Anda di halaman 1dari 123

Models to explain out-of-plane

bending mechanism in mooring chain


links

By Nilakash Das
Delft University of Technology

Models to explain out-of-plane bending


mechanism in mooring chain links

Committee:
Prof.dr. A. Metrikine
Author: Dr.ir. J. De Oliveira Barbosa
Nilakash Das Ir. P.J. Aalberts
Ir. R. Hageman
Dr. F. Pisano

A thesis presented for the degree of


Master of Science
In
Offshore and Dredging Engineering

July 25, 2016


Acknowledgements

After almost nine months of blood, sweat and tears, I am extremely pleased to submit my final
thesis report. I had never worked on a research project before and the experience has been exhil-
arating. A lot of people were instrumental in this endeavour to whom I would like to convey my
heartfelt gratitude.
This thesis would not have been possible if not for my parents. They have selflessly supported
me in my quest for undertaking a masters program in a different country. I am completely in
debt to my parents for trusting my capabilities. I also would also like to thank my sister, who has
always been a pleasant friend.
I would like to acknowledge my committee chairman, Professor Andrei Metrikine who gave
many invaluable suggestions during the course of my thesis. Without the courses of Structural
Dynamics which he taught, and Computational Dynamics, which he introduced last year, this
thesis would have extended far beyond the stipulated time. He taught me to understand the
underlying physics of a problem without going into the mathematics. I consider myself extremely
lucky to have been supervised by a man of Professor Metrikine’s calibre.
I would like to express my sincere gratitude to my daily university supervisor, Dr. Joao Barbosa.
What is remarkable is that in-spite of supervising so many students, he remembers every minute
detail of my research. I am simply awed by his enormous passion towards solving a problem. I
am also humbled by his response as he entertained many of my silly ideas. He would always make
time for my doubts, even when he was on a vacation! I also cannot thank him enough for agreeing
to supervise me on the Research Exercise coursework. A big Obrigado to him for all his efforts.
I would like to express my special thanks to my company (MARIN) supervisors Dr. Pieter
Aalberts and Ir. Remco Hageman, who gave me a golden opportunity to work in a world renowned
research institute . They have been extremely supportive during my thesis, providing me with all
the resources that I needed. I am extremely thankful to their suggestions as well their effort in
proof-reading my final report.
All the research in this thesis was carried out at MARIN. It was a sheer pleasure in being
around so many scientists and students. I learnt about exciting developments not only in the field
of structural mechanics, but also in the fields of hydrodynamics, CFD and model testing. I would
like to give a shout-out to all my friends in MARIN for making my stay a memorable experience.
Finally, my journey during the masters program has been a long and arduous one. A journey
becomes an enjoyable experience when one shares it with a few good people. I want to express my
special gratitude to my partner Miss. Priyanaka Ganatra, who has always been encouraging and
supportive. I also made some really good friends over the course of time who were always there
for me and I would like to acknowledge them.
Abstract

In 2002, several mooring chains of Girassol Off-loading buoy which was installed offshore Angola
ruptured just after 8 months of service. A new failure mechanism called out-of plane bending(OPB)
fatigue in mooring chain links was identified in addition to tension induced fatigue after a series of
experiments.
Currently, there are no models in literature which can comprehensively explain OPB mechanism
in mooring chain links. Models to explain OPB mechanism are required to calculate OPB stresses
in chain links for fatigue damage evaluation. These models need to be ”simple” and ”adequate”
in the sense that it should idealize the complicated geometry of chain links while giving accurate
predictions of OPB stresses when subjected to rotation and tension, at a fraction of time used
by solid finite element models or full scale experiments. The research carried out in this master’s
project deals with development of such models to explain OPB mechanism in mooring chain links.
In this thesis, a new interlink stiffness model is developed to describe the nonlinear hysteritic
relationship between OPB moment and interlink angle. The interlink stiffness model is then
applied to a system of chain links using two different methods: a physics based approach and a
semi-empirical approach. Each of these methods present a new approach to idealize a system of
chain links. Then, a case study is presented in which a methodology to calculate OPB stresses
for one sea-state is described. This methodology follows a de-coupled approach in which results
of coupled floater-mooring analysis are taken as an input for the simplified models of chain links.
To verify this methodology, a comparative study is performed at the end in which coupled floater-
mooring analysis is carried out with and without including the interlink stiffness of chain links.
Contents

1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Literature Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Problem definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Research Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5 Scope of Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.6 Research Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Development of a FE Model 7
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2.1 Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 FE model development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.3.1 Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.4.1 Verification of FE Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.4.2 Comparison with raw experimental data . . . . . . . . . . . . . . . . . . . . 11
2.4.3 Effect of pretension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.4.4 Effect of friction coefficient . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.4.5 Sensitivity to interlink angle definition . . . . . . . . . . . . . . . . . . . . . 13
2.4.6 Comparison with ANSYS Transient Analysis . . . . . . . . . . . . . . . . . 14
2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3 FE Studies on OPB Mechanism 17


3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.3 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.4 FE studies and observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.4.1 Rate Independency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.4.2 Memory effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.4.3 Stick Slip . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.4.4 FE studies on a 7 link chain . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.5 Inference and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.5.1 On FE studies on a 7 link chain . . . . . . . . . . . . . . . . . . . . . . . . 24
3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

4 A New Interlink Stiffness Model 25


4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.2 Hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.3 Maxwell Slip Model: A friction model . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.4 Interlink Stiffness Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.4.1 Constant Tension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.4.2 Varying Tension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.4.3 Additional Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
CONTENTS

5 A Physics Based Approach 37


5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.2 Idealization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.3 Equivalent Link Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.4 Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.6 Alternative approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5.6.1 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

6 A Semi-Empirical Approach 47
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
6.2 Idealization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
6.3 Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
6.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6.4.1 3 link model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6.4.2 7 link model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

7 A Case Study On Calculating OPB Stress 55


7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
7.2 Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
7.3 Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
7.4 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
7.4.1 Step 1: Post-processing after coupled floater-mooring analysis . . . . . . . . 58
7.4.2 Step 2: Derive interlink stiffness model . . . . . . . . . . . . . . . . . . . . . 60
7.4.3 Step 3: Derive equivalent link geometry . . . . . . . . . . . . . . . . . . . . 63
7.4.4 Step 4: Prepare simplified model . . . . . . . . . . . . . . . . . . . . . . . . 65
7.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
7.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

8 A Comparative Study 69
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
8.2 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
8.3 Description of the numerical simulation . . . . . . . . . . . . . . . . . . . . . . . . 70
8.3.1 Equation of motion of the floater . . . . . . . . . . . . . . . . . . . . . . . . 70
8.3.2 Equation of motion of the mooring line . . . . . . . . . . . . . . . . . . . . 72
8.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
8.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

9 Critical Analysis and Recommendation 79


9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
9.2 Critical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
9.2.1 On FE modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
9.2.2 On interlink stiffness model . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
9.2.3 On Simplified Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
9.2.4 On the Methodology for estimating OPB stress . . . . . . . . . . . . . . . . 82
9.3 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

A ANSYS Nonlinear Quasi-Static Analysis 83


A.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
A.2 Solution Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
A.2.1 Tangent Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
A.2.2 Newton-Raphson Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

B ANSYS Transient Dynamic Analysis 87


B.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
B.2 Solution Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
CONTENTS

C ANSYS frictional contact 89


C.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
C.2 Penalty Based Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
C.3 Augmented Langrangian Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
C.4 Friction and Elastic Slip . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

D Genetic Algorithm 91
D.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
D.2 Outline of Genetic Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

E Fixed-point iteration 93

F Coupled Floater-Mooring Analysis Input Details 95


F.1 Environmental details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
F.2 Floater details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
F.3 Mooring line details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
F.3.1 Chain links . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
F.3.2 Polyester rope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
F.4 Frequency dependent parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
F.4.1 Added Mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
F.4.2 Radiation Damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
F.4.3 First-Order Wave Force Transfer Function . . . . . . . . . . . . . . . . . . . 99
F.4.4 Quadratic Transfer Function . . . . . . . . . . . . . . . . . . . . . . . . . . 99
F.4.5 Steady Drift Transfer Function . . . . . . . . . . . . . . . . . . . . . . . . . 100

List of Figures 107

List of Tables 111


The important thing is to not stop questioning. Curiosity has its own
reason for existence. One cannot help but be in awe when he contemplates
the mysteries of eternity, of life, of the marvelous structure of reality. It
is enough if one tries merely to comprehend a little of this mystery each
day.
Albert Einstein
Chapter 1

Introduction

1.1 Background
In 2002, several mooring chains of Girassol Off-loading buoy which was installed offshore Angola at
a depth of 1400 m, ruptured just after 8 months of service. These chains were designed according to
conventional tension fatigue assessment using API RP 2SK T-N curves to a fatigue life of 20 years
with a factor of safety equal to 3. It was observed that four out of five mooring lines had failed at the
first free link that was connected to the floater. A new phenomenon, called Out-of Plane Bending
(OPB) fatigue was identified in addition to tension fatigue after a series of experiments.([6], [23]
and [22])

Figure 1.1: Girassol buoy chainhawse: location of failed link [6]

In the case of Girassol Buoy, the connection between the mooring line and the floater was
located inside a chainhawse (figure 1.1). Failure was observed to occur at Link 5 which was the
first free link of the mooring lines. The relative rotation between the buoy and the mooring line
caused Link 5 to rotate with respect to Link 4, which was fixed to the floater. Generally, it was
assumed that chain links do not posses any rotational stiffness at their contact and they can rotate
freely with respect to one another. However, due to proof testing of links which involves application
of a large tensile load at 70% of Minimum Breaking Load(MBL) induces plastic strains at the link
contacts. When the mooring chains operate under high pretension (>10% MBL), the links behaves
in a complex way by getting ”locked” at their contact and by resisting high moments. This caused
Link 5 to bend out of its plane and therefore it was called out of plane bending (OPB). The high
alternating bending stresses or OPB stresses in Link 5 was attributed as the root cause for its
failure. This mechanism is different from tension induced fatigued which uses stress histogram of
the mooring chain without considering chain rotations (figure 1.2).
To demonstrate the OPB mechanism, consider figure 1.3. in which two links are in pretension
and a tension angle is imposed at the link that can bend out of its plane. This link is the OPB
link. The moment MA that is generated at the links’ contact is the OPB moment and the angle
between the two links is called the interlink angle. The connection between the two links can
be idealized as a rotational spring which is a measure of the interlink stiffness. The rotational
spring can be a complex function of proof loading, interlink angle, tension, frictional coefficient,

1
2 CHAPTER 1. INTRODUCTION

Figure 1.2: Tension and OPB fatigue location in mooring chain links[6]

link material and link shape. The flexural stresses in the link that bends in its plane (IPB Link) are
benign as its sectional modulus is much higher than the OPB link. The terms OPB Moments and
OPB stresses will be used interchangeably as both quantities only differ by a factor of geometry
or section modulus of the link.

Figure 1.3: OPB mechanism

1.2 Literature Overview


After the Girassol Buoy incident, several experiments were carried out to understand the OPB
mechanism in mooring chain links. The set-up of the experiments varied but all had the same
underlying principle, i.e impose large tension and rotation to a system of links and measure the
stresses in the links that bends out of its plane. Figure 1.4 illustrates one of the experimental set-
ups which was used in [6] and [22]. In these experiments, full scale tests of studless and studded
chain links were carried out and an empirical relationship between observed OPB moment/stress,
interlink angle and applied tension was developed. This became the first interlink stiffness model
to be proposed. The model can be stated as follows:

d
Mopb = min{ aT b dc αint , µT } (1.1)
2
where Mopb is the OPB moment, T is the tension, alphaint is the interlink angle, µ is the friction
coefficient for steel (0.5-0.7 in air and 0.3 in sea water), d is the chain diameter and a,b and c
are empirical constants. Equation 1.1 represents two regimes in the OPB stress/moment-interlink
angle curve: the ”sticking” regime which is a linear function of interlink angle and the sliding
regime. Finite element analysis (FEA) was carried out to reproduce the experimental results in
[23].
A calculation methodology was proposed in [5] to determine OPB stress as a function of interlink
angles, friction coefficient and tension by approximating the relationship with polynomial functions.
The interlink angles were approximated as polynomial functions of friction coefficient, applied
1.2. LITERATURE OVERVIEW 3

Figure 1.4: Experimental set-up of a system of chain links: Chainhawse pushes down link T4 to
generate interlink angles between T4 and T5 and OPB stresses in T5.

tension and rotation. Another methodology was proposed in [9] in which an empirical relationship
between OPB stresses and imposed tension and rotation was developed.
A Joint Industry Project(JIP) aptly named Chain JIP [25] was initiated to improve the models
developed in [6] by performing full scale tests for a range of link diameters and conducting sensitivity
studies with respect to chain grade, link shape and manufacturing effect using FEA. In the same
test program, it was observed that the OPB stresses decreases in magnitude as one moves away
from the loading source thereby confirming the hypothesis that the first free link connected to the
floater is the most crucial link with regards to OPB fatigue. A fatigue damage criteria for the
chain links due to OPB was also proposed.
A guidance note to calculate OPB stresses for fatigue evaluation was released by Bureau Ver-
itas(BV) [20] which incorporates some of the results from Chain JIP. To asses OPB stresses in
time domain, it advises to first preform coupled floater mooring calculations without considering
bending or interlink stiffness in the mooring chain links. Then a beam finite element (FE) model
of a system of links has to be prepared which should be ”calibrated” with an interlink stiffness
model obtained from chain JIP or full scale tests. The tensions at the top of the mooring line
and the relative rotation between the floater and the mooring line obtained from coupled floater
mooring calculations are given as an input to this model and time domain analysis is performed
quasi-statically. The final outcome of this model are the OPB stresses. Figure 1.5 illustrates the
model suggested by BV.

Figure 1.5: Beam FE model suggested in BV guidance note [20]


4 CHAPTER 1. INTRODUCTION

1.3 Problem definition


Currently, there is no clear methodology in literature to asses OPB stresses in time domain. The
methodology suggested by BV has several flaws and the reliability of the beam FE model it suggests
has not been demonstrated. The suggestion that each link be represented as a beam with the same
dimensions does not faithfully represent a system of chain links whose planes are perpendicular
to each other. It also does not specify how to ”calibrate” the model with an interlink stiffness
model. Moreover, the interlink stiffness model it recommends, which was first developed in [25]
and later improved in [25], describes only a small linear region in the relationship between observed
OPB stresses and interlink angles. The sliding limit it proposes is an overestimation as it does not
take into account the deformation of the links due to proof loading. In fact, the derivation of the
sliding limit which is given in Page 3 of [24] is incorrect as it directly substitutes αi = π/2 and
not by calculating αi by equilibrating applied tension with limiting friction at the links’ contact.
The dependence of OPB moments on interlink angles has been observed to be nonlinear and it
exhibits a hysteritic relationship (figure 1.6). The models proposed in [5] and [9] are devoid of any
physical meaning as they are just empirical relationships between OPB moments/stresses, interlink
angles, imposed rotations and tensions. These models also do not take into account the hysteritic
relationship between OPB moments/stresses and interlink angles.

Figure 1.6: OPB mechanism

In the lack of any reliable model to asses OPB stresses in time domain, an alternative solution
could be to use a solid FE model of a system of chain links. But this is a computationally expensive
procedure as the mesh has to be extremely refined especially at the connection between two links.
It was also pointed out in [23], [25] and [9] that FEA gives under-conservative results as compared
to experimental observations and therefore, it should be used with caution.
This has necessitated the development of simplified models to explain OPB mechanism for a
system of chain links. In this thesis, a simplified model will be defined as one that should perform
the following tasks:

1. It should be able to reproduce the physics of an actual system of links when subjected to
rotation and tension.

2. It should give a reasonable prediction of OPB stresses and interlink angles in time domain
compared to a solid FE model of a system of chain links.

3. It should take less time to calculate than a solid FE model.

Finally, the decoupled approach to asses OPB stresses also raises some questions. In every
research relating to OPB of mooring chain links, the problem has been characterised by only
imposing tension and rotations obtained from coupled floater-mooring analysis to a system of
chain links. There is no comparative study done in literature to determine whether including
interlink stiffness of chain links in a coupled floater-mooring analysis will change the line tensions
and floater rotations.
1.4. RESEARCH QUESTIONS 5

1.4 Research Questions


This thesis will attempt to answer the following research questions:
1. Can a mathematical model be developed that will describe the hysteritic relationship between
OPB moments/stresses and interlink angles?

2. Using this mathematical model, can a simplified model of a system of chain links be developed
which when subjected to tension and rotation give good predictions of OPB moments/stresses
and interlink angles?
3. If the simplified model is successful, how can it be incorporated into a methodology to asses
OPB stresses in time domain?

4. If the mathematical model describing the relationship between OPB moment/stress and
interlink angle is included in coupled floater-mooring analysis, will the line tensions and
floater rotations change from a case of coupled floater-mooring analysis carried out without
including this mathematical model?

1.5 Scope of Research


This research is mainly focused on capturing the OPB phenomenon of the first free link in a system
of chain links as it is the most crucial one with regards to OPB fatigue. This will be demonstrated
in the course of this thesis. The in-plane bending (IPB) effect is ignored as it is quite small in
magnitude compared to the OPB effect. Each chain link will have their planes perpendicular to
each other and torsion will be ignored. This will ensure a worst case scenario for OPB fatigue.
Only studless chain links are considered. Due to lack of an experimental set-up, a verified and
validated solid FE model of chain links will be considered as a representation of reality. For the
comparative study, a simple model consisting of one mooring line and one cylindrical floater is
considered.

1.6 Research Methodology


The research presented in this thesis was carried out in the following steps:
1. Verify and validate a solid FE model of a system of chain links (Chapter 2)

2. Carry out FE studies on OPB mechanism of mooring chain links (Chapter 3)


3. Develop a new interlink stiffness model to describe the relationship between OPB stresses/moments
and interlink angles (Chapter 4)
4. Develop a simplified model of a system of chain links using a physics based approach (Chapter
5)
5. Develop a simplified model of a system of links using a semi-empirical approach (Chapter 6)
6. Demonstrate a methodology to asses OPB stresses for one sea-state using a case study.(Chapter
7)

7. Perform a comparative study of coupled floater-mooring analysis with and without including
the interlink stiffness model. (Chapter 8)
8. Critical analysis and recommendations. (Chapter 9)
6 CHAPTER 1. INTRODUCTION
Chapter 2

Development of a FE Model

2.1 Introduction
In the absence of an experimental set-up to observe the OPB mechanism in mooring chain links, it
was decided to develop a finite element(FE) model. This chapter gives details on the development
of a solid FE model of a three link chain system subjected to tension and rotation. The model
has been verified and validated with experimental results from [6] and [23]. This FE model will be
used to derive the interlink stiffness model in Chapter 3 and to validate the simplified models in
Chapters 4 and 5.

2.2 Experimental Setup


This section explains the experimental setup which was used in [6] and [23]. A studless 1ink with
124 mm diameter is considered (figure 2.1). The material grade is R4 and Minimum Breaking
Load (MBL) is 1465 tons, calculated according to [21].

