Вы находитесь на странице: 1из 80

Falculty V

Institute of Land and Sea Transport Systems


Chair of Dynamics of Maritime Systems

Modellingofacranevesselondynamicpositioning
andassessmentofitsoperationalcapability
Master's Thesis

Author Gerrit Aÿbrock

Degree course Master of Science


Naval Architecture and Ocean Engineering
Student registration no. 371413
Supervisors Prof. Dr.-Ing. Andrés Cura Hochbaum
Dr.-Ing. Florian Stempinski (GeoSea NV)
Sebastian Uharek M.Sc.

Berlin, 25 June 2018


Eidesstattliche Erklärung

Die selbständige und eigenhändige Anfertigung versichere ich an Eides statt.

Berlin, 25. Juni 2018

Gerrit Aÿbrock
Abstract

Construction of oshore wind parks is a major contribution to energy revolution. Jack-Ups


are entrenched vessels to install the foundations. As future parks attain deeper waters,
oating crane vessels are gaining ground. Utilising these saves the time to jack the vessel
up. Nevertheless, motions during installation restrict their operational capability. This fore-
grounds the vessel's Dynamic Positioning (DP) system. To ensure save operations, time
domain simulations are a reliable tool as long as environmental loads and vessel dynamics
are properly modelled. Both is outlined here, with emphasis on vessel dynamics and the
signicance of modelling the vessel's DP system. However, the importance of considering
slowly varying wave forces is proven as well.
A developed external function comprises a feedback lter, a PID controller and an optim-
ised thrust allocation by quadratic programming. It relieves a previously used analytical
spring-damper system for positioning. Results from model tests are analysed and validate
calculations in irregular sea states. Most Probable Maximum (MPM) vessel motions are the
performance indicators to assess response based operability. Simulations with a crane vessel
show, that save operations are possible up to a signicant wave height of three meters in
bow quartering waves. The developed external function and derived operational limits serve
as a basis to investigate upcoming installation methods.
Keywords: dynamic positioning; external function; slowly varying wave forces

Kurzfassung (Abstract)

Die Errichtung von Oshorewindparks leistet einen erheblichen Beitrag zur Energiewende.
Meist werden Jack-Up Schie zur Installation der Fundamente eingesetzt. Da neue Parks
in mit unter tieferen Gewässern geplant sind, werden schwimmende Kranschie immer rel-
evanter. Der Einsatz dieser erspart die nötige Zeit zur Errichtung des Jack-Up Schies.
Allerdings ist die Einsatzfähigkeit des schwimmenden Schies durch die Bewegungen be-
grenzt. Dies rückt dessen System zur Dynamischen Positionierung (DP) in den Vorder-
grund. Solange Umweltlasten und die Dynamik des Schies korrekt modelliert sind, dienen
Simulationen der Planung sicherer Oshore Operationen. Beide Aspekte werden behandelt,
obgleich der Fokus auf der Modellierung des Schies und seines DP Systems liegt.
Die Berücksichtigung von Wellenlasten im Bereich der Dierenzenfrequenzen erweist sich
als wesentlich. Die entwickelte Routine zur Modellierung des DP Systems umfasst einen
geschlossenen Regelkreis sowie eine Schnittstelle zur Allokation der gesamt Kraft mittels
eines Optimierungsalgorithmus. Ein herkömmliches Feder-Dämpfer System zur Positionier-
ung wird dadurch abgelöst. Ergebnisse aus Modellversuchen dienen der Validierung von
Seegangssimulationen. Anhand von wahrscheinlich höchsten Bewegungen ("most probable
maximum") werden Indikatoren zur Bemessung der Einsatzfähigkeit abgeleitet. Zur Planung
sicherer Oshore Operationen mit einem Kranschi kann die Einsatzgrenze mit drei Metern
HS in schräg von vorn einlaufenden Wellen beziert werden. Das modellierte DP System
sowie ermittelte Einsatzgrenzen dienen der Überprüfung neuer Installationsmethoden.
Contents

Contents

Abstract III

List of Figures VII

List of Tables VII

List of Programme Code VIII

List of Abbreviations VIII

Nomenclature IX

1 Introduction 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Discussion of Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Objective and Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2 Background Hydrodynamic Analysis 4


2.1 Coordinate Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Wave Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3 Vessel Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3.1 Response in Waves . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3.2 Second Order Wave Forces . . . . . . . . . . . . . . . . . . . . . . 11
2.3.3 Wind Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3.4 Current Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3.5 Roll Damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.4 Dynamic Positioning System . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.4.1 Feedback Filter . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.4.2 Controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.4.3 Allocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4.4 Optimisation Problem . . . . . . . . . . . . . . . . . . . . . . . . 22
2.4.5 Solver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.5 Equation of Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3 Numerical Model 25
3.1 Panel Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.2 Mesh and Grid Dependency . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.3 Results Frequency Domain Calculation . . . . . . . . . . . . . . . . . . . . 27
3.4 Current and Wind Coecients . . . . . . . . . . . . . . . . . . . . . . . . 31
3.5 Time Domain Calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.6 Postprocessing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

V
Contents

4 Implementation of Dynamic Positioning System 38


4.1 Feedback Filter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.2 Controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.3 Allocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.4 Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

5 Results 48
5.1 Inuence Filter Setting . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.2 Inuence Second Order Wave Forces . . . . . . . . . . . . . . . . . . . . . 49
5.3 DP Footprint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.4 Statistics Thrust Allocation . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.5 DP Envelope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

6 Conclusion 58
7 Perspective 59
References 61
Appendix 64
A Comparison Environmental Loads in Time Domain 64
B Comparison Commanded and Allocated Forces 65
C Source Code External Function 66
C.1 Basic Structure of the External Function . . . . . . . . . . . . . . . . . . . 66
C.2 Part of Allocation in the Main Routine . . . . . . . . . . . . . . . . . . . . 67

VI
List of Figures

List of Figures

1.1 Illustration of the crane vessel. . . . . . . . . . . . . . . . . . . . . . . . . 1


2.1 Coordinate systems and unit vectors. . . . . . . . . . . . . . . . . . . . . . 4
2.2 Local coordinate systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.3 Absolut QTF of dierence frequencies in yaw at. . . . . . . . . . . . . . . 13
2.4 General view of the computational grid above the SWL. . . . . . . . . . . 15
2.5 Total roll damping derived from model tests at load case one. . . . . . . . 17
2.6 Block diagram of time domain calculation. . . . . . . . . . . . . . . . . . . 19
2.7 Bode plot open controller. . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.1 Mesh convergence in frequency domain. . . . . . . . . . . . . . . . . . . . 26
3.2 Visualisation and comparison of panel mesh. . . . . . . . . . . . . . . . . . 27
3.3 RAOs of motions in six DoFs. . . . . . . . . . . . . . . . . . . . . . . . . 29
3.4 Dimensionless mean wave forces. . . . . . . . . . . . . . . . . . . . . . . . 30
3.5 Current coecients. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.6 Comparison of wind coecients. . . . . . . . . . . . . . . . . . . . . . . . 34
3.7 Convergence of forces and motions for a decreasing time step. . . . . . . . 36
4.1 Dynamic thrust region and overall thrust limits. . . . . . . . . . . . . . . . 42
4.2 Step test in surge direction. . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.3 Step test in sway direction. . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.4 Step test in yaw direction. . . . . . . . . . . . . . . . . . . . . . . . . . . 47
5.1 Comparison of feedback lter settings. . . . . . . . . . . . . . . . . . . . . 49
5.2 Power spectra of yaw motion. . . . . . . . . . . . . . . . . . . . . . . . . 50
5.3 DP footprint in irregular sea state 2. . . . . . . . . . . . . . . . . . . . . . 52
5.4 Comparison of mean and standard deviation of bow tunnel thrust. . . . . . 53
5.5 Comparison of thrust allocation statistics in sea state irregular 2. . . . . . 54
5.6 DP Envelope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
A.1 Comparison of yaw moments in time domain. . . . . . . . . . . . . . . . . 64
B.1 Comparison of forces by the PID controller and allocator. . . . . . . . . . . 65

List of Tables

2.1 Classication of roll damping contributions with respect to critical damping. 18


3.1 Mesh parameters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2 References values wind and current coecients. . . . . . . . . . . . . . . . 32
3.3 Comparison current coecients. . . . . . . . . . . . . . . . . . . . . . . . 33
3.4 Convergence of forces and motions with dt by standard deviations. . . . . 36
4.1 Main particulars and parameters of the vessel's Load Case. . . . . . . . . . 43
4.2 Overview tank step tests. . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.1 Overview model tests in irregular waves. . . . . . . . . . . . . . . . . . . . 48
5.2 Comparison of statistics in irregular waves. . . . . . . . . . . . . . . . . . 51
5.3 MPM motion in surge and sway at the PoI . . . . . . . . . . . . . . . . . 56

VII
List of Abbreviations

List of Programme Code

4.1 Position Filter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40


4.2 PID Controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.3 Limiting PID output. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
C.1 Header . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
C.2 Allocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

List of Abbreviations

AP Aft Perpendicular. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .43


CFD Computational Fluid Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
CoG Center of Gravity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
CoR Center of Rotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
DoF Dimension of Freedom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
DLL Dynamic Link Library . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
DP Dynamic Positioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . III
FRA Fixed Reference Axes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
GM Metacentric Height . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
LC Load Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
LCG Longitudinal Center of Gravity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
LSA Local Ship Axes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
MARIN Maritime Research Institute Netherlands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
MPM Most Probable Maximum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . III
OCV Oshore Construction Vessel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
PDF Probability Density Function
PID Proportional Integral Derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
PoI Point of Installation
QTF Quadratic Transfer Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
RANSE Reynolds Averaged Navier Stokes Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
RAO Response Amplitude Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
SLSQP Sequential Least Squares Quadratic Programming. . . . . . . . . . . . . . . . . . . . . . . . . . .22
SWL Still Water Level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
TLP Tension Leg Platform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
VCG Vertical Center of Gravity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
VLCC Very Large Crudeoil Carrier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

VIII
Nomenclature

Nomenclature

Symbol Denotation Unit

Latin Symbols
[x 0; y 0; z 0]T Coordinates in hybrid/heading xed system at CoG [m ]
[x; y; z ]T Coordinates in hybrid/heading xed system at PoI [m ]
[x ; y ; z ]T Coordinates in vessel xed system (LSA) [m]
B Matrix of damping coecients [kg=s ] and [kgm=s ]
F Vector of forces and moments [N ] and [Nm]
MA Matrix of added masse and inertia [kg ] and [kgm2 ]
MRB Matrix of mass and inertia [kg ] and [kgm2 ]
P Matrix of real part of QTF [N=m2 ] and [Nm=m2 ]
Q Matrix of imaginary part of QTF [N=m2 ] and [Nm=m2 ]
s Vector of slock forces and moments [N ] and [Nm]
T Matrix of absolute QTF [N=m ] and [Nm=m2 ]
2

x Vector to CoG in FRA and Euler angles [m][rad ]


B Breadth [m]
D Draught [m ]
d Water depth [m ]
fx Single thruster force heading xed x direction [N ]
fy Single thruster force heading xed x direction [N ]
Hs Signicant wave height [m ]
LP P Length between perpendiculars [m ]
T Period [s ]
t Time variable [s ]
Tp Peak period [s ]
TZ Zero up-crossing period [s ]
U Absolut relative velocity [m=s ]
T j;t Thrust of single actuator j 2 f1; :::; 8g at time t [N ]

Greek Symbols
[; ;  ]T Coordinates in inertial/earth xed system(FRA) [m]
Azimuth thruster angle [ ]
Relative angle of encounter [ ]
^W Wave amplitude [m]
 Angle of encouter [ ]
r Displacement [m3 ]
! Circular frequency [rad=s ]
!p Circular peak frequency [rad=s ]
 Velocity potential [m2 =s ]
7 Velocity potential due to diraction [m2 =s ]
j Velocity potential due to radiation j 2 f1; :::; 6g [m2 =s ]

IX
Nomenclature

Symbol Denotation Unit


'; ; Euler angles [ ]
'a Roll amplitude [ ]
W Surface elevation [m ]
2 Variance [ 2 ]

Constants
 Density sea water 1025:00 [ mkg3 ]
g Acceleration due to gravity 9:81 [ sm2 ]

X
1 Introduction

For the installation of oshore wind farm foundations, GeoSea NV is currently building a new
oating crane vessel. Operability of these vessels is restricted by many indicators, whereas
the ability to maintain position and heading is a minimum requirement. Within the scope of
this thesis an external function for an existing time domain simulation is developed. Taking
the vessel's DP system into account, assessment of the its operational capability is done.
An early design of the vessel in operational condition is illustrated in Figure 1.1.

Figure 1.1: Illustration of the crane vessel. (Source: GeoSea NV)

So far transport and installation as well as maintenance work on turbines and blades is
done with Jack-Up vessels. The process of jacking a vessel up is complex and asks a lot
of preliminary calculation considering soil conditions and the vessel's hull structure (see
DNVGL [8]). Those vessels are always a compromise. On the one hand a hydrodynamic
optimised hull regarding sea keeping performance and transit speed is asked and on the
other hand a stationary platform that has to withstand rough environmental conditions is
needed. Since oshore installations seek deeper waters and rougher soil conditions the need
of oating crane vessels grows. In opposite to cranes on semi-submersible platforms, they
have the advantage of higher transit speeds. Time lost in transit is one point that needs to
be considered when planning oshore operations. During installation at a xed location, the
vessel's ability to maintain that position needs particular examination. Oshore Construction
Vessels (OCVs) in general operate either moored, assisted by tugs, or do have a DP system
that takes the task of stationing and correcting the heading.

1.1 Motivation

The analysed crane vessel will mainly operate on DP. Disregarding other operational limits,
the vessel motions in the horizontal plane can already restrict the operational capability. To

1
1 Introduction

project save oshore operations time domain simulations are conducted. These calculations
strongly depend on considered environmental loads and the vessel model. Loads are modelled
by forces due to irregular waves as well as almost static wind and current velocities. Knowing
the characteristics of the vessel's DP system is crucial for successful operations, hence
the vessel model needs to incorporate that. Therefore, an external function is developed
that comprises the implementation of a position feedback to a controller and an allocation
algorithm that distributes total forces and the yawing moment to the individual thrusters.

1.2 Discussion of Literature

DP systems help the crew to keep their vessel in place since years. What all systems do
have in common are a position feedback and the controlled allocation of forces and mo-
ments generated by single thrusters. Following Pinkster and Nienhuis [26] DP systems are
introduced rst in the sixties and have been intensively used and developed in the oshore
industry. They presented an approach to minimize the consumed power by thrusters to the
price of reducing the number of controlled Dimension of Freedoms (DoFs). One thruster is
located at the vessel's bow or aft at the point of environmental load incident. The vessel's
heading is unrestrained and governed by the environmental loads. Although, for purposes
of foundation installation a dedicated heading has to be maintained, since the operation
can benet from shielding eects on the vessel's lee side. A more complex time domain
model has been developed by van den Boom, H. J. J. and Nienhuis [32] and [22]. Thruster
- thruster interactions are considered to compensate for their wake eld and its inuence on
other actuators in close proximity. Wichers et al. [35] promoted research on DP system and
explained that observers are not necessary for time domain simulation since the entire state
vector of the vessel is available. A comprehensive study is provided by Ruth [30]. Dier-
ent allocation algorithms are compared as well as an approach to take thruster ventilation
into account when vessels operate in harsh conditions. Moreover the inuence of dierent
types of controllers is addressed by Rabanal et al. [27]. A way to combine the control and
allocation task trying to avoid infeasible thruster commands is analysed by Veksler et al.
[34] contrasting advantages and disadvantage of model predictive control. Considering roll
motion as being aected by a DP system is done by Rudaa et al. [29]. Additionally classi-
cation societies publish standards on how to model and test DP systems for time domain
calculations (see DNV GL [6]). Most of those subjects are far beyond this thesis. But it
helped a lot to distinguish necessary features for a reliable model from complex full scale
system considering e.g. noises on position signals and observer models like summarized by
Fossen [11].
Regarding the model of environmental loads van Oortmerssen [33] and Pinkster [25] ex-
plained the need to consider slowly varying wave forces of second order that cause motions
at dierence frequencies. A propulsion system cannot counter act high frequent wave forces,
thus that second order forces at dierence frequencies are the governing dynamic load for
a DP system. Slowly varying forces occur in irregular sea state like described by Newman
[20]. Lately Pessoa et al. [24] measured second order motions of an ordinary oating body
in bi-chromatic waves in good agreement with numerical calculations.

2
1.3 Objective and Approach

Following Gutsch et al. [12] assessing response-based operability of OCVs can be summarized
by performance indicators. Here MPM vessel motions indicate the vessels operability.

