predicts the performance of ships in the future.
76
CHAPTER 6
Conclusions
The overall goal of the project has been to develop and validate a CFD model of a full scale
self-propelled ship. This goal has been fulfilled.
The sea trial results of the four sister ships have shown similar performance. A comparison of
the sea trial results and monitored in-service performance data has shows that the delivered power
in-service is on average approximately 8% higher than the sea trial results. This is expected since
fouling on the hull is present in service but not at the sea trial. The combination of the similar
measured performance for the sister ships and the reasonable offset to the monitored in-service
performance indicates that the sea trial results are a trustworthy reference for the extrapolated
towing tank data and full scale CFD.
The development of the CFD model has gone through several stages. First a resistance setup
has been validated. The resistance setup has been developed prior to this project. In parallel, an
open water setup has been created and validated. The results of the resistance and open water
CFD calculations have been compared with towing tank results and the results showed small
discrepancies of similar magnitude to the measurement uncertainties.
The resistance setup and the open water setup have been combined into a CFD setup that
simulates a self-propulsion test. First, the self-propulsion test has been simulated in model scale.
The results from the model scale self-propulsion simulations have shown reasonable similarity to
the towing tank results.
Based on the model scale setup, self-propulsion test have been simulated in full scale and
compared to sea trial data and the extrapolated self-propulsion test from the towing tank. The
comparison has shown that the full scale self-propulsion CFD simulations overestimate the deliv-
ered power by approximately 2% on average compared to the sea trial results. This discrepancy
is reasonable since the simulations have been performed with the stock propeller instead of the
actual propeller and the fact that air resistance, bilge keel resistance and hull roughness resistance
are calculated using the ITTC procedure instead of being modelled in CFD.
The provided towing tank test results have been extrapolated to full scale using the ITTC
procedure. It was found that this prediction overestimated the delivered power with 6% in average
compared to the sea trial results.
The strengths, weaknesses and uncertainties related to the sea trial, CFD and towing tank
predictions have been discussed. This thesis has shown that both the traditional towing tank
approach and full scale CFD can make reasonable estimations of the full scale performance of a
ship. In the case studied in this thesis, the CFD approach gave an estimation closer to the sea
trial results. To conclude whether the full scale CFD or the towing tank approach is superior in
prediction of the performance of a ship, more similar studies are required. However, this thesis has
indicated that CFD has great potential to be the method for estimating full scale performance in
the future.
78
CHAPTER 7
Future Work
As mentioned in the conclusion more studies like this are required in order to determine the
accuracy of full scale CFD and extrapolated towing tank tests.
It is recommended for future studies of similar kind, that the availability of and access to
key data (e.g. the ship and propeller geometry, sea trial results, towing tank results) are checked
before the project start. A full list of required and optional data can be found in Appendix A. The
difference in delivered power from sea trial, extrapolated towing tank test and full scale CFD is
relatively small so low uncertainties are required for all data in order to be be able to make strong
conclusions.
A disadvantage of the CFD approach is that the turbulence is modelled using a turbulence
model. Therefore it could be interesting to study the influence of using a different turbulence
model(e.g. k − ω SST) in the self-propulsion simulations.
Furthermore, it would also be very interesting to include roughness, bilge keel and air resistance
in the CFD simulation. Thereby, the CFD approach will be completely independent of the ITTC
procedure.
80
 Bibliography
[1] Satu K. Hänninen, Tommi Mikkola, and Jerzy Matusiak. Computational and experimental
study on local ship loads in short and steep waves. Journal of Marine Science and Technology
(Japan), 19(1):103–115, 2014.
[2] J. Wackers, B. Koren, H. C. Raven, A. van der Ploeg, A. R. Starke, G. B. Deng, P. Queutey,
M. Visonneau, T. Hino, and K. Ohashi. Free-Surface Viscous Flow Solution Methods for Ship
Hydrodynamics. Archives of Computational Methods in Engineering, 18(1):1–41, 2011.
[3] Lars Larson. Gothenburg 2010, a workshop on numerical ship hydrodynamics.
[4] IMO. ISO 15016:2015. MEPC 68/INF.14, page 1, 2015.
[5] A. Molland, S. Turnock, and D. Hudson. Ship Resistance and Propulsion. Cambridge Univer-
sity Press, 2011.
[6] John Newman. Marine Hydrodynamics. MIT Press, 1977.
[7] Milton Van Dyke. An Album of Fluid Motion. 1982.
[8] W. Froude. The Papers of William Froude. 1955.
[9] G. Hughes. Friction and form factor resistance in turbulent flow and a proposed formulation
for use in model and ship correlation. Transactions of the Royal Institution of Naval Architects,
96:314–376, 1954.
[10] ITTC. Recommended Procedures and Guidelines, 1978 ITTC Performance Prediction Method,
2005.
[11] J. Carlton. Marine Propellers and Propulsion, 3. edition. Butterworth-Heinemann, 2012.
[12] 26th ITTC Resistance Committee. The Resistance Committee, Final Report and Recommen-
dations to the 26th ITTC. 26th International Towing Tank Conference, I:17–71, 2011.
[13] L. Eça. On the accuracy of the numerical prediction of scale effects on ship viscous resistance.
International Conference on Computational Methods in Marine Engineering, pages 1–10, 2005.
[14] Karl E. Schoenherr. Resistance of flat surfaces. Trans SNAME, pages 40:279–313, 1932.
[15] C. W. B. Grigson. A planar friction algorithm and its use in analysing hull resistance. RINA,
1999.
[16] Tahara Y. Katsui T., Asai H., Himeno Y. The Proposal of a New Friction Line. Fifth Osaka
Colloquium on Advanced CFD Applications to Ship Flow and Hull Form Design, 2005.
82 Bibliography

[17] ITTC. ITTC – Recommended Procedures, Testing and Extrapolation Methods, Resistance
Resistance, Test, 7.5-02 -02-01. ReVision, pages 266 – 273, 2002.