Figure 2.1: Nominal dimensions of a 124 mm studless link [23]

2.2.1 Description
The SBM Chain Test Facility is configured to simulate the relative motion between two adjacent
links in a mooring chain. Depending on the chain size, different length of chain links could be
accommodated. Figure 2.2 shows the test schematic for a 81 mm chain but the principle is same for
all link sizes. The chain tension is maintained constant by a hydraulic cylinder while a chainhawse
fixed with a rig shoe is pushed down very slowly, by another hydraulic cylinder which causes link
labeled T4 in figure 2.2 to rotate relative to the link labeled T5. This generates significant interlink
angles and OPB stresses in link T5.
The test set up for the 124 mm link chain is shown in figure 2.3. Link 1 is pushed down by the
rig shoe to create interlink angles between Link1 and Link2, which is monitored for OPB stresses
while tension is kept constant. Link2 is mounted with strain gauges as shown in figure 2.4. The

7
8 CHAPTER 2. DEVELOPMENT OF A FE MODEL

Figure 2.2: Test schematic for a 81 mm chain link: Chainhawse pushes down T4 to generate
interlink angles between T4 and T5 and OPB stresses in T5.

location of the rosettes was in the transitionqof the straight and bend part of the link. The quantity
that was reported was the stress range of σxx 2 + σ 2 where σ
yy xx and σyy are the normal stresses
in x and y direction

Figure 2.3: Test schematic for a 124 mm chain links

Figure 2.4: Location of reported stresses

Inclinometers were used to measure relative angles between the links although their exact
location is not clear. Figure 2.5 gives the definition of interlink angles according to [22], which was
SBM’s FE work to reproduce experimental results. Again the interpretation of this definition is
not unambiguous. Figure 2.6 gives one possible interpretation which has been used to validate the
experimental results. A sensitivity study with respect to other possible interpretations will also be
carried out as part of this chapter
2.2. EXPERIMENTAL SETUP 9

Figure 2.5: SBM definition of interlink angle

Figure 2.6: Interpretation 1 of interlink angle definition

An important point to note that SBM has not published the values of the stress amplitudes
but rather the trends. But they have published an empirical formula [6] which approximates the
linear region in the OPB stress vs interlink angles curve (figure 2.7), which is as follows:

σopb = aT 0.3 d−1 αint (2.1)

where a is a constant, T is the pretension in the chain, d is the diameter of the links and αint
is the interlink angle. This empirical fit is used to reproduce the values of experimental results.
Thus, the experimental values that have been used for validation may not be exact.

Figure 2.7: Empirical fit after OPB tests for 81 mm studded, 106 mm and 124 mm studless links
10 CHAPTER 2. DEVELOPMENT OF A FE MODEL

2.3 FE model development


2.3.1 Description
The solid model (figure 2.8) is a 3 link model created in ANSYS to simulate experimental loading
and boundary conditions. The contact between links is modeled as stiff springs which does not
allow penetration but allows sliding by taking into account the frictional coefficient. Refer appendix
C for a description on the formulation of frictional contact used in ANSYS. Link 1 is fixed and
rotations around y axis are applied on the upper straight part of Link 3. Tension is applied on
Link 3 in x direction. OPB stresses are monitored in the transition between straight and bent part
of Link 2. Additional supports are applied to restrain the link in y direction.

Figure 2.8: Solid model of 124 mm chain links

A non linear material model called the Ramberg-Osgood Law (figure 2.9) is used with isotropic
hardening. Information on the material model was obtained from [22]. The contact of the chain-
hawse rig with Link 3 and the movement of the chainhawse is not modeled directly but instead,
rotations are applied to the upper straight part of link 3. Nonlinear quasi-static analysis (refer
appendix A) is carried out due to the application of proof load which induces plastic strains and
due to the effect of frictional contact between the links.

Figure 2.9: Ramberg Osgood true stress strain curve:Yield Stress=580 MPa, Ultimate Stress=860
MPa, Ultimate Strain=11%, α = 0.002, n=10.3

The sequence of loading is as follows:

1. Apply 70% MBL (14360 tons) in tension and release to simulate chain proof testing.

2. Apply constant tension of 4.4% MBL (60 tons) or 6.4% MBL (94 tons).

3. Apply rotations to upper straight part of link 3.


2.4. RESULTS 11

2.4 Results
The OPB stresses are monitored on Link 2 at the location mentioned in the experiments (See
figure 2.4) and reported in the same way. Since, exact conversion of strain data to stresses is not
mentioned, the stress ranges are directly reported after the analysis. The definition of interlink
angles has been assumed to be the angle between lines joining the extremities (figure 2.6). Friction
coefficient is taken as 0.5 unless explicitly specified.

2.4.1 Verification of FE Model


To verify the FE model a full solid model is constructed (see figure 2.8) and the mesh sizes are
changed until a converged solution is obtained. In all the simulations, one complete cycle of rotation
is applied without initiating any sliding in the links. Once the full solid model is verified (labeled
as Grid 1 Full model in figure 2.10), symmetric boundary conditions are applied. Symmetric
boundary conditions are applied to reduce computation time. The mesh size of the symmetric
model is changed until a converged solution is obtained which is exactly similar to that of the full
solid model. The corresponding grid size is noted and this model used in further simulations. In
this case, the model termed as Grid 3 symmetry (figure 2.11) is chosen which is a solid model of
the chain links with symmetric boundary conditions.

Figure 2.10: Verification of FE model

2.4.2 Comparison with raw experimental data


Figure 2.12 shows the comparison of experimental data with FEA. The inner and the outer envelope
represent the set of all possible OPB stress ranges obtained after a set of 50 stabilized cycles of
chainhawse movement. Values from FEA are published for 3 cycles of applied Link 3 rotations.
The applied rotations are between +2 and -2 degrees and the tension is fixed at 60 tons. The
results of the FE model shows a good agreement with experimental observations.
12 CHAPTER 2. DEVELOPMENT OF A FE MODEL

Figure 2.11: Grid 3 symmetry model

Figure 2.12: Comparison between FEA and experimental data

2.4.3 Effect of pretension


The effect of pretensions is clear according the empirical fit (equation 2.1). The same exercise
was repeated in FEA for two different pretension values (60 and 94 tons) and a similar trend was
observed (figure 2.13).

2.4.4 Effect of friction coefficient


The friction coefficient between the links is around 0.5 in air and around 0.3 in sea water. In the
experiments, it was observed that sliding initiates faster in seawater although the initial slope of
the lines is similar. The same has been observed in FEA which is shown in figure 2.14. In this
simulation , rotation was applied in only one direction until sliding of links occurs.
2.4. RESULTS 13

Figure 2.13: Effect of pretension

Figure 2.14: Effect of friction coefficient

2.4.5 Sensitivity to interlink angle definition


The definition of interlink angles used by SBM (See figure 2.5) is not clear. A sensitivity study is
carried for three possible interpretations of the definition of interlink angles. The three possible
interpretations are as follow:
14 CHAPTER 2. DEVELOPMENT OF A FE MODEL

1. Interpretation 1 (figure 2.6)

2. Interpretation 2 (figure 2.15)

3. Interpretation 3 (figure 2.16)

Figure 2.15: Interpretation 2 of interlink angle definition

Figure 2.16: Interpretation 3 of interlink angle definition

Figure 2.17 shows a comparison of OPB stress-interlink angle curve for the three interpretations.
From this figure, it is clear that the sensitivity with respect to interlink angle definition is large.

2.4.6 Comparison with ANSYS Transient Analysis


Transient analysis (dynamic analysis) in ANSYS should give similar results as ANSYS quasi-static
analysis as the loading in the experiments was carried out at a very low frequency. In this case,
dynamic FEA was carried out by applying rotations at 0.25 Hz. Figure 2.18 shows the comparison
between a quasi static analysis and transient analysis at 0.25 Hz. This comparison proves that the
inertial effects are absent at low frequencies.

2.5 Conclusion
The results from FEA show quite similar trends to what is obtained from experimental results.
However, the values are not a perfect match because of several possible reasons:

1. As stated before, the experimental results that are used for comparison has not been directly
obtained but back calculated using the empirical formula. There can be a good variation in
the raw data itself. The exact values of the raw data are unknown but calibrated using the
empirical formula. This can have a big impact in the OPB stress-interlink angle diagrams.
2.5. CONCLUSION 15

Figure 2.17: Sensitivity to interlink angle definition

Figure 2.18: Quasi-static vs dynamic analysis

2. The exact loading condition of the chainhawse pushing down the link to create interlink
rotations has not been exactly reproduced in FEA due to insufficient information. The
loading has been included in a representative manner by applying rotations directly.

3. The material law used is a ”static” material law which does not account for cyclic strain
rates. In the chain JIP OPB [25], it was shown that FEA results could be improved with
a combined isotropic and kinematic hardening model that takes into account plastic strain
rate.

4. The link geometry used in FEA was a nominal one without taking into account exact fabri-
16 CHAPTER 2. DEVELOPMENT OF A FE MODEL

cation dimensions. In the Chain OPB it was demonstrated that minor changes in the bend
radius of the link can create a discrepancy in the stress values.
5. The OPB stress diagrams are also sensitive to the definition of interlink angles.

The 3 link symmetric model which has been developed in this chapter will be used in the next
chapter (Chapter 3) to develop an interlink stiffness model and later to develop simplified models
of chain links (Chapters 4 and 5). Transient analysis will be used because in the frequencies of
our interest (10-20 s for floater rotation) the model behaves quasi-statically. Moreover, ANSYS
transient analysis is faster as it uses an implicit integration scheme which is unconditionally stable
(refer appendix B for more information). It does not have to perform multiple iterations for each
time step which is done in ANSYS quasi-static structural analysis (refer Appendix A).
Chapter 3

FE Studies on OPB Mechanism

3.1 Introduction
After the 3 link solid FE model is verified and validated, it is used to carry out several studies on
OPB mechanism in mooring chain links. This chapter explains the FE studies that are carried out
to understand the OPB mechanism. At the end, several conclusions are drawn which will enable
the development of an interlink stiffness model.

3.2 Definitions
A few definitions are required before the FE study is presented. Figure 3.1 explains some of the
definitions that are used consistently in this thesis. The x and z axes are shown in the figure while
y axis goes inside the plane of the figure. The connection between links A and B and between
links B and C will be termed as Connection 1 and Connection 2 respectively. Points A and B lie
diametrically opposite to each other on Link A while points C and D lie diametrically opposite to
each other on Link B in the xz plane with y = 0. The angle that forms between line AB and line
CD will be defined as the interlink angle (αint,F E ) between Link A and Link B.
The OPB stresses (σopb,F E ) on Link B will be calculated as follows:
σxx,P − σxx,Q
σopb,F E = (3.1)
2
where P and Q are two points at the centre of Link B with P lying diametrically opposite to Q
along z axis. σxx,P is the normal stress along x at P and σxx,Q is normal stress at Q. Equation 3.1
isolates the stresses due to OPB from tension and it is valid when the bending curvature in the
link is small ([20]). OPB Moment can be easily calculated by multiplying the OPB stress with
sectional modulus of the link.
Mopb = σopb,F EA ∗ Z (3.2)
3
where Z = π ∗ Dnom /64, Dnom is the nominal diameter of the link. It is important to note that
any variation in the OPB stresses or moments within link B is small and will be ignored. The
terms OPB stress and OPB moment will be used interchangeably as both differ only by a factor
of geometry (Z).

3.3 Methodology
This section describes the methodology followed in conducting a FE study of the chain links.
Rotation (φopb ) around y axis and tension (T) is applied at Link A in the face colored as yellow.
Link C is fixed in the face colored as orange. The sequence for tension loading is described in figure
3.2. Between 0 s and 5 s, a load at 70% MBL is applied and released and at the 6th second, the
tension is brought to initial tension in the chain link system. This is done to simulate the effect
of proof testing and initial loading of chain links. Rotations are applied only after the 6th second.
The analysis is performed quasi-statically til the 6th second and dynamically after the 6th second.
ANSYS dynamic analysis is carried out because the frequencies are quite low for floater rotations
(below 0.25 Hz) and it is faster than nonlinear quasi-static analysis.

17
18 CHAPTER 3. FE STUDIES ON OPB MECHANISM

Figure 3.1: Definitions used

Figure 3.2: Tension history

From this point onwards, all loading sequences (tension and rotation) will include the time
elapsed during proof loading and initial loading while not explicitly mentioning about it. For
instance, if the statement says tension is kept constant, it implies that tension is kept constant
only after proof load and initial load is applied to the system. Similarly, if the statement says
rotation is kept constant, it implies that rotation is zero during the time elapsed for proof loading
and initial loading and then, it is kept constant at a certain value.
The state of the system at the beginning of the seventh second will be the initial state and this
will be taken as the starting point of every simulation for the simplified models.

3.4 FE studies and observations


The following observations are made after FEA of the 3 link chain system by keeping tension
constant at 60 tons. Additionally, FE studies are also performed for a 7 link chain and they are
mentioned in this section.

3.4.1 Rate Independency


Figure 3.3 shows a plot of OPB stress vs interlink angle for two different simulations. One simula-
tion was executed by applying one cycle of input rotation quasi-statically while the other simulation
was executed by applying the same rotation at 0.25 Hz. The floater rotation amplitudes are chosen
in such a way that the sliding limit is not reached. Figure 3.3 proves that in the presiding regime,
the relationship between OPB stress and interlink angle is rate independent. This relationship is
also rate independent in the post sliding regime as evident from figure 2.18.
3.4. FE STUDIES AND OBSERVATIONS 19

Figure 3.3: Rate independency in pre-sliding

3.4.2 Memory effect


Figure 3.4 shows the rotation input while figure 3.5 shows the resulting relationship between OPB
stress and interlink angle. The rotation is applied in such a way that two inner loops form inside one
outer hysteresis loop. When the inner loop closes, it follows the trajectory of the outer loop. This
indicates the presence of memory in the hysteritic relationship between OPB stress and interlink
angle [3].

Figure 3.4: Rotation input to check memory effect

3.4.3 Stick Slip


The vertical (z) displacements for two points (B and C) as shown in figure 3.1 are plotted in figure
3.6. The two points belong to link A and B respectively and they lie in close vicinity of each other.
Figure 3.7 reveals that the two points stick when the OPB stress-interlink angle curve is within
the sliding limits while they tend to slip away from each other when the curve reaches the sliding
20 CHAPTER 3. FE STUDIES ON OPB MECHANISM

Figure 3.5: Existence of memory

threshold.

Figure 3.6: Vertical displacement of points B and C

3.4.4 FE studies on a 7 link chain


Variation of OPB stress across different links
A 7 link solid FE model of the chain links (figure 3.8) is prepared to study the variation of OPB
stress and interlink angles across various links. The goal was to observe if the first link is indeed
the most crucial link with regards to OPB fatigue. Varying tension and rotation is applied to the
3.5. INFERENCE AND DISCUSSION 21

Figure 3.7: Stick slip behaviour

FE model.

Figure 3.8: 7 link FE model

Figure 3.9 shows a the variation of the OPB stress across the OPB links in the chain and
figure 3.10 shows the variation of interlink angles. The OPB stress at Link B is maximum while
the interlink angle at Connection 1 (between A and B) is far greater than the interlink angles at
the other connections. The interlink angle decreases in magnitude as one moves towards the fixed
boundary.

Effect of boundary condition


The goal of this study was to observe if OPB stress in the first free link is dependent on the
boundary condition. 3 link, 4 link, 7 link and 15 link solid FE models are prepared. All the FE
models are subjected to the same constant tension (60 tons) and one cycle of rotation between +2
and -2 degrees. The OPB stress in the first free link (link B) for all the models is plotted in figure
3.11.
As one can observe, the OPB stress in the first free link becomes almost independent of the
number of links of boundary condition when more than 7 links are used.

3.5 Inference and discussion


The observations from the FE study leads to a conclusion that the hysteritic relationship between
OPB stress and interlink angle follow a friction model. The frictional behaviour arises due to a
complex interaction between the link surfaces. It displays two regimes: a pre-sliding (sticking)
regime and a sliding regime (figure 3.7).
22 CHAPTER 3. FE STUDIES ON OPB MECHANISM

Figure 3.9: Variation of OPB stress(7 link model)

Figure 3.10: Variation of interlink angles (7 link model)

In the sticking regime, the relative displacement between the two contacting surfaces is in-
finitesimal. The friction force in this regime is mainly a function of displacement, and is due to
the adhesive forces derived from the asperity junction elasto-plastic deformation [8]. The force-
displacement curve (or as in this case OPB stress- interlink angle curve) exhibits a nonlinear
hysteritic relationship.
As the displacement increases, more and more junctions break, and finally there is a breakaway
displacement beyond which gross sliding (the sliding regime) begins. In the sliding regime the
asperity junctions have been broken, and a macroscopic relative displacement of the surfaces in
contact takes place. Usually, the friction force is a function of the relative velocity in this regime
[8]. In the case of OPB stresses/moments-interlink angles, any dependence on velocity is absent as
the analysis is performed at very low frequencies. Both the sticking and sliding regimes are rate
3.5. INFERENCE AND DISCUSSION 23

Figure 3.11: Effect of boundary condition on OPB stress

Figure 3.12: Friction behaviour of a block resting on a rough surface [8]

independent.
An analogy for the OPB stress-interlink angle relationship can be obtained by studying the
frictional behaviour of a block resting on a rough surface excited by a force Fapp (figure 3.12). The
two frictional regimes are shown in the figure. In this case, the gross sliding regime is a function
of the block’s velocity. The existence of memory effect and rate independency is also evident from
the figure when the force Fapp is applied at a frequency 50 times higher than the previous case.
24 CHAPTER 3. FE STUDIES ON OPB MECHANISM

3.5.1 On FE studies on a 7 link chain


The FE studies on the 7 link model proves that the first free OPB link next to the loading source
is most crucial with regards to OPB fatigue. The OPB stress is also observed to be independent
of the boundary condition when more than 7 links are used. When rotation is applied at one end
of the system of links, it actually transmits a combination of shear force and moment. This shear
force increases towards the fixed boundary of the chain link system. This results in an increase in
resistance towards interlink rotation. Since, OPB stress is dependent on interlink rotation, it also
decreases in magnitude.

3.6 Conclusion
In this chapter, FE studies were performed on a three link and seven link chain. The observations
from FE study on three link model leads to a conclusion that a friction model will be required to
idealize the connection between two links. This friction model should be capture the sticking and
sliding regimes as well as the memory of the system in the hysteretic relationship. Such a model
will be developed in the next chapter (Chapter 4).
Chapter 4

A New Interlink Stiffness Model

4.1 Introduction
This chapter will introduce a new interlink stiffness model which will describe the observed non-
linear hysteritic relationship between OPB stresses or moments and interlink angle. This model
has been developed after performing FE studies on the three link model.

4.2 Hypothesis
The hysteritic relationship between OPB stresses or moments and interlink angles is due to friction
at the link contacts.

4.3 Maxwell Slip Model: A friction model


A model that will faithfully simulate the relationship between OPB stress/moment and interlink
angle should posses the following characteristics:

1. It should represent the sticking and sliding regimes.

2. It should take the system’s memory into account while simulating the hysteritic relationship.

There is a well known model in friction modeling known as the Maxwell Slip Model, which can
simulate the hysteritic relationship between frictional force and displacement ([26] [4]) by taking
into account the memory effect. It utilizes a piecewise approximation of the nonlinear hysteresis
curve through the use of a number of elementary operators subjected to Coulomb friction. In this

Figure 4.1: Idealization of a block resting on a surface

section, a modified version of Maxwell Slip Model is presented for conceptual understanding. One
can understand this model by considering the simple case of block resting on a surface (figure 3.12).
This scenario can be idealized in figure 4.1 where the block is connected to the surface through four
springs or Maxwell elements. Each spring i has a characteristic stiffness ki and a sliding threshold
Wi . When a displacement x is applied, forces Fi will be generated in each spring according to its
characteristic stiffness. This is the sticking region and it will continue until the sliding limit Wi
is reached. After this point, only constant force equal to the sliding limit will be generated in the

25
26 CHAPTER 4. A NEW INTERLINK STIFFNESS MODEL

springs. When the direction of displacement is reversed, spring forces will reduce in magnitude
until the sliding limit −Wi is reached after which the forces become constant again. The total
frictional force F is a summation of the forces generated in each spring. Mathematically, it can be
expressed as follows:
4
X
F (t) = Fi (t), (4.1)
i=1

where the individual spring forces Fi are expressed as:


 ∂Fi (t−δt)
min{Fi (t − δt) +
 ∂t δt, Wi }, if x(t) − x(t − δt) ≥ 0
Fi (t) = (4.2)
 ∂Fi (t−δt)
max{Fi (t − δt) + δt, −Wi } otherwise

∂t

with
∂Fi (t) ∂ki x(t)
= (4.3)
∂t ∂t

A first order Taylor’s Series Expansion in Equation 4.2 about time t − δt. A hysteresis plot
generated using one Maxwell Element is shown in figure 4.2

Figure 4.2: Hysteresis plot generated with one Maxwell Element(k1 = 100,W1 = 55)

This type of modeling is called grey-box modeling because it is a partial theoretical structure
which needs data to complete the model. The parameters ki and Wi can be obtained by using a
suitable system identification technique. A

4.4 Interlink Stiffness Model


The Maxwell Slip model can be used to explain the hysteritic relationship between OPB stress and
interlink angle. In this section, this model will be first adapted to a case of constant tension and
later modified to a case of varying tension.
4.4. INTERLINK STIFFNESS MODEL 27

4.4.1 Constant Tension


Methodology

A random low frequency input rotation (φopb ) is generated (figure 4.3) for 55 s while the tension is
kept constant at 60 tons. Using these input signals, FEA of the three link chain is performed. The
OPB stresses (σopb,F E ) obtained are used for validation while the interlink angles (αint,F E ) that
are generated will be used as an input to the interlink stiffness model. The goal is to accurately
predict the OPB stresses if the interlink angles at Connection 1 are known.

Figure 4.3: Rotation input at constant Tension

Formulation

Figure 4.4 shows an idealization of the connection between Link A and Link B (Connection 1). 4
rotational springs are placed between Link A and Link B. According to [8], 4 springs are sufficient
for satisfactory results.

Figure 4.4: Idealization of the connection between Links A and B

Each rotational spring has a characteristic stiffness kri and sliding threshold Wi . Link C is
not considered because the moments generated at Connection 2 will be equal and opposite to the
moments generated at Connection 1 when only interlink angle is imposed between A and B. Since
OPB stresses and moments differ only by a factor of geometry (see equation 3.2), equation 4.1 and
4.2 can be modified without any loss of generality:

4
X
σopb (t) = σopb,i (t), (4.4)
i=1
28 CHAPTER 4. A NEW INTERLINK STIFFNESS MODEL

where
∂σopb,i (t−δt)

min{σopb (t − δt) +
 ∂t δt, Wi }, if αint,F E (t) − αint,F E (t − δt) ≥ 0
σopb,i (t) = (4.5)
 ∂σopb,i (t−δt)
max{σopb (t − δt) + δt, −Wi } otherwise

∂t

and
∂σopb,i (t) ∂ kri αint,F E (t)
= (4.6)
∂t ∂t
σopb is the OPB stress calculated by the interlink stiffness model and σopb,i is analogous to
OPB moment generated by each rotational spring. In equation 4.6, kri is a constant. The unit of
tension (T) is tons and the unit of interlink angle (αint,F E ) is degrees. It will be later shown how
these equations can easily be modified to calculate OPB moments.
In the above expression, there are eight unknowns, 4 kri and 4 Wi . To determine these un-
knowns, an optimization scheme has to be used based. The first 10 seconds of σopb,F EA will be
used as a training set to obtain the unknowns. The cost function J is first formulated as:
10
X
J= {σopb,F E (t) − σopb (t)}2 , (4.7)
t=0

The optimization problem is formulated as:

{k̂r, Ŵ} = arg min J(kr, W) (4.8)


kr,W

where kr = [kr1 , kr2 , kr3 , kr4 ] and W = [W1 , W2 , W3 , W4 ]. k̂r,Ŵ is a solution to the optimization
problem.
A non-linear regression based on Genetic Algorithm (see Appendix D) is used to solve the
optimization problem which searches a large parameter space to locate points where a global or
local minima may exist.

Results
The optimization problem 4.8 is solved using MATLAB’s Genetic Algortithm toolbox and the
solution obtained is shown in table 4.1.

k̂r Ŵ
95.158 88.058
0.039 0.216
1.741 7.794
16.491 0

Table 4.1: Parameters obtained after solving 4.8

Once the parameters are obtained, they are plugged into equations 4.5 and 4.6 and the OPB
stress is generated using equation 4.4. Figure 4.5 shows a comparison between OPB stresses
generated by the interlink stiffness model for constant tension and the OPB stresses from FEA.
The plots have been transformed such that the lower sliding threshold is 0.
The stresses generated by the interlink stiffness model are in excellent agreement with the
ones obtained fromm FEA. The goodness of fit (figure 4.6) is based on Normalized Mean Square
Error (NMSE) ([18]) which gives a number between 1 and -infinity with 1 being an excellent fit
and -infinity being a very bad fit. The goodness of fit in this case is 0.977. This shows that
the interlink stiffness based on the Maxwell Slip Model works quite well for the case of constant
tension. Figure 4.7 shows a comparison of the hysteresis plots between FEA and interlink stiffness
model for constant tension. The piecewise linear approximation of the nonlinear hysterisis curve
is evident from this figure.

4.4.2 Varying Tension


In the previous case, an interlink stiffness model is developed by considering tension as constant.
However, a complete interlink stiffness model must take into account the varying tension input.
4.4. INTERLINK STIFFNESS MODEL 29

Figure 4.5: Comparison of OPB stress between FEA and interlink stiffness model(constant Tension)

Figure 4.6: Goodness of Fit: Interlink Stiffness Model(constant Tension) vs FEA

Methodology
A random low frequency rotation (φopb ) (figure 4.8) and tension (T ) (figure 4.9) signal is generated
for 100 s. Using these input signals, FEA of the three link chain is performed. The OPB stresses
(σopb,F E ) obtained are used for validation while the interlink angles (αint )(figure 4.10) from FEA
will be used as an input to the interlink stiffness model. The goal is to accurately predict the OPB
stresses if the interlink angles at Connection 1 are known.

Formulation
The idealization is similar to the previous case (see figure 4.4) except now, the rotational springs
(kri ) and the sliding thresholds(Wi ) will change due to varying tension. When the tension increases,
the springs will become stiffner and the sliding threshold will increase. This is based on the interlink
30 CHAPTER 4. A NEW INTERLINK STIFFNESS MODEL

Figure 4.7: Interlink stiffness model Hysteresis vs FEA

Figure 4.8: Rotation input

Figure 4.9: Tension input


4.4. INTERLINK STIFFNESS MODEL 31

Figure 4.10: Interlink angles generated after FEA

stiffness model (see equation 1.1) which was developed in [6] and [23]. One can also think of the
increase in sliding threshold as a consequence of Coulomb’s law of friction which states that limiting
friction is directly proportional to normal force [29]. Due to plastic deformation at the link contacts,
the proportionality constant will not be the friction coefficient, but some unknown factor which
needs to be determined.
Using these ideas, an attempt will be made to modify the rotational springs and the sliding
thresholds to take into account the effect of changing tension:

kri = ai T (t)bi (4.9)

Wi = ci T (t) + di (4.10)

The formulation for kri is inspired from equation 1.1 while the formulation for Wi is based on
Coulomb’s law.
Equations 4.9 and 4.10 are inserted into equation 4.6 to give the following expression:

∂σopb,i (t) dT dαint,F E


= ai {αint,F E (t)bi T bi −1 + T bi (t) } (4.11)
∂t dt dt
Equations 4.11 and 4.10 are inserted in 4.5 and the OPB stress is calculated using 4.4. There
are now 16 unknowns,4 for each ai , bi , ci and di . The first 20 seconds of OPB stress from FEA
is used as a training set to obtain the unknown parameters. The cost function is formulated as
follows:
X20
J= {σopb,F E (t) − σopb (t)}2 , (4.12)
t=0

The optimization problem is formulated as follows:

{â, b̂, ĉ, d̂} = arg min J(a, b, c, d) (4.13)


a,b,c,d

where a = [a1 , a2 , a3 , a4 ], b = [b1 , b2 , b3 , b4 ], c = [c1 , c2 , c3 , c4 ] and d = [d1 , d2 , d3 , d4 ]. â,b̂,ĉ,d̂ is


a solution to the optimization problem. Genetic Algorithm (see Appendix D) is used to solve the
above optimization problem. A time step (δt) of 0.05 s is chosen which is the same time-step used
by FEA.
32 CHAPTER 4. A NEW INTERLINK STIFFNESS MODEL

â b̂ ĉ d̂
0.027 0.037 320.256 41.95
0.659 0.002 170.11 45.62
0.012 0.08 260.545 48.08
12.789 0.41 0.622 5.14

Table 4.2: Parameters obtained after solving 4.13

Results

Table 4.2 shows the values of the parameters obtained solving the optimization problem.
The values obtained from table 4.2 are plugged into equations 4.9 and 4.10 and the OPB
stresses are obtained using equations 4.4, 4.5 and 4.6. The comparison between the OPB stresses
predicted by the interlink stiffness model and FEA is shown in figure 4.11 and it shows an excellent
agreement. The goodness of fit using the NMSE criteria is 0.94 (figure 4.12). This shows that the
parametrisation the rotational stiffness and sliding threshold is probably correct.

Figure 4.11: Comparison of OPB stress between FEA and interlink stiffness model

Figure 4.13 shows a comparison of the hysteresis plots between FEA and the interlink stiffness
model.

4.4.3 Additional Results


Some additional studies were carried out to determine the robustness, sensitivity to time-step and
reproducibility of the interlink stiffness model. They are described in this section.

Robustness

To evaluate the robustness of the interlink stiffness model, the model parameters were changed by
1% and the results were compared to the original model as well as results from FEA (figure 4.14).
The results are invariant with respect to any minor change in the values of the parameters. Thus,
it is concluded that the interlink stiffness model is very robust.
4.4. INTERLINK STIFFNESS MODEL 33

Figure 4.12: Goodness of Fit: Interlink Stiffness Model(constant Tension) vs FEA

Figure 4.13: Interlink stiffness model Hysteresis vs FEA

Sensitivity to time-step

Since, Taylor’s expansion was used to construct the interlink stifnees model, as sensitivity study
was carried with respect to the size of the time-step (δt). The OPB stress is plotted for four
different time- steps and compared to the results of FEA (figure 4.15). The original time step for
the interlink stiffness model is taken was 0.05 s. From the results, it can be concluded that model
is not very sensitive to any change in time-step.
34 CHAPTER 4. A NEW INTERLINK STIFFNESS MODEL

Figure 4.14: Robustness of interlink stiffness model

Figure 4.15: Sensitivity to time-step

Reproducibility

The aim of this study was to check if the interlink stiffness model can give correct predictions of
OPB stress for different inputs, if the initial state of the system is kept same. A different rotation
and tension signal for 50 seconds is generated and FEA is performed without changing the proof
load and initial tension in the system. Using the interlink angles generated by FEA, OPB stress
is generated using the interlink stiffness model and the results are compared to OPB stress from
FEA (figure 4.16). From the comparison, it can be observed that the interlink stiffness model can
reproduce results for a different input when the initial state is kept the same.
4.5. CONCLUSION 35

Figure 4.16: Reproducibility of interlink stiffness model

4.5 Conclusion
In this chapter, a new interlink stiffness model is developed. This model gives a relationship of
OPB stresses and interlink angles. It captures the sticking and sliding regimes of OPB stresses
while at the same time it includes the system’s memory in simulating the hysteritic relationship.
The model has been demonstrated to work quite well for both rotation and varying tension as
input. The model is quite robust and insensitive to time-step. It can be used for any other input
time-traces provided the initial state of the system (chain link geometry and material, proof load
and initial tension) is kept same. As stated earlier, this model can easily be modified to calculate
the OPB moments(Mopb ) by changing equations as follows:
Mopb (t) = σopb (t)Z (4.14)
wgere Z is the section modulus (see equation 3.2). The interlink stiffness model can also be used
in case of studded links. The parameters of the model can easily be derived from an actual
experimental set-up.
A major disadvantage of the interlink stiffness model is that it cannot calculate the stresses
on a IPB link (link C). It is because this model is derived assuming that OPB link behaves like
an Euler-Bernoulli beam ([20]). This is true for small deflections when the thickness is small
compared to the length of the beam. However, this does not hold true for the IPB link because
the thickness is comparable to the length of the beam. Shear deformations cannot be ignored in
such a case. Figure 4.17 illustrates a deformation in an IPB link magnified by a factor of 2.

Figure 4.17: Deformations in an IPB link

In the current chapter, the interlink angles were known from FEA. In the next two chapters,
simplified models idealizing the 3 chain links will be developed by including the interlink stiffness
model. The simplified models will predict both the interlink angles and OPB stress.
36 CHAPTER 4. A NEW INTERLINK STIFFNESS MODEL
Chapter 5

A Physics Based Approach

5.1 Introduction
In the Chapter 3, an interlink stiffness model is developed. This model can predict the OPB
moments or stresses if the interlink angles are known. The model is stated as follows:
4
X
Mopb (t) = Mopb,i (t) (5.1)
i=1

where
∂Mopb,i (t−δt)

min{Mopb (t − δt) +
 ∂t δt, (ci T + di ) Z}, if αint (t) − αint (t − δt) ≥ 0
Mopb,i (t) =
 ∂Mopb,i (t−δt)
max{Mopb (t − δt) + δt, −(ci T + di ) Z} otherwise

∂t
(5.2)
and
∂Mopb,i (t) dT dαint
= ai Z{αint (t)bi T bi −1 + T bi (t) } (5.3)
∂t dt dt
In this chapter, a simplified model idealizing the three links as beams is developed. The beams
are connected to each other by a rotational spring which is given by the derivative of the interlink
stiffness model. This approach is termed as ’physics based’ because the simplified model will
attempt to reproduce the physics of an actual three link chain system subjected to tension and
rotation. Additionally it should perform the following tasks:
1. It should give a reasonable prediction of OPB stresses and interlink angles in time domain
compared to a solid FE model.
2. It should take less time to calculate than a solid FE model.

5.2 Idealization
Figure 5.1 shows the simplified model of the three link chain system. The links are connected to
each other by a rotational spring kopb . Rotation φopb (t) (figure 4.8 ) and Tension T (t) (figure 4.9
) is applied at the end of Link C while Link A is fixed. The links have been simplified to beams
with uniform circular cross-section. The length and diameter of the IPB link are lipb and dipb
respectively while that of the OPB link are lopb and dopb respectively . The process of simplifying
the link geometry is explained in the next section. The axial and the vertical degrees of freedom
are shared between the links at their connection but the rotational degree of freedom is not shared.
The interlink angle at Connection 1 (αint,c1 ) is given as:
αint,c1 = θ2 − θ1 ; (5.4)
The interlink angle at Connection 2 (αint,c2 ) is given as:
αint,c2 = θ3 − θ4 ; (5.5)
The OPB stress (σopb ) is noted at location P which is in the middle of the link B.

37
38 CHAPTER 5. A PHYSICS BASED APPROACH

Figure 5.1: Idealization of a 3 link chain

5.3 Equivalent Link Geometry


Due to the complicated geometry of the links, especially the curved part, it is difficult to discretize
the links and use in a numerical method. The method which is proposed here reduces the links to
simple beams with uniform circular cross-section of equivalent bending stiffness. This is done by
performing cantilever bending tests of the solid FE model of IPB link (figure 5.2) and OPB link
(figure 5.3). A range of forces (P) from 100 N to 5000 N is applied and displacement at the tip
(∇FE ) is obtained. If the length of the simple beam is denoted as l and the diameter is denoted
as d, the displacement of the cantilever beam at the tip can be analytically calculated as:

Pl3
∇= ; (5.6)
3EI
where
πd4
I= (5.7)
64

Figure 5.2: Cantilever bending of IPB link

Figure 5.3: Cantilever bending of OPB link


5.4. FORMULATION 39

The equivalent length (l) and diameter (d) can be obtained by optimizing the beam to give
similar displacements as the FE model. The cost function is formulated as:
n
X
Jlink = (∇FE − ∇)2 ; (5.8)
i=1

where n represents the number of test cases. The optimization problem is formulated as follows:
(l, d) = arg min Jlink (l, d) (5.9)
l,d

The solution (l, d) is sought such that it they are in the neighbourhood of the nominal dimensions
of 124 mm link. The optimization problem 5.9 is solved for both the IPB link and the OPB link
using Genetic Algorithm and the results are summarized in tables 5.1 and 5.2. It should be noted
that the equivalent diameter of the OPB link is smaller than IPB link. This justifies the fact
that the IPB link has a greater bending stiffness than the OPB link. Another point to note that
the dimensions obtained by this method is only used to discretize the equations in the governing
equation of motion. One cannot use these dimensions directly to calculate the stresses or moments.

l (m) d (m)
0.665 0.145

Table 5.1: Equivalent beam dimension of IPB link

l(m) d(m)
0.724 0.112

Table 5.2: Equivalent beam dimension of OPB link

5.4 Formulation
A finite element method is used to formulate the 2D beam model of the chain links. The stiffness
matrix (K) is formulated as follows:
K = Kbeam + Kopb (5.10)
where Kbeam denotes the global stiffness matrix of the beams [10]. It is assembled from
elemental stiffness matrix (kbeam ) of a beam (of length L and moment of inertia I) given as follows
 
12 6L −12 6L
EI  6L 4L2 −6L 2L2 
kbeam = 3   (5.11)
L −12 −6L 12 −6L
2 2
6L 2L −6L 4L
Kopb is the global stiffness matrix due of the rotational springs. The local stiffness matrix of
the rotational spring (kopb ) is given as
 
kopb −kopb
kopb = (5.12)
−kopb kopb
where kopb is derived by taking a derivative of equation 5.1.
∂Mopb
kopb = (5.13)
∂αint
Kopb is assembled from the local stiffness matrices of the two rotational springs. It must be borne
in mind that Kopb is not only a function of the displacement at any point of time, but also on the
path taken by Kopb as a function of displacement until that point of time. This is because of the
memory effect which was incorporated in deriving the interlink stiffness model. If X represents
the displacement of the system, the governing equation will be given as:
K(Kt−1 , Kt−2 . . . K0 , X) X = F(t) (5.14)
40 CHAPTER 5. A PHYSICS BASED APPROACH

where
Kt−i = K(X(t − iδt)) (5.15)
The force matrix (F) contains forces due to applied rotation and tension. It is given as follows:
 

 0  
 0 

 

 
F(t) = .
.. (5.16)
 
M (t)
 
 φ 

 

T (t)

where Mφ (t) is a force matrix due to input rotation φopb (t) and T(t) is the applied tension.
5.14 is a non-linear equation which describes a quasi-static process. The links are discretised
into 10 elements each. The initial displacement is calculated as:

K0 X0 = F(0) (5.17)

where F(0) is the force matrix due to initial tension and zero rotation. K0 is the initial stiffness
matrix and it is does not include the nonlinear rotational springs. Proof loading is not applied
directly because but its effect is included in Kopb matrix. A numerical method based on fixed
point iteration (Appendix E) is used to solve the non-linear problem. The equation is solved in a
sequence of time steps to obtain the solution X in time domain. The process is quasi-static because
of the absence of any inertial effects in the frequencies of interest.

5.5 Results
Due to convergence issues, equation 5.14 is solved iteratively using a variable convergence criteria.
The convergence criteria is relaxed at certain time steps to enable when convergence difficulty
arises. Figure 5.4 shows a comparison of the plot of interlink angles between the simplified physics
based model and FEA at Connection 1 and figure 5.5 shows the same plot for Connection 2.
Although they show some agreement, there is a substantial amount of noise.

Figure 5.4: Time domain plot of interlink angles at Connection 1

The numerical noise is seen because the interlink stiffness model uses a piecewise linear approx-
imation of the nonlinear OPB moment-interlink angle curve. As a result, there are several points
in the curve where the derivative (kopb ) does not exist. Figure 5.6 illustrates this point. One
such point appears when the links slide with respect to each other. Figure 5.7 shows a plot of the
normalized kopb as a function of time when the interlink angles from FEA is used as an input. As
5.5. RESULTS 41

Figure 5.5: Time domain plot of interlink angles at Connection 1

one can observe, there are several instances where discontinuity in (kopb ) exists. In fixed-point it-
eration, the function f (Appendix E) has to be Lipschitz continuous in order to obtain a converged
solution.

Figure 5.6: Discontinuity of kopb

The OPB stress is directly calculated using the following relationship:

Mopb (t)
σopb (t) = (5.18)
Z
where Z is the sectional modulus of the OPB link given by equation 3.2. This formulation is
possible because the effect of shear force across the OPB link is negligible. Figure 5.8 shows a
comparison of the OPB stress between the simplified model and FEA.
42 CHAPTER 5. A PHYSICS BASED APPROACH

Figure 5.7: Normalized interlink stiffness as a function of time

Figure 5.8: Time domain plot of OPB stress

In the next section, an alternative approach is suggested model so that the obtained solutions
are smooth.

5.6 Alternative approach


Due to the non-smoothness of the solution, an alternative model is proposed. This model is shown
in figure 5.9. The rotational spring between link B and C is dropped and a rigid connection is
5.6. ALTERNATIVE APPROACH 43

made. The axial and vertical degrees of freedom at the connection between A and B are shared.
The stiffness matrix due the rotational spring at Connection 1 is removed from the system’s global
stiffness matrix and adjusted as an external force in right hand side of equation 5.14. The system’s
stiffness matrix is now given as follows:

K = Kbeam (5.19)

The governing equation is given as:


K X = F(t) (5.20)
where the force matrix F(t) is formulated as follows:
 
 0 
..

 

 
.

 


 

−Mopb (t)

 

 

Mopb (t)
 
F(t) = (5.21)

 0 


 .. 

.

 


 

M (t)

 


 φ 

T (t)
 

where Mφ (t) is a force matrix due to input rotation φopb (t) and T(t) is the applied tension.
The OPB moment (Mopb (t)) is given by equations 5.1, 5.2 and 5.3. It is applied at the rotational
degree of freedom corresponding to θ1 and θ2 (figure 5.1)
The initial displacement is calculated as:

K0 X0 = F(0) (5.22)

where F(0) is the force matrix due to initial tension, zero rotation and zero OPB moment..

Figure 5.9: Idealization of a 3 link chain

Although equation 5.20 is nonlinear, iteration is not be required. One can consider it as a
system excited by a position dependent force, which is the OPB moment in this case. The system
is stable at all time due to the rigid connections.

5.6.1 Results
3 link model
Figure 5.10 is a comparison of the interlink angle at Connection 1 between FEA and the alternative
model suggested and it shows an excellent agreement. The OPB stress is calculated using equation
5.18. The comparison of OPB stress between FEA and the alternative model is shown in figure 5.11
which also shows an excellent agreement. The legend ’Physics based model’ should be interpreted
as the alternative physics based model.

7 link model
A comparison is also made with a 7 link FE model (see figure 3.8). The 7 link model is idealized in
figure 5.12. The rotation and tension applied to the 7 link FE model is same as that for the 3 link
FE model, but they are applied for 100 seconds instead of 200 seconds due to limited computational
44 CHAPTER 5. A PHYSICS BASED APPROACH

Figure 5.10: Time domain plot of interlink angles(3 link model)

Figure 5.11: Time domain plot of OPB stress (3link model

Figure 5.12: Idealization of a 7 link chain

resources. A comparison of interlink angle at Connection 1 (between Links A and B) between FEA
and the proposed alternative model is shown in figure 5.13and a comparison of OPB stress on Link
B is shown in figure 5.14. The comparisons are not as good as that of the three link model.
Although the proposed alternative model gives smooth solutions and predicts OPB stress rea-
5.7. CONCLUSION 45

Figure 5.13: Time domain plot of interlink angles (7 link model)

Figure 5.14: Time domain plot of OPB stress (7 link model)

sonably well, the model’s usefulness is limited to the first OPB link only. The model assumes
that OPB moments will be constant across the rest of the chain which is not the case. It also
cannot predict the interlink angle between Links B and C. In fact, The comparison with the 7
link model suggests that the interlink stiffness between the other links needs to be considered for
a more accurate calculation of the OPB stress.

5.7 Conclusion
In this chapter, a physics based model is first proposed. This model attempts to describe the
physics of an actual system of links. It does somewhat explain the observed behaviour but the
solution (interlink angles) obtained are characterized by numerical noise. The numerical noise is
due to the discontinuity in the interlink stiffness of the system. If the obtained solutions were
46 CHAPTER 5. A PHYSICS BASED APPROACH

smooth, a truly simplified model of a system of links would have been obtained which would have
justified the definition of a simplified model adopted in Chapter 1.
To overcome the problem of non-smooth solution, an alternative approach is proposed in which
the rotational spring between the first two links is included as an external force while the rotational
springs between rest of the links is replaced by a rigid connection. The alternative model gives
smooth and excellent predictions for the first OPB link at a fraction of time compared to a solid
FE model. But it does not satisfy the definition of a simplified model as it cannot be used to
predict stresses or interlink angles across other links.
Since the highest stresses occur in the first OPB link (see figure 3.9), this model can still be
applied to estimate OPB fatigue. A case study using this alternative approach is elucidated in
Chapter 7. From this point onwards, a reference to the physics model will imply the alternative
model which is proposed in this chapter.
Chapter 6

A Semi-Empirical Approach

6.1 Introduction
In Chapter 5, a system of chain links was idealized as a system of simple beams connected to each
other by a rotational hysteritic spring. In this chapter, a different approach to idealize and simplify
a system of chain links is presented. This approach is termed as a ’semi-empirical’ because the
simplified model will try to reproduce only the interlink angles and OPB stress using some physical
consideration while ignoring the overall behaviour of a system of links.

6.2 Idealization
Figure 6.1 shows an idealization of the three links. Each link is considered as a rigid bar. k1
and k3 are linear springs representing the bending stiffness of links A and C respectively. k2 is a
linear spring which is representative of the bending stiffness of link B and the interlink stiffness
between links B and C. Linear springs are used because the interlink angles at other connections
are quite small as compared to Connection 1 (see figure 3.10). The interlink stiffness between links
A and B is modeled as an external moment. This external moment is generated by four rotational
springs which takes interlink angle and tension as input. Each rotational spring has a characteristic
stiffness kri and a sliding threshold Wi .

kri = Ai T (t)Bi (6.1)

Wi = Ci T (t) (6.2)

Figure 6.1: Idealization of a 3 link chain (semi-empirical approach)

This idealization is quite similar to physics based model described in Chapter 4. Although
tension is not directly applied, it’s effect is included in the OPB moments. The interlink angle

47
48 CHAPTER 6. A SEMI-EMPIRICAL APPROACH

between links A and B is given as:


αint = θ2 − θ1 ; (6.3)
It is important to note that this idealization cannot calculate OPB stresses as the links are
considered rigid. However, if the interlink angles at connection 1 are known one can use the stress
version of interlink stiffness model to calculate the OPB stress.

6.3 Formulation
The governing equation of motion for the rigid link model is given as:
    
k2 + k3 −k2 0 θ3  0
 −k2 k2 0  θ2 =  −Mopb  (6.4)
0 0 k1 θ1 Mopb + k3 φopb (t)
 

Mopb is given by the interlink stiffness model as follows:


4
X
Mopb (t) = Mopb,i (t) (6.5)
i=1

where
∂Mopb,i (t−δt)

min{Mopb (t − δt) +
 ∂t δt, Ci T }, if αint (t) − αint (t − δt) ≥ 0
Mopb,i (t) = (6.6)
 ∂Mopb,i (t−δt)
max{Mopb (t − δt) + δt, −Ci T } otherwise

∂t

and
∂Mopb,i (t) dT dαint
= Ai {αint (t)Bi T Bi −1 + T Bi (t) } (6.7)
∂t dt dt
The unit of tension (T) is tons and the unit of interlink angle (αint ) is degrees in the above
expressions. The unknown parameters in the model are 3 linear springs given by k1 , k2 and k3 and
12 OPB moment parameters given by Ai ,Bi and Ci . These parameters are obtained by optimizing
the system with respect to observed interlink angles at Connection 1 from FEA. The first 20
seconds of the interlink angles obtained from FEA is used for training and the rest 180 s is used
for validation. The input rotation and tension is shown in figures 4.8 and 4.9 respectively.
Genetic Algorithm is employed to solve the optimization problem. The cost function is first
formulated as follows:
X20
J= {αint,F E (t) − αint (t)}2 , (6.8)
t=0
The optimization problem is formulated as follows:
{k̂, Â, B̂, Ĉ} = arg min J(k, A, B, C) (6.9)
k,A,B,C

where k = [k1 , k2 , k3 ], A = [A1 , A2 , A3 , A4 ], B = [B1 , B2 , B3 , B4 ] and C = [C1 , C2 , C3 , C4 ]


It is important to note that these OPB moment parameters are different from the ones used
in the interlink stiffness model because in this case, the system is optimized with respect to the
observed interlink angles only. Once the parameters are obtained, the governing equation 6.4 can
be solved to to calculate the interlink angles in time domain. The interlink angles are obtained
using equation 6.3. After the interlink angles are calculated, the stress version of the interlink
stiffness model (equations 4.4, 4.5 and 4.6) is used to calculate the OPB stress. The stress version
of the interlink stiffness model is stated again for the reader’s benefit.
4
X
σopb (t) = σopb,i (t), (6.10)
i=1

where
∂σopb,i (t−δt)

min{σopb (t − δt) +
 ∂t δt, (ci T + di )}, if αint (t) − αint (t − δt) ≥ 0
σopb,i (t) = (6.11)
 ∂σopb,i (t−δt)
max{σopb (t − δt) + δt, −(ci T + di )} otherwise

∂t
6.4. RESULTS 49

and
∂σopb,i (t) dT dαint
= ai {αint (t)bi T bi −1 + T bi (t) } (6.12)
∂t dt dt
The parameters ai , bi , ci and di are given in table 6.1

6.4 Results
6.4.1 3 link model
The parameters obtained after solving the optimization problem (equation 6.9)is shown in table
6.1 .

k̂ Â B̂ Ĉ
271704.26 30.588 0.284 37096.67
776532.4 60.025 0.201 16718.21
636131.54 34.104 0.271 13140.379
896.928 0.032 34290.86

Table 6.1: Parameters obtained after solving 6.9

Once the parameters are obtained, equations 6.4 and 6.3 are used to generate the interlink
angles. Then equations 6.10, 6.11, 6.12 and table 6.1 are used to calculate the OPB stress. Figure
6.2 shows a comparison of the interlink angles obtained using the semi emprical approach and the
interlink angles obtained from FEA. The agreement between FEA and the semi-empirical model
is quite good.

Figure 6.2: Time domain plot of interlink angles (3 link model)

Figure 6.3 shows a comparison of the OPB stress between FEA and the semi-empirical model.
The agreement in the comparison is excellent.

6.4.2 7 link model


This approach is also applied to idealize a 7 link chain. The 7 link chain is idealized as a system
of 3 rigid links as shown in figure 6.4. In this model, links C to G in the FE model is represented
by a single link C with an equivalent bending and interlink stiffness denoted by k3 . k2 is a linear
50 CHAPTER 6. A SEMI-EMPIRICAL APPROACH

Figure 6.3: Time domain plot of OPB stress (3 link model)

spring which is representative of the bending stiffness of link B and the interlink stiffness between
links B and rest of the chain from Links C to G. k1 denotes the bending stiffness of link A. The
interlink stiffness between links A and B is modeled as an external moment. This external moment
is generated by four rotational springs which takes interlink angle and tension as input. Each
rotational spring has a characteristic stiffness kri and a sliding threshold Wi .
The rotation and tension applied to the 7 link FE model is same as that for the 3 link FE model,
but the simulation time is taken as 100 seconds instead of 200 s due to limited computational
resources.

Figure 6.4: Idealization of a 7 link model

First, the unknown parameters are obtained in exactly the same way as they were obtained for
the 3 link model. A 10 s time trace of interlink angles at Connection 1 (between links A and B)
from FEA is used for training and the rest of the time trace is used for validation. The parameters
obtained after solving the optimization problem is given in table 6.3. The presence of the extra
column corresponding to D̂ is because the sliding threshold Wi for the 7 link model has been
expressed as follows:
Wi = Ci T (t) + Di (6.13)
6.4. RESULTS 51

k̂ Â B̂ Ĉ D̂
523352.775 4883.704 2132311.652 34101.545 1797.162
792101.367 13214.944 564723.294 34895.228 1167.798
935616.97 1296.319 915793.54 136933.686 1189.825
6853.139 460696.929 111426.146 1160.754

Table 6.2: Parameters obtained after solving 6.9 for 7 link model

Once the parameters are obtained equations 6.4 and 6.3 are used to generate the interlink angles.
Then equations 6.10, 6.11, 6.12 and table 6.1 are used to calculate the OPB stress. Figure 6.5
shows a comparison of the interlink angles obtained using this approach and the interlink angles
obtained from FEA. Figure 6.6 shows a comparison of the OPB stress between FEA and the semi-
empirical model. The agreement is not as good as the one obtained for the 3 link model. This is
because the interlink stiffness has been completely ignored between links C to G.

Figure 6.5: Time domain plot of OPB stress (7 link model)

Improving predictions for 7 link model


The prediction of OPB stress by the semi-empirical model for a 7 link chain is quite poor as
compared to the predictions for a 3 link chain. This is because the interlink stiffness has been
completely been ignored for links C to G. Due to the poor quality of predictions, an improvement
is suggested in this section. Instead of using the interlink stiffness model obtained from FEA of a
3 link chain (see equations 6.10, 6.11, 6.12 and table 6.1) to calculate OPB stress, a new interlink
stiffness model will be derived from FEA of a 7 link chain. This procedure is exactly same as
mentioned in section 4.4.2. The interlink stiffness model given by equations 6.10, 6.11 and 6.12 is
stated as follows:
4
X
σopb (t) = σopb,i (t), (6.14)
i=1

where
∂σopb,i (t−δt)

min{σopb (t − δt) +
 ∂t δt, (ci T + di )}, if αint (t) − αint (t − δt) ≥ 0
σopb,i (t) = (6.15)
 ∂σopb,i (t−δt)
max{σopb (t − δt) + δt, −(ci T + di )} otherwise

∂t
52 CHAPTER 6. A SEMI-EMPIRICAL APPROACH

Figure 6.6: Time domain plot of OPB stress (7 link model)

and
∂σopb,i (t) dT dαint
= ai {αint (t)bi T bi −1 + T bi (t) } (6.16)
∂t dt dt
The parameters of this model will be obtained from FEA of a 7 link model. Using 10 s of
the 100 s simulation as a training set, the above model is optimized with respect to OPB stress
obtained at Link B of the 7 link model (see figure 3.8). The cost function can be formulated as :
10
X
J= {σopb,F E (t) − σopb (t)}2 , (6.17)
t=0

The optimization problem is formulated as follows:

{â, b̂, ĉ, d̂} = arg min J(a, b, c, d) (6.18)


a,b,c,d

Genetic Algorithm is employed to solve the optimization problem given by equation 6.18. Table
displays the values of the parameters obtained.

â b̂ ĉ d̂
0 2.03 10.322 70.819
0 0.948 4.915 48.822
3.943 0.604 0.662 3.088
0.637 0 5.395 15.35

Table 6.3: Parameters obtained after solving 6.18

The whole process is summarized in as follows. First the interlink angles are generated using
equations 6.4, 6.3 and 6.3. Then OPB stress is calculated using the interlink stiffness model derived
from FEA of a 7 link model which is given by equations 6.14, 6.15, 6.16 and table 6.3. Figure 6.7
shows a comparison of OPB stress predicted by the semi-empirical approach and FEA. The results
show a much better agreement then what is observed in figure 6.6.

6.5 Conclusion
A semi-empirical approach is presented in this chapter. This approach idealizes the links as rigid
bars connected by linear rotational springs while the OPB moments is modeled as an external
force. The semi-empirical model does not satisfy the definition of a simplified model although it
6.5. CONCLUSION 53

Figure 6.7: Time domain plot of OPB stress (7 link model)

gives near excellent predictions of OPB stress and interlink angle. The main advantage of this
model lies in the fact that it uses just 3 degrees of freedom to describe a large system of links. As
a result, it is a much faster method of calculating the OPB stresses if the overall physics of the
chain is not of concern.
The difference between the semi-empirical model and the physics-based model is that the former
does not calculate the OPB stresses. The stress version of the interlink stiffness model is utilized
separately to calculate the OPB stresses. Also, the parameters of the semi-empirical model model
change if the number of links in the reference model changes. This happens even if dynamic and
kinematic initial conditions are kept the same.
54 CHAPTER 6. A SEMI-EMPIRICAL APPROACH
Chapter 7

A Case Study On Calculating


OPB Stress

7.1 Introduction
This chapter presents a case study on estimating nominal OPB stresses for one sea-state. The
models that have been developed in the previous chapters will be applied here. The goal of this
chapter is to demonstrate a new methodology to asses OPB fatigue.

7.2 Description
Moho Nord is an exploration and production project launched off the Congolese coast in March
2013 (figure 7.1). It is located in 750 to 1050 metres water depth . Moho Nord Field Development
consists of a TLP (Tensioned Leg Platform) type dry tree unit for the development of Albian
reserves and a subsea development for the production of Miocene reserves. Productions are sent
on board a FPU (Floating Production Unit), for separation and export. This case study presents
a methodology to determine the OPB stresses in the Moho Nord FPU mooring system for one sea
state. The OPB stresses will be used to estimate the fatigue damage of a mooring line in addition
to tension induced fatigue.

Figure 7.1: Location of Moho Noord field

The Moho Noord FPU mooring layout is shown in figure 7.2. The mooring system consists of
4 clusters, each cluster has three mooring lines. The global and local coordinate system is shown
in figure 7.3. The subscript ’G’ refers to the global coordinate system and the subscript ’L’ refers
to the local coordinate system. Figure 7.3 also shows the positive sign conventions adopted for

55
56 CHAPTER 7. A CASE STUDY ON CALCULATING OPB STRESS

Figure 7.2: Moho Noord FPU mooring layout

floater rotation. All floater motions are measured in the local coordinate system. Details of the top
chain are listed in table 7.2. The connection between the floater and the mooring line is situated
on the deck of the FPU (figure 7.6). The sea-state is described by a JONSWAP spectrum with a
significant height (Hs ) of 2 m and a peak period (Tp ) of 7 seconds as shown in figure 7.4.

Figure 7.3: Moho Noord FPU coordinate system and positive sign convention for floater motion

7.3 Assumptions
Before the methodology to calculate OPB stresses is presented, some assumptions are made. They
are as follows:

1. The OPB mechanism in mooring chain links occur if tension in the mooring line is more than
7.4. METHODOLOGY 57

Figure 7.4: JONSWAP spectrum describing the 1 hour sea-state

Length from H-link to stopper (m) 68m – 118m


Type / Quality R4 Studless
Intact diameter (mm) 152
Corroded diameter (mm) 137
MBL (intact) (kN) 16405
MBL (corroded – 15mm) (kN) 13829
Weight in air (kg/m) 462
Weight in sea water (kg/m) 401.7
Axial Stiffness EA (MN) 1355

Table 7.1: Characteristics of top chain of Moho Noord FPU mooring lines

5% of MBL (Minimum breaking Load).

2. The first free OPB link connected to the floater is the most crucial link with regards to OPB
fatigue.

3. The orientation of the chain links that represents the top chain will be such that their plane
of geometry are perpendicular to each other. This configuration will be fixed so that a worst
case scenario is ensured.

4. The relative angular rotation between the mooring line and the floater is small.

5. Only relative rotations out of plane of the first link are considered. Torsion and relative
rotation in the plane of the links are ignored.

6. Influence of floater translation is ignored.

7. OPB of chain links does not affect the dynamics of the mooring line (tension and shape of
the line). It dissipates after the first few links of the top chain.

8. Current and wave forces contribute to negligible bending in the chain links as compared to
relative rotation between the floater and the mooring line.

7.4 Methodology
The methodology is summarized in the flow chart shown in figure 7.5. A decoupled approach is
followed in which coupled floater-mooring analysis is performed first for one sea state (figure 7.4).
The mooring line where OPB stresses need to be assessed is isolated. A simplified model of a
58 CHAPTER 7. A CASE STUDY ON CALCULATING OPB STRESS

system of links based on physics based approach (Chapter 5) is prepared which is representative
of the first few links of the top chain (figure 7.6). The mooring line restoring force that acts on the
floater and the relative rotation between the floater and the mooring line are obtained from coupled
floater-mooring analysis and they are used as an input to the simplified model. The output of the
simplified model will be the OPB stresses and interlink angles for that sea state. This methodology
will be described in detail in relation to the case study in the following sections.

Figure 7.5: Flow chart depicting the methodology for estimating nominal OPB stress

Figure 7.6: Simplified OPB model representing the first few links of the top chain

7.4.1 Step 1: Post-processing after coupled floater-mooring analysis


Coupled floater-mooring calculations of Moho Nord FPU is performed in ANYSIM for a 1 hour
sea state (figure 7.4). The wind and wave direction is taken along positive xG in the global
coordinate system (figure 7.3) for this sea-state. Line S6 (figure 7.2) is chosen for OPB fatigue
assessment as it will have the highest tension for this sea state. The output of the coupled floater
mooring calculations are the mooring forces in line S6 and floater rotations. These quantities will
be post-processed to prepare inputs for evaluating OPB stresses using a simplified model.

Tension
The components of the mooring line force are Fx which acts along xL direction and Fy which acts
along yL direction, at any time t. Since the calculations are performed in 3D, there will be an
additional component Fz which acts along a line perpendicular to the plane of figure 7.3(zL ). The
mooring line force F~ is given as:

F~ = Fx î + Fy ĵ + Fy k̂ (7.1)
7.4. METHODOLOGY 59

î, ĵ and k̂ are unit vectors along positive xL , yL and zL axes respectively.
The tension T can thus be calculated as:
q
T = |F~ | = Fx2 + Fy2 + Fz2 (7.2)

Since tension T changes in time, it will be written as T (t). The tension in the mooring line is
shown in figure 7.7.

Figure 7.7: Relative rotation between FPU and mooring line

Relative rotation
The concept of relative rotation is illustrated in figure 7.7. Relative rotation (φopb (t)) between the
floater and the mooring line is required as an input to calculate OPB stress. Let u~f denote a unit
vector along xL (see figure 7.3). u~f rotates with yaw (ψ) of the vessel. It can be expressed as
follows:
u~f = cos(ψ)î + sin(ψ)ĵ (7.3)

The projection of mooring line force F~ in the xL yL plane is expressed as follows:

F~xy = Fx î + Fy ĵ (7.4)

The angle γ between Fx y and uf in the xy plane is calculated as:

F~xy · u~f
γ = cos−1 (7.5)
|F~xy |

where |F~xy | is given as


q
|F~xy | = Fx2 + Fy2 (7.6)

Let ~v be a unit vector in zL direction which is fixed at the connection of the mooring line and
the floater (figure 7.3). It can be expressed as follows:

~v = −k̂ (7.7)

Once γ is known, the rotation(β) of the vessel in a plane defined by the mooring line force F~
and the fixed vertical unit vector ~v can be calculated as follows:

β = θ cos γ + φ sin γ (7.8)


60 CHAPTER 7. A CASE STUDY ON CALCULATING OPB STRESS

where θ is the pitch of the vessel and φ is the roll of the vessel. The change in β with respect to
the initial rotation is calculated as:

∆f = β(t) − β(0) (7.9)

The angle (λ) between the mooring line force F~ and the fixed vertical vector ~v at the connection
of the mooring line and the floater is can obtained as:

F~ · ~v
λ = cos−1 (7.10)
|F~ |

The change in λ with respect to to the initial rotation will be given as:

∆l = λ(t) − λ(0) (7.11)

The relative rotation between the floater and the mooring line can finally be calculated as
follows:
φopb = ∆f − ∆l (7.12)
Since φopb changes in time, it will be denoted as φopb (t).

7.4.2 Step 2: Derive interlink stiffness model


After T (t) and φopb (t) is post-processed from coupled floater-mooring calculation, the interlink
stiffness model needs to be derived. The nominal diameter of the links is taken as the corroded
diameter of the link (137 mm). The nominal dimensions of the link are shown in figure 7.8. A
solid FE model of a system of links will be created to develop the interlink stiffness model.

Figure 7.8: Nominal dimensions of a 137 mm studless link [23]

FE model
A solid FE model of a 3 link chain having a nominal diameter(Dnom ) of 137 mm is created in
ANSYS to derive the interlink stiffness model. The contact between links is modeled as stiff
springs which does not allow penetration but allows sliding by taking into account the frictional
coefficient (See Appendix ??). The friction coefficient is taken as 0.5 which is the friction coefficient
of steel in air, as the connection between the mooring line and the floater is on the deck of the
floater. The material law used for the FE model is based on R4 mooring chain grade and it is
shown in figure 7.9.
Figure 7.10 explains some of the definitions that are used in the FE model. The x and z axes are
shown in the figure while y axis goes inside the plane of the figure. Points A and B lie diametrically
opposite to each other on Link A while points C and D lie diametrically opposite to each other on
Link B in the xz plane with y = 0. The angle that forms between line AB and line CD will be
7.4. METHODOLOGY 61

Figure 7.9: Ramberg Osgood true stress strain curve:Yield Stress=580 MPa, Ultimate Stress=860
MPa, Ultimate Strain=11%, α = 0.002, n=10.3

Figure 7.10: Definitions used in the case study

defined as the interlink angle (αint,F E ) between Link A and Link B.


The nominal OPB stresses (σopb,F E ) on Link B will be calculated as follows:
σxx,P − σxx,Q
σopb,F E = (7.13)
2
where P and Q are two points at the centre of Link B with P lying diametrically opposite to Q
along z axis. σxx,P is the normal stress along x at P and σxx,Q is normal stress at Q. Equation
3.1 isolates the stresses due to OPB from tension and it is valid when the bending curvature in
the link is small ([20]). It is important to note that any variation in the OPB stresses or moments
within link B is small and will be ignored. The connection between links A and B and between
links B and C will be termed as Connection 1 and Connection 2 respectively. If the OPB stresses
are known, the OPB moments can be calculated as follows:
Mopb = σopb,F EA ∗ Z (7.14)
Z is the section modulus of the OPB link given by:
3
Z = π ∗ Dnom /64 (7.15)
A random low frequency rotation signal and tension signal is generated for 10 s. The rotation
signal is generated such a way that both the sticking and sliding regimes of OPB stress are observed.
These signals are applied to the FE model after it has been subjected to proof-load (at 70% corroded
MBL) and initial load (given by T (0)). The complete tension loading signal is shown in figure 7.11
and the complete rotation signal is shown in figure 7.12. FEA of the chain links is performed
quasi-statically during proof-loading and initial loading while it is performed dynamically after
initial loading. The interlink angles (αint,F E ) and the OPB stresses (σopb,F E ) are generated after
FEA.
62 CHAPTER 7. A CASE STUDY ON CALCULATING OPB STRESS

Figure 7.11: Tension loading applied to FE model

Figure 7.12: Rotation applied to FE model

Interlink Stiffness Model


Figure 7.13 shows an idealization of the connection between Link A and Link B (Connection
1). 4 rotational springs are placed between Link A and Link B. Each rotational spring i has a
characteristic stiffness kri and sliding threshold Wi . kri and Wi are given as follows:

kri = ai T (t)bi (7.16)

Wi = ci T (t) + di (7.17)

Figure 7.13: Idealization of the connection between Links A and B


7.4. METHODOLOGY 63

Link C is not considered because the moments generated at Connection 2 will be equal and
opposite to the moments generated at Connection 1 when only interlink angle is imposed between
A and B. The interlink stiffness model can now be stated as follows:
4
X
σopb (t) = σopb,i (t), (7.18)
i=1

where
∂σopb,i (t−δt)

min{σopb (t − δt) +
 ∂t δt, (ci T + di )}, if αint,F E (t) − αint,F E (t − δt) ≥ 0
σopb,i (t) =
 ∂σopb,i (t−δt)
max{σopb (t − δt) + δt, −(ci T + di )} otherwise

∂t
(7.19)
and
∂σopb,i (t) dT dαint
= ai {αint (t)bi T bi −1 + T bi (t) } (7.20)
∂t dt dt
The interlink stiffness model given by equations 7.18,7.19 and 7.20 imply that if the interlink angles
are known, OPB stress can be calculated. The parameters ai , bi , ci and di need to be determined.
This will be done by optimizing the interlink stiffness model with respect to σopb,F E . The interlink
angles from FEA(αint,F E ) will be used as an input to the model. The cost function is formulated
as follows:
X10
J= {σopb,F E (t) − σopb (t)}2 , (7.21)
t=0
The optimization problem is formulated as follows:
{â, b̂, ĉ, d̂} = arg min J(a, b, c, d) (7.22)
a,b,c,d

where a = [a1 , a2 , a3 , a4 ], b = [b1 , b2 , b3 , b4 ], c = [c1 , c2 , c3 , c4 ] and d = [d1 , d2 , d3 , d4 ]. â,b̂,ĉ,d̂ is a


solution to the optimization problem. Nonlinear regression based on Genetic Algorithm is used to
solve this optimization problem. The parameters obtained after solving the optimization problem
are given in table 7.4.2.

a b c d
0.027 0.037 320.256 41.95
0.659 0.002 170.11 45.62
0.012 0.08 260.545 48.081
12.789 0.41 0.622 5.137

Equations 7.18, 7.19 and 7.20 are generalized and modified as follows to calculate OPB mo-
ments.
X4
Mopb (t) = Mopb,i (t) (7.23)
i=1
where
∂Mopb,i (t−δt)

min{Mopb (t − δt) +
 ∂t δt, (ci T + di ) Z}, if αint (t) − αint (t − δt) ≥ 0
Mopb,i (t) =
 ∂Mopb,i (t−δt)
max{Mopb (t − δt) + δt, −(ci T + di ) Z} otherwise

∂t
(7.24)
and
∂Mopb,i (t) dT dαint
= ai Z{αint (t)bi T bi −1 + T bi (t) } (7.25)
∂t dt dt
Equations 7.23, 7.24 and 7.25 along with table 7.4.2 is the interlink stiffness model for the 137 mm
chain link system.

7.4.3 Step 3: Derive equivalent link geometry


The complicated geometries of IPB and OPB links will be reduced to simple beams of uniform
circular cross-section of equivalent bending stiffness. This will make the process of discretization
for a simplified model easier. FE calculations are used to derive the equivalent link geometry.
64 CHAPTER 7. A CASE STUDY ON CALCULATING OPB STRESS

FE model
Cantilever bending test are performed using a solid FE model of the IPB link (figure 7.14) and
OPB link (figure 7.15). The links are subjected to a range of forces (P) and the displacement at
the tip (∇FE ) corresponding to each force is obtained. In this case study, P is taken as as follows:
 
100
 500 
 
P= 1000 N (7.26)

2500
5000
The above process is repeated for both the IPB link and OPB link.

Figure 7.14: Cantilever bending of IPB link

Figure 7.15: Cantilever bending of OPB link

Equivalent link geometry


If the length of the simple beam is denoted as l and the diameter is denoted as d, the displacement
of the cantilever beam at the tip can be analytically calculated as:
Pl3
∇= ; (7.27)
3EI
where
πd4
I= (7.28)
64
The equivalent length (l) and diameter (d) can be obtained by optimizing the beam to give similar
displacements as the FE model. The cost function is formulated as:
n
X
Jlink = (∇FE − ∇)2 ; (7.29)
i=1

where n represents the number of test cases. The optimization problem is formulated as follows:
(l, d) = arg min Jlink (l, d) (7.30)
l,d
7.4. METHODOLOGY 65

The solution (l, d) is sought such that it they are in the neighbourhood of the nominal dimensions
of 137 mm link. The optimization problem 7.30 is solved for both the IPB link and the OPB link
using Genetic Algorithm and the results are summarized in tables 7.2 and 7.3.

l (m) d (m)
0.782 0.165

Table 7.2: Equivalent beam dimension of IPB link

l(m) d(m)
0.79 0.127

Table 7.3: Equivalent beam dimension of OPB link

7.4.4 Step 4: Prepare simplified model


Once the interlink stiffness model and equivalent link geometries are determined, a simplified model
of a system of links will be developed to predict OPB stresses. A 7 link simplified model based on
the physics based approach (Chapter 5) is shown in figure 7.16. The choice on the number of links
is based on the reasoning that as the number of links increases, OPB stresses become independent
of the boundary condition (see Section 3.4.4). In figure , dipb and lipb are the equivalent dimensions
of the IPB link and dopb and lopb are the equivalent dimensions of the OPB link. The interlink
stiffness model is included as an external moment between links A and B. The axial and vertical
degrees of freedom at the connection between A and B are shared. T (t) and φopb (t) are inputs to
the simplified model.

Figure 7.16: Idealization of a 7 link chain

The interlink angle at Connection 1 (between A and B) will be calculated as follows:

αint = θ2 − θ1 ; (7.31)

The governing equation is formulated as follows:

K X = F(t) (7.32)

where the global stiffness matrix K is assembled from the stiffness matrices of the beams.

K = Kbeam (7.33)

Kbeam can be obtained from Appendix. The force matrix F(t) is formulated as follows:
 
 0 
..

 

 
.

 


 

−M (t)
 
opb

 

 
Mopb (t)
 
F(t) = (7.34)

 0 


 .
.


.

 


 

 Mφ (t) 

 
 

T (t)
 
66 CHAPTER 7. A CASE STUDY ON CALCULATING OPB STRESS

where Mφ (t) is a force matrix due to input rotation φopb (t) and T (t) is the applied tension.
The OPB moment (Mopb (t)) is given by equations Equations 7.23, 7.24, 7.25 and table 7.4.2. It is
applied at the rotational degree of freedom corresponding to θ1 and θ2 (figure).
The initial displacement is calculated as:

K0 X0 = F(0) (7.35)

where F(0) is the force matrix due to initial tension (T (0)) and initial rotation (φopb (0)).
The governing equation 7.32 is solved in a series of time steps using the initial solution X0 to
obtain a solution in time domain. Using equation 7.31, the interlink angles can be calculated.The
OPB stress can now be calculated using the following relationship:

Mopb (t)
σopb (t) = (7.36)
Z
where Z is the sectional modulus of the OPB link given by equation 7.15. Figure 7.17 shows a plot
of the interlink angles and figure 7.18 shows a plot of the OPB stresses for the given sea-state.

Figure 7.17: Time domain plot of interlink angles for 1 hour sea state

7.5 Discussion
From figures 7.17 and 7.18, one can observe that the trend of interlink angle and OPB stress
is exactly similar. Both plots have the same shape. This is because the interlink angles in the
case study are quite small. This makes OPB stress to lie in the linear sticking regime of OPB
stress-interlink angle relationship. In order to validate if the current model can capture sliding
regime, the input rotations are multiplied by a factor of 10. A comparison of interlink angles
(figure 7.20) and OPB stress (figure 7.21) is carried out to see if there is a change in the trend
of OPB stress and interlink angle. One can conclude of from the observations that if the input
rotations are increased, sliding regimes are observed which is indicated by the flattening of the
OPB stress curve. Differences in the trend of interlink angles and OPB stress are also seen in the
comparison.

7.6 Conclusion
In this chapter, a methodology is presented to calculate nominal OPB stresses for a particular
sea-state. Once the OPB stresses are known, they can be converted into hot-spot stresses by
7.6. CONCLUSION 67

Figure 7.18: Time domain plot of OPB stress for 1 hour sea state

Figure 7.19: OPB crack location

Figure 7.20: Comparison of interlink angles

multiplying with stress concentration factors (SCF) at the location of interest. On such location
where fatigue cracks occur due to OPB in mooring links is shown in figure 7.19. The hot-spot
stresses can be converted into stress ranges and Rainflow Counting can be used to determine the
68 CHAPTER 7. A CASE STUDY ON CALCULATING OPB STRESS

Figure 7.21: Comparison of OPB stress

number of cycles corresponding to each stress range. A suitable SN curve ([25]) can be employed
to evaluate the desired strength corresponding to each stress range. The fatigue damage due to
OPB can be finally estimated by Miner’s rule ([27]).
Chapter 8

A Comparative Study

8.1 Introduction
The methodology adopted in the current work follows a de-coupled approach, in which, coupled
floater-mooring analysis is performed first. The tension and relative rotation are post-processed
after coupled floater-mooring analysis and taken as an input for the simplified OPB models of
chain links. Currently, there is no study in literature to indicate whether including interlink and
bending stiffness in coupled floater-mooring analysis changes tensions and floater rotations. This
chapter attempts to answer this question by performing a comparative study.

8.2 Methodology
For the comparative study, two scenarios are chosen in which a simple cylindrical floater with
identical environmental conditions are considered. The mooring line comprises of a polyester rope,
which is 90% of the total line length, and steel chain links. The polyester rope is idealized as a
string in both the scenarios. The chain links are idealized as a single string in the first scenario,
hence it is called a ”String-String Model” (figure 8.1).

Figure 8.1: String-String Model

In the second scenario, steel chain links is considered as a composition of two parts: the first
part idealizes the floater link as a simple beam and the second part idealizes rest of the chain links
as another simple beam. This model is called a ”String-Beam Model” (figure 8.2). The connection
between the floater link beam and the beam representing rest of the links is taken as the interlink

69
70 CHAPTER 8. A COMPARATIVE STUDY

stiffness. The interlink stiffness is modeled as an external moment and not as an rotational spring.
The link considered in this comparative study is the 137 mm chain link which was used the case
study (Chapter 7). More details regarding environmental conditions, floater and mooring chain
links can be found in Appendix F.
The main objective is to compare tension and floater rotation in time domain for each of these
scenarios.

Figure 8.2: String-Beam Model

8.3 Description of the numerical simulation


The numerical time-domain simulation for the moored floater follows the approach of CALM buoy
coupled floater-mooring analysis, as described in [7]. The numerical model used in the time-domain
simulations for the moored floater consists of two separate models, which are linked as shown in
figure 8.3. The two main components of the simulation model are a numerical model for the floater
and a numerical model for the dynamic behaviour of the mooring lines. The interaction between
the two models is explained as follows:for each time step, the equation of motion of the floater is
solved. The resulting floater motions are used as a kinematic boundary condition for the mooring
model. The equations of motion for the mooring model are solved and the resulting line tension
at the top of the chain is taken as an external force on the floater for the next time step. In
this comparative study, equations of motion for the floater remain the same while the equation of
motion for the mooring line changes for the two scenarios.

8.3.1 Equation of motion of the floater


The equation of motion for the floater in time-domain, in the absence of any linear or quadratic
damping, is given by the convolution integral
Z t
[m + A∞ ]ẍ + Kf x + h(t − τ )ẋ(τ )dτ = q(t, x, ẋ) (8.1)
0

where m is the mass matrix of the floater, x is a three degree of freedom displacement vector of
the center of gravity of the cylindrical floater composed of surge (xf ) , heave (zf ) and pitch (θf ,
positive anticlockwise)  
x f 
x = zf (8.2)
θf
 
8.3. DESCRIPTION OF THE NUMERICAL SIMULATION 71

Figure 8.3: Numerical simulation flow chart [7]

A∞ is the added mass at infinite frequency (frequency independent added mass), Kf is the hydro-
static stiffness and q is the exciting force given by
(1) (2)
q(t, x, ẋ) = qW A + qW A + qext (8.3)
(1) (1)
where qW A is the first order wave excitation force, qW A is the second order wave excitation force
and qext is the mooring line force on the floater.
Calculation of first-order and second order wave excitation forces is explained later.
h(τ ) in equation 8.1 is the retardation function computed by a transform of the frequency
dependent added mass and damping
Z ∞
1
h(τ ) = [c(ω) + iωa(ω)]eiωt dω (8.4)
2π −∞
Frequency dependent added mass (A) and radiation damping matrix(C) is given as follows

A(ω) = A∞ + a(ω) , A∞ = A(ω = ∞) (8.5)

C(ω) = C∞ + c(ω) , C∞ = C(ω = ∞) (8.6)

Separation of motions
Solving the integral in equation 8.1 is very time consuming, so another method is used which is
based on separation of motions [1]. The separated motion method is a common approach by using
a multiple scale approach. This method separates the wavefrequency (or high frequency) part
from the lowfrequency part. The exciting force is separated in a high-frequency part q (1) and a
low-frequency part q (2)

q(t, x, ẋ) = q (1) + q (2)


(1)
q (1) = qW A (8.7)
(2)
q (2) = qW A + qext
The position vector X can then be separated into

x = xHF + xLF (8.8)

The high-frequency motions xHF is solved in frequency time domain expressed by


1
[m + A(ω)]xHF
¨ + C(ω)xHF
˙ + Kf xHF = qW A (ω) (8.9)

The high-frequency motion xHF (ω) can be converted into time domain as follows
t
X
xHF (t) = xHF (ω)cos(ωt − kx + φ(ω)) (8.10)
0
72 CHAPTER 8. A COMPARATIVE STUDY

where k and φ(ω) are frequency dependent wave-number and phase respectively.
The low frequency motions are solved in time-domain is expressed as follows:
(2)
[m + A(ω = 0)]xLF
¨ + Kf xLF = qW A + qext (8.11)

Wave forces
The total wave forces that act on the floater is given as
(1) (2)
qW A = qW A + qW A (8.12)

(1)
If the sea-spectrum S(ω) is known for an irregular sea, the first order wave forces qW A can be
calculated as
N
(1)
X
qW A (t) = Aj |H1 (ωj )|cos(ωj t + j + δ(ωj )) (8.13)
j=1

where H1 (ω) is the frequency dependent transfer function for the first order wave force, j is the
random phase angle and δ(ω) is the phase of H1 (ω). Aj is the wave amplitude given by
q
Aj = 2 S(ωj )∆ω (8.14)

(2)
The second order wave forces qW A is a composition of steady drift forces qst and a low frequency
2
component qLF
(2) (2)
qW A = qst + qLF (8.15)
The steady wave drift forces are independent of time and is calculated as
N
X
qst (t) = A2j |Hs (ωj )| (8.16)
j

The low frequency component of second order wave forces calculated as


N
N X
(2)
X
qW A (t) = Aj Ak |HQT F (ωj )|cos{(ωk − ωj )t + (k − j ) + φQT F (ωj ))} (8.17)
j=1 k=1

where HQT F (ω) is the quadratic transfer function and φQT F is the phase angle of the quadratic
transfer function.

Frequency dependent terms


The frequency dependent terms involved in the equation of motion of the floater are A, C, H1 (ω),
Hs (ω) and HQT F (ω). These terms are calculated for the floater before the start of the simulation
without considering the mooring line. In this thesis, ANSYS AQWA is used to generate the
frequency dependent terms. ANSYS AQWA takes a sea-spectrum and the direction of the spectrum
as an input and performs diffraction analysis [2]. The frequency dependent terms used in this
comparative study can be referred from Appendix F

8.3.2 Equation of motion of the mooring line


The governing equation of motion of the mooring line can be written as

M Ẍ + KX = F (t, X) (8.18)

The above equation uses a finite element discretization for the mooring line with M as the
mass matrix and K as the stiffness matrix of the mooring line. Both these matrices differ for the
string-string model and string-beam model. F (t, X) is a force matrix which includes wave and
current forces on the mooring line as well the forces due to the motion of the floater.
8.3. DESCRIPTION OF THE NUMERICAL SIMULATION 73

String-String Model
In the string-string model, the mooring line is discretized into NC elements for the chain part and
NP elements for polyester rope. Each node has two degrees of freedom; axially along z direction
and horizontally along positive x direction. The global mass matrix M can be written as

M = Mchain + Mpolyester (8.19)

The global mass matrix is assembled from elemental mass matrices given by
 
140 0 70 0
ρ p Ap l p 
 0 156 0 54 
Melement =  , p = chain, polyester (8.20)
420  70 0 140 0 
0 54 0 156

where ρ is the density, A is the area of cross section calculated using nominal diameters, l is the
length of each element and the subscript p denotes whether the element lies within the chain or
the polyester part.
Similarly, global stiffness matrix K is written as

K = Kchain + Kpolyester (8.21)

The global stiffness matrix is assembled from elemental stiffness matrices given by
 (EA) (EA)

p
lp 0 − lp p 0
T
− lTp 
 
 0 lp 0
kelement =  (EA)p

(EA)p
 , p = chain, polyester (8.22)
− lp 0 0 
lp 
T T
0 − lp 0 lp

(EA)p denotes the axial stiffness of the chain or polyester rope. T is the tension developed in each
element. The elemental matrix formulations can be found in [19]. If the global coordinates of the
nodes of an element at any time t is given as (Xi , Zi ) and (Xj , Zj ), tension can be calculated as
q 
(EA)p
T = (Xi − Xj )2 + (Zi − Zj )2 − lp,org (8.23)
lp,org

where lp,org is the un-stretched original length of an element in the chain or polyester part. Due
to changing tension, the global stiffness matrix also changes. Hence, equation 8.18 is a non-linear
equation. The forces due to the floater can be given as

Ff (t) = −MX1 Ẍ1 − KX1 X1 − MZ1 Z¨1 − KZ1 Z1 (8.24)

X1 and Z1 is a displacement matrix given by


   

 x1 
 
z1 
x1 z1 
 
   
X1 = . , Z 1 = .. (8.25)
 .. 

  
 .
  
 
x1 z1
 

where x1 and z1 are surge and heave motions of the floater which are applied at the top node of
the mooring line, MX1 , MZ1 and KX1 , KZ1 are corresponding column matrices.

String-Beam Model
The formulation of the string-beam model is different the mass and stiffness matrix changes due
to the beam representation of of chain part of the mooring line. The chain is discretized into
NC + 1 elements which includes the additional floater link and the polyester rope is discretized
into NP elements. The nodes that lie in the polyester part has two degrees of freedom as explained
previously, whereas the nodes that lie in the beam part has three degrees of freedom: axial, lateral
as well as rotational.
74 CHAPTER 8. A COMPARATIVE STUDY

The global mass matrix M can be obtained by referring equation 8.19. The elemental mass
matrix for the polyester part remains as it is but the elemental mass matrix of the chain part is
now given as
 
140 0 0 70 0 0
 0
 156 22lq 0 54 −13lq 
2
ρq Aq lq  0
 22lq 4lq 0 13lq −3lq2 
 , q = f loater − link, chain (8.26)
Melement =
420   70 0 0 140 0 0  
 0 54 13lq 0 156 −22lq 
0 −13lq −3lq2 0 −22lq 4lq2

Similarly, the elemental stiffness matrix for the polyester remains unchanged whereas the ele-
mental stiffness matrix for the chain part is now given as
 (EA)q (EA)

lq 0 0 − lq q 0 0
12EI T 6EI
 0 lq3 + lq 0 − 12EI
lq3 − lq
T 6EI 

 lq2 lq2  
6EI 4EI 6EI 2EI 
 0
lq2 lq 0 − l2 l q
kelement =  (EA)q q
 , q = f loater−link, chain
 
(EA)q
− l 0 0 lq 0 0 
 q 
 0
 − 12EI
lq3 − lq
T
− 6EI
lq3 0 12EI
lq3 + lq
T
− 6EI 
lq3 
6EI 2EI
0 lq2 lq 0 − 6EIlq2
4EI
lq
(8.27)
where EI denotes the bending resistance of the the beam. The elemental matrix formulations can
be found in [19] which uses a Euler-Bernoulli tensioned beam formulation. The tension is updated
for each time step and it is given by equation 8.23.
To obtain the force from the floater on the mooring line, one can refer equation 8.24 which will
include extra terms due to the floater rotation. In the string-beam mode, the interlink stiffness is
included as an external moment between the floater-link and rest of the chain. The OPB moment
formulation is stated as follows
X4
Mopb (t) = Mopb,i (t) (8.28)
i=1

where
∂Mopb,i (t−δt)

min{Mopb (t − δt) +
 ∂t δt, (ci T + di ) Z}, if αint (t) − αint (t − δt) ≥ 0
Mopb,i (t) =
 ∂Mopb,i (t−δt)
max{Mopb (t − δt) + δt, −(ci T + di ) Z} otherwise

∂t
(8.29)
and
∂Mopb,i (t) dT dαint
= ai Z{αint (t)bi T bi −1 + T bi (t) } (8.30)
∂t dt dt
The interlink angle αint can be referred from figure 8.2. The parameters of the OPB moment
have already been determined in Chapter 8 and they can be referred from table 7.4.2. The OPB
moment Mopb is included in a column matrix corresponding to θ2 and θ3 degrees of freedom in the
right hand side of equation 8.18. It is given as follows
 
 0 
..

 

 
.

 


 

0

 

 
Mopb = −Mopb (t) (8.31)
 Mopb (t) 

 
 


 .. 

0 .

 


 

0
 

Environmental loads on mooring line


The forces on the mooring line (F (t, X)) i.e the right hand side of equation 8.18, is given as

F (t, X) = Ff + Fenv (8.32)


8.4. RESULTS 75

Equation 8.32 will include an extra term due to OPB moment in case of string-beam model. The
calculation of Ff has already been described for both the models. For the calculation of the
environmental loads Fenv , Morrison’s equation is used which is given as

πD2 D
F = cm ρ (u̇ − ẅ) + Cd ρ (u + uc − ẇ)|u + uc − ẇ| (8.33)
4 2

where

• F :force per unit length

• Cm :inertia coefficient

• Cd :drag coefficient

• D:hydrodynamic diameter of the mooring line

• ρ:density of water

• w:horizontal displacement of the mooring line

• u:horizontal velocity of waves

• uc :horizontal velocity of current

A deep water approximation of airy wave theory is used to calculate the wave velocities along the
vertical length of the mooring line. The horizontal wave velocity at depth z for an irregular sea is
given as
N
X
u(z, t) = Aj ωj ekz cos(kx − ωj t) (8.34)
j=1

A time invariant current profile is considered for the calculation of current velocity and it is
given as
 1/7
z+d
uc (z) = uc,0 (8.35)
d

where uc,0 is the current velocity on the water surface and d is the depth of the water surface.
Wind loads are not considered in this study.

8.4 Results
Time domain simulation for 100 s is carried out for the string-string model and string-beam model
considering a unidirectional wave spectrum as shown in figure 8.4. The starting point of the
simulation as the hydrostatic condition where initial tension in the mooring line is due to the
difference in the wight of the floater and buoyancy.

KX0 = Buoyancy − W eight (8.36)

where X0 is the initial displacement of the mooring line.


The equations of motions of the floater (equation 8.1) and the mooring line (equation 8.18)
are solved in time domain using MATLAB ODE45 solver [17], according to the method described
in figure 8.3. The floater rotations obtained for both the model are compared in figure 8.5. The
tension at the top chain, which is obtained as the mooring line force that acts on the floater, is
compared in figure 8.6 . From the comparisons, it can be observed there is a change in the tension
while floater rotation remains unchanged.
76 CHAPTER 8. A COMPARATIVE STUDY

Figure 8.4: Wave spectrum

Figure 8.5: Comparison of floater rotation

8.5 Conclusion
From this comparative study, it can be concluded that tensions changes when bending and interlink
stiffness is included in the mooring line. In coupled floater-mooring analysis, bending stiffness is
not included in mooring chains because the curvatures are very small such that the entire chain
behaves like a string. However, on inclusion of interlink stiffness, local curvatures are developed in
the mooring line in the vicinity of the floater, which changes the tension. Floater rotations do not
change because the external moment due to interlink stiffness that act on the floater is very small
as compared to the wave forces and moments. For instance, the moments due to first order wave
force is 8 orders of magnitude while highest OPB moments are of 4 orders of magnitude.
8.5. CONCLUSION 77

Figure 8.6: Comparison of tension


78 CHAPTER 8. A COMPARATIVE STUDY
Chapter 9

Critical Analysis and


Recommendation

9.1 Introduction
The goal of this Master’s thesis was to develop simple models that can explain the OPB mechanism
in a system of mooring chain links when they are subjected to both tension and rotation. In this
thesis, a simplified model was defined as one that performs the following tasks:
1. It should be able to reproduce the physics of an actual system of links when subjected to
rotation and tension.
2. It should give a reasonable prediction of OPB stresses and interlink angles in time domain
compared to a solid FE model of a system of links.
3. It should take less time to calculate than a solid FE model.
The key to develop such a model required that the non-linear hysteritic relationship between
OPB moment-interlink angle or the interlink stiffness model is properly described. A new interlink
stiffness model (figure 9.3) based on Maxwell Slip Model for friction modeling [15] was developed
as part of this thesis. This model was capable of explaining the observed nonlinear hysteritic
relationship between OPB moment and interlink angle. Extensive use of FEA was made to derive
this model.
The interlink stiffness model was applied to a system of chain links using two different ap-
proaches (figure 9.4). In one approach, the links were idealized as beams of uniform circular
cross-section with equivalent bending stiffness. This model was termed as ”physics based” because
it intended to describe the overall physics of a system of links. In another approach, the links
were idealized as rigid bars and the model was termed as ’Semi-empirical’ because it intended
to reproduce the interlink angles and OPB stress using some physical considerations but without
describing the overall behaviour of a system of links. Both the models give accurate predictions of
OPB stress and interlink angles at a fraction of time compared to a FE model.
A case study is also presented in this thesis in which a methodology to calculate OPB stresses
for a sea-state is described. This methodology follows a de-coupled approach in which results
of coupled floater-mooring analysis are taken as an input to the simplified models. To verify this
methodology, a comparative study is performed in which coupled floater-mooring analysis is carried
out with and without including the interlink stiffness of chain links. An interesting conclusion was
drawn at the end that tensions indeed change when interlink stiffness is included.
In this chapter, a critical analysis of the research is presented so that the reader is aware of the
limitations and shortfalls.

9.2 Critical Analysis


9.2.1 On FE modeling
In this master’s thesis, extensive use of FE modeling in ANSYS has been made. A solid FE
model of chain links (see figure ) was verified and validated for one load case when tension in the

79
80 CHAPTER 9. CRITICAL ANALYSIS AND RECOMMENDATION

chain was kept constant. This FE model was used to develop an interlink stiffness model and
subsequently, to develop the simplified models for OPB stress prediction. It is very important to
note the FE model may not be an accurate representation of reality. [9] and [10] has reported that
FE modeling usually produces under conservative results. This raises some concern especially in
the case of interlink stiffness model for varying tension. There is no way of confirming whether the
results from FE analysis correspond to reality.
Another point to note that there is a discrepancy in the application of boundary condition in
the simplified model and in FEA. Although ANSYS allows application of rotation as a boundary
condition, it does so by using the concept of remote node which is illustrated in figure 9.1. To
apply rotation at the specified nodes in the FE model, a remote node is created which does not
lie on the FE model but has a rigid connection with the nodes on the FE model where rotation
is to be applied. A combination of force and moment is applied at the remote node which gets
converted in to required rotation. Thus, applying rotations directly to the simplified model is not
an accurate representation of kinematic boundary conditions.

Figure 9.1: Force and moment applied at remote node to produce rotation at required nodes in
FE model

9.2.2 On interlink stiffness model

An interlink stiffness model is developed in this thesis which accurately describe the non-linear
hysteritic relationship between OPB moments/stress and interlink angles in mooring chain links. It
utilizes a piecewise approximation of the hysteresis curve through the use of a number of elementary
operators (spring elements). While extremely good results for OPB moments/stress are obtained
using this model, one cannot use it to derive the interlink stiffness (kopb ) which is a derivative of
OPB moment with respect to interlink angle. kopb is a discontinuous function because derivatives
do not exist at certain points in the OPB moment-interlink angle see graph (figure 9.2). This
makes it difficult to construct a truly simplified model to accurately describe the behaviour of a
system of chain links.
Another shortcoming of the interlink stiffness model is that it can predict stresses only on
the OPB link and not on the IPB link. It is because this model has been derived assuming that
the OPB link behaves like an Euler-Bernoulli beam. This is true for small deflections when the
thickness is small compared to the length of the beam. However, this does not hold true for the
IPB link because the thickness is comparable to the length of the beam.
Finally, the author would like to raise an important question on the nature of the interlink
stiffness model itself. The formulation of interlink stiffness model is quite similar to the frictional
contact formulation used in ANSYS which is based on Return Mapping Algorithm ([12]). Both
formulations imply that evolution of a force in time is dependent on the path taken by the force
till that point of time in the force-displacement curve. As a result, there is nothing unique about
the interlink stiffness model because one can consider it as a consequence of the formulation that
ANSYS uses. A much greater confidence can be placed in the model only if it is tested against an
actual experiment.
9.2. CRITICAL ANALYSIS 81

Figure 9.2: Discontinuity of kopb

Figure 9.3: Interlink Stiffness Model

9.2.3 On Simplified Models


Physics based model
In this model, the links were idealized as simple beams of uniform circular cross section with
equivalent bending stiffness. Rotational springs(kopb ) were placed between the links. Due to the
discontinuity of kopb , the model gave non-smooth solution for interlink angles. If this model had
worked, one would have obtained a truly simplified model which could describe the behaviour of a
system of links.
An alternative model was proposed in which the interlink stiffness is included as an external
moment between the first two links while a rigid connection is made between rest of the links. By
doing so, the model gave near perfect values of OPB stress and interlink angles for a 3 link model
at a fraction of a time compared to a solid FE mode. However, such a model is limited in its use
till the first OPB link only. It cannot be used to predict stresses or interlink angles across any
other link. Hence, this model does not satisfy the definition of a simplified model. The discrepancy
in the results increase for a larger chain link model(for instance the 7 link model). This indicates
that interlink stiffness should be considered between all the chain links.

Semi-empirical model
The main idea behind this model was not to reproduce the physics of a system of links, but to
create an equivalent mechanical system which will reproduce just two quantities. Those quantities
are OPB stress and interlink angle between the first two links. This model does not satisfy the
definition of a simplified model. But an interesting feature of this model is that it can give near
82 CHAPTER 9. CRITICAL ANALYSIS AND RECOMMENDATION

accurate prediction of these quantities using just three degrees of freedom. The models also works
quite well for a 7 link chain. Such a model can be used for quick calculations if one wishes to
ignore the overall behaviour of a system of links.

Figure 9.4: Physics based model and Semi-empirical model

9.2.4 On the Methodology for estimating OPB stress


The de-coupled approach followed in this thesis to estimate OPB stress raises some concern. In
the comparative study (Chapter 10) in which coupled floater-mooring analysis is carried out with
and without including interlink stiffness, it is observed that while floater rotations do not change,
tension does change. .

9.3 Recommendations
This work should be seen as a starting point in developing models to explain OPB mechanism in
mooring chains comprehensively. Although the models developed during this research is a vast
improvement over previous models in literature, they are still incomplete. A few recommendations
to improve the models are listed below which can be seen as a scope for future work.
1. An attempt can be made towards smoothening of the interlink stiffness model to allow for
existence of higher order derivatives.

2. A more robust numerical technique such as arc-length method can be applied to solve the
governing nonlinear equation of the physics based simplified model.
3. The interlink stiffness model as well as the simplified models should be validated extensively
against experiments.
4. The simplified physics based model can be modified by adopting a comprehensive interlink
stiffness model that takes into account out-of plane bending, in-plane bending and torsion
and that makes kopb a smooth function for all values of interlink angle.
5. Methodology to incorporate interlink stiffness of mooring chain links in coupled floater-
mooring analysis should be investigated for more complicated cases of mooring layout.
Appendix A

ANSYS Nonlinear Quasi-Static


Analysis

A.1 Introduction
A nonlinear analysis is needed if the loading on a structure causes significant changes in stiffness.
Typical reasons for stiffness to change significantly are:
• Strains beyond the elastic limit (plasticity)
• Large deflections in the structure
• Contact between two bodies
When a load causes significant changes in stiffness, the load-deflection curve becomes nonlinear
(figure A.1). The challenge is to calculate the nonlinear displacement response using a linear set
of equations.

Figure A.1: Nonlinear load-displacement graph

The governing equation for nonlinear quasi-static analysis can be stated as follows:
K(X) X = F (A.1)
where K(X) is the stiffness matrix which is dependent on the displacement X. A nonlinear analysis
is an iterative solution because this relationship between load (F) and response (X) is not known
beforehand. Time dependent effects are not considered. ANSYS uses Newton-Raphson algorithm
to solve a non-linear problem.

A.2 Solution Method


The solution method used by ANSYS to solve nonlinear quasi-static problem is described in the
following sections.

83
84 APPENDIX A. ANSYS NONLINEAR QUASI-STATIC ANALYSIS

A.2.1 Tangent Matrix


The solution method for a geometrically non-linear problem begins with a linearization of the
non-linear equilibrium equations, such as the general force equilibrium equations given by:
Pext = Fint (u) (A.2)
where Fint is a vector of nodal internal member forces, which are functions of the nodal degree
of freedom (DOF) displacements u . Pext is a vector of externally applied loads. Assuming there is
a known set of nodal displacement DOFs, u0 , that satisfies equation A.2, the equilibrium equations
can be linearized by perturbing the force about this known solution point. A small perturbation
of the externally applied load corresponds to a perturbation in the nodal DOF displacements and
equation A.2 becomes:
Pext + dP = Fint (u0 + dU) (A.3)
A first-order taylor series expansion of the right hand side of equation A.3 results in:
∂Fint (u)
Pext + dP = Fint (u0 ) + |u=u0 du (A.4)
∂u
Since u0 is a solution that satisfies equation A.2, equation A.4 reduces to:
∂Fint (u)
dP = |u=u0 du = KT (u0 )du (A.5)
∂u
where
∂Fint (u)
KT (u) = (A.6)
∂u
Tangent stiffness matrix KT is essentially a matrix of sensitivities. In particular, it is the sensitivi-
ties of the internal member forces to perturbations in the nodal displacement DOFs of the system.
The tangent stiffness matrix contains information regarding both the linear elastic and geometric
stiffness of the structure.

A.2.2 Newton-Raphson Method


As with any set of non-linear equations, the most effective way to solve the non-linear problem is
to use an incremental-iterative technique. ANSYS uses an iterative technique based on Newton-
Raphson method is used. The externally applied load is divided into increments ∆P and at each
increment a Newton-Raphson iteration is run in order to converge on the equilibrium condition,
as illustrated in figure A.2. Given an initial configuration for the structure, and thus the initial
nodal DOF displacement vector u0 , the initial tangent stiffness matrix KT (u0 ) can be calculated.
The externally applied load is then incremented by some small increment ∆P and the linearized
system of equations from equation A.5 is solved for the nodal incremental displacement due to the
external load:
−1 −1
δu11 = [KT (u0 ] ∆P = [KT (u0 ] Pext,1 (A.7)
Here the subscripts represent the increment number, and the superscripts represent the iteration
number. The internal member forces at the updated configuration (Fint (u0 + δu11 )) are calculated,
and the equilibrium condition from equation A.2 is checked. Since a linear approximation was used
to solve for the nodal displacements, it is likely that there is an unbalance in the force equilibrium,
otherwise known as a residual force R. The equilibrium equation is then:
Pext,1 − Fint (u0 + δu11 ) = R11 (A.8)
and iteration proceeds to converge on a solution for the nodal DOF displacements that results in
the residual being as close to zero as possible. This is done by calculating the updated tangent
stiffness matrix from the updated nodal displacements and solving for the nodal displacement
increments due to the residual force.
−1
δu21 = [KT (u0 + δu1 ] R11 (A.9)
The updated internal member forces (Fint (u0 + δu11 + δu21 )) are again calculated, and the equi-
librium equation becomes,
Pext,1 − Fint (u0 + δu11 + δu21 ) = R21 (A.10)
A.2. SOLUTION METHOD 85

Figure A.2: Newton-Raphson incremental-iterative solution method

If the residual is not close to zero within an acceptable tolerance, then the process is repeated. If
the residual is approximately zero, then the equilibrium condition has been achieved and the new
updated equilibrium configuration due to the externally applied load of Pext,1 is

u1 = u0 + δu11 + δu21 + · · · + δui1 = u0 + ∆u1 (A.11)

where i is the number of iterations required to achieve equilibrium condition.


The solution process continues by again incrementing the externally applied load such that
Pext,2 = Pext,1 + ∆P, and proceeding with the iteration on the linear system of equations. This is
repeated until the full applied load is reached, with the final solution being uj = u0 + ∆u1 + · · · + ∆uj ,
where j is the total number of load increments. In some nonlinear static analyses, if one uses the
Newton-Raphson method alone, the tangent stiffness matrix may become singular (or non-unique),
causing severe convergence difficulties. Such occurrences include nonlinear buckling analyses in
which the structure either collapses completely or ”snaps through” to another stable configura-
tion. For such situations, one can activate an alternative iteration scheme, the arc-length method,
to help avoid bifurcation points and track unloading.
NOTE: The content in this appendix is taken from [13] and [14].
86 APPENDIX A. ANSYS NONLINEAR QUASI-STATIC ANALYSIS
Appendix B

ANSYS Transient Dynamic


Analysis

B.1 Introduction
Transient dynamic analysis (sometimes called time-history analysis) is a technique used to deter-
mine the dynamic response of a structure under the action of any general time-dependent loads.
One can use this type of analysis to determine the time-varying displacements, strains, stresses,
and forces in a structure as it responds to any combination of static, transient, and harmonic
loads. The time scale of the loading is such that the inertia or damping effects are considered to
be important. If the inertia and damping effects are not important, one could use a static analysis
instead. The basic equation of motion solved by a transient dynamic analysis is
Mü + Cu̇ + Ku = F (B.1)
At any given time, t, these equations can be thought of as a set of ”static” equilibrium equations
that also take into account inertia forces (Mü) and damping forces (Cu̇). ANSYS uses the New-
mark time integration method to solve these equations at discrete time points. The time increment
between successive time points is called the integration time step.

B.2 Solution Method


Equation B.1 can be rewritten for a time t + ∆t as follows:
Mü(t + ∆t) + Cu̇(t + ∆t) + Ku(t + ∆t) = F(t + ∆t) (B.2)
ANSYS use a Newmark time integration scheme to solve B.2, also referred to as constant accela-
ration method. In the Newmark method, it is assumed that the acceleration during an integration
time step ∆t is constant and an average value. For constant acceleration, one can write the
kinematic relations:
∆t2
u(t + ∆t) = u(t) + u̇(t)∆t + u¨av (B.3)
2

u̇(t + ∆t) = u̇(t) + u¨av ∆t (B.4)


The constant, average accelaration is:
ü(t + ∆t) + ü(t)
u¨av = (B.5)
2
Combining equations B.3 and B.5 yields:
∆t2
u(t + ∆t) = u(t) + u̇(t)∆t + [ü(t + ∆t) + ü(t)] (B.6)
4
which is solved for the acceleration at t + ∆t to obtain:
4 4
ü(t + ∆t) = 2
[u(t + ∆t) − u(t)] − u̇(t) − ü(t) (B.7)
∆t ∆t

87
88 APPENDIX B. ANSYS TRANSIENT DYNAMIC ANALYSIS

If equations B.5 and B.7 are substituted into equation B.4, one can find the velocity at time t + ∆t
to be given by:
2
u̇(t + ∆t) = [u(t + ∆t) − u(t)] − u̇(t) (B.8)
∆t
Equations B.7 and B.8 express acceleration and velocity at t + ∆t in terms of known conditions
at the previous time step and the displacement at t + ∆t. Thes equations can be substituted
into equation B.2 and after a bit of algebric manipulation, the recurrence relation for the Newark
Method can be obtained as follows:

K̃X(t + ∆t) = Feff (t + ∆t) (B.9)

where
4 2
K̃ = M+ C+K (B.10)
∆t2 ∆t
with
   
4 4 2
Feff (t + ∆t) = F(t + ∆t) + M ü(t) + u̇(t) + u(t) + C u̇(t) + u̇(t) (B.11)
∆t ∆t2 ∆t

The system of algebraic equations represented by equation B.9 can be solved at each time step
for the unknown displacements. For a constant time step ∆t, matrix K̃ is constant and need be
computed only once. The right-hand side Feff (t + ∆t) is updated at each time step. At each
time step, the system of algebraic equations is solved to obtain displacements. For this reason, the
procedure is known as an implicit method. By back substitution through the appropriate relations,
velocities and accelerations can also be obtained.
The Newmark method is known to be unconditionally stable. This does not mean, however, that
the results are independent of the selected time step. Accuracy of any finite difference technique
improves as the time step is reduced, and this phenomenon is a convergence concern similar to
mesh refinement in a finite element model. For dynamic response of a finite element model, one
must be concerned with not only the convergence related to the finite element mesh but also the
time step convergence of the finite difference method selected. ANSYS transient analysis requires
the user to specify “load steps,” which represent the change in loading as a function of time. The
software then solves the finite element equations as if the problem is one of static equilibrium at
the specified loading condition. It is very important to note that the system equations represented
by equation B.9 are based on the finite element model, even though the solution procedure is that
of the finite difference technique in time.
NOTE: The content in this appendix is taken from Chapter 10 in [10]
Appendix C

ANSYS frictional contact

C.1 Introduction
In finite element analysis, if two independent parts are present, there is no stiffness relationship
defined between them, and the resulting stiffness matrices will be uncoupled. Consequently, one
part may pass through the other during the course of the simulation. Contact elements are required
to define the interaction of two or more sets of meshes to prevent such penetration. ANSYS
contact elements typically support four different algorithms: Augmented Lagrangian, Pure Penalty,
Multipoint constraint, and Lagrange Multiplier Methods. The default and most commonly-used
option is the Augmented Lagrangian formulation, which can be thought of as a variation of the
pure penalty method. This appendix will describe the Pure Penalty and Augmeneted Lagrangian
methods.

C.2 Penalty Based Methods


When two parts come into contact, ideally, no penetration will occur. The pure penalty method
can be thought of as placing stiff springs between the two parts that have come into contact with
each other:
p = kn xn (C.1)

One can see from equation C.1 that as the contact stiffness kn is increased, the resulting
penetration xn decreases for a finite amount of contact pressure p. Ideally, the penetration xn
should be zero, but equation C.1 would result in contact stiffness kn being infinite. However, if
the penetration xn is small, the results are still very accurate. The choice of the contact stiffness
kn affects both accuracy and convergence. Too low of a value of kn results in large penetration
xn ; one can imagine that if the penetration is on the same order of magnitude as the calculated
displacements, the results will be quite suspect. On the other hand, selection of kn that is too
high leads to convergence difficulties; the reason for this is because any variation of penetration
xn leads to a large change in contact pressure p. ANSYS automatically calculates contact stiffness
kn , based on the underlying solid element’s size and material.

C.3 Augmented Langrangian Method


As noted earlier, the Augmented Lagrangian method is a penalty-based approach. Specifically,
equation C.1 is modified as follows during contact:

p = kn xn + λ (C.2)

During the Newton-Raphson iterations, the contact penetration xn is checked against an


automatically-calculated maximum allowable penetration tolerance. λ is increased automatically
when xn exceeds the tolerance limit. The benefit of the augmented Lagrangian method is that the
results are less sensitive to the value of contact stiffness kn .

89
90 APPENDIX C. ANSYS FRICTIONAL CONTACT

C.4 Friction and Elastic Slip


When friction is present, an analogous situation exists for behavior in the tangential direction.
Until the frictional shear stress τ exceeds the limiting shear stress τlim , the contact points should
be “sticking”. While zero slip is desired, in a penalty-based method, the slip xt is related to the
frictional stress τ by the tangential contact stiffness kt as follows:

τ = kt xt ifτ ≤ τlim (C.3)

where τlim is given as


τlim = µp (C.4)
Similar to the case in the normal direction with penetration xn , if the elastic slip xt is small,
it will not compromise accuracy in frictional models. In ANSYS, the tangential contact stiffness
kt is automatically calculated. A more advanced formulation for frictional contact can be found
in [12] which makes use of Return Mapping Algorithm.
NOTE: The content in this appendix is taken from [11].
Appendix D

Genetic Algorithm

D.1 Introduction
A genetic algorithm (GA) is a method for solving both constrained and unconstrained optimization
problems based on a natural selection process that mimics biological evolution. The algorithm
repeatedly modifies a population of individual solutions. At each step, the genetic algorithm
randomly selects individuals from the current population and uses them as parents to produce the
children for the next generation. Over successive generations, the population ”evolves” toward an
optimal solution.
One can apply the genetic algorithm to solve problems that are not well suited for standard
optimization algorithms, including problems in which the objective function is discontinuous, non-
differentiable, stochastic, or highly nonlinear.
The genetic algorithm differs from a classical, derivative-based, optimization algorithm in two
main ways, as summarized in the following table.

Classical Algorithm Genetic Algorithm


Generates a single point at each it- Generates a population of points at
eration. The sequence of points ap- each iteration. The best point in the
proaches an optimal solution population approaches an optimal solu-
tion
Selects the next point in the sequence Selects the next population by compu-
by a deterministic computation tation which uses random number gen-
erators

D.2 Outline of Genetic Algorithm


A detailed information on GA can be found in [16] and [30]. In this section, an outline of the
algorithm is presented.
1. The algorithm begins by creating a random initial population.
2. The algorithm then creates a sequence of new populations. At each step, the algorithm uses
the individuals in the current generation to create the next population. To create the new
population, the algorithm performs the following steps:
(a) Scores each member of the current population by computing its fitness value.
(b) Scales the raw fitness scores to convert them into a more usable range of values.
(c) Selects members, called parents, based on their fitness.
(d) Some of the individuals in the current population that have lower fitness are chosen as
elite. These elite individuals are passed to the next population.
(e) Produces children from the parents. Children are produced either by making random
changes to a single parent—mutation—or by combining the vector entries of a pair of
parents—crossover.

91
92 APPENDIX D. GENETIC ALGORITHM

(f) Replaces the current population with the children to form the next generation.
3. The algorithm stops when one of the stopping criteria is met.

NOTE: The content in this appendix is taken from [15] and [16].
Appendix E

Fixed-point iteration

In numerical analysis, fixed-point iteration is a method of computing fixed points of iterated


functions [28]. More specifically, given a function f defined on the real numbers with real values
and given a point x0 in the domain of f , the fixed point iteration is

xn+1 = f (xn ), n = 1, 2, . . . (E.1)

which gives rise to the sequence x0 , x1 , x2 , . . ., which is hoped to converge to a point x. If f is


Lipschitz continuous, then one can prove that the obtained X is a fixed point of f , i.e.,

x = f (x), n = 1, 2, . . . (E.2)

More generally, the function f can be defined on any metric space with values in that same space.
NOTE: The content in this appendix is taken from [28].

93
94 APPENDIX E. FIXED-POINT ITERATION
Appendix F

Coupled Floater-Mooring Analysis


Input Details

F.1 Environmental details


A unidirectional wave-spectrum, as shown in figure F.1 is chosen for the coupled floater-mooring
analysis. Other environmental details are shown in table F.1

Figure F.1: Wave spectrum

Environmental Parameter Value


Significant wave height (Hs ) 2m
Peak period (T0 ) 9s
Water depth (d) 800 m
Current velocity on surface (uc,0 ) 0.2 m/s

F.2 Floater details


A cylindrical floater is considered for the comparative study. The details of the cylindrical floater
are shown in table F.2

95
96 APPENDIX F. COUPLED FLOATER-MOORING ANALYSIS INPUT DETAILS

Floater Parameter Value


Diameter 25 m
Height 125 m
Hydrostatic draft 100 m
Mass 5.5 × 107 kg
Center of gravity (xcog , ycog ) 0m
Center of gravity zcog -61.63 m
Radius of gyration in x (kxx ) and y 75.29 m
(kxx )
Radius of gyration in z (kzz ) 62.72 m
Metacentric height (GM) 6.98 m

The hydrostatic stiffness matrix Kf is given as


 
0 0 0
Kf = 0 4928553.5N/m 0  (F.1)
0 0 66090396N.m/deg

F.3 Mooring line details


F.3.1 Chain links
The details of the chain links are shown in table F.3.1

Chain link Parameter Value


Chain length (un-stretched) 38.1 m
Link diameter 137 mm
Link length 822 mm
Hydrodynamic diameter 273.6 mm
Density 7850 kg/m3
Axial stiffness ((EA)c ) 1.35 × 109 N
Bending stiffness ((EI)c ) 4.46 × 106 N m2
Inertia coefficient (Cm ) 1
Drag coefficient (Cd ) 0.8

The OPB parameters are shown in table F.3.1

a b c d
0.027 0.037 320.256 41.95
0.659 0.002 170.11 45.62
0.012 0.08 260.545 48.081
12.789 0.41 0.622 5.137

F.3.2 Polyester rope


The details of the polyester rope are shown in table F.3.2

Chain link Parameter Value


Polyester rope length (un-stretched) 650 m
Diameter 111 mm
Hydrodynamic diameter 133.5 mm
Density 1370 kg/m3
Axial stiffness ((EA)c ) 1.19 × 109
Inertia coefficient (Cm ) 1.2
Drag coefficient (Cd ) 0.7
F.4. FREQUENCY DEPENDENT PARAMETERS 97

F.4 Frequency dependent parameters


The frequency dependent parameters are generated for the cylindrical floater in ANSYS AQWA
before the start of coupled floater-mooring simulation. These parameters are reported in this
section.

F.4.1 Added Mass


The added mass for surge in x (Axx ), in z (Axz ) and in θ (Axθ ) direction is shown in figure F.2

Figure F.2: Added mass for surge

The added mass for heave in x (Azx ), in z (Azz ) and in θ (Azθ ) direction is shown in figure F.3

Figure F.3: Added mass for heave

The added mass for pitch in x (Aθx ), in z (Aθz ) and in θ (Aθθ ) direction is shown in figure F.4
98 APPENDIX F. COUPLED FLOATER-MOORING ANALYSIS INPUT DETAILS

Figure F.4: Added mass for pitch

F.4.2 Radiation Damping

The radiation damping for surge in x (Cxx ), in z (Cxz ) and in θ (Cxθ ) direction is shown in figure
F.5

Figure F.5: Radiation damping for surge

The radiation damping for heave in x (Czx ), in z (Czz ) and in θ (Czθ ) direction is shown in
figure F.6
The radiation damping for pitch in x (Cθx ), in z (Cθz ) and in θ (Cθθ ) direction is shown in
figure F.7
F.4. FREQUENCY DEPENDENT PARAMETERS 99

Figure F.6: Radiation damping for heave

Figure F.7: Radiation damping for pitch

F.4.3 First-Order Wave Force Transfer Function


The first order wave force transfer function for surge (H1x ), heave (H1z ), and pitch (H1θ ) are
shown in figures F.8, F.9 and F.10 respectively.

F.4.4 Quadratic Transfer Function


The quadratic transfer function for surge (HQT F x ), heave (HQT F z ), and pitch (HQT F θ ) are shown
in figures F.11, F.11 and F.13 respectively.
100 APPENDIX F. COUPLED FLOATER-MOORING ANALYSIS INPUT DETAILS

Figure F.8: First order wave force transfer function for surge

Figure F.9: First order wave force transfer function for heave

F.4.5 Steady Drift Transfer Function


The steady drift transfer function for surge (Hsx ), heave (Hsz ), and pitch (Hsθ ) is shown in figure
F.14.
F.4. FREQUENCY DEPENDENT PARAMETERS 101

Figure F.10: First order wave force transfer function for pitch

Figure F.11: Quadratic transfer function for surge


102 APPENDIX F. COUPLED FLOATER-MOORING ANALYSIS INPUT DETAILS

Figure F.12: Quadratic transfer function for heave

Figure F.13: Quadratic transfer function for pitch


F.4. FREQUENCY DEPENDENT PARAMETERS 103

Figure F.14: Steady drift transfer function


104 APPENDIX F. COUPLED FLOATER-MOORING ANALYSIS INPUT DETAILS
Bibliography

[1] Rika Afriana. Coupled Dynamic Analysis of Cylindrical FPSO, Moorings and Riser Based
on Numerical Simulation (Master’s thesis). 2011.
[2] ANSYS AQWA. AQWA User’s Manual. url: http://148.204.81.206/Ansys/150/Aqwa%
20Users%20Manual.pdf.
[3] Farid Al-Bender. “Fundamentals of friction modeling”. In: ().
[4] V. Lampaert, J. Swevers, & F. Al-Bender. “Modification of Leuven Intergrated Friction
Model Structure”. In: IEEE Trans. on Automatic Control (2002).
[5] E.ter Brake, J van der Cammen & R. Uittenbogaard. “Calculation Methodology of Out Of
Plane Bending of Mooring Chains”. In: OMAE2007-29178 (2007).
[6] P.Jean, K.Goessens & D.L’Hostis. “Failure of Chains By Bending On Deepwater Mooring
Systems”. In: OTC 17238 (2005).
[7] Ir.J.L.Cozijn & Dr.Ir.T.H.J.Bunnik. “Coupled Mooring Analysis for a Deep Water Calm
Buoy”. In: OMAE2004-51370 (June 2004).
[8] Demosthenis D. Rizos & Spilios D. Fassois. “Presliding Friction Identification Based Upon
the Maxwell Slip Model Structure”. In: (2004).
[9] Tom Lassen, Jan Aarsnes& Einar Glomnes. “Fatigue Design Methodology for Large Mooring
Chains Subjected to Out-Of-Plane Bending”. In: OMAE2014-23308 (2014).
[10] David V. Hutton. Fundamentals of Finite Element Analysis.
[11] Sheldon Imaoka. “Contact Analysis Tips”. In: ANSYS Release 11.0 (January 2009).
[12] A.Mijar & J.Arora. “Return Mapping Procedure for Frictional Force Calculation: Some In-
sights”. In: J. Eng. Mech.,10.1061/(ASCE)0733-9399(2005)131:10(1004), 1004-1012 (2005).
[13] Sonia Lebofsky. Numerically Generated Tangent Stiness Matrices for Geometrically Non-
Linear Structures (Master’s thesis). 2013.
[14] ANSYS Training Manual. Introduction to Nonlinear Analysis. url: www2.kuas.edu.tw/
prof/me06/part-2/2_08-nonlin.ppt.
[15] MathWorks. Genetic Algorithm. url: http://nl.mathworks.com/discovery/genetic-
algorithm.html.
[16] MathWorks. How the Genetic Algorithm Works. url: http://nl.mathworks.com/help/
gads/how-the-genetic-algorithm-works.html.
[17] MathWorks. ode45. url: http://nl.mathworks.com/help/matlab/ref/ode45.html.
[18] Mathworks. Evaluating Goodness of Fit. url: http://nl.mathworks.com/help/curvefit/
evaluating-goodness-of-fit.html.
[19] Achilleas Mina. Modeling the vortex-induced vibrations of a multi-span free standing riser
(Master’s thesis). 2013.
[20] BV Guidance note. “Fatigue of Top Chain of Mooring Lines due to In-Plane and Out- of-
Plane Bending”. In: NI 604 DT R00 E (2014).
[21] Orcaflex. Chain mechanical properties. url: https://www.orcina.com/SoftwareProducts/
OrcaFlex/Documentation/Help/Content/html/Chain,MechanicalProperties.htm.
[22] P.Vargas & P.Jean. “FEA of Out-Of-Plane Fatigue Mechanism Of Chain Links”. In: OMAE2005-
67354 (2005).

105
106 BIBLIOGRAPHY

[23] C.Melis, P.Jean & P.Vargas. “Out-Of-Plane Bending Testing Of Chain Links”. In: OMAE2005-
67353 (2005).
[24] L.Rampi & P.Vargas. “Fatigue Testing Of Out-Of-Plane Bending Mechanism Of Chain
Links”. In: OMAE2006-92488 (2006).
[25] L.Rampi, F.Dewi & P.Vargas. “Chain Out of Plane Bending (OPB) Joint Industry Project
(JIP) Summary and highlights”. In: OTC-25779-MS (2015).
[26] V. Lampaert & J. Swevers. “On-Line Identification of Hysteresis Function with Nonlocal
Memory”. In: IEEE/ASME International Conference on Advanced Intelligent Mechatronics
(2001).
[27] Weibull. Miner’s Rule and Cumulative Damage Models. url: http://www.weibull.com/
hotwire/issue116/hottopics116.htm.
[28] Wikipedia. Fixed-point iteration. url: https://en.wikipedia.org/wiki/Fixed- point_
iteration.
[29] Wikipedia. Friction. url: https://en.wikipedia.org/wiki/Friction.
[30] Wikipedia. Genetic Algorithms. url: https://en.wikipedia.org/wiki/Genetic_algorithm.
List of Figures

1.1 Girassol buoy chainhawse: location of failed link [6] . . . . . . . . . . . . . . . . . . 1


1.2 Tension and OPB fatigue location in mooring chain links[6] . . . . . . . . . . . . . 2
1.3 OPB mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.4 Experimental set-up of a system of chain links: Chainhawse pushes down link T4
to generate interlink angles between T4 and T5 and OPB stresses in T5. . . . . . 3
1.5 Beam FE model suggested in BV guidance note [20] . . . . . . . . . . . . . . . . . 3
1.6 OPB mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2.1 Nominal dimensions of a 124 mm studless link [23] . . . . . . . . . . . . . . . . . . 7


2.2 Test schematic for a 81 mm chain link: Chainhawse pushes down T4 to generate
interlink angles between T4 and T5 and OPB stresses in T5. . . . . . . . . . . . . 8
2.3 Test schematic for a 124 mm chain links . . . . . . . . . . . . . . . . . . . . . . . 8
2.4 Location of reported stresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.5 SBM definition of interlink angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.6 Interpretation 1 of interlink angle definition . . . . . . . . . . . . . . . . . . . . . . 9
2.7 Empirical fit after OPB tests for 81 mm studded, 106 mm and 124 mm studless links 9
2.8 Solid model of 124 mm chain links . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.9 Ramberg Osgood true stress strain curve:Yield Stress=580 MPa, Ultimate Stress=860
MPa, Ultimate Strain=11%, α = 0.002, n=10.3 . . . . . . . . . . . . . . . . . . . . 10
2.10 Verification of FE model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.11 Grid 3 symmetry model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.12 Comparison between FEA and experimental data . . . . . . . . . . . . . . . . . . . 12
2.13 Effect of pretension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.14 Effect of friction coefficient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.15 Interpretation 2 of interlink angle definition . . . . . . . . . . . . . . . . . . . . . . 14
2.16 Interpretation 3 of interlink angle definition . . . . . . . . . . . . . . . . . . . . . . 14
2.17 Sensitivity to interlink angle definition . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.18 Quasi-static vs dynamic analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3.1 Definitions used . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18


3.2 Tension history . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.3 Rate independency in pre-sliding . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.4 Rotation input to check memory effect . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.5 Existence of memory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.6 Vertical displacement of points B and C . . . . . . . . . . . . . . . . . . . . . . . . 20
3.7 Stick slip behaviour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.8 7 link FE model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.9 Variation of OPB stress(7 link model) . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.10 Variation of interlink angles (7 link model) . . . . . . . . . . . . . . . . . . . . . . 22
3.11 Effect of boundary condition on OPB stress . . . . . . . . . . . . . . . . . . . . . . 23
3.12 Friction behaviour of a block resting on a rough surface [8] . . . . . . . . . . . . . 23

4.1 Idealization of a block resting on a surface . . . . . . . . . . . . . . . . . . . . . . . 25


4.2 Hysteresis plot generated with one Maxwell Element(k1 = 100,W1 = 55) . . . . . . 26
4.3 Rotation input at constant Tension . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.4 Idealization of the connection between Links A and B . . . . . . . . . . . . . . . . 27

107
108 LIST OF FIGURES

4.5 Comparison of OPB stress between FEA and interlink stiffness model(constant Ten-
sion) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.6 Goodness of Fit: Interlink Stiffness Model(constant Tension) vs FEA . . . . . . . . 29
4.7 Interlink stiffness model Hysteresis vs FEA . . . . . . . . . . . . . . . . . . . . . . 30
4.8 Rotation input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.9 Tension input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.10 Interlink angles generated after FEA . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.11 Comparison of OPB stress between FEA and interlink stiffness model . . . . . . . 32
4.12 Goodness of Fit: Interlink Stiffness Model(constant Tension) vs FEA . . . . . . . . 33
4.13 Interlink stiffness model Hysteresis vs FEA . . . . . . . . . . . . . . . . . . . . . . 33
4.14 Robustness of interlink stiffness model . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.15 Sensitivity to time-step . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.16 Reproducibility of interlink stiffness model . . . . . . . . . . . . . . . . . . . . . . . 35
4.17 Deformations in an IPB link . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

5.1 Idealization of a 3 link chain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38


5.2 Cantilever bending of IPB link . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.3 Cantilever bending of OPB link . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.4 Time domain plot of interlink angles at Connection 1 . . . . . . . . . . . . . . . . . 40
5.5 Time domain plot of interlink angles at Connection 1 . . . . . . . . . . . . . . . . . 41
5.6 Discontinuity of kopb . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.7 Normalized interlink stiffness as a function of time . . . . . . . . . . . . . . . . . . 42
5.8 Time domain plot of OPB stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5.9 Idealization of a 3 link chain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.10 Time domain plot of interlink angles(3 link model) . . . . . . . . . . . . . . . . . . 44
5.11 Time domain plot of OPB stress (3link model . . . . . . . . . . . . . . . . . . . . . 44
5.12 Idealization of a 7 link chain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.13 Time domain plot of interlink angles (7 link model) . . . . . . . . . . . . . . . . . . 45
5.14 Time domain plot of OPB stress (7 link model) . . . . . . . . . . . . . . . . . . . . 45

6.1 Idealization of a 3 link chain (semi-empirical approach) . . . . . . . . . . . . . . . 47


6.2 Time domain plot of interlink angles (3 link model) . . . . . . . . . . . . . . . . . . 49
6.3 Time domain plot of OPB stress (3 link model) . . . . . . . . . . . . . . . . . . . . 50
6.4 Idealization of a 7 link model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
6.5 Time domain plot of OPB stress (7 link model) . . . . . . . . . . . . . . . . . . . . 51
6.6 Time domain plot of OPB stress (7 link model) . . . . . . . . . . . . . . . . . . . . 52
6.7 Time domain plot of OPB stress (7 link model) . . . . . . . . . . . . . . . . . . . . 53

7.1 Location of Moho Noord field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55


7.2 Moho Noord FPU mooring layout . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
7.3 Moho Noord FPU coordinate system and positive sign convention for floater motion 56
7.4 JONSWAP spectrum describing the 1 hour sea-state . . . . . . . . . . . . . . . . . 57
7.5 Flow chart depicting the methodology for estimating nominal OPB stress . . . . . 58
7.6 Simplified OPB model representing the first few links of the top chain . . . . . . . 58
7.7 Relative rotation between FPU and mooring line . . . . . . . . . . . . . . . . . . . 59
7.8 Nominal dimensions of a 137 mm studless link [23] . . . . . . . . . . . . . . . . . . 60
7.9 Ramberg Osgood true stress strain curve:Yield Stress=580 MPa, Ultimate Stress=860
MPa, Ultimate Strain=11%, α = 0.002, n=10.3 . . . . . . . . . . . . . . . . . . . . 61
7.10 Definitions used in the case study . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
7.11 Tension loading applied to FE model . . . . . . . . . . . . . . . . . . . . . . . . . . 62
7.12 Rotation applied to FE model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
7.13 Idealization of the connection between Links A and B . . . . . . . . . . . . . . . . 62
7.14 Cantilever bending of IPB link . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
7.15 Cantilever bending of OPB link . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
7.16 Idealization of a 7 link chain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
7.17 Time domain plot of interlink angles for 1 hour sea state . . . . . . . . . . . . . . . 66
7.18 Time domain plot of OPB stress for 1 hour sea state . . . . . . . . . . . . . . . . . 67
7.19 OPB crack location . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
7.20 Comparison of interlink angles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
LIST OF FIGURES 109

7.21 Comparison of OPB stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

8.1 String-String Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69


8.2 String-Beam Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
8.3 Numerical simulation flow chart [7] . . . . . . . . . . . . . . . . . . . . . . . . . . 71
8.4 Wave spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
8.5 Comparison of floater rotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
8.6 Comparison of tension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

9.1 Force and moment applied at remote node to produce rotation at required nodes in
FE model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
9.2 Discontinuity of kopb . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
9.3 Interlink Stiffness Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
9.4 Physics based model and Semi-empirical model . . . . . . . . . . . . . . . . . . . . 82

A.1 Nonlinear load-displacement graph . . . . . . . . . . . . . . . . . . . . . . . . . . . 83


A.2 Newton-Raphson incremental-iterative solution method . . . . . . . . . . . . . . . 85

F.1 Wave spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95


F.2 Added mass for surge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
F.3 Added mass for heave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
F.4 Added mass for pitch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
F.5 Radiation damping for surge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
F.6 Radiation damping for heave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
F.7 Radiation damping for pitch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
F.8 First order wave force transfer function for surge . . . . . . . . . . . . . . . . . . . 100
F.9 First order wave force transfer function for heave . . . . . . . . . . . . . . . . . . . 100
F.10 First order wave force transfer function for pitch . . . . . . . . . . . . . . . . . . . 101
F.11 Quadratic transfer function for surge . . . . . . . . . . . . . . . . . . . . . . . . . . 101
F.12 Quadratic transfer function for heave . . . . . . . . . . . . . . . . . . . . . . . . . . 102
F.13 Quadratic transfer function for pitch . . . . . . . . . . . . . . . . . . . . . . . . . . 102
F.14 Steady drift transfer function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
110 LIST OF FIGURES
List of Tables

4.1 Parameters obtained after solving 4.8 . . . . . . . . . . . . . . . . . . . . . . . . . 28


4.2 Parameters obtained after solving 4.13 . . . . . . . . . . . . . . . . . . . . . . . . 32

5.1 Equivalent beam dimension of IPB link . . . . . . . . . . . . . . . . . . . . . . . . 39


5.2 Equivalent beam dimension of OPB link . . . . . . . . . . . . . . . . . . . . . . . . 39

6.1 Parameters obtained after solving 6.9 . . . . . . . . . . . . . . . . . . . . . . . . . 49


6.2 Parameters obtained after solving 6.9 for 7 link model . . . . . . . . . . . . . . . . 51
6.3 Parameters obtained after solving 6.18 . . . . . . . . . . . . . . . . . . . . . . . . 52

7.1 Characteristics of top chain of Moho Noord FPU mooring lines . . . . . . . . . . . 57


7.2 Equivalent beam dimension of IPB link . . . . . . . . . . . . . . . . . . . . . . . . 65
7.3 Equivalent beam dimension of OPB link . . . . . . . . . . . . . . . . . . . . . . . . 65

111

Anda mungkin juga menyukai