1.3 Objective and Approach

The main goal of this thesis is the development of an external function that models the
crane vessel's DP system in time domain. The function is written in Fortran and interferes
with the simulation framework of ANSYS Aqwa1 . Although, it is compiled to an Dynamic
Link Library (DLL), hence it can be integrated in other software.2 To counteract the slow
varying motions of the vessel a moving time average lters the position signal derived from
the time integration. Single PID controllers for the three DoFs predict forces in surge and
sway and the yawing moment based on the negative feedback from the lter. After limiting
the maximum commanded forces of the controller, an optimised allocation algorithm is asked
to distribute the total forces and the moment. The implementation of maximum azimuth
rates and a linear acceleration of rpm considers the DP systems dynamics.
Aiming to obtain a reliable vessel model, the response to waves is calculated preliminary in
frequency domain. The inuence of slowly varying second order wave forces is discussed,
and found to have a signicant contribution to the vessel's response. Loads due to wind
and current are considered by force coecients. Although, the DP system acts only in the
three horizontal DoFs roll damping is properly corrected by results from decay tests at model
scale.
The developed model is validated against model tests. Analysing the new build crane vessel,
predictions of MPM motions are calculated in dierent sea states.

1.4 Outline

First, assumptions and the theoretical background are outlined (see section 2) before the
developed numerical model (see section 3) and the implemented external function are in-
troduced in section 4. Validation of the model is done in section 4.4 before results are
presented, conclusions drawn and perspectives given.
A digital version of this thesis comprises the source code of the external function, a script
to compile the function and an exemplary input le.

1 Introduction: https://ansys.com
2 Integration can be done using libraries like python's ctypes that can make Fortran subroutines available.
https://docs.python.org

3
2 Background Hydrodynamic Analysis

2 Background Hydrodynamic Analysis

Assumptions modelling the crane vessel and its DP system to simulate motion in time
domain are summarised in the following sections. As soon as coordinate systems are given,
theory and assumption on the three environmental disturbances, wave, wind and current are
documented.

2.1 Coordinate Systems

The earth xed, inertial system is a right hand orientated and is in the following referred to
as the Fixed Reference Axes (FRA). Its positive longitudinal axis points to the bow of the
vessel and the transverse axis points positive to port-side, thus the vertical axis is orient-
ated positive upward. The origin of the FRA is located in the Still Water Level (SWL) at
the vessels transom on the centre line. The DP system works in the hybrid reference axis,
which has the same orientation as the FRA but is rotated by the yaw angle about the
vertical axis, thus orientation of ~k and the vertical inertial axis are identical (see Figure 2.1
comparing the green hybrid and the black FRA system). Therefore the DP system will only
return forces and moments in the DoFs of the horizontal plane. These are surge with unit
vector ~i, sway with unit vector ~j and yaw indicated with the angle about the axis along
the unit vector ~k . The hybrid system is hereafter also referred to as the heading xed system.

Transformation between coordinate systems are made using the common rigid body trans-
formations about the Euler angles ['; ; ]. The angle ' describes the vessels roll angle
about the ship xed longitudinal axis with the unit vector ~i ,  the pitch angle about the
transverse axis in the hybrid system ~j.

270 180 [~i;~j; ~k ]T hybrid/heading xed


 0 [~i ;~j ; ~k  ]T ship xed/LSA

 z0
inertial/FRA
90
y0 z
 = 0
[0 ; 0 ; 0 ]T ' 0
y
0
x0
CoG  x

Figure 2.1: Coordinate systems and unit vectors. (Source: author)

4
2.1 Coordinate Systems

vessel fore
BTF
BTA fx

0
RFF
T ( )
y
0
RFA
x
0

CoG 90 270



0
fy
y

ld
x

e
ke
PS

wa
P oI
PSM SB
SBM vessel aft
(a) (b)

Figure 2.2: Local coordinate systems. (a) Local heading xed coordinate sys-
tem at PoI and sketch of thruster arrangement. (b) Orientation and
denition of positive thrust direction for azimuth thrusters.

The position vector of the Local Ship Axes (LSA) is given by [0 ; 0 ; 0 ] and points to the
vessel's Center of Gravity (CoG). Values for the analysed Load Case (LC) are given in
Table 4.1. The DP system's set and reference point is dened at a remote location, the
Point of Installation (PoI) (see Figure 2.2 (a)). Presented vessel motions are transferred
to the PoI. Deviations are always given in that heading xed coordinate system. This axis
system is still parallel to the FRA, thus nor roll or pitch are considered. The DP systems
Center of Rotation (CoR) is identical to the PoI.
Figure 2.2 (a) additionally provides a sketch of the foreseen thrusters and their arrangement.
Only azimuth and tunnel thrusters will equip the vessel, thus speaking of a thruster refers to
one of the actuators. Azimuth thrusters are indicated by trapezia and bow tunnel thruster
(BT) by squares. Abbreviations are given according to location on the vessel. E.g. PSM
refers to the azimuth thruster at port-sid mid and RFF to the retractable azimuth thruster at
fore-fore location. Both azimuths at fore-fore and fore-aft location are retractable. Delivered
forces by bow tunnel thrusters are dened positive to port-side.
Figure 2.2 (b) provides the translation of local thruster forces in Cartesian coordinates fx
and fy to the polar coordinate system. Thrust T j ( ) of a single thruster j is referred to as
the resulting pulling force on the vessel. E.g. T P S (90 ) would pull the vessel to port side
and mainly cause a negative yawing moment.
Angles of encounter for environmental loads are dened in the FRA (see Figure 2.1). E.g.
waves hitting on starboard-side correspond to  = 90 , since the wave force points in the
same direction as the just explained pulling thrust force. Considering that deviation from
the set heading in operations on DP can be expected to be small , dierent systems become
only relevant performing step tests.

5
2 Background Hydrodynamic Analysis

2.2 Wave Theory

Waves are dened in the FRA. The assumption of linear Airy wave theory are given below.

ˆ The waves are present, the ocean surface is already perturbed.


ˆ The sea bed is horizontal and impermeable.
ˆ The uid's density is constant, it is incompressible.
ˆ The uid is ideal, it is inviscid.
ˆ The surface tension is not considered.
ˆ The wave is two dimensional, it is long-crested.
ˆ The analysed wave is not aected by other water movement.
ˆ The wave motion is assumed to be a potential ow.
ˆ The wave amplitude ^W is considerable small in comparison to the wave length .
ˆ The wave steepness is considered to be small. Ratios of wave height H and wave
length  of H < 0:05 are assumed.

The Laplace equation r2  = 0 can be simplied by applying these assumptions and the
boundary conditions, thus non-linear components in the boundary conditions will be disreg-
arded. The velocity potential  is introduced to describe water particle motions and the
wave elevation. It is derived from the Laplace equation which is
@ 2 @ 2
@2 + @ 2 = 0 (2.1)

for the two dimensional case. The following boundary conditions are set to solve this
dierential equation:

ˆ The bottom boundary condition


@
w = @ = 0 on = d (2.2)

The particle velocity w orthogonal to the sea bed at z= d is zero. This condition
states the impermeability of the bottom.
ˆ Kinematic free surface boundary condition
@W @
@t @ = 0 with W (; t )  = 0 (2.3)

This statement denes that no particle leaves the free surface. Particles that are part
of the free surface at (;  ) at time t are still part of the free surface at ( + d;  + d )
at t + dt .

6
2.2 Wave Theory

ˆ Dynamic free surface boundary condition


@
@t + g  W = 0 on =0 (2.4)

It is stated by the dynamic free surface condition that the pressure at the bottom
 = 0 is constant. Neglecting the convective terms of the Bernoulli equation this
expression can be derived.
ˆ Finally, the linearised and generalised free surface boundary condition:
@  !2
@ + g  = 0 on =0 (2.5)

Combining the kinematic, dynamic surface condition and the presence of an harmonic-
ally oscillating velocity potential  with the circular frequency ! yields to the general-
ized free-surface boundary condition (see Faltinsen [9]). With the complex, harmonic
approach for the velocity potential  = 'e i!t it is one of the conditions used to solve
Equation 2.1.

Considering these boundary conditions together with the Bernoulli approach the Laplace
equation can be solved and derives:
^
 = W k!(k )  cosh( k (d +  ))
sinh(kd )  sin(k !(k )t ) (2.6)

with the regular wave amplitude ^W , the circular frequency


!(k ) = kg  tanh(kd ) (2.7)
p

and the wave number


2
k=  (2.8)

Introducing the velocity potential  to the dynamic free surface boundary condition the form
of the water surface of a regular linear wave can be described as a function of location x
and time t :

W (; t ) = ^W  cos(k !(k )t ) (2.9)

Further references and a thorough description of the linearisation of the Laplace equation is
provided by Clauss et al. [5].

Irregular waves - sea states Surface elevations oshore usually do not have the shape of
one regular linear wave. It is rather an entire sea state consisting of several superimposed
regular waves. This is called an irregular sea state. It is a composition of multiple, linear
sinusoidal waves, each individual wave with its own amplitude, frequency and phase. The

7
2 Background Hydrodynamic Analysis

surface elevation at an arbitrary location can be described as the sum of dierent regular
waves:
n
 W i (t ) = ^W cos(!i t i )
X
i
(2.10)
i =1

A Fourier analysis enables the identication of individual values of the wave components
such as the wave amplitude ^Wi from a time trace of an irregular wave. Amplitudes and
frequencies of individual wave components of an entire sea state can be summarised by
an energy density spectrum where the energy density S (! ) is a function of the circular
frequency ! . The wave energy density S (! ) multiplied with the constants as there are the
uid's density  and the acceleration due to gravity g results in the energy per unit area of
the wave within the frequency interval ! . As ! ! 0 the wave energy density is dened
to:

S W W
(!i )d! = 21 ^W2 i
(2.11)

An important statistical parameter is the spectral moment mn .


Z 1
mn = !n  S W W
(!)d! (2.12)
0

The zeroth moment m0 is directly related to the surface elevation in time. The variance of
the surface elevation at a particular location  2 (x;t ) equals the zeroth moment.
Pn
 ( x; t ) 
 (x; t )
2 Z 1
  (x;t ) =
2
W
i =1 W
n 1
i W
= m0 = S  (!)d! W W
(2.13)
0

The mean of the surface elevation is given by (x; t ) where  (x; t )i indicates one time step.
Parameters which are commonly used to dene a sea state are the peak period and the
signicant wave height. The signicant wave height is directly related to the zero spectral
moment and is dened as:

HS = 4 m0
p (2.14)

At the frequency where the energy density spectrum peaks the circular peak frequency !p is
dened. The more commonly used peak period Tp = 2=!p can be easily converted. Other
periods used to dene a sea state are the averaged wave period T1 = 2 m 0 and the mean
m1
zero up-crossing period TZ = 2 m
q
0.
m2

8
2.3 Vessel Model

Analytical Sea States Simulated sea states are modelled using the analytical formulation
of the JONSWAP spectrum by Houmb and Overvik [14].The spectrum's ordinate denotes
to
g2 a
SW W (!) =  5   exp
5!P4
 
(2.15)
! 4
4!
where !P is the peak frequency, the peak enhancement factor, being a constant de-
pending on the signicant wave height HS and
(! !P )2
 
a = exp 22!P2 (2.16)

0:07 where !  !P
(
= (2.17)
0:09 where ! > !P
The constant factor can be calculated by the spectrums zeroth moment (see Equa-
tion (2.12)).

(
= m
HS =4)2
(2.18)
0 W W

To meet the required HS , scales the discrete spectrum. The peak enhancement factor
is chosen according to [7] if it is not dened by conducted model tests used for validation.

2.3 Vessel Model

Modelling of environmental loads and corresponding vessel responses is explained in the


following sections.

2.3.1 Response in Waves

Assuming a linear superposition of vessel response per frequency, the common distinction
between loads due to incident waves and loads due to the structure's motion is made (see
Faltinsen [9, ch. 3, pp.39]).

ˆ Wave excitation loads on structures being restrained from oscillating, are calculated
according to Froude-Krylo and diraction forces and moments. Froude-Krylo forces
FKr are derived from the velocity potential of the incident wave 0 . Integrating the
pressure on the structures surface Sb in the incident velocity eld yields the Froude-
Krylo forces.

FKr = 
Z
0 n~  dS (2.19)
@t
Sb

9
2 Background Hydrodynamic Analysis

Here the vector n~ points into the uid and the potential 0 corresponds to Equa-
tion (2.6). Loads due to the presents of a hydrodynamic compact3 structure, so
called diraction forces and moments caused by deection and reection of incident
waves, are considered introducing the diraction potential 7 . Following Chakrabarti
[4] the diraction potential is also called scattering potential.
ˆ Forces and moments acting on the structure, when it is obliged to oscillate in any DoF
at excitation frequency, are described by the additional and individual velocity potential
per DoF, called radiation potential (see Chakrabarti [4, ch.8.1, pp.330]). As soon as
these potentials are derived for a given structure, they are considered as the added
mass and potential damping per exciting wave frequency.

Summarizing the total velocity potential at a point within the uid due to the presents of a
compact moving structure in waves is given by
6

 = 0 + j +  7
X
(2.20)
j =1

where j indexes the six DoFs (see Chakrabarti [4, p.331]).


Additionally to boundary conditions mentioned for 0 in section 2.2 the boundary on the
structures wetted surface is needed
@
s_ T n = (2.21)
@n
where the velocities of the structure's CoG are given by s_ T = [x;_ y;_ z;_ p; q; r ] and the
vector
n
 

n = (2.22)
rn
with r the vector from the LSA origin (CoG) to the surface. In order to solve the dened
boundary value problem the direction of propagation needs to be dened for the potentials
one to seven. This is done considering the Sommerfeld radiation condition
p @
lim R @R  i j = 0 j 2 1; :::6; 7

(2.23)
R!1

p
where  = eigenvalues and i = 1. It states that at a innite, radial distance R from
the structures centre (origin LSA) the scattered/diracted and radiated potential need to
vanish.

Linear Transfer Functions Decomposition of linear forces due to the radiated waves of
a oscillating structure delivers hydrodynamic mass and damping coecients. Since these
inertia and damping forces depend on the structures relative acceleration and velocity to the
3 The structures characteristic diameter is larger than 20% of the wave length.

10
2.3 Vessel Model

uid they are considered as the structures response. Obtaining Response Amplitude Oper-
ators (RAOs) this response (radiation potential) is divided by the excitation (incident wave
and the diraction potential). Such RAOs describe the structure's response per unit wave
amplitude and summarize the solution of the rst order boundary value problem. Results
for the six DoF and dierent angles of encounter are presented in section 3.3.

2.3.2 Second Order Wave Forces

As long as viscous damping is properly considered, wave induced ship motions can be reas-
onably predicted by linear potential theory. But in order to design a mooring or DP system,
forces of second order need to be considered. Second order forces and motions can be
approximated by a Taylor expansion, the pressure can also be expressed as a rst order term
p(1) and a second order term p(2) oscillating about a mean p(0) .
The pressure at a point within the uid of a known velocity potential  can be determined
using the Bernoulli equation. The second order term expansion denotes to

p =  r
1 ~ @ 
2 (2)
 2  X r ~ ~ @  (1)
 
:
2
(2) (1)
(2.24)
2 @t @t
According to Journée and Massie [16, Eq. 9.37] this is just one contribution to the calcu-
lation of second order forces. Following Chakrabarti [4, pp. 260] wave induced forces of
second order have dierent causes. Combination of rst order terms like the relative wave
height r as well as the squared velocity within the Bernoulli equation contribute to second
order forces. Using the method of direct pressure integration on the hull Journée and Massie
[16, Eq. (9.52)] distinguish two main contributions.
1. Integration of the second order pressure p(2) (see Equation 2.24) about the mean
wetted surface S0 .
2. Integrating the rst order pressure over the instantaneous wetted surface s results in
a force proportional to the relative wave height r squared.

F~ =
I
1  n~  dl
2 g r
(2) (1) 2
(2.25)

r

wl

Here, the normal vector n~ is pointing from the structure into the uid in the FRA.
The initial surface integral about s degrades to an integral about the waterline (see
Equation 2.25). Drawing on rst order RAOs, r in Equation (2.25) can be substituted
by the transfer function, thus the wave amplitude ^ and the force depends just on the
incident wave amplitude squared.
Pinkster [25] showed that the main contribution to second order forces are to be expected
from the relative surface elevation along the hull. Based on this, contribution from second
order potential (see Equation 2.24 second term) is neglected here on, where as for situation
in shallow water and consideration of forces also in the vertical plane the second order
potential is signicant. According to Faltinsen [9, p. 145] added resistance is equivalent

11
2 Background Hydrodynamic Analysis

to the longitudinal second order wave force component. The added resistance in waves is
caused by products of two rst order terms (see Boese [3]). The time invariant mean is of
this second order forces is commonly known as the mean wave force of individual regular
waves or the wave drift force.