[18] A. Garcia-Gomez. On the form factor scale effect. Ocean Engineering, 26(1):97–109, 2000.

[19] D. Park. A study on the effect of flat plate friction resistance on speed performance prediction
of full scale. 2015.

[20] Keh S. Min and Seon Hyung Kang. Study on the form factor and full-scale ship resistance
prediction method. Journal of Marine Science and Technology, 15(2):108–118, 2010.

[21] H. C. Raven, A. Van Der Ploeg, A. R. Starke, and L. Eça. Towards a CFD-based prediction
of ship performance - progress in predicting full-scale resistance and scale effects. Transac-
tions of the Royal Institution of Naval Architects Part A: International Journal of Maritime
Engineering, 150(4):31–42, 2008.

[22] K. Minsaas. Grunnlag for fartsprognoser. Technical report, Marintek (former: Norges Hydro-
dynamiske Laboratorier) , Trondheim, Norway., 1982.

[23] D. Wilcox. Turbulence Modelling for CFD, 2nd edition. DCW. Industries, 1998.

[24] Cuong Nguyen. Turbulence Modeling. MIT, 1(8):1–6, 2005.

[25] D. C. Wilcox. Turbulence Modeling for CFD, 3rd edition, 2006.

[26] D. C Wilcox. Formulation of the k-omega Turbulence Model Revisited. AIAA Journal,
46(11):2923–3838, 2008.

[27] F. R. Menter. Two-Equation Eddy-Viscosity Turbulence Models for Engineering Applications.

American Institute of Aeronautics and Astronautics, 32(8):1989–1605, 1994.

[28] CD-Adapco. STAR-CCM+ Documentation, Version 10.04, 2015.

[29] A.E.P. Veldman. BOUNDARY LAYERS IN FLUID DYNAMICS. Lecture Notes in Applied
Mathematics, University of Groningen., 2011.

[30] Salim M. Salim. Wall y+ strategy for dealing with wall-bounded turbulent flows. Proceedings
of the International MultiConference of Engineers and Computer Scientists, II, 2009.

[31] R. Azcueta. Computation of Turbulent Free-Surface Flows Around Ships and Floating Bodies.
PhD Thesis, 2001.

[32] ITTC. Recommended Procedures and Guidelines, Testing and Extrapolation Methods Propul-
sion, Propulsor Open Water Test, 7.5-02 -03-02.1, 2011.

[33] ITTC. Recommended Procedures and Guidelines, Propulsion/Bollard Pull Test, 7.5-02-03-
01.1. – Recommended Procedures and Guidelines, page 16, 2011.

[34] Stephan Helma. An Extrapolation Method Suitable for Scaling of Propellers of any Design.
Fourth International Symposium on Marine Propulsors, 2015.
Bibliography 83

[35] Eirik Eisinger. A Method for Describing Ocean Environments for Ship Design. Masters Thesis,
University of Oslo, 2016.

[36] Paul Berrisford, D Dee, K Fielding, M Fuentes, P Kallberg, S Kobayashi, and S Uppala. The
ERA-Interim Archive. ERA report series, 1(1):1–16, 2009.

[37] M Kume, K. I.; Hirata, N.; Hasegawa, J.; Tsukada, Y.; Hinatsu. Uncertainty Analysis Method
for Ship Performance Test. Senpaku Gijutsu Kenkyusho Hokoku/papers of Ship Research
Institute, 37(5):163–189, 2000.

[38] G. Kuiper. Cavitation inception on ship propeller models. Ph.D. thesis, Delft University of
Technology, 1981.

[39] Douwe R. Rijpkema, J. Baltazar, and J.A.C. Falcão de Campos. Viscous flow simulations
of propellers in different Reynolds number regimes. Proceedings of the 4th International
Symposium on Marine Propulsors, I, 2015.

[40] Ridha Abid. Evaluation of two-equation turbulence models for predicting transitional flows.
International Journal of Engineering Science, 31(6):831–840, 1993.

[41] Xiao Wang. Computational Analysis of Marine-Propeller Performance Using Tranition-

Sensitive Turbulence Modeling. Journal of Fluid Engineering, 134(1), 2012.

[42] G. Kuiper. The Wageningen Propeller Series, 1992.

84
APPENDIX A
Required Data
The following is a list of data required to perform a study similar to the study of this thesis.
Required data:

• Principal dimensions

• Speed with main engine characteristics

• General arrangement (G/A)

• High quality 3D geometry of hull, rudder, propellers (the one used in model test and actual
on ship) and other appendages.

• Exact location of propeller and appendages.

• Full towing tank report with model scale results in sea trial condition. Including results from
resistance test, open water test and self-propulsion. Water temperature at each test. Infor-
mation on used corrections to the results (e.g. corrections to the open water characteristics).

• Sea trial report with results corrected for current, wind and waves.

Optional data:

• Sea trial report with raw measurements of speed, rpm, engine power, current, wind and
waves.

• Extrapolated results from a towing tank test.

• 3D geometry of superstructure.

• In-service performance data.

• Added resistance test from towing tank.

• Wind tunnel test report.

DTU Mechanical Engineering
Section of Fluid Mechanics, Coastal and Maritime Engineering
Technical University of Denmark

Nils Koppels Allé, Bld. 403

DK-2800 Kgs. Lyngby
Denmark
Phone (+45) 4525 1360
Fax (+45) 4588 4325

www.mek.dtu.dk