Mscthesis DAOkret

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 134

Investigation of the Fatigue

Behaviour of Well Systems

D.A. (Daniel) Okret


Delft University of Technology
Investigation of the Fatigue Behaviour
of Well Systems

Master of Science Thesis

For the degree of Master of Science in Offshore Engineering at Delft


University of Technology

D.A. (Daniel) Okret

November 24, 2016

Chairman: Prof. dr. A.V. Metrikine TU Delft


Thesis committee: Dr. ir. A. Tsouvalas TU Delft
Dr. ir. M.A.N. Hendriks TU Delft
Ir. J. McConochie Shell
Ir. M. van der Kraan Shell

Faculty of Mechanical, Maritime and Materials Engineering (3mE) · Delft University of


Technology
The work in this thesis was supported by Royal Dutch Shell. Their cooperation is hereby
gratefully acknowledged.

Copyright
c Offshore and Dredging Engineering
All rights reserved.
Table of Contents

Preface xi

Abstract xiii

1 Introduction 1
1-1 Background information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1-2 Problem Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1-3 Research Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1-4 Thesis Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 System Description 5
2-1 Overall Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2-2 Upper systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2-2-1 Mobile Offshore Drilling Unit . . . . . . . . . . . . . . . . . . . . . . . . 7
2-2-2 Tensioning System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2-2-3 Drilling fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2-2-4 Drilling Riser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2-2-5 Flex Joints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2-3 Lower systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2-3-1 Lower Marine Riser Package . . . . . . . . . . . . . . . . . . . . . . . . 8
2-3-2 Blowout Preventer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2-3-3 Wellhead . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2-3-4 Well System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

3 Loads acting on the Well System 13


3-1 Sources of Cyclic Loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3-1-1 Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3-1-2 Vessel motions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3-1-3 Currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3-2 Fatigue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

Master of Science Thesis D.A. (Daniel) Okret


ii Table of Contents

4 Riser and Well System Dynamics 19


4-1 Structural Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4-1-1 Discrete Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4-1-2 Continuous Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4-2 Riser Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4-2-1 Pipe Conveying Fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4-2-2 Riser Tension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4-2-3 Material Damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4-2-4 Riser equation of motion . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4-3 Well System Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4-3-1 Soil Stiffness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4-3-2 Soil Damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4-4 Challenges of Analytical Modelling . . . . . . . . . . . . . . . . . . . . . . . . . 27
4-4-1 Coupled modes due to spatial varying tension . . . . . . . . . . . . . . . 27
4-4-2 Time varying boundary conditions . . . . . . . . . . . . . . . . . . . . . 28
4-4-3 Time vs. Frequency Domain . . . . . . . . . . . . . . . . . . . . . . . . 28

5 Frequency Domain Analysis 31


5-1 Model description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5-2 Governing Equation of Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5-3 Interface and Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . 34
5-4 Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5-4-1 Numerical validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5-4-2 Validation with FEM . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5-4-3 Comparison with Case Study . . . . . . . . . . . . . . . . . . . . . . . . 40
5-5 Sensitivities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5-5-1 Water Depth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5-5-2 Effects of drilling fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5-5-3 Effect of BOP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5-5-4 Soil Stiffness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5-5-5 Soil Damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5-5-6 Drag Coefficient and Diameter . . . . . . . . . . . . . . . . . . . . . . . 50
5-5-7 Placement of the connector . . . . . . . . . . . . . . . . . . . . . . . . . 51

6 Time Domain Analysis 53


6-1 Model description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
6-2 Post processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
6-3 Effect of Tension variation across time . . . . . . . . . . . . . . . . . . . . . . . 55
6-3-1 Source of fatigue damage . . . . . . . . . . . . . . . . . . . . . . . . . . 56
6-4 Verification of tensioning model . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
6-4-1 Mitigation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

D.A. (Daniel) Okret Master of Science Thesis


Table of Contents iii

7 Conclusions and Recommendations 63


7-1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
7-1-1 Frequency domain model . . . . . . . . . . . . . . . . . . . . . . . . . . 63
7-1-2 Time domain model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
7-1-3 Main identified improvements . . . . . . . . . . . . . . . . . . . . . . . . 65
7-2 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

Bibliography 69

Glossary 73
List of Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

A Hydrodynamics 77
A-1 Wave Spectrum, Amplitudes and Velocity . . . . . . . . . . . . . . . . . . . . . 77
A-2 Wave Steepness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
A-3 Spectra and Time Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
A-4 Vessel Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
A-5 Linearisation Of Hydrodynamic Forces . . . . . . . . . . . . . . . . . . . . . . . 81

B Dynamics 83
B-1 Beam Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
B-2 Experimental Modal Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
B-3 Soil Stiffness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
B-4 Soil Damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
B-5 Fourier Transforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

C Fatigue 91
C-1 Fatigue mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
C-2 Fatigue Analysis in Time Domain . . . . . . . . . . . . . . . . . . . . . . . . . . 92
C-3 Fatigue Analysis in Frequency Domain . . . . . . . . . . . . . . . . . . . . . . . 93
C-4 SCF’s, SN curves and Safety Factors . . . . . . . . . . . . . . . . . . . . . . . . 95
C-5 Nonzero mean stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

D Riser Tensioners 99
D-1 Tensioning systems and components . . . . . . . . . . . . . . . . . . . . . . . . 99
D-2 Source of time varying tension . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
D-3 Modelling of tensioner system . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

E Programming 105
E-1 Frequency Domain Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
E-2 Time Domain Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

F Drilling Rig Information 113

Master of Science Thesis D.A. (Daniel) Okret


iv Table of Contents

D.A. (Daniel) Okret Master of Science Thesis


List of Figures

1-1 Drilling set-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2-1 Drilling riser configuration and components . . . . . . . . . . . . . . . . . . . . 6


2-2 Schematic of typical well construction . . . . . . . . . . . . . . . . . . . . . . . 10
2-3 Example of connector (Leopard SD by Oilstate) . . . . . . . . . . . . . . . . . . 10
2-4 Drilling riser configuration and components . . . . . . . . . . . . . . . . . . . . 11
2-5 Secondary well barrier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

3-1 JONSWAP Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14


3-2 Strouhal Number as a function of Reynolds Number . . . . . . . . . . . . . . . . 16

4-1 Mechanical scheme of SDOF . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20


4-2 Analyse beam as a spring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4-3 Simply supported beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4-4 Timoshenko vs Euler-Bernoulli beam models . . . . . . . . . . . . . . . . . . . . 22
4-5 Beam on viscoelastic foundation . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4-6 Example of p-y curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

5-1 Scheme of Riser and Well System model . . . . . . . . . . . . . . . . . . . . . . 32


5-2 Scheme of conductor in 143 m water depth . . . . . . . . . . . . . . . . . . . . 37
5-3 Comparison of Semi-Analytical model and FREERISE results . . . . . . . . . . . 38
5-4 Frequency Response Function of semi-analytical model . . . . . . . . . . . . . . 39
5-5 Comparison of Semi-Analytical model and USFOS results . . . . . . . . . . . . . 40
5-6 Unfactored fatigue lifetime of riser system for well in New Zeeland . . . . . . . . 41
5-7 Frequency Response Function for different water depths . . . . . . . . . . . . . . 42
5-8 Fatigue lifetime for different water depths . . . . . . . . . . . . . . . . . . . . . 43
5-10 Effect of drilling mud density on fatigue lifetime . . . . . . . . . . . . . . . . . . 44

Master of Science Thesis D.A. (Daniel) Okret


vi List of Figures

5-9 Effect of drilling mud density on curvature Frequency Response Function 5 metres
below the mudline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5-11 Effect of mud flow velocity for T p = 6s and Hs = 3m . . . . . . . . . . . . . . 45
5-12 Effect of BOP size on curvature Frequency Response Function 5 metres below the
mudline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5-13 Effect of sea state on fatigue lifetime for 270 metres water depth . . . . . . . . . 46
5-14 Effect of soil stiffness on curvature Frequency Response Function 5 metres below
the mudline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
5-15 Fatigue lifetime for different soil stiffness parameters . . . . . . . . . . . . . . . 48
5-16 Effect of soil damping on curvature Frequency Response Function 5 metres below
the mudline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5-17 Fatigue lifetime for varying soil damping ratios . . . . . . . . . . . . . . . . . . . 50
5-18 Fatigue lifetime for varying damping ratios . . . . . . . . . . . . . . . . . . . . . 50
5-19 Effect of riser drag coefficient on fatigue lifetime . . . . . . . . . . . . . . . . . 51
5-20 Fatigue lifetime for different connector placements . . . . . . . . . . . . . . . . . 52