Quadratic Transfer Function In a long-crested irregular sea state the second order force
occurs as a time invariant function, mainly due to the relative wave elevation. Looking at
pairs of waves with frequency !1 and !2 the analogy of a bi-chromatic wave and its envelope
can be drawn. A wave representing that envelope oscillates at a lower frequency than !1 or
!2 . This low frequency waves cause additional forces and following linear superposition also
signicant responses in the low frequency band. This contributions mainly draw on dierence
of frequency pairs !1 !2 called dierence frequencies. The second order force due to this
contribution can be calculated by discrete summation over all frequency combination.
N X
N
F~ (2) = i(1) j(1) P~ij cos((!i !j )t + (i j ))
X

i =1 j =1
N
N X
+ i(1) j(1) Q~ ij sin((!i !j )t + (i j ))
X
(2.26)
i =1 j =1

The vectors P~ij and Q ~ ij are known as Quadratic Transfer Functions (QTFs) since they
are independent of the wave amplitude. For example the second order force component
due to relative wave elevation of the QTFs is determined solving the integral along the
waterline (see Equation (2.25)). While substituting a time dependent formulation for r(1)
into Equation (2.25), addition theorem of trigonometry are needed to express the second
order force Equation 2.26 as a function of dierence frequencies (!i !j ) and actually sum
frequencies. Thus originally also Pij and Pij+ need to be distinguished. Similar developments
can be made for the other contributions to the second order forces (see Chakrabarti [4,
ch.7.2.1.1.2, pp.260] for detailed derivation of contributions). As soon as P and Q consider
all contributions, Equation (2.26) is valid for the total second order force. The QTF matrix
can be calculated in frequency domain after the boundary value problems are solved for the
rst odrer potential and if necessary also the second order potential. In directional sea states
the QTFs also depend on the incident wave direction. Equation (2.26) does not consider
sum frequencies, since these high frequency excitations are more relevant analysing e.g. a
Tension Leg Platform (TLP) where resonance frequencies are higher due to the systems
stiness (see Faltinsen [9, p. 133]). Forces calculated at these dierence frequencies, are
often referred to as slowly varying forces.
The double summation in Equation (2.26) is relatively time consuming and can be simplied.
In order to compute the QTF, ANSYS Aqwa makes use of the approximation by Newman
[20], as long as the contribution of the second order potential is not required. As explained
by Faltinsen [9, pp. 157] Newman's approximation implies that Qij = Qji = 0 and the
summations decrease to one since Pij = Pji = 0:5(Pii + Pjj ), thus low frequency forces
can be calculated by means of the main diagonal of the QTF matrix. Since mean wave
forces and moments do not have a phase information Pii = Tii is required.

12
2.3 Vessel Model

During a time domain simulation ANSYS Aqwa calculates in-stationary second order forces
based on this transfer functions and the instantaneous wave amplitude. Figure 2.3 b) gives
an example of the QTF in yaw, which qis subject to !i and !j and therefore plotted with
isolines of absolute QTF value Tij = (Pij )2 + (Qij )2 .Since the calculation takes advant-
age of the approximation by Newman the plotted matrix is symmetric.

a) b) 1e 4
1.0 4.8

0.9 4.2
[rad/s]

0.8
2

3.6
0.7

2]
W
3.0

0.5 g LPP
Tij
0.6

[
2.4

QTF Yaw Tij in


0.5
1e 4 c)
1.8
0.4 0.00
0.25 1.2
[1/ 2W]

0.3
0.50 Pii
10T1, i 0.75 M3FF 0.6
0.2
|M3FK|
|M3 | 0.2 0.4 0.6 0.8 1.0
7
1= 2 [rad/s]
4 2 0 0.10.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0.0
[1/ 2W] 1e 3 1 [rad/s]

Figure 2.3: Absolute QTF of dierence frequencies in Yaw at angle of encounter


 = 210 .a) Comparison of the matrix's edge !1 = 0:1 rad=s with linear
moment. b) symmetric matrix of absolute QTF. c) Comparison of matrix
main diagonal and absolute mean wave moment by far eld method. All
moments are made dimensionless by 0:5g rLP P .

Averaging Equation (2.26) in time only the matrix main diagonal contributes, and the mean
wave force, is extracted by taking the QTF real part Pii . Using the QTF to calculate
mean wave forces is referred to as the near eld method, since it makes use of the direct
pressure integration along the hull. Although, mean wave (drift) forces are usually calculated
by means of the far eld method, as long as only components in the horizontal plane are
needed. In Figure 2.3 c) the QTF real part Pii is plotted together with the mean wave
force calculated by the far eld method, since the mean does not has a phase delay. Here
the time derivative of the momentum conservation within the structures surrounding uid
is considered. The boundary surfaces are the structure's surface Sb , the lateral surface of
the bounding cylinder S1 at R ! 1, the free surface SF and the sea bed SB . Applying the

13
2 Background Hydrodynamic Analysis

generalized Gauss theorem the volume integral of the momentum G decreases to a surface
integral over the boundaries described above.
@G @
@t n~ + rUn ds
ZZ  

@t =  (2.27)
S

n~ denotes the positive normal pointing outside of the uid and Un the normal velocity of the
boundary surface (see Faltinsen [9, p. 134 ]). The denoted velocity potential is subject to
the incident wave as well as the diracted and radiated waves by the structure.
Figure 2.3 c) compares both methods. Good agreement is obtained, thus that in section
3.3 only mean wave forces calculated by the far eld method are presented. In case !1
or !2 approaches zero the QTF can be compared against the linear excitation moment
in form of Froud-Krylo M3F K and diraction moments M37 (see Figure 2.3 a)). This
examples indicates up to almost ten times higher linear forces than second order forces.
Although, already Remery and van Oortmerssen [28] showed that slow varying forces should
be considered designing mooring spreads. Since DP systems are only capable to counter
act motions at low frequencies, wave excitation forces in time domain will be considered
applying QTFs here after.

2.3.3 Wind Model

Wind loads are determined using force coecients in the three DoFs of the horizontal plane
in the hybrid reference axes. These are given as functions of the relative angel of encounter
=  and the absolute relative wind velocity U . According to DNV GL [6] a one minute
mean value for stationary wind loads is considered.
F ( ) F ( ) M ( )
C1 ( ) = jU 1 jU
W ind
C2 ( ) = jU 2 jU
W ind
C6 ( ) = jU 3 jU
W ind
(2.28)
W ind W ind W ind W ind W ind W ind

Coecients are available for a load case with seven monopiles fastened orthogonal to the
longitudinal axis on the main deck. As illustrated in Figure 2.4 seven transition pieces nd
their place upright on the aft deck for installation of oshore wind farm foundations. A set
of coecients is determined using numerical calculations based on the Reynolds Averaged
Navier Stokes Equations (RANSE) by Maximiano and Schrijvers [19]. The method and
results are presented in section 3. The fact that the crane boom is in rest position, is
expected to have less inuence on the resulting yawing moment and side force due to
wind.

2.3.4 Current Model

When the vessel's DP system can maintain position and currents are low, relative velocities
are small, according to Fossen [11, p.138] jU j < 2 m=s . Available current coecients are
determined by Maximiano and Schrijvers [19] solving the RANSE. These are provided for a

14
2.3 Vessel Model

Figure 2.4: General view of the computational grid above the SWL. Blue lines
show the grid resolution on the SWL and black lines the surface mesh on
the geometry. [19]

reference velocity of 2 kn and have a quadratic dependency on the velocity. For time domain
simulations they are implemented following Equation (2.28). Validating the numerical model
against model test results from step tests (see section 4.4) have shown that damping in all
three DoFs is underestimated. To match step tests presented in Figure 4.2,4.3 and 4.4 linear
damping is added (see Equation 2.29). Entries of B(1) and B(2) are compared in Table 3.3.
The additional damping is necessary since current coecients by Maximiano and Schrijvers
[19] do not provide forces depending on the yaw rate r . Adding also linear damping in surge
and sway contributes to stabilize the model already at small velocities.

Xu 0    0 Xjuju Xjv jv    0
   
 0 Yv Yr   Yju ju Yjv jv 0
B =
(1)
 ..

 . ... ..  B
(2)
=  
.. . . . ..  (2.29)
. . .
0 Nv    Nr Njuju Njv jv    0
For simulations of step tests in calm water the software does not take the frequency depended
damping or added mass into account, since no waves are present. Therefore also a frequency
independent added mass is introduced, to take inertia eects into account.

Xu_ 0    0
 
 0 Yv_ Yr_ 
M = 
A

.. ... ..  (2.30)
. .
0 Yv_    Nr_

15
2 Background Hydrodynamic Analysis

Added masses and linear damping contributions are taken from calculations in frequency
domain (see section 2.3.1) from the largest available period T = 60 s . Values are given
in section 3.4. Overshoots could be reduced taking inertia forces into account where as
settling times elongate.

2.3.5 Roll Damping

Although, the implemented DP system acts only in three DoFs roll damping in time domain
calculations, that are based on potential theory, needs to be properly corrected. That way
calculated accelerations are still available and reasonable, since the system of dierential
equations (see section 2.5 Equation 2.50) is solved in six DoFs.

Since roll damping is underestimated by potential theory it needs to be adjusted by means of


additional calculations or test results. Here the damping of role motion by radiated waves is
computed by the potential panel code Aqwa Line (see section 3). To consider the missing
viscous damping, results from roll decay model tests are used. Fitting model test data, a
quadratic approach is made to distinguish between governing linear phenomena like damping
by wave radiation and quadratic ones like damping by viscous drag forces or eddy creation.
Therefore, the work done by the damping force acting on the vessel in one roll period is
considered as the sum of a linear and quadratic part and is set to equal the equivalent work
just by a linear force.
ZT ZT
_ (t}) '_ (t )dt = [b| L'{z_ (t}) + b| Qj'_ ({z
b| '{z t )j'_ (t})]'_ (t )dt (2.31)
0 Fequv : 0 FL FQ

The roll motion and its velocity are modelled as harmonics with amplitude 'a and angular
frequency ! .

'(t ) = 'a  cos(!t ) (2.32)


'_ (t ) = !'a  sin(!t ) (2.33)

Substituting Equation 2.32 and 2.33 into Equation 2.31 the work done to damp the roll
motion within one period is described by
ZT ZT ZT
! 'a b sin (!t )dt = ! 'a bL
2 2 2 2 2
sin (!t )dt +! 'a bQ j sin(!t )j sin2(!t )dt
2 3 3
(2.34)
|0 {z } |0 {z } |0 {z }
=! =! 8=(3! )

16
2.3 Vessel Model

Solving the integrals Equation 2.35 gets linearised. Dividing both sides by !'2a the equi-
valent damping value b, which is obtained from model test, is isolated.
8
b = bL + 3 !'a  bQ (2.35)
| {z }
FIDP

The total damping derived from model tests is presented in Figure 2.5. Since additional
damping for the diraction analyses will be considered frequency independent b is only a
function of the roll amplitude 'a . For purposes discussed in this thesis only small roll

1e10
1.2
Total damping b in [kgm2/s]

1.0

0.8

0.6

0.4

0.2
2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0
Roll amplitude a in[ ]

Figure 2.5: Total roll damping derived from model tests at load case one.

angles will be relevant, hence a roll damping of the model tests for an amplitude of 'a = 2
is considered. In order to modify the roll damping used during the sea keeping analyses with
Aqwa Line the quadratic part as already marked in Equation 2.35, is taken as a additional
frequency independent damping (FIPD).

kgm2
b('a = 2 ) bL (!' ) = (1:9  109 8:64  108 ) s (2.36)
9 kgm
2
= 1:036  10 s (2.37)

This is a suitable approach, since the inuence in the low frequency part, where hydrostatic
moments related to the vessels Metacentric Height (GM) and also in the high frequency

17
2 Background Hydrodynamic Analysis

range, out of the resonance, where radiation damping and inertias are the governing factors,
is negligible. Usually damping values are given as portion of the linear critical damping bcr it .
2g  r  GM
bcr it = (2.38)
!'
= 2gH2O  91765 m2 12:44 m  14:8 s
3

= 5:408  1010 kgm


2

s
Fraction of critical damping are given in the Table 2.1. Note that the quadratic part is
linearised.

Table 2.1: Classication of roll damping contributions with respect to critical


damping for the LC dened in Table 4.1.

Damping %=bcr it
linear 1.598
quadratic 1.916
total (model test) 3.513

2.4 Dynamic Positioning System

The vessel will be controlled in three DoFs - surge, sway and yaw. In automation and control
theory the vessel and its hydrodynamic behaviour are referred to as the plant and the DP
system is evaluated separately as the controller of the plant. The control loop is closed
by a negative feedback of the plant's output, the vessel's position (see Figure 2.6). The
plant is indicated by the green box on the right called ANSYS Aqwa since forces due to
the environment are considered using the existing framework. The other box named DP
user_force shows the comprises the external function with its subroutines.

For a DP system ltering the feedback signal is important to save the actuators from wear
and tear due to oscillating loads in the band of wave frequencies where !  !P . To maintain
position the controller also needs to distribute forces to dierent actuators of the propulsion
units. The considered vessel has in total eight actuators being six azimuth propellers and
two tunnel thrusters. The arrangement is illustrated in Figure 2.2 (a). At the bow azimuth
thrusters are retractable, since enough power is installed at the aft for transit. More detailed
information on the propulsion per thruster regarding open water diagram or comparable has
not been available for this thesis. No machinery or mechanical dynamics are considered.
Commanded thrusts by the allocator are assumed to be available immediately. Nevertheless,
azimuth rates and a constant acceleration of the rpm is considered in both directions and
implemented in the external force.

18
2.4 Dynamic Positioning System

Setpoint Const. added Wind


Position & Heading mass and damping Coe s.

r
e FPID Thrust Time domain y Position
Controller Integration [; ; ]
Allocation

_ jmax ; T_ max
j

yf Thruster Wave forces


Dynamics QTF
ANSYS Aqwa

Filter

DP user force
Figure 2.6: Block diagram of time domain calculation indicating the interface
for the external function.

2.4.1 Feedback Filter

The actual DP system on the vessel will have an observer to calculate the prevalent posi-
tion and velocity, based on position measurements and an appropriate mathematical model
following control theory (see Fossen [11]). Since position and velocity are available in a
time domain simulation by integrating the dierential equation explained in section 2.5 no
estimates are needed. Only the time step and the modelling assumptions in the previous
sections inuence the results. An online transformation to frequency domain every time
step to apply dedicated band lters would slow down the computation and is not required,
since the eect of a moving time average is similar to a low pass lter [23]. For a moving
time average the time step and the time constant are necessary to dene the number of
time samples to be averaged. This denitely causes a phase delay between actual position
and average position, which is compensated by a sti controller.

2.4.2 Controller

Three individual controllers are considered to provide the two required forces and the yawing
moment to correct for the error between ltered position and set point. The error e (t ) is
dened in the hybrid system for each DoF by adding the negative feedback y (t ) to the
set point r .

ej (t ) = rj yj (t ) for j 2 fg1; 2; 6g (2.39)

19
2 Background Hydrodynamic Analysis

The set point can be considered as a function in time r (t ) for step tests or to approach an
installation side with a load already on the hook. This mode comparable to a joystick mode
is not further developed. Here the set point is meant to be constant and dened at the PoI
including the yaw angle = 0.

The controllers are chosen to have proportional, integral and derivative gain, thus repres-
enting PID controllers. In a rst attempt the controller was designed to have only a P and
D part comparable to a spring damper system with natural period and a percentage critical
damping. But the back draw of the stationary deviation from the set point remained and an
I part had to be introduced. Although the plant itself already provides an integrative part,
since its transfer function can be modelled by an IT1 element, controllers with additional
integrative contribution are necessary. Characteristic for an IT1 element is a linear increase
of the output for t ! 1 as a response to a step input. A so rough comparison of con-
trollers being capable to work in dierent dedicated modes on a full scale vessel is provided
by Rabanal et al. [27]. The gains of the here implemented PID-Controller are determined
setting the proportional gain and two time constants, the reset time ("Nachstellzeit") I
and the rate time ("Vorhaltezeit") D (see Lunze [18]).
K
KI =  P
j
j
; K D = KP   D
j j j
(2.40)
I j

The control law denotes to


Zt
dej (t )
FP ID = KP ej (t ) + KI
j j j
ej (t )dt + KD dt
j
for j 2 f1; 2; 6g (2.41)
0

where the controllers output is either a force in [N ] or the yawing moment in [Nm], thus
the unit of the corresponding gain is dened as [N=m] or [Nm=rad ] respectively.

KP = diag 3:2  105 5:7  105    0    1:5  109 (2.42)




 D = diag 30:0 30:0    1    30:0 [s ] (2.43)




 I = diag 150:0 150:0    1    160:0 [s ] (2.44)




The gains and time constants are scaled to full scale from the controller used during model
tests. In case no gains are available they can be estimated following Nienhuis [22]. and van
den Boom, H. J. J. and Nienhuis [32] who suggest a procedure based on the vessel's added
mass in the controlled DoFs. The amplitude and phase response of the open controllers are
shown in Bode4 plot Figure 2.7. In the low frequency range the controller shows the typical
integrative behaviour, compensating a stationary set point deviation well. The derivative part
is causing the behaviour in the high frequency range. An innite gain for high frequencies
is usually not desirable, but in combination with the feedback lter, the gain approaches a
constant value.
4 Hendrik Wade Bode (1905  1982), electrical engineer

20
2.4 Dynamic Positioning System

1011
Magnitude
109

107

10 4 10 3 10 2 10 1 100
90
[ ]

60 Surge
30 Sway
0 Yaw
30
60
90
10 4 10 3 10 2 10 1 100
[rad/s]
Figure 2.7: Bode plot comparing open controller's frequency response for the
three individually controlled DoFs. Gains are taken from model test, trans-
lated to full scale.

2.4.3 Allocation

The previous described controller provides forces and a moment, that are necessary to keep
the vessel in place. Now these are distributed to the actuators, which is called allocation.
This task usually needs to be described by an optimisation problem, since there are more
free variables than constraints. Here 6  2 + 2  2 + 3 = 17 free variables are chosen.

ˆ In total 12 individual forces for x and y direction per azimuth thruster: fxj , fyj with
j 2 f1; :::; 6g,

ˆ one force in y direction per tunnel thruster: fxj with j 2 f7; 8g

ˆ and three slack variables s = [sx ; sy ; s ]T to allow the optimiser a deviation to the
commanded force by the Proportional Integral Derivative (PID) controller.

The choice of variables is some how predened by the choice of the actual optimiser. It needs
to be considered beforehand what kind of constraints are required, and what a reasonable
cost function should penalise.

21
2 Background Hydrodynamic Analysis

2.4.4 Optimisation Problem

An optimisation problem is dened in order to nd the individual thrust and azimuth angle
commands per actuator (see section 2.4.3).

min
u;s
u Wu + s Qs
T T
(2.45)
subject to Bu = FP ID + s (2.46)
Au  b (2.47)
1s1 (2.48)

It is the optimisers task to minimize the cost function Equation (2.45) with respect to 14
sub forces summarized by u 2 <141 and the three slack variables denoted by s 2 <31 .
The matrices Q 2 <33 and W 2 <1414 are weights to ensure a high cost for missing
the commanded force, by Q >> W. No weights are applied between individual thrusters.
Minimizing the squared forces, has the advantage of achieving a solution that does not
favour uneven distribution of forces. Usually all thrusters should be used in the same range
of workload. The equality constraints are given by Equation 2.46 and represent the need
to match the commanded forces and yawing moment by the controller. Matrix B 2 <1414
contains the thrusters lever arms with respect to the CoR and can reallocate all individual
thruster forces to the two forces and one moment of, FAllocated 2 <61 . The inequality
constraint is introduced by Equation 2.47, in order to stay with in the physical boundaries
of each thruster. Therefore b 2 <141 and A 2 <1414 respectively contain the solution of
the last time step and the thrust rate to ensure a new optimum considering the maximum
turning speed and change in thrust per time step and actuator. The slack variable instead
is unbounded, but remembering the high weight Q it contributes to the cost function quiet
a lot. This kind of problem formulation is solved by an optimiser that can handle quadratic
cost functions.

2.4.5 Solver

The implemented solver uses quadratic programming methods combined with disjunctive
programming techniques. Choosing quadratic programming methods ensures to nd the
global minimum. Likewise it enables using quadratic cost functions that can minimize the
consumed power of all thrusters. For simplicity the power consumption is approximated
by the sum of squared thruster forces. Nevertheless, constraints can only be linear, thus
that the absolute maximum thrust needs to be linearised. This is done approximating the
circle with the radius fx2l + fy2l by a polygon. The easiest approach is done setting four
p

constraints per thruster (see Figure 4.1). Moreover the partial derivatives with respect to all
parameters of cost and constraint functions are needed since the optimiser uses a line search
method. Additional constrains are required to always dene a convex room of solution, that
saves convergence of the solver
The implemented solver is based on the Sequential Least Squares Quadratic Program-
ming (SLSQP) method and is developed by Kraft [17]. Before the actual problem is handed

22
2.5 Equation of Motion

over to the solver it is scaled by the maximum available thrust squared of an azimuth thruster
at the aft of the vessel. The accuracy of the solvers solution as well as the step size is ad-
justed to the scaled problem formulation. Regarding details on the solvers convergence
reference is given to Kraft [17].

Previously to the implementation of the quadratic programming method, the a simplex


optimiser following the algorithm by Nelder-Mead method was tested. Subroutines are
available from the numerical recipes.5 In principle it has the advantage of approaching the
minimum geometrically. Although constraints can only be implemented modifying the cost
function which became soon complex as more and more actuators were considered.
Therefore the quadratic programming algorithm is chosen, not least because deriving the
global minimum is always preferred.

2.5 Equation of Motion

Although only three DoFs in the hybrid reference system are controlled by the DP system
the equation of motion is solved in all six DoFs.

Mx = F
X
(2.49)
[MRB MA + MA(!)] x = FW av e + FW ind + FDP B + B(1)(!) + B(2)jx_ j x_ (2.50)
 (1) 

where

x_ = [u; v; w; p; q; r ]T (2.51)

Following Fossen [10, p.33] for surface ships at U  0, MA = MA T and MA strictly positive
denite. To acknowledge common sign convention of hydrodynamic derivatives for man-
oeuvring simulations, where e.g. Xu_ is generally negative MA is subtracted in Equation 2.50.
MA(!) denotes the frequency depended added mass due to the radiation potential. The ri-
gid bodies mass matrix with given inertias following Table 4.1 is given by MRB . No moments
of deviation are taken into account. The right hand side has the following contributions:
FW av e denotes the sum of second and rst order wave force. Either excitation forces on
the vessel are calculated using the full QTF of dierence frequencies as explained in
section 2.3.2 page 12 and the rst order forces, or Froud-Krylo and diraction forces
as well as the corresponding mean wave forces calculated by the far eld method are
considered.
FW ind considers the loads due to coecients dened in section 2.3.3. The coecients are
based on the relative wind velocity and loads are calculated multiplying the coecient
with the velocity and its absolute value.
5 http://numerical.recipes

23
2 Background Hydrodynamic Analysis

FDP represents the allocated forces in surge and sway as well as the yawing moment by the
DP system.
B contains linear, frequency independent damping (see section 2.3.4),
(1)

B the quadratic, frequency independent damping (see section 2.3.4) and


(2)

B (!) the linear, frequency dependent damping derived from the six radiation potentials
(1)

(see section 2.3.1).

24
3 Numerical Model

The solution of problem statements above either boundary value problems in section 2.3.1
or integration of the system of dierential equations in section 2.5 or the solution of the
optimisation problem in section 2.4.3 is calculated by the potential code Aqwa version 17
which is part of the ANSYS Inc. software framework. The developed external function is
also tested for compatibility with version 18.1 and works with both versions.
Aqwa oers to perform the hydrodynamic analysis in stages.
Stage 1-3 are called Aqwa Line and solves the boundary value problem by calculating source
strengths on a panelled surface. Results are given in form of the rst and second order
excitation forces and the hydrodynamic added mass and damping coecients.
Stage 4-5 comprise the analyses of hydrodynamic responses in sea states, called Aqwa Drift.
At this stage the developed external function to model the DP system is included. The
forces per time step are calculated and Equation (2.50) is integrated in time with a
predictor-corrector method.

3.1 Panel Method

Forces on the structure are calculated following potential theory as introduced by Newman
[21]. Considering the software's manual [1, ch. 4.1.2, pp.36], a database of pulsating
Green's function is used to calculate the sought potentials. The lower limit of evaluated
Green's function is ! = 0:001 g=d . To obtain the source distribution over the mean wetted
p

surface, the Hess-Smith constant panel method is used on quadrilateral and triangular panels.
This implies constant potential and source strength on each panel. For detailed derivation
of the numerically solved Fredholm Integrals reference is made to [1].

3.2 Mesh and Grid Dependency

To analyse the crane vessel for its hydrodynamic response the hull surface needs to be
meshed. The geometry is provided as a .step le by the new building department of
GeoSea NV. Minor simplication are performed. Tunnels for bow thrusters are closed.
The displaced water mass is included in the total displacement mentioned in Table 4.1.
Additionally small single faces at the aft part of the vessel are merged with neighbouring
faces. This allows the meshing algorithm to calculate a homogeneous mesh. This steps
are done with DesignModeller which is part of the Ansys Inc. framework. The surface
mesh, dening the spacial distribution of source strengths is generated by Ansys Workbench.
Mesh settings are documented in Table 3.1. Figure 3.2 gives an impression of the mesh's
resolution. Most of the panels are rectangular. Some interfaces are bridge with triangles.
Force calculation in frequency domain is analysed for convergence with the amount of panels
to discretise the hull. Figure 3.1 shows convergence of mean wave forces and moments with
an increasing number of panels in bow and aft quartering seas heading on port side. Forces

25
3 Numerical Model

Table 3.1: Mesh parameters.

No. Description No. of di. Panels max. Panel size


01 coarse 1816 5m
02 medium 2812 4m
03 med. ne 6617 3m
04 ne 10107 2m

are compared against results of the next coarser mesh. The mesh no. 02 is compared
against no. 01 and so on. The criterion is calculated averaging over a frequency range of
! = [0:1; 1:0]rad=s .
Z1:0
Fj (; !)d! for j 2 f1; 2; 6g (3.1)
0:1

The relative error documented in Figure 3.1 is always based on the next coarser grid. Except

F1 at 240.00 F2 at 240.00 M3 at 240.00


F1 at 300.00 F2 at 300.00 M3 at 300.00
% difference to coarser grid

10

20

30

40

50

60
2812 6617 10107
No. of diffracting panels.
Figure 3.1: Mesh convergence of mean wave forces in frequency domain in bow
and aft quartering waves for the three controlled DoFs.

26
3.3 Results Frequency Domain Calculation

for mean forces in sway and moments in yaw at waves inciding from 0 and 180 (not
plotted here) convergence could be shown. In that direction diverging forces are acceptable,
since absolute values in that DoFs are small. Although Figure 3.2 does not show obvious

02 medium fine mesh

01 coarse mesh

Figure 3.2: Mesh comparison with mesh no. 01 at the bottom and 02 on top
of the picture

dierences the mean wave forces dier signicantly. The mesh used for calculation further
on is chosen to be the medium ne one no. 04 since another renement does not change
mean wave forces signicantly.

3.3 Results Frequency Domain Calculation

The background on wave induced forces is summarized in section 2.3.1. Running Aqwa from
stage one to three the linear response, mean wave forces and the QTFs are calculated. This
hydrodynamic basis is present in the following.

Linear Wave Forces Calculating the frequency dependent added mass and damping of the
vessel by estimating the radiation potential on each panel, RAOs based on linear superposi-
tion of the frequency components are derived for the motion in six DoFs. Figure 3.3 shows
the vessel's responses per unit wave amplitude ^W for the LC of Table 4.1 in operational
condition at 10:5 m draught and a water depth of 40 m. Responses are plotted in steps of
 = 30 . Similar motion RAO can be obtained for pairs of bow/aft quartering directions.
The roll resonance frequency at !' = 0:42 rad=s corresponds well to the period derived
from model tests (see Table 4.1).

27
3 Numerical Model

ˆ Highest responses are obtained in bow and aft quartering waves.


ˆ Yawing moments reach a maximum for 60 and 120 .
ˆ Surge and sway motion decline fast to almost half the wave amplitude for frequencies
smaller than 0:5 rad=s

Mean Wave Forces Since rst order inciding waves and rst order vessel motion cause
second order wave forces with a mean 6= 0 resulting forces and yawing moments are shown
below. Loads are computed using the method of momentum conservation as described in
section 2.3.2.Here M3 as the mean wave moment in yaw is equivalent to the main diagonal
jTii j in Figure 2.3. Ordinates in Figure 3.4 are made dimensionless, dividing by the waters
density , acceleration of gravity g and the corresponding power of the displacement r.
The oberservations are:
ˆ In bow and aft quartering conditions almost symmetric forces and moments act on
the hull.
ˆ Yawing moments in bow quartering waves at  = 150 result in larger loads, compared
to waves inciding at aft quartering waves (  = 30 ).
ˆ In all three DoFs a rst peak is reached in the area of roll resonance, before the mean
forces and moments approach a constant value.
ˆ Oscillations at ! > 0:1rad=s are of numerical nature, due to the far eld method.
ˆ Even in beam seas the hull encounters a mean force 6= 0 in surge direction (see
Figure 3.4 a) ). 0:6 rad=s .

28
3.3 Results Frequency Domain Calculation

0 60 120 180
30 90 150
1e1
1.6

Roll in [ /m]
Surge in [m/m]

4 1.4
1.2
3 1.0
0.8
2
0.6
1 0.4
0.2
0 0.0
5 1.0
Sway in [m/m]

Pitch in [ /m]

4 0.8

3 0.6

2 0.4

1 0.2

0 0.0
1e 1
1.2
Yaw in [ /m]
Heave in [m/m]

5
1.0
4
0.8
0.6 3

0.4 2

0.2 1
0.0 0
0.25 0.50 0.75 1.00 1.25 0.25 0.50 0.75 1.00 1.25
in [rad/s] in [rad/s]
Figure 3.3: RAOs of motions in six DoFs for dierent angles of encounter  at
10:5 m draught and 40 m water depth.
29
3 Numerical Model

0 60 120 180
30 90 150
1e 4 a)
]

4
g 2W
2 F1
[
Surge in

4
1e 3 b)
2.0
g 2W]
2 F2
[

1.5
Sway in

1.0

0.5

0.0

c)
1.5 1e 4
2]
W
2 M3

1.0
g LPP
[
Yaw in

0.5
0.0
0.5
1.0
1.5 0.2 0.4 0.6 0.8 1.0 1.2
in [rad/s]
Figure 3.4: Dimensionless mean wave forces a) surge b) sway and c) yaw mo-
ment per squared wave amplitude ^W for dierent angle of encounter  at
2

10:5 m draught and 40 m water depth.


30
3.4 Current and Wind Coecients

3.4 Current and Wind Coecients

Coecients used to determine wind and current loads as explained in section 2.3.3 and 2.3.4
are calculated by means of the open-usage CFD code ReFRESCO 6 solving the RANSE.
Maximiano and Schrijvers [19] decided for the following assumptions to derive the coe-
cients.

ˆ The vessel is xed.

ˆ A stationary ow and forces are considered, thus that unsteady eects in the wake
are neglected.

ˆ The inow velocity is uniform; for the wind calculations the atmospheric boundary
layer is foreseen in vertical direction.

ˆ Eects of a free deforming water surface are not considered, thus a at slip wall is
implemented.

ˆ The water depth is meant to be large enough that no shallow water eects are expec-
ted.

The grid for current calculations has 33 million cells and is rened for the centre skeg at
the aft and the bilge keels. The maximum y + is smaller than ve, thus no wall function is
used. Almost 31 million cells are necessary to mesh the structures above the SWL. The
deck layout is mainly shaped by six monopiles and corresponding transition pieces as well as
the deckhouse, crane pedestal and boom. The maximum y + is smaller than two. Numerical
settings are:

ˆ Time scheme - steady state calculation

ˆ Spacial discretization - higher order Quick scheme

ˆ Turbulence model - SST k -!

ˆ The small y + values allow calculations without wall functions.


The boundaries of the cylindrical domain (radius with respect to vessels main frame 4 LP P ,
bottom for at 500 m or atmosphere at 4 LP P respectively) are dened as follows - inow:

ˆ uniform velocity [u; v; w ]T = Ur ef [cos(); sin(); 0]T ,


ˆ zero normal gradient for the pressure,

ˆ eddy viscosity ratio equals one and the turbulence intensity is set to 1%.
6 http://refresco.org

31
3 Numerical Model

At the outow boundary a constant zero pressure and for all other quantities a zero gradient
boundary is applied. At the water surface a free-slip boundary with a normal velocity of zero
is set. For wind calculations the water surface is meant to be a no-slip boundary. The wind
prole is dened by a reference wind speed Ur ef = 30 m=s at a height of ten meter above
the SWL.
z
 0:135
U (z ) = Ur ef z10 (3.2)

Initial conditions are chosen according to the inow boundary settings. Iterative convergence
is understood to be reached considering the L2 norm where residuals decrease by third order
and more. Following Maximiano and Schrijvers [19] this is commonly considered to be
sucient convergence for steady state calculations.

Determination of Coecients The CoR to determine the yawing moments is located on


the vessel's centre line at the main frame. Reference values of coecient determined by
Computational Fluid Dynamics (CFD) are summarized in Table 3.2.
F
Cj () = 0:5  U 2 j L  z with j 2 f1; 2; 6g (3.3)
r ef PP r ef

Applied quadratic coecients depending on the absolute, relative current velocity and its
angle of encounter Cl () are presented in Figure 3.5 and are calculated following Equa-
tion (3.3). Results of currents from  > 180 on port-side are mirrored and shown in
dashed lines and empty markers. Symmetry along the vessel's longitudinal centre is proven.

Table 3.2: References values wind and current coecients.

Environment Ur ef zr ef
Wind 30 m=s 51 m
Current 2 kn 10:5 m

Current Since coecients calculated by CFD do not supply loads due to yaw motion,
alternatives are sought. The idea is to take the radiation damping in yaw due to yaw motion
B66
(1)
at the lowest available frequency. In order to categorise this value, comparison is also
presented in the other two DoFs. The third reference is taken from the SIMMAN 7 workshop
2008. A Very Large Crudeoil Carrier (VLCC) has been investigated for its manoeuvring
performance. The carriers main dimensions are LP P = 320 m, B = 58 m,D = 20:8 m and
a block coecient of cB = 0:81. A summary of the workshop is given by Stern et al. [31].
Table 3.3 compares the coecients of dierent dependency. all are calculated following
Equation (3.3) and are multiplied by 103 . However, comparison is dicult, since dierent
7 Reference tanker beeing investigated in the SIMMAN workshop 2008. http://simman2008.dk

32
3.4 Current and Wind Coecients

10 C1(0 180 ) 10 C1(195 345 )


C2(0 180 ) C2(195 345 )
10 C6(0 180 ) 10 C6(195 345 )
1.00
C( ) [ ]

0.75
0.50
0.25
0.00
0.25
0.50
0.75
1.00
0 20 40 60 80 100 120 140 160 180
Angle of encounter [ ]
Figure 3.5: Quadratic current coecients calculated by CFD.

Table 3.3: Comparison of linear and quadratic current coecients per unit ve-
locity. All coecients are multiplied by 103.
DoF X Y N
Order u ju ju v jv jv vvv r jr jr rrr
Bll(1) (0:1 r ads ) 68:90 16:60 34:45
MARIN CFD 23:00 734:00
KVLCC1 2:20 1:50 24:00 2:23 74:7 3:32 0:27 1:25

assumptions lead to that coecients. Data for the VLCC are given for a quasi stationary
situation at a transit speed of 15:5 kn. Even if they are multiplied by the transit speed, thus
all values in Table 3.3 are based on a unit velocity of 1 m=s , coecients in any DoF or by
any method do not seem related. Anyhow, a damping due to yaw was missing and is applied
by adding the radiation damping at ! = 0:1 rad=s . Moreover the linear damping in surge
and sway is also added frequency independent, since the vessel on DP turns about a remote
CoR it starts to drift, thus that u and v are 6= 0. A comparison of the dierent applied
coecients, additionally to the quadratic coecients from Maximiano and Schrijvers [19],
is done for the yaw step test (see Figure 4.4). It is observed that linear damping and added
mass result in a numerical model closer to model tests.

33
3 Numerical Model

The corresponding added mass coecient are taken into account as well like described in
section 2.3.4 page 14. The main diagonal in Equation (3.4) is given dimensionless, relative
to the vessel's displacement in kg or the inertia kgm2 (see Table 4.1).

0:082 0  0
 

0 0:623 2:0  108 kg 


M = 

A .. ... .. (3.4)
. .

2:0  108kgm2   

0 0:385
The performance during step test could be improved by minimizing overshoots considering
the frequency independent added mass and damping.
Finally it needs to be concluded, that largest velocities and acceleration are observed during
step tests. Since the main goal is the simulation of the DP system in irregular waves, where
only much smaller velocities occur, the choice of current coecients is not outstanding
(compare appendix A, Figure A.1).

Wind Wind coecients resulting from CFD calculations are presented in Figure 3.6 and
derived from [19]. A wind Moment due to yawing like for relatives velocities in water, is
not added. Comparison is made to coecients calculated following Blendermann [2]. The

C1 DRI0103 C2 DRI0103 10 C6 DRI0103


C1 CFD C2 CFD 10 C6 CFD
1.00
C( ) [ ]

0.75
0.50
0.25
0.00
0.25
0.50
0.75
1.000 20 40 60 80 100 120 140 160 180
Angle of encounter [ ]
Figure 3.6: Comparison of wind coecients.

chosen vessel for comparison is identied by its number DRI0103 [2, p. 25] and has the deck

34
3.5 Time Domain Calculation

layout of a drilling vessel with the tower around the mainframe. The equivalent wind prole
is applied. Loads predicted following Blendermann [2] are generally larger except for the
surge direction. Here the inuence of the monopiles (length>BW L ) orientated orthogonal
to the longitudinal axis make the dierence. Calculations regarding station keeping are only
made with coecients taken from CFD calculations.

3.5 Time Domain Calculation

Equation (2.50) is integrated in two steps following a predictor corrector method (see AN-
SYS Inc. [1, ch.13.8, p.161].

Predictor Forces on the right hand side are calculated considering the initial position and
velocity or the ones of the last time step. Initial position is given by [0 ; 0 ; 0 ] also all angles
set to zero. Velocities are set to zero as well, thus the calculations starts at rest position.
prevalent acceleration is won substituting the total force in Equation (2.49) and solving for
x. An intermediate new velocity x_  and position x is derived.
x_  (t + dt ) = x_ (t ) + x(t )  dt (3.5)
dt 2
x (t + dt ) = x + x_ (t )  dt + x(t )  (3.6)
2
(3.7)

The intermediate result in predictor step are denoted by  .

Corrector Again the total force is calculated but now based on the intermediate position
and velocity and an intermediate force F (t + dt ) is derived. This is again substituted in
Equation (2.49) to obtain accelerations at t + dt . Finally the new position and velocity can
be derived integrating the new acceleration.
x(t ) + x(t + dt )  dt
x_ (t + dt ) = x_ (t ) + (3.8)
2
x(t + dt ) = x + x_ (t )  dt +
2x(t ) + x(t + dt )  dt 2 (3.9)
6
(3.10)

Time Step Dependency For time domain simulations it is essential to use a dedicated
time step, that is proven to do not inuence the results. A set of time steps is investigated
and standard deviations of the allocated force in surge and sway as well as the yawing
moment are compared in Table 3.4. Standard deviations of forces and moments are made
dimensionless by g; ; r and LP P . Additional standard deviations of the resulting position of
the PoI in the hybrid axes as well as of the heading are given for decreasing time step. Time
step dt = 0:2 s is chosen as accurate enough since standard deviation of allocated forces by

35
3 Numerical Model

Table 3.4: Convergence of forces and motions with dt by standard deviations.

dt x y  F 1 F 2  M3
[s ] [m] [m ] [m ] r]
[ 104
g r]
[ 104
g
[ r
g LP P
]
105

0.020 0.313 0.410 0.284 1.534 3.677 9.147


0.050 0.313 0.410 0.284 1.538 3.679 9.148
0.100 0.309 0.413 0.284 1.546 3.688 9.229
0.150 0.307 0.416 0.284 1.540 3.700 9.303
0.200 0.308 0.420 0.286 1.546 3.738 9.341
0.300 0.308 0.447 0.291 1.555 3.951 9.236
0.400 0.314 0.498 0.307 1.557 4.010 9.100

the user_force and resulting positions do not change much, as the dt is further decreased.
Figure 3.7 proves this choice since with time steps dt < 0:2 s standard deviations of motions
and most forces/moment, do not change signicantly compared to the next longer time
step. Note that this is an exemplary study for approximately half an hour duration, since
the maximum result le size is easily reached with short time steps. The computation time
% difference to next longer time step

6
x
y
8
F1
10 F2
M3
12 0.30 0.20 0.15 0.10 0.05 0.02
dt [s]
Figure 3.7: Convergence of forces and motions for a decreasing time step in-
dicated by standard deviations  for x; y; (solid) and allocated forces
F1 ; F2 ; M3 (dashed) in an irregular sea state of HS = 3 m, TP = 8 s ,
= 1:55 and an angle of encounter  = 210 .

36
3.6 Postprocessing

increased from approximately some minutes for the largest time step to almost a quarter of
an hour for the shortest one.

3.6 Postprocessing

Methods used to assess the vessel's operability are summarized below.

Footprints of Trajectory Trajectories in time domain calculations are presented in section


5 by a footprint of the deviation in x and y and a distribution of the heading angle. The
footprint is calculated transforming the Cartesian coordinates in a polar system. The nal
footprint is the result of a two dimensional histogram with 36 bins in angular direction and
a radial step of 2 mm. The last bin ends at r = 2 m. Summing the lled bins along each
direction and stopping e.g when 99% of all trajectory points in that direction are summed
up, gives the probable radius per angular bin that will not be exceeded in 99% of time. Since
this is signicantly inuenced by outliers and events are also evaluated independent of their
occurrence. The MPM is used as the criterion for DP envelopes (see section 5.5).

Most probable maximum Denition is exemplary shown for the surge motion.

xmax = 2  x ln2Nx
r
(3.11)

Here x denotes the standard deviation of x (t ) and Nx the number of cycles determined
analysing the corresponding time trace for its zero up crossings TZx . Signals analysed for
their zero crossings do have a zero mean.

37
4 Implementation of Dynamic Positioning System

4 Implementation of Dynamic Positioning System

The DP routine is written to interfere at the user_force interface of the ANSYS Inc. Aqwa
time domain simulation called Aqwa Drift. The external function or user_force needs to
be provided as a DLL which is called twice per time step by the Aqwa main routine, since
it uses a predictor corrector time integration scheme. The steps are called stages and are
distinguished by the user_force input variable integer::stage 2 f1; 2g. Just before the
system of equations is solved the user_force is asked for forces to be considered that step
of integration. Generally the user_force runs in three dierent modi. The integer::mode
2 f0; 1; 99g is passed on from the main routine. Initially mode=0 is only called once, mode=1
every time step twice and mode=99 once after the last time step is completed.
The header of the external function's template by the software developer provides the
declared variables and is provided by programme code C.1 in the appendix. The main
routine of the developed external function is only provided on the digital appendix called
uf_sourceCode.f90. The variables I_CONTROL,R_CONTROL can be used to pass on integers
and real values from the Aqwa Drift conguration le to the user_force. The following
parameters need to be set. A template AD_template.dat is provided on the appended disc

R_CONTROL[19] sets the time constant F for the feedback lter in seconds.
R_CONTROL[93,...,96] set the PoI by three coordinates in the FRA and the heading
that has to be maintained.
R_CONTROL[87,...,92] set the delta for a step test in the FRA. Entry 89
provides the dt relative to the rst time step.

and is structured in decks. Within deck zero options to use the full QTF matrix and to
call the external function are set. Deck one to eight are not used before in deck nine the
frequency independent added masses and damping values are given. Deck ten holds the
interface variables I_ and R_CONTROL as well as the current and wind coecients of second
order. Wind and current speeds are dened in deck 11 before deck 13 takes the information
about wave spectra. The time step is set in deck 16. The other pre declared variables
denote as follows. POSITION contains the vector to the prevalent CoG's position and the
structures Euler angles in the FRA. VELOCITIES are also passed on by the main routine in
the FRA and are given for the vessels CoG. The CoG's coordinates [0 ; 0 ; 0 ] are stored
for t = 0 and provided by COG in the FRA relative to the vessel's transom, the centre line
and the SWL.
Returned to the main routine are the variables FORCE, ADDMASS and ERRORFLAG. The re-
turned added mass is always zero where as the force passed back to the main routine has
three entries 6= 0 for the rst two DoFs and the sixth, representing the allocated yawing
moment. As position and velocity are provided for the vessel's CoG also forces, meant to
be returned to the main routine, need to be translated to that point and act along the FRA.
Declarations of variables are documented in programme code C.1.
Additional declaration for implemented subroutines are explained by comments within the
les on the digital appendix. To compile and link the source code written in Fortran the

38
4.1 Feedback Filter

gfortran gcc-7.2.0 compiler, part of the mingw-64 project8 (version 5), is used. The
script to compile and link the function built-user_force64.bat is also provided on the
appended disc.
The results calculated by the external function are saved to the hard drive each time step.
Files are rst temporary located at C:\temp\user_force64\ before they are copied to the
current working directory at the end of each simulation. It is strongly recommended to run
every computation in a dedicated directory, hence existing les will be overwritten. The
vessel's position by the CoG, the set point, the vessel xed velocity and the ltered motion
of the CoG in the FRA is stored in pos.txt. Values are given in m or rad respectively.
Forces calculated by the controller, and allocated forces by the optimiser routine are saved
to forces.txt. Values are given in N or Nm respectively. The allocated thrusts and
azimuth angles per thruster, as dened in Figure 2.2 (b), are stored in thruster.txt each
time step. Angles are given in deg between 180 and +180 . The last two columns of
that le provide also the number of iterations of the optimiser and the status. Regarding
dierent statuses reference is given to Williams [36]. The le log.txt provides even more
information on the gains and optimiser runs.

4.1 Feedback Filter

According to section 2.4.1 the position lter is implemented as a moving average in time. In
a temporary le a list with the inertial trajectory and Euler angles [; ; ; '; ; ] is stored.
The list's length corresponds to the lter's time constant and the chosen time step. In
predictor stage, the ltered position from the last time step is used. Each corrector step,
the list gets updated with the new position. Before the list gets updated, the actual ltered
position is calculated by averaging the list's entries unweighted in time. Programme code
4.1 shows the lter's implementation.

1 IF (STAGE.EQ.2) THEN ! Update Moving average list only in corrector stage


,! (2)
2 ! Just for initializing
3 POSITION_avg(:,1) = (/ (0.), (0.), (0.), (0.), (0.), (0.) /)
4 ! Kick first entry in the list.
5 last_Positions(1:avg_timesteps-1,:) = last_Positions(2:avg_timesteps,:)
6 ! Append current position at the end of the list.
7 last_Positions(avg_timesteps,:) = POSITION(:,1)
8 ! Write updated list to file, to save for next time step
9 open(unit=50,file="c:\\temp\\user_force64\\lastFRA.dat", &
10 form='formatted',status='REPLACE')
11 ! Summation over all entries in list
12 DO i=1,avg_timesteps
13 WRITE(50,*) (last_Positions(i,j),j=1,DOF)
14 POSITION_avg(:,1) = POSITION_avg(:,1) + last_Positions(i,:)
15 ENDDO
8 Project reference: MinGW-64 original project MinGW for 64 bit. Ready build gcc compiler are hosted on
sourceforge.net

39
4 Implementation of Dynamic Positioning System

16 close(50)
17 ! Average
18 POSITION_avg(:,1) = POSITION_avg(:,1)/avg_timesteps
19 ELSEIF (STAGE.EQ.1) THEN ! Moving average predictor stage (1)
20 ! List is not updated, but position is always averaged.
21 POSITION_avg(:,1) = (SUM(last_Positions, DIM=1))/avg_timesteps
22 ENDIF

Programme Code 4.1: Position Filter.

4.2 Controller
Commanded forces for the allocator are calculated by a PID Controller according to section
2.4.2. Gains are hard coded and chosen like explained above. Forces and the yawing moment
are returned to the main routine as well as the error in all DoFs and the controllers integral
part. Where as the subroutine PID is called in both stages, error and integral are only
updated after the corrector stage. Although the set point in the conguration le is given
at the PoI it is translated to the CoG, thus the error is calculated there and in the FRA.
This way the calculated position by the Aqwa main routine does not need to be translated to
the PoI. Since no big deviations from the set point in yaw are expected this is a reasonable
approach. Programme code 2.41 shows the way of implementation.

1 SUBROUTINE PID(sub_POSITION, sub_Setpoint, &


2 sub_Error, sub_Error_old, &
3 sub_Integral, sub_Integral_old, &
4 sub_FORCE, sub_dt, sub_PID_Gains)
5 IMPLICIT NONE
6 REAL, INTENT(IN) :: sub_dt
7 REAL, DIMENSION (6), INTENT(IN) :: sub_POSITION, &
8 sub_Setpoint, sub_Error_old, sub_Integral_old
9 REAL, DIMENSION (6), INTENT(OUT) :: sub_FORCE, sub_Error
10 REAL, DIMENSION (6), INTENT(OUT) :: sub_Integral
11 REAL, DIMENSION (6,3), INTENT(OUT) :: sub_PID_Gains
12 REAL, DIMENSION (6) :: sub_Derivative, sub_Kp, sub_Kr, sub_Ki
13 !!!!!!!!!!!!!!!!!!!!!!!! GAINS !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
14 sub_Kp = [3.203e+05,5.678e+05,0.,0.,0.,1.522e+09] ! P
15 sub_Ki = [2.142e+03,3.797e+03,0.,0.,0.,9.499e+06] ! I
16 sub_Kr = [9.655e+06,1.712e+07,0.,0.,0.,4.918e+10] ! D
17 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
18 sub_Error = (sub_Setpoint - sub_POSITION) *sub_Kp
19 sub_Integral = sub_Integral_old + sub_Error*sub_dt*sub_Ki
20 sub_Derivative = (sub_Error-sub_Error_old)/sub_dt *sub_Kr
21 sub_FORCE = sub_Error + sub_Derivative + sub_Integral
22 END SUBROUTINE PID

Programme Code 4.2: PID Controller

40
4.3 Allocation

Limiting the commanded force and moment is necessary since the allocator, which is
connected in series to the controller is bounded to maximum thruster forces and a max-
imum yawing moment. Summing up an absolute, maximum force ahead, transverse and the
absolute, maximum moment by all thrusters can raise is used to limit the PID command.
Necessity is tested by two independent if statements (see row 3 and 8).

1 FORCE_pre_abs(1,1) = ((FORCE_pre(1,1)**2 + FORCE_pre(2,1)**2))**0.5


2 FORCE_pre_ang(1,1) = ATAN2(FORCE_pre(2,1),FORCE_pre(1,1))
3 IF (FORCE_pre_abs(1,1).GE.FORCE_max(1,1)) THEN
4 WRITE(99,*) 'Lidding abs. force :', FORCE_max(1,1), ' < ', FORCE_pre_abs
5 FORCE_pre(1,1) = FORCE_max(1,1) * COS(FORCE_pre_ang(1,1))
6 FORCE_pre(2,1) = FORCE_max(1,1) * SIN(FORCE_pre_ang(1,1))
7 ENDIF
8 IF (ABS(FORCE_pre(6,1)).GE.MOMENT_max(1,1)) THEN
9 WRITE(99,*) 'Lidding abs. moment:', MOMENT_max(1,1), ' < ', FORCE_pre(6,1)
10 FORCE_pre(6,1) = SIGN(MOMENT_max(1,1),FORCE_pre(6,1))
11 ENDIF

Programme Code 4.3: Limiting PID output.

4.3 Allocation
Commanded forces from the controller are distributed to the individual actuators. Two
free variables per azimuth thruster and one free variable for a tunnel thruster are chosen
(see section 2.4.3). The optimisation algorithm collaborates all these in one vector x. To
implement the dynamic thrust regions also x_old is introduced but misses the information
for the slack variables. Like it is implemented in programme code C.2 row four, the slack
variables are initialized with zero each optimiser run. More detail on the used optimiser and
its implementation can be found in section 2.4.5 and here below on page 42. The optimiser's
source is provided in the digital appendix in the modules beginning with slsqp_ ...f90 or
can be downloaded at reference [36]. Since the optimiser is expected to match allocated
and the commanded force tau_pid, it is proven to save simulation time by running the
allocation algorithm only in the corrector stage (stage=2), maintaining results unaected.
In appendix B the controller and allocator forces and moments are compared for a time
domain simulation of an exemplary sea state. The resulting footprint stays unaected.
Only in sway the allocator can not aord the commanded force by the controller and shows
slightly lower forces in the low frequency range (see Figure B.1). Before the optimiser is
initialized the problem is scaled by the squared maximum available thrust of azimuth thruster
one. Either the input or the step size needs to be scaled to the problem. Scaling the input
aims numerical stability of the solver.

Thruster dynamics Thruster rates in angular as well as in radial direction are considered,
adjusting the optimiser's over all bounds each run. Initial values are set in the source le
craneVessel_62_thrustutils.f90. The optimiser oers the x_u and x_l variable for
upper and lower bounds for each component of the free variable vector x. As shown in row

41
4 Implementation of Dynamic Positioning System

10 to 20 of programme code C.2 in the appendix, either the thrust rate or the thrusters
maximum thrust sets the upper bound x_u. Like illustrated in Figure 4.1 e.g. the lower
limits per azimuth thruster x_l[j] and x_l[j+1] are set to fxjmin and fyjmin . This is done
for all actuators up to index 14 of x. The azimuth rate is checked after the optimiser has
found the best solution. It is checked and probably reset if change rates are exceeded in
the subroutine checkAzimuthRates dened in craneVessel_62_thrustutils.f90. The
subroutine in row 51 translates x and y forces to the polar notation and compares azimuth
angles of last and current time steps (see subroutine fxfy2rpmAlpha also implemented
in module craneVessel_62_thrustutils.f90). It then resets the new angle if neces-
sary, but keeps the new allocated thrust command. The sum of thruster forces and the
yawing moment is recalculated in sumAllocated, which is agian provided by the module
craneVessel_62_thrustutils.f90.

fxj
j
T m ax

fxjmax = Tmax
j
= 2
p
t +1
fxj;max

t +1
fyj;max T j; t fyj;mint +1
fyjmax = fxjmax

fxj;mint +1

fyj

Figure 4.1: Dynamic thrust region and overall thrust limits. (Source: author)

Optimiser The optimiser is based on quadratic programming techniques (see sec-


tion 2.4.5). It can handle a quadratic cost function and linear constraints. A mod-
ern Fortran version of the SLSQP algorithm is provided by Williams [36]. The open
source code comes in modules, which are compiled by the gfortran compiler like the
other code of the external function. The cost function is implemented as shown by
slsqp_costfunc_craneVessel_62_slack.f90 on the digital appendix.Equality constraints
need to be implemented rst, before inequality constraints. Since this optimiser already
provided the overall bound x_u and x_l no inequality constraints are needed. In case a
better denition of the maximum thrust region as shown in Figure 4.1 is needed linear
constraints can be implemented after the rst three equality constraints to meet the com-
manded controller forces and the yawing moment. A detailed description of how to dene
those constraints and translate them from the polar coordinate system to the here chosen

42
4.4 Validation

free variables per azimuth thruster in x and y direction is provided by Ruth [30, pp.74, ch.
4.4.2]. Since the optimiser works only with constraints that dene a convex environment
where it can converge and nd the minimum, easily sub allocation problems occur. As soon
as a thruster has two possible convex sub environments, where valid solutions for x can be
found combinations of sub allocation problems have to be optimised after each other and
nally evaluated for the global minimum.

4.4 Validation
System properties are investigated best performing step tests in the three DoFs the
system is controlled (see DNV GL [6, ch.6.1.3.10; p.42]). Here the controller's set
point resembles a step function with dened deltas. Step tests shown in Figure 4.2 to
4.4 show the behaviour of the entire closed control loop with negative feedback (section 2.4).

To judge simulation results model test data is analysed. Tank tests were conducted at the
Maritime Research Institute Netherlands (MARIN) by Heerink and Beks [13]. At a geometric
scale of 1 to 40 the water depth in the shallow water basin corresponds to 40 m full scale.
The position measurement was done by an optical system consisting of three cameras fol-
lowing a target mounted on the vessel. Control of the actuators was done by the DP system
of the actual propulsion supplier of the analysed crane vessel. Table 4.1 presents the tested
LC.

Table 4.1: Main particulars and parameters of the vessel's Load Case.

Particular Symbol Magnitude Dimension


Draught T 10.50 [m]
Displacement r 91765 [m3]
Length between perpendiculars LP P 206.30 [m]
Breadth moulded on SWL BW L 49.00 [m]
Longitudinal Center of Gravity (LCG) 0 102.28 [m]
Vertical Center of Gravity (VCG) 0 2.27 [m]
Metacentric height GM 1.94 [m]
Mass radius of gyration around x -axis Kxx 0 0:49 [BW L]
Mass radius of gyration around y -axis Ky y 0 0:27 [LP P ]
Mass radius of gyration around z -axis K 0 zz 0:28 [LP P ]
Natural period in roll motion T' 14.80 [s ]

The mass matrix for calculations with the external function is based on the given
particulars, the CoG and the radii of gyration from Table 4.1. The location of the Aft
Perpendicular (AP) is dened at 8:477 m in front of the transom.

43
4 Implementation of Dynamic Positioning System

Results from these tests are analysed to judge the performance of the external function and
the numerical model in general. Here, regarding the vessel's step response (see section 4.3)
and in section 5 the vessel's station keeping performance in waves (see section 5).
Step test parameters are summarised in Table 4.2. Out of rest position the DP set point is

Table 4.2: Overview tank step tests.

Test No. 02BT_03_ Description 


001_007_01 Surge step 1:0 m
001_008_01 Sway step +1:0 m
001_009_01 Yaw step +5:0 

set immediately to 1 m in surge or sway or to 5 in yaw direction. Steps have always been
conducted independently for every DoF. Corresponding test numbers are given in Table 4.2.
The vessel's response is analysed for overshoot and settling time.
All gures present the time traces of the heading xed motions of the PoI x and y and
the vessel's yaw angle. Model test show dierent raise and settling times depending on
the step size. One step test is depicted for validation and shown in the following gures.
This non-linear behaviour can be provoked by additional lters, virtual moving set points
that approach the actual set point or other limitations to the controllers output that are
not modelled by the external function. As long as relative velocities are small, step tests
computed with the developed DP system, show a linear response disregarding the step delta
(see section 2.3.4).
The yaw step shown in Figure 4.3 is performed around the PoI since position of x and y
almost do not change. Although, the vessel turns much slower during model test, larger
deviation from the set point in both DoFs are observed.
Speeds to approach the new set point in x or y direction shown in Figure 4.2 and 4.3 match
model test results best. Gains from model test work well with the numerical model, even
though not all characteristics of the full scale DP could be implemented in the numerical
model. In surge the settling time agrees well with derived times at model tests, although
the overshoot is predicted smaller by simulation. This could indicate a too soft model in
surge or sway. However in yaw direction the numerical model tends to be way to sti.
Generally, step responses with the controller settings explained in section 2.4.2 are satisfying,
and prove a reasonable model. Although, it is still to be shown that the numerical model is
not too sti or too soft in all DoFs, to match also model test results in waves, since step
test do not explain the model response to high frequent disturbances.

44
4.4 Validation

a)
1
x [m]

MARIN
AQWA
0

20 100 200 300 400 500 600 700 b)


y [m]

2 MARIN
AQWA
1
0
1
0 100 200 300 400 500 600 700 c)
[ ]

MARIN
6 AQWA
4
2
0
0 100 200 300 400 500 600 700
t [s]
Figure 4.2: Time traces of PoI in hybrid axes (a), b)) and the vessels yaw angle
during surge step test comparing model test (blue) and simulation results
(green).

45
4 Implementation of Dynamic Positioning System

a)
1
x [m]

1
MARIN
AQWA
20 100 200 300 400 500 600 700 b)
y [m]

2 MARIN
AQWA
1
0
1
0 100 200 300 400 500 600 700 c)
[ ]

MARIN
6 AQWA
4
2
0
0 100 200 300 400 500 600 700
t [s]
Figure 4.3: Time traces of PoI in hybrid axes (a), b)) and the vessels yaw angle
during sway step test comparing model test (blue) and simulation results
(green).

46
4.4 Validation

a)
1
x [m]

0
MARIN
AQWA
1 AQWA VLCC
AQWA B(1) = 0
20 100 200 300 400 500 600 700 b)
y [m]

2 MARIN
AQWA
1 AQWA VLCC
AQWA B(1) = 0
0
1
0 100 200 300 400 500 600 700 c)
8
[ ]

6
4 MARIN
AQWA
2 AQWA VLCC
AQWA B(1) = 0
0
0 100 200 300 400 500 600 700
t [s]
Figure 4.4: Time traces of PoI in hybrid axes (a), b)) and the vessels yaw angle
during yaw step test comparing model test (blue) and simulation results
(green). For the step test in yaw additional dierent current coecients
are tested. No linear coecients dashed dotted, and coecients from the
tanker by dashed lines.

47
5 Results

5 Results

The analysed crane vessel is supposed to install foundations or do decommissioning works


oshore. In order to assess the vessels operational capability at a given sea state, dierent
scenarios are simulated with the developed external function. The indicators that evaluate
the capability are the MPM motions in the horizontal plane. Inuences of the assumptions
on the station keeping performance are discussed and criteria to judge the operational
capability are given.
The considered LC is described in Table 4.1. Presented trajectories, envelopes or footprints
are always given for the PoI in the heading xed coordinate system. For installation
purposes it is easier to judge operability by deviation from the PoI, orthogonal to and inline
with the vessel's centre line. Therefore results are translated from CoG to the PoI and are
corrected only for the yaw angle , to obtain heading xed trajectories.

To evaluate simulations on DP in waves, model test results are analysed from Heerink
and Beks [13] (see also section 4.4). The conducted sea state, that serves for validation,
is summarised in Table 5.1. For this test campaign the vessel has only been exposed
Table 5.1: Overview model tests in irregular waves.

Test No. Duration HS Tp 


Description Type
02BT_03_ [h] [m] [s] [] [-]
006_001_01 Irregular 2 JONSWAP 3.0 3.0 8.0 210.0 1.55

to waves. In the 220 m long and 15:8 m wide basin, waves are generated at the far
short end. The vessel was located at approximately three quarters of the basin length
before the lattice type beach that absorbs the waves generated at the other short end
by a piston type wave maker. To realise dierent angles of encounter the vessel was rotated.

5.1 Inuence Filter Setting


In section 2.4.1 the feedback lter is introduced. To avoid oscillating commands to the
actuators at high frequencies, a reasonable time constant F is sought for the time average.
For simplicity the same time constant is applied for all three DoFs.
Three settings are investigated, where F = 1 s corresponds to almost no ltering of the
position signal. Sea state 2 from model tests is simulated for approximately 30 min with
F set to 8 and 16 s . Figure 5.1 shows the overall allocated forces in Figure a) and c)
and the yawing moment in d) transformed back from time domain to frequency domain. It
is observed that the time constant and the lter in general have a signicant inuence on
the allocated forces. For F = 8 s almost all energy in the region of the wave frequency
is ltered out. Figure 5.1 b) shows the nal inuence on the DP footprint. It is obvious
that a too long time constant causes delays, that let the actuators react too slowly, thus
that the station keeping performance decreases. For all other simulations F = 8 s is proven
to contribute to a stable controller-plant behaviour. During calculation of sea states with

48
5.2 Inuence Second Order Wave Forces

1e12 a) b)
2
Power spectrum F1 [N 2/s]

x in [m]
F =1s 99% Footprint F = 1 s
3.0 F =8s 99% Footprint F = 8 s
F = 16 s 99% Footprint F = 16 s
2.5 1
2.0
0
1.5
1.0
1
0.5
0.00 25 50 75 100 125 150 175 200 22 1 0 1 2
Period T [s] y in [m]
1e13 c) 1e16 d)

Power spectrum M3 [Nm2/s]


Power spectrum F2 [N 2/s]

2.00 F =1s 3.5 F =1s


F =8s F =8s
1.75 F = 16 s 3.0 F = 16 s
1.50 2.5
1.25
2.0
1.00
1.5
0.75
0.50 1.0
0.25 0.5
0.000 25 50 75 100 125 150 175 200 0.00 25 50 75 100 125 150 175 200
Period T [s] Period T [s]

Figure 5.1: Comparison of feedback lter settings by force power spectra in the
three controlled DoFs and DP footprints for three dierent time constants
F in sea state 2 (JONSWAP HS = 3 m; Tp = 8 s ; = 1:55).

durations comparable to tank tests, the chosen time constant and the lter technique in
general are conrmed.

5.2 Inuence Second Order Wave Forces


The inuence of calculating excitation forces due to waves with QTFs is already indicated
in section 2.3.2.
Both, time domain simulations considering wave forces using QTFs and just applying Froud-
Krylo and Diraction forces in combination with mean wave forces are conducted. The
sea state 2 is used as an example and simulated for three hours. The resulting yaw motion
of the vessel is transformed back to frequency domain, thus the power spectrum is shown in
Figure 5.2. Considering wave forces by QTFs results in larger yaw amplitudes at long periods
above T  40 s . The gures legend provides the standard deviation of yaw motion. Since
it corresponds to the zeroth moment of the plotted power spectra, larger over all motions
are conrmed, when wave forces due to QTFs are taken into account. It can clearly be
observed that considering wave forces at dierence frequencies is denitely conservative if
position keeping is addressed. Remembering that an azimuth thruster is usually restricted to

49
5 Results

0.0025

S ( ) in [rad 2/s]
With QTF = 0.25
Without QTF = 0.06
0.0020

0.0015

0.0010

0.0005

0.00000 20 40 60 80 100 120


T in [s]
Figure 5.2: Power spectra of yaw motion in bow quartering waves at 210 ,
comparing the inuence of wave forces by QTF in sea state 2 (JONSWAP
HS = 3 m; Tp = 8 s ; = 1:55).

rates of 720 per minute, a lot of energy of the yaw motion is located in that low frequency
region above 40 s (see Figure 5.2).

5.3 DP Footprint
For given environmental conditions like at conducted model tests presented in section 4.4,
DP footprints are the means of visualising the vessels trajectory. Background on how
these are determined can be obtained from subsection 3.6, page 37. Additionally, the
combination of the MPM motion in heading xed longitudinal and transverse direction give
a rectangular, which indicates the maximum expected motion.

Figure 5.3 compares footprints of model tests in irregular sea state 2 and simulation results
(see Table 5.1, page 48). Subgure 5.3 a) gives indication on the probability density function
of the vessel's heading.
ˆ As the result of a smaller standard deviation psi during model test, the modelled DP
seems less sti in yaw.
Statistics of the vessel's trajectory are compared in Figure 5.3 b). The following is ob-
served.
ˆ Generally, statistical predictions, by means of simulations with the external function
to maintain position, match model test results well.

50
5.3 DP Footprint

ˆ The rectangular of MPM motion in x and y of model test results in sea state 2 predicts
less motion in sway than in surge.
ˆ The trajectory calculated (green) indicates an opposing behaviour.
As per Table 5.2 the mean zero up-crossing period and standard deviation for the three
controlled DoFs are summarised for sea state 2. The zero crossing periods are determined

Table 5.2: Comparison of statistics of footprints in irregular waves given by


standard deviations and averaged zero up-crossing periods of horizontal vessel
motion and yaw angle in sea state 2.

HS = 3 m x y psi TZ x TZ y TZ
Tp = 8 s [m] [m] [m] [s ] [s ] [s ]
AQWA 0.34 0.40 0.26 30.6 65.3 43.5
MARIN 0.43 0.39 0.18 55.3 66.0 23.8

directly from individual time traces with a zero mean. Interpreting the periods of a single
DoF as individual natural periods for that particular sea state observations from Figure 5.3
b) are conrmed, since larger natural periods correspond to softer systems. The numerical
vessel model is less sti in yaw compared to model test results but stier in surge. Motions
in sway direction generally match, considering the MPM motion in Cartesian coordinates.

51
5 Results

PDF( ) [ ] 1e 2 a)
2 = 0.18
= 0.26
1

0
2.0 1.5 1.0 0.5 0.0 0.5 1.0 1.5 2.0
[ ]
b)
2
x [m]

99% Footprint MARIN


99% Footprint AQWA

22 1 0 1 2
y [m]
Figure 5.3: DP footprint in irregular sea state 2 (JONSWAP HS = 3 m; Tp =
8 s ; = 1:55) at  = 210 , bow quartering seas. Green and blue lines
compare simulation and model test results based on a three hours sea state.
Subgure a) compares the heading's probability density function and Sub-
gure b) 99% condence footprints of the heading xed position of the PoI.
Dashed lines indicate the combination of MPM motion in x and y .

52
5.4 Statistics Thrust Allocation

5.4 Statistics Thrust Allocation


Since a stable thrust allocation method is implemented, that corresponds to the full scale
system, thruster behaviour simulated by the external function and during model test is
confronted in the following. Presented thrust is given with respect to the maximum
capacity of each actuator.

Figure 5.4 and 5.5 compare statistics of thrust allocation from model tests and simulation.
Sea state 2 is simulated for approximately three hours. From model tests only the feedback
signal of each thruster is analysed, since the external function does not distinguish between
allocated command and feedback.
Mean values of thrust and the mean orientation are indicated by markers. In radial
direction simulated mean thrust per actuators is in good agreement with the model test.
However, the mean thrust direction diers for ve of six thrusters by almost 15 . Since,
resulting motion of the vessel is still comparable to model tests (see Figure 5.3), the bow
thrusters are used more during simulation as indicated by about 30% higher mean thrusts
in Figure 5.4.

T[%] BTF AQWA T[%] BTF MARIN


T[%] BTA AQWA T[%] BTA MARIN

40 20 0 20 40
% of Tmax
Figure 5.4: Comparison of mean and standard deviation of bow tunnel thrust in
percentage of maximum available thrust from simulation (green) and model
test (blue) in sea state 2 (JONSWAP HS = 3 m; Tp = 8 s ; = 1:55.

Areas represent the corresponding standard deviation in percentage of maximum thrust and
angular direction. For comparison red areas mark the spoil zones, that are only considered
at tank tests. A spoil zone denes an angle range that is unfavourable regarding thruster
hull or thruster-thruster interaction. Both is not considered by the external function, since
the process of allocation would need much more development. Reference is given to Ruth
[30]. Since that green and red angle ranges overlap only slightly for the fore retractable
thruster RFF and the port-side thruster PS, results of the external function are still reliable.
All six simulated thrusters show a wide angular standard deviation compared to the model
test records. Although, the area indicated by feedback signals from model tests always lays
within the simulated one.

53
5 Results

MARIN: PDF( ) 100 % [ , T] [± , ± T]


AQWA: PDF( ) 100 % [ , T] [± , ± T]
RFF vessel fore a) RFA vessel fore b)
30 330 30 330
60 300 60 300

90 270 90 270
5% 5%
10 % 10 %
120 15 % 240 120 15 % 240
20 % 20 %
PS 150 210 c) SB 150 210 d)
30 330 30 330
60 300 60 300

90 270 90 270
5% 5%
10 % 10 %
120 15 % 240 120 15 % 240
20 % 20 %
PSM 150 210 e) SBM 150 210 f)
30 330 30 330
60 300 60 300

90 270 90 270
5% 5%
10 % 10 %
120 15 % 240 120 15 % 240
20 % 20 %
150 210 150 210
vessel aft vessel aft
Figure 5.5: Comparison of thrust allocation statistics for azimuth thrusters in
sea state 2 (JONSWAP HS = 3 m; Tp = 8 s ; = 1:55). Comparison of
simulation (green) and model test (blue) by mean values (dots). Areas
give  the standard deviation of azimuth angle [ ] and thrust in percentage
of maximum thrust. Red areas indicate spoil zones. Lines show PDFs of
azimuth angles .
54
5.5 DP Envelope

Additionally, Figure 5.5 provides PDFs of the thrusters azimuth angle. Presented distribu-
tions are scaled to full '( )d = 100. Mean angle and the angle with the maximum
R

probability indicate skewed distributions.


Generally the propulsion system is not used to capacity. The areas given by the standard
deviation of the azimuth angle and the feedback thrust signal show no saturation of any
thruster. Since MPM vessel motion in the horizontal are already close to the operational
limit, it can be concluded, that the propulsion system will not be the limiting factor.

5.5 DP Envelope
In order to evaluate the vessel's operational capability referring to horizontal motions, it is
a common approach to calculate the maximum sea state, where the vessel can maintain
position within a given tolerance. Such studies can also be asked by the classication society
or the marine warranty surveyor and are summarised by a DP envelope plot.

180 170 16
12.0 0 15
01
10.5 40
MPM x 9.0
13

MPM y 7.5
01

vWmin 6.0
20

vWmax 4.5
11

3.0
0 100 90 80 70 6

1.5
0.0
[ ]
vW [m/s]

0
50

40
30
20 1
0 0

Figure 5.6: DP Envelope for a current velocity of 1:03 m=s . Envelopes ensure
a MPM surge (blue) or sway motion (orange) of less then 2 m at that sea
state dened by the radial plotted wind speed vW .

55
5 Results

Parameters dening the sea state are related. Here relations according to the table by
IMCA [15, p.2] are considered. The wave spectrum is a JONSWAP spectrum and follows
the formulation in section 2.3.1. The peak enhancement factor is chosen to 3:3. To
calculate the maximum sea state, wind wave and current are aligned. A DP envelope for a
constant current of 1:03 m=s is shown by the maximum one minute mean wind velocity in
Figure 5.6.
The criteria are the MPM motion smaller than 2 m in x and y , that ensure safe installation
operations. For a symmetric vessel only half of the envelop is calculated. Even though the
deck layout might not be symmetric along the longitudinal axis, the wind coecients for
loads from port and starboard side do not dier much (see section 3.4). Wind loads are
based on a constant one minute mean velocity.

Table 5.3: MPM motion in surge and sway at the PoI for a maximum one
minute wind velocity corresponding to Figure 5.6.

 MPM x MPM y
[] [m] [m ]
0.0 1.727 0.038
15.0 2.436 1.174
20.0 1.977 0.963
30.0 1.962 1.446
45.0 1.542 1.854
60.0 1.642 1.960
75.0 1.775 1.733
90.0 0.789 1.729
105.0 1.713 1.923
120.0 1.631 1.967
135.0 1.545 1.772
150.0 1.563 1.794
160.0 1.692 1.170
165.0 1.968 1.364
180.0 1.687 0.052

The blue envelope gives the maximum sea state with MPM surge motion of less than 2 m.
Respectively, the orange line ensures a MPM sway motion of less then 2 m. Obviously
the vessels surge motion is less in beam seas and the vessel's sway motion is less in head
or following seas, thus at those angle of encounter good operability can be predicted.
However, the maximum wind velocity for both motions dene the performance indicator
to assess operational capability. Since a shielding eect of the vessel can be expected at a
draught of 10:5 m, installation on the vessel's lee side might be favourable when objects
need to be lifted through the splash zone. Short waves with length in the range of the
lifted object's characteristic length might be diracted on the vessel's lu side. For small
deviations from head seas up to   150 the sway motion is not the limiting factor. At
this angle of encounter the same maximum wind velocity is derived for both motions in

56
5.5 DP Envelope

the horizontal plane, hence it is found to be the favourable corresponding relative angle of
encounter =  = 30 for oshore installations with environmental loads from one
direction. Although, in an environment heading in from beam direction, the allowable wind
speed decreases to 4 m=s which corresponds to a sea state of HS = 1:58 m TP = 5:86 s .
The corresponding MPM motions are summarised in Table 5.3. For  = 15 the
minimum wind velocity for the set of simulation was chosen to fast, thus that the criterion
in surge can not be fullled. The actually operational capability regarding MPM hori-
zontal motion of the vessel in that direction will be smaller, speaking of a less high sea state.

In order to get reliable data per direction the wind speed should only be increased by small
steps. Moreover a duration of at least three hours needs to be simulated per wind speed.
Although, it occurs that the criterion is missed in a particular sea state but is again met
in a slightly higher wind speed. Here the maximum wind speed is obtained as soon as the
allowable MPM is exceeded. Actually, also dierent seeds have to be simulated or even
longer duration than three hours are necessary to get a statistically better database. The
amount of simulation is already reduced, since the range of wind speeds is adjusted for
the angle of encounter. The wind speed increment is constant but for loads around beam
conditions only smaller wind speeds are simulated. Therefore the maximum and minimum
expected wind speed is plotted in Figure 5.6, to indicate if the performance could even be
better or worse if the minimum wind speed was considered too slow or too fast at a particular
angle of encounter.

57
6 Conclusion

6 Conclusion

A crane vessel operating on dynamic positioning is modelled. The time domain simulation
software Aqwa Drift as part of the ANSYS Inc. framework is extended by an external
function in the form of a DLL written in Fortran. The library provides the developed model
of the vessel's DP system for three DoFs. Subroutines are implemented for
ˆ the position feedback lter,
ˆ the PID controller,
ˆ and an interface to an optimised allocation algorithm by quadratic programming.
Previous to the development of the external function, the vessel is analysed for its response
in waves and sea states. Frequency domain calculations with the panel code Aqwa Line are
conducted. Convergence of mean wave forces is obtained with increasing mesh resolution.
Considering QTFs of dierence frequencies, in order to calculate the second order wave
forces, is proven to be necessary, since larger motions are observed in later time domain
simulations.
Environmental inuences by wind and current are modelled using a system of coecients
derived from CFD calculations by Maximiano and Schrijvers [19]. In order to perform step
tests in calm water, frequency independent damping and added mass is utilised to account
for linear derivatives. This is found to be necessary, since it stabilises the vessel model in
calm water.

Results in time domain are investigated for their time step dependency. Convergence of
forces and resulting motions is derived for a time step of dt = 0:2 s .
Step tests in calm water as well as simulations on DP in waves are validated against model
tests. Comparing step test results show a well behaved numerical model in all three DoFs.
Since the actual set point during model tests remains unclear, larger overshoots and settling
times are observed in yaw step tests.
Regardless those uncertainties at step tests the controller behaves well in irregular
disturbances. The model's response on DP in sea state 2 from model tests, is in good
agreement with the test results. DP footprints in sea state only simulations match model
test results well. In a three hours sea state of HS = 3 m and TP = 8 s the PoI alongside
the vessel stays within a square of 3 m edge length. However, simulations show a less sti
numerical model in yaw compared to tank tests and oppositely stier in surge direction.
The feedback lter of the external function is identical for the three DoFs. Since this might
not be the case for the model test DP, individual time constants for the external function
could solve this problem of dierent natural periods.

The external function comprises a thrust allocation for six azimuth thrusters and two bow
tunnel thrusters. Statistics of simulations in waves on DP are compared against feedback
signals from thrusters at tank tests. Except for the bow tunnel thrusters, mean thrust can
be reliably predicted but standard deviations of azimuth angles are larger than during model
tests. For the conducted sea state the propulsion system's workload does not seem to be
the limiting factor, since horizontal vessel motions are close to the operational limit.

58
From calculations where the vessel is exposed to directionally aligned disturbances due to
wind, current and waves a DP envelope is derived. This holding study, evaluates MPM
motions of the vessel in sea states dened by increasing wind velocities.
Assessing the vessel's operational capability, is based on its DP envelope (see Figure 5.6).
The MPM horizontal motions in x and y at the PoI are the performance indicators that
are evaluated to calculate the maximum one minute wind velocity for each angle of en-
counter. For save operations the MPM is restricted to a amplitude of 2:0 m. Combining
both indicators results in a maximum wind velocity of 10 m=s at bow quartering seas (see
Figure 5.6  = 160  ). Following IMCA [15] it corresponds to a sea state with HS = 3:1 m
and TP = 8:2 s .

7 Perspective

The developed external function is dedicated to model the DP system of the specic crane
vessel. Both the numerical vessel model and the external function are coordinated to
represent a reliable closed control loop. This is the basis for further analyses of installation
methods. Next to the introduced indicators of motion in surge and sway, the consideration
of the MPM yaw motion could make predictions for save installation operations more reliable.

Since the vessel is currently build, predictions of expected motions during operation are
meant to be conservative. In case future operations or safety standards set stricter limits
for horizontal motion, more advanced DP modes are necessary, to allow less vessel motion.
A known feature, that can improve the accuracy of a DP system, is the direct collaboration
of thrusters in groups. The power supply for the actuators will anyhow be provided for single
groups of thrusters, to full redundancy requirements. Thrusters in groups, work against
each other,thus only the dierence in thrust of both results in overall forces and moments.
This serves a system that does not need much azimuth angle changes and works up to a
limit of absolute environmental loads. Since at a point, more than half of the DP capacity
is needed, azimuth thruster would need to rotate and the DP mode has to change.
To include such a feature the external function would need further development. Regarding
the optimiser's cost function a more complex allocation problem needs to be dened. As
already indicated in section 4 sub allocation problems for dierent convex rooms of local
minimums need to be optimised before the global minimum can be derived from the results
of sub solutions.
Another aspect is the consideration of roll motion. Since this model already incorporates
proper correction of viscous roll damping (see section 2.3.5) the inuence of the DP system
on roll motion can be investigated. The developed user force needs to be expanded by
a second moment. As shown in section 5.4 unused resources of thrust could be used for
additional roll damping due to single dedicated thrusters operating in a roll damping mode.

59
References

[1] ANSYS Inc. (2017). Aqwa: Theory Manual.


[2] Blendermann, W. (Dezember 1996). Wind Loading of Ships - Collected Data from Wind
Tunnel tests in Uniform Flow: Bericht Nr. 574.
[3] Boese, P. (1970). Eine einfache Methode zur Berechnung der Widerstandserhöhung
eines Schies im Seegang. Technical Report 258, Institut fuer Schibau der Universitaet
Hamburg, Hamburg.
[4] Chakrabarti, S. K. (1994). Hydrodynamics of oshore structures. Computational Mech-
anics, Southampton.
[5] Clauss, G., Lehmann, E., and Östergaard, C. (1992). Oshore Structures: Volume I:
Conceptual Design and Hydromechanics. Springer, London.
[6] DNV GL (07.2016). Assessment of station keeping capability of dynamic positioning
vessels. Standard DNVGL-ST-0111.
[7] DNV GL (08.2017). Environmental Conditions and Environmental Loads. Recommen-
ded Practice DNVGL-RP-C205.
[8] DNVGL (10.2016). Site specic assessment of mobile oshore units for marine warranty.
Standard DNVGL-ST-N002.
[9] Faltinsen, O. M. (1990). Sea loads on ships and oshore structures. Cambridge ocean
technology series. Cambridge University Press, Cambridge, paperback ed., reprinted. edi-
tion.
[10] Fossen, T. I. (1994). Guidance and control of ocean vehicles. Wiley, Chichester.
[11] Fossen, T. I. (2011). Handbook of marine craft hydrodynamics and motion control.
Wiley, Chichester, West Sussex, U.K and Hoboken N.J.
[12] Gutsch, M., Sprenger, F., and Steen, S. (2016). Inuence of Design Parameters on
Operability of Oshore Construction Vessels. Jahrbuch der Schibautechnischen Gesell-
schaft, 110:230241.
[13] Heerink, R. and Beks, D. (2018). Orion HLV model tests for the validation of the
time domain aNySIM model. Technical Report 30384-3-BT, Maritime Research Institute
Netherlands, Wageningen, The Netherlands.
[14] Houmb, O. G. and Overvik, T. (1976). Parameterization of wave spectra. In Technische
Hogeschool Delft and Norges tekniske høgskole, editors, First International Conference on
the Behaviour of O-shore Structures (BOSS '76), volume 1, pages 144169, Trondheim,
Norway. Norwegian Institute of Technology.
[15] IMCA (06.2000). Specication for DP Capability Plots. Technical Report IMCA M I
40.

61
References

[16] Journée, J. M. J. and Massie, W. W. (2001). Oshore Hydrodynamics. Delft University


of Technology, rst edition.
[17] Kraft, D. (1994). Algorithm 733: TOMPFortran Modules for Optimal Control Cal-
culations. ACM Transactions on Mathematical Software, 20(3):262281.
[18] Lunze, J. (2013). Regelungstechnik 1. Springer, Berlin, Heidelberg.
[19] Maximiano, A. and Schrijvers, P. (2016). Current and Wind Loads on a Wind Tur-
bine Installation Vessel. Technical Report 29817-5-CPO, Maritime Research Institute
Netherlands, Wageningen, The Netherlands.
[20] Newman, J. N. (1974). Second-order, slowly-varying forces on vessels in irregular
waves. In Bishop, R. E. D. and Price, W. G., editors, International Symposium on the
Dynamics of Marine Vehicles on Structures in Waves, pages 193197, London, UK.
Mechanical Engineering Publications Ltd.
[21] Newman, J. N. (1979). The theory of ship motions. Advances in Applied Mechanics,
18:221283.
[22] Nienhuis, U. A. (1987). Simulations of Low Frequency Motions of Dynamically Posi-
tioned oshore Structures. In Royal Institution of Naval Architects, editor, RINA Trans-
actions and Annual Report, pages 112, London. RINA.
[23] Oppenheim, A. V., Schafer, R. W., and Buck, J. R. (1998). Discrete-time signal
processing. Prentice Hall and Prentice-Hall International (UK), Upper Saddler River, NJ
and London, 2nd ed. edition.
[24] Pessoa, J., Fonseca, N., and Soares, C. G. (2013). Analysis of the First Order and
Slowly Varying Motions of an Axisymmetric Floating Body in Bichromatic Waves. Journal
of Oshore Mechanics and Arctic Engineering, 135(1):111.
[25] Pinkster, J. A. (1980). Low Frequency Second Order Wave Exciting Forces on Floating
Structures. Doctor, Technische Hogeschool Delft, Delft.
[26] Pinkster, J. A. and Nienhuis, U. A. (1986). Dynamic Positioning Of Large Tankers
At Sea. In American Institute of Mining, Metallurgical, and Petroleum Engineers, Inc.,
editor, Proceedings of the 18th Annual Oshore Technology Conference, volume OTC
5208, pages 459467. Oshore Technology Conference.
[27] Rabanal, O. M., Brodtkorb, A. H., and Breivik, M. (2016). Comparing Controllers for
Dynamic Positioning of Ships in Extreme Seas. IFAC-PapersOnLine, 49(23):258264.
[28] Remery, G. and van Oortmerssen, G. (1973). The Mean Wave, Wind And Current
Forces On Oshore Structures And Their Role In The Design Of Mooring Systems.
In American Institute of Mining, Metallurgical, and Petroleum Engineers, Inc., editor,
Proceedings of the 5th Annual Oshore Technology Conference, pages 169184. Oshore
Technology Conference.
[29] Rudaa, S. E., Steen, S., and Hassani, V. (2016). Use of Tunnel- and Azimuthing
Thrusters for Roll Damping of Ships. IFAC-PapersOnLine, 49(23):546551.

62
[30] Ruth, E. (2008). Propulsion control and thrust allocation on marine vessels. Doctor,
Norwegian University of Science and Technology, Trondheim.
[31] Stern, F., Agdrup, K., Kim, S. Y., Cura-Hochbaum, A., Rhee, K. P., Quadvlieg,
F., Perdon, P., Hino, T., Broglia, R., and Gorski, J. (2011). Experience from SIMMAN
2008The First Workshop on Verication and Validation of Ship Maneuvering Simulation
Methods. Journal of Ship Research, 55(2):135147.
[32] van den Boom, H. J. J. and Nienhuis, U. A. (1983). Hydrodynamic Analysis of Dy-
namically Positioned Vessels. In Yeung, R. W., editor, The proceedings: International
Workshop on Ship and Platform Motions, pages 363383, Berkeley, CA. Dept. of Naval
Architecture and Oshore Engineering, University of California.
[33] van Oortmerssen, G. (1977). The Behaviour of Moored Ships in Waves. In American
Institute of Mining, Metallurgical, and Petroleum Engineers, Inc., editor, Proceedings of
the 9th annual Oshore Technology Conference, volume OTC1977-2882, Houston.
[34] Veksler, A., Johansen, T. A., Borrelli, F., and Realfsen, B. (2016). Dynamic Positioning
With Model Predictive Control. IEEE Transactions on Control Systems Technology,
24(4):13401353.
[35] Wichers, J., Bultema, S., and Matten, R. (1998). Hydrodynamic Research on and
Optimizing Dynamic Positioning System of a Deep Water Drilling Vessel. In American
Institute of Mining, Metallurgical, and Petroleum Engineers, Inc., editor, Proceedings of
the 30th annual Oshore Technology Conference, volume OTC1998-8854, pages 112,
Houston. Oshore Technology Conference.
[36] Williams, J. (2018). Modern Fortran Edition of the SLSQP Optimizer. ht-
tps://github.com/jacobwilliams/slsqp. [Accessed: April 2018].

63
Appendix

Appendix

A Comparison Environmental Loads in Time Domain

Figure A.1 compares the three applied environmental loads in an exemplary time domain
simulation by derived power spectra. Wind and current loads are applied following section
3.4. It can be observed that wave loads are the governing environment since the standard
deviation is more than ten times larger.

1e 8 = 5.41e-05 [-]
[ ]

4 Wave
g LPP
M3

0
1e 9 = 5.65e-06 [-]
1.5 Current
1.0
0.5
0.0
1e 13 = 1.03e-07 [-]
3 Wind
2
1
0
1e 7 = 6.92e-05 [-]
2 DP userForce

0
0 25 50 75 100 125 150 175 200
Period T [s]

Figure A.1: Comparison of environmental loads in yaw due to waves, current


and wind. A sea state corresponding to a wind speed of 9:0 m=s (JONSWAP
HS = 2:9 m; Tp = 7:9 s ; = 3:3) is simulated as well as loads due to
wind and a current of 1:03 m=s . Time domain results are presented by
power spectra. All loads are aligned at  = 150 and made dimensionless.
Additional comparison is made to the moment allocated by the external DP
function at the bottom.

64
B Comparison Commanded and Allocated Forces

In Figure B.1 the sum of allocated forces and the commanded force by the controller are
compared in the sea state irregular 2.

a) b)
1.0 1e12 2.0
Power spectrum F1 [N 2/s]

x in [m]
PID command 99% Footprint PID command
of Allocated 1.5 99% Footprint of Allocated
0.8
1.0
0.6 0.5
0.0
0.4 0.5
1.0
0.2
1.5
0.00 25 50 75 100 125 150 175 200 2.02 1 0 1 2
Period T [s] y in [m]
c) d)
1.0 1e13 1e16
Power spectrum M3 [Nm2/s]
Power spectrum F2 [N 2/s]

PID command 1.4 PID command


of Allocated of Allocated
0.8 1.2
1.0
0.6
0.8
0.4 0.6
0.4
0.2
0.2
0.00 25 50 75 100 125 150 175 200 0.00 25 50 75 100 125 150 175 200
Period T [s] Period T [s]

Figure B.1: Comparison of forces and the yawing moment by the PID controller
and the allocator in the sea state irregular 2 (JONSWAP HS = 3 m; Tp =
8 s ; = 1:55). In green forces as commanded by the PID in predictor and
corrector stage and in black the sum of all allocated forces calculated by
the optimizer applied in corrector stage. Figure a) and b) compare the
power spectra of forces, c) the spectrum of the yawing moment and d) the
corresponding footprint.

The green line shows a simulation without allocation. The black the method like describe
in section 4 performing the allocation only in the corrector step. The force in sway and the
yawing moment show some slack (see c) and d)) where as the footprint stays unaected.
This comparison is necessary, since the optimiser is only run in the correcter stage of time
integration. To avoid divergence of the optimiser when sudden high force gradients to the

65
Appendix

previous time steps are calculated in predictor stage it is enough to run the optimiser ones.
It is exemplary shown that the duration of computations can be shortened.

C Source Code External Function

The developed external function is provided by the digital appendix together with this doc-
umentation on a compact disc. It comprises:
ˆ The batch script built-user_force64.bat to compile the function with the
gfortran compiler. If the script is edit with the prevalent installation directory of
the installed Aqwa version the compiled user_force.dll is copied, as the user has
permission to that directory.
ˆ The source code of the external function with its main routine
uf_sourceCode.f90. Modules and subroutines are located in the sub direct-
ory .\externalFunction_craneVesselDP_v182\src\.
ˆ An example of a conguration le for a time domain simulation with Aqwa Drift
and the developed external function (.\aqwaDrift_exampleRun\AD_template.dat).
Exemplary result les saved by the external function, are provided as well.
In the following section parts of source code are appended, that are referenced and explained
in section 4.

C.1 Basic Structure of the External Function


Programme code C.1 documents the basic structure of the user_force (see section 4) and
the initial declaration of variables to interfere with the Aqwa main routine.

1 SUBROUTINE USER_FORCE(MODE,I_CONTROL,R_CONTROL,NSTRUC,TIME, &


2 TIMESTEP,STAGE, POSITION,VELOCITY,COG, &
3 FORCE,ADDMASS, &
4 ERRORFLAG) bind(C, name="USER_FORCE")
5 !DECLARATION TO MAKE USER_FORCE PUBLIC WITH UN-MANGLED NAME
6 !DEC$ attributes dllexport , STDCALL , ALIAS : "USER_FORCE" :: user_force
7 !BL
8 !DEC$ ATTRIBUTES REFERENCE :: I_CONTROL, R_CONTROL
9 !DEC$ ATTRIBUTES REFERENCE :: POSITION, VELOCITY, COG, FORCE, ADDMASS
10 !DEC$ ATTRIBUTES REFERENCE :: MODE, NSTRUC, TIME, TIMESTEP, STAGE
11 !DEC$ ATTRIBUTES REFERENCE :: ERRORFLAG
12 ! DECLARATION - USER_FORCE
13 IMPLICIT NONE
14 INTEGER MODE, NSTRUC, STAGE, ERRORFLAG
15 REAL TIME, TIMESTEP
16 INTEGER, DIMENSION (100) :: I_CONTROL
17 REAL, DIMENSION (100) :: R_CONTROL
18 REAL, DIMENSION (3,NSTRUC) :: COG
19 REAL, DIMENSION (6,NSTRUC) :: POSITION, VELOCITY, FORCE
20 REAL, DIMENSION (6,6,NSTRUC) :: ADDMASS

66
21 IF (MODE.EQ.0) THEN
22 !...
23 ELSEIF (MODE.EQ.1) THEN
24 !...
25 ELSEIF (MODE.EQ.99) THEN
26 !...
27 ENDIF
28 RETURN
29 END SUBROUTINE USER_FORCE

Programme Code C.1: Header

C.2 Part of Allocation in the Main Routine

1 IF (STAGE.EQ.2) THEN
2 ! Initialize solver with solution from last time step
3 x(1:14) = x_old
4 x(15:17)= [0.0_wp,0.0_wp,0.0_wp]
5 write(99,*) 'Initial solution', x
6 ! Handover commanded force/moment from pid to solver.
7 tau_pid = (/FORCE_pre(1,1), FORCE_pre(2,1), FORCE_pre(6,1)/)
8 tau_pid = MATMUL(TRANSPOSE(TRAFO3),tau_pid) !! FRA to LSA
9 write(99,*) 'PID', tau_pid
10 ! Dynamic thrust regions
11 do i = 1,nAT,1
12 xl(i*2-1) = max(-max_Thrust(i),x_old(i*2-1) -
,! thrustRate(i)*max_Thrust(1)**2/sqrt(2.))
13 xl(i*2 ) = max(-max_Thrust(i),x_old(i*2 ) -
,! thrustRate(i)*max_Thrust(1)**2/sqrt(2.))
14 xu(i*2-1) = min( max_Thrust(i),x_old(i*2-1) +
,! thrustRate(i)*max_Thrust(1)**2/sqrt(2.))
15 xu(i*2 ) = min( max_Thrust(i),x_old(i*2 ) +
,! thrustRate(i)*max_Thrust(1)**2/sqrt(2.))
16 enddo
17 xl(13) = max(-max_Thrust(7),x_old(13) - thrustRate(7)*max_Thrust(1)**2)
18 xu(13) = min( max_Thrust(7),x_old(13) + thrustRate(7)*max_Thrust(1)**2)
19 xl(14) = max(-max_Thrust(8),x_old(14) - thrustRate(8)*max_Thrust(1)**2)
20 xu(14) = min( max_Thrust(8),x_old(14) + thrustRate(8)*max_Thrust(1)**2)
21 ! Scaling the problem
22 x = x / max_Thrust(1)**2
23 x_old = x_old / max_Thrust(1)**2
24 xl = xl / max_Thrust(1)**2
25 xu = xu / max_Thrust(1)**2
26 tau_pid = tau_pid / max_Thrust(1)**2
27 call solver%initialize(n,m,meq,max_iter,acc,test_func,test_grad,&
28 xl,xu,linesearch_mode=linesearch_mode,iprint=0,st c
,! atus_ok=status_ok,&
29 report=report_iteration,&
30 alphamin=0.1_wp, alphamax=1.0_wp) !to limit
,! search steps
31 if (status_ok) then

67
Appendix

32 call solver%optimize(x,istat,iterations)
33 ! Rescaling the problem
34 x = x * max_Thrust(1)**2
35 x_old = x_old * max_Thrust(1)**2
36 xl = xl * max_Thrust(1)**2
37 xu = xu * max_Thrust(1)**2
38 tau_pid = tau_pid * max_Thrust(1)**2
39 write(99,*) 'aziRate', aziRate*180.0/pi
40 write(99,*) 'xu ', xu
41 write(99,*) 'Solution: ', x
42 write(99,*) 'xl ', xl
43 write(99,*) 'Iterations: ', iterations, 'Status: ', istat
44 if (istat.NE.0 .AND. istat.NE.8 .AND. istat.NE.9) then
45 write(*,*) 'Iterations: ', iterations, 'Status: ', istat
46 end if
47 else
48 write(99,*) ' # Error c01: Solver could not initialize!'
49 write(*,*) ' # Error c01: Solver could not initialize!'
50 endif
51 CALL checkAzimuthRates(x(1:nAT*2), x_old(1:nAT*2))
52 CALL sumAllocated(loc_Thrust, x(1:14), FORCE_alloc_LSA)
53 write(99,*) 'ALC', FORCE_alloc_LSA
54 CALL fxfy2rpmAlpha(x(1:14), rpm, Alpha)
55 ENDIF
56 ! LSA to FRA
57 FORCE_alloc_LSA(:,1) = MATMUL(TRAFO3, FORCE_alloc_LSA(:,1))

Programme Code C.2: Allocation

68

Вам также может понравиться