6-1 USFOS representation of model . . . . . . . . . . . . . . . . . . . . . . . . . . . 54


6-2 USFOS beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
6-3 Comparison of fatigue lifetime including and excluding tension variation for a sea
state with Hs = 5m and T p = 12s . . . . . . . . . . . . . . . . . . . . . . . . . 56
6-4 Comparison of fatigue lifetime including and excluding the fatigue lifetime reduction
caused by axial stresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
6-5 Moments as a function of time at mudline . . . . . . . . . . . . . . . . . . . . . 57
6-6 Eigen modes of riser configuration . . . . . . . . . . . . . . . . . . . . . . . . . 58
6-7 Moments as a function of period at the mudline . . . . . . . . . . . . . . . . . . 58
6-8 Effect of riser drag coefficient on fatigue lifetime . . . . . . . . . . . . . . . . . 59
6-9 Vertical displacement of top node . . . . . . . . . . . . . . . . . . . . . . . . . 59
6-10 Heave motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
6-11 Effect of feedback loop duration on time varying tension . . . . . . . . . . . . . 60
6-12 Uncompensated time varying tension . . . . . . . . . . . . . . . . . . . . . . . . 61

A-1 Wave Record Analysis and Recreation . . . . . . . . . . . . . . . . . . . . . . . 79


A-2 Heave Motions of a Vertical Cylinder . . . . . . . . . . . . . . . . . . . . . . . . 80

B-1 Examples of boundary conditions of a beam . . . . . . . . . . . . . . . . . . . . 84


B-2 Half-power bandwidth method . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
B-3 Static P-Y curve given by API . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
B-4 Dynamic P-Y curve given by API . . . . . . . . . . . . . . . . . . . . . . . . . . 87
B-5 Stress-strain behaviour implied by viscous damping . . . . . . . . . . . . . . . . 88
B-6 Damping ratio profiles for several displacement amplitudes . . . . . . . . . . . . 90

C-1 Phases of fatigue life . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91


C-2 Rainflow plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

D.A. (Daniel) Okret Master of Science Thesis


List of Figures vii

C-3 Why Bendat’s Method is Conservative . . . . . . . . . . . . . . . . . . . . . . . 93


C-4 SN curves in sea water with cathodic protection . . . . . . . . . . . . . . . . . . 95
C-5 Effect of nonzero mean stress on fatigue . . . . . . . . . . . . . . . . . . . . . . 97
C-6 Fatigue resistance reduction because of non-mean stress . . . . . . . . . . . . . 98

D-1 Schematic overview for wireline (left) and DAT (right) systems . . . . . . . . . . 99
D-2 Detailed scheme of DAT system . . . . . . . . . . . . . . . . . . . . . . . . . . 100
D-3 Example of tensioner transfer functions . . . . . . . . . . . . . . . . . . . . . . . 102
D-4 Comparison of tension variation for both methods in 250m water depth . . . . . 103
D-5 Comparison of tension variation for both methods in 1000m water depth . . . . . 103
D-6 Comparison of tension variation for both methods in steep sea states . . . . . . . 104
D-7 Comparison of tension variation for both methods in mild sea states . . . . . . . 104

E-3 Pinned-pinned beam with distributed load . . . . . . . . . . . . . . . . . . . . . 106


E-1 Cantilever beam with point load . . . . . . . . . . . . . . . . . . . . . . . . . . 107
E-2 Cantilever beam with distributed load . . . . . . . . . . . . . . . . . . . . . . . 107
E-4 Overview of Matlab routine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
E-5 Effect of simulation duration for Hs = 1 and T p = 12 . . . . . . . . . . . . . . 109
E-6 Overview of time domain model routines . . . . . . . . . . . . . . . . . . . . . . 111

F-1 Heave RAO’s of vessels used in analysis . . . . . . . . . . . . . . . . . . . . . . 114


F-2 Surge RAO’s of vessels used in analysis . . . . . . . . . . . . . . . . . . . . . . . 114

Master of Science Thesis D.A. (Daniel) Okret


viii List of Figures

D.A. (Daniel) Okret Master of Science Thesis


List of Tables

2-1 Riser Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8


2-2 BOP stack size and weight . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2-3 BOP and LMRP Hydrodynamic Properties . . . . . . . . . . . . . . . . . . . . . 9

5-1 Comparison of first three natural frequencies . . . . . . . . . . . . . . . . . . . . 37


5-2 Effect of drilling mud density on unfactored fatigue lifetime at 5 metres below
mudline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5-3 Effect soil stiffness on the critical modes 5 metres below mudline . . . . . . . . . 47
5-4 Effect of soil damping on modal damping of critical modes 5 metres below mudline 49
5-5 Effect of soil damping on modal damping of critical modes 5 metres below mudline 51
5-6 Effect of soil damping on modal damping of critical modes 5 metres below mudline 52

6-1 Natural periods of riser configuration . . . . . . . . . . . . . . . . . . . . . . . . 57

B-1 Point-wise cyclic p-y curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87


B-2 Representative properties for normally consolidated clay . . . . . . . . . . . . . . 87
B-3 Representative properties for overconsolidated clay . . . . . . . . . . . . . . . . . 88
B-4 Representative values of initial stiffness for overconsolidated clay . . . . . . . . . 88

C-1 Cycle counting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93


C-2 Fatigue Safety Factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
C-3 SCF’s, SN curves and safety factors . . . . . . . . . . . . . . . . . . . . . . . . 96
C-4 Mechanical properties of steel . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

E-1 Input values for bvp4c validation . . . . . . . . . . . . . . . . . . . . . . . . . . 106


E-2 Fatigue lifetime 2 metres under the mudline . . . . . . . . . . . . . . . . . . . . 109

F-1 Types of Floating Drilling Rigs - Drilling Depth . . . . . . . . . . . . . . . . . . 113


F-2 Types of Floating Drilling Rigs - Generations . . . . . . . . . . . . . . . . . . . . 113

Master of Science Thesis D.A. (Daniel) Okret


x List of Tables

D.A. (Daniel) Okret Master of Science Thesis


Preface

This report in front of you describes the research conducted for the final project to conclude
my master degree in Offshore and Dredging Engineering at Delft University of Technology.
In this preface I would like to acknowledge the people who made me achieve this possible. I
would like to thank Andrei Metrikine, chairman of the committee, for inspiring me, awakening
my passion for structural dynamics and guiding me throughout the project. Furthermore, I
thank Anderi van der Stap for giving me the opportunity to do this project at Shell.
During this project I have been supported by many people. At Shell I would like to thank
my supervisors Jason McConochie, Micha van der Kraan and Jan-Willem van der Graaf for
their support, time and sharing of valuable experience. I thank Emiel van Twillert, who was
always there for me and showed me the ins and outs of the company. At the TU Delft I would
like to thank Apostolos Tsouvalas who shared his deep understanding of structural dynamics
with me and for critically reviewing my work.
A special thanks to my parents and family, who were always there for me and provided me
with the opportunities to get this far. Finally, I thank all my friends who made my journey
so much fun.

Delft, November 24, 2016 D.A. (Daniel) Okret

Master of Science Thesis D.A. (Daniel) Okret


xii Preface

D.A. (Daniel) Okret Master of Science Thesis


Abstract

Subsea wells at water depths larger than 100 metres are usually drilled using a Mobile Offshore
Drilling Units (MODU). During drilling operations, the MODU is attached to the well system
through a marine drilling riser. The motions of the drilling rig and riser due to environmental
conditions cause cyclic loads on the well system, which may lead to its fatigue failure. Most
of the easily recoverable reservoirs have already been developed, and new explorations take
place at more remote locations with challenging conditions (e.g. deeper waters and harsher
environments). Moreover, technological advances and the low oil prices enable and stimulate
higher recovery rates of existing wells, leading to a longer connection time between the drilling
rig and well system. These two developments result in an increased magnitude and duration
of the loads acting on the well system, making it more vulnerable to fatigue failure. The well
system is part of the primary barrier that contains reservoir fluids, and therefore a failure
could have severe consequences such as an oil spill.
Currently, there are no codes or international standards on how the fatigue calculations of the
well system should be performed. Companies use different methods for this assessment, and
results can vary significantly depending on the applied method and input parameters. The
large spread of fatigue calculation results makes it challenging to evaluate which method is
more reliable, and gives a better estimation of the actual fatigue damage to be expected. The
purpose of this study is to provide a better understanding of the way in which the fatigue
behaviour of the well system can be modelled and what critical parameters affect it. Two
models were developed to analyse the fatigue behaviour of the well system.
The first method is a semi-analytical frequency domain model where all nonlinear effects are
linearised. The most important linearisations are the soil stiffness and hydrodynamic drag
acting on the riser. This model is very fast and easy to set up, and gives accurate fatigue
estimates for the scenarios considered in this study. Another advantage of this model is that
it can include a soil damping representation that has been pointed out by literature as being
the most realistic for soil-conductor interaction. However, due to the applied linearisation, it
cannot capture important loads acting on the system, such as the time variation of tension.
To simulate these loads, a finite element time domain model is developed with the USFOS
software. This software has been extensively used for the assessment of offshore structures,
particularly soil structure interaction. It is able to represent nonlinearities and can therefore
include more features in a more accurate manner. Its main disadvantages are that simulations
are time consuming and more complex to set up. Moreover, soil damping has not been

Master of Science Thesis D.A. (Daniel) Okret


xiv Abstract

included in this model.


The most important findings are the effects of soil damping and time variation of tension. Soil
damping is currently not used for the fatigue assessment of a well system. The soil damping
method applied in the frequency domain model showed to increase the modal damping for
critical modes to the well fatigue. Modal damping can increase by a factor larger than 5.
Tension in the well system is not constant in time, because the tensioning system typically
cannot compensate all motions of the MODU. This effect is typically neglected by the indus-
try. However, including the time variation tension can severely decrease the fatigue lifetime of
the well system, by a factor exceeding 100. This occurs because critical modes of the system
are excited which contribute significantly to the well system fatigue. An active compensation
system could mitigate these effects. The most efficient way to implement such a system is by
only compensating the time varying tension, and not providing the full tension for the riser.
Further improvements to the fatigue lifetime can be achieved by increasing the drag diameter
of the riser and placing the fatigue sensitive elements, such as connectors, away from fatigue
prone areas also significantly improves the fatigue performance of the system. A larger drag
diameter can be achieved by applying buoyancy neutral elements (which could be buoyancy
elements filled with water) to the riser.

D.A. (Daniel) Okret Master of Science Thesis


”Learn from yesterday, live for today, hope for tomorrow. The important thing is
not to stop questioning.”
— Albert Einstein
Chapter 1

Introduction

This chapter will start with the background information of the project in Section 1-1. The
problem definition is given in Section 1-2 and the research questions are formulated in Section
1-3. Further, the structure of this thesis is elaborated in Section 1-4.

1-1 Background information

The production of hydrocarbons has been going on for decades, and most of the easily recov-
erable reservoirs have been found and developed. The majority of new explorations occur at
more remote locations with challenging conditions (e.g. deeper waters and harsher environ-
ments). As a result, these new developments require heavier equipment, longer drilling times
and the emergency response time is increased. At the same time, technology has enabled
higher recovery rates of existing wells, which makes life time extensions of existing wells more
profitable. This leads to an incensed amount of maintenance (workover) and a longer period
of connection between the well system and the drilling rig.

Subsea wells are usually drilled by a Mobile Offshore Drilling Unit. The drilling rig is attached
to the well system through a marine drilling riser and a subsea stack. The dynamic response
of the MODU and drilling riser due to environmental conditions (e.g. current and waves)
impose dynamic loads on the well system. Figure 1-1 gives an overview of the drilling set-up.

Master of Science Thesis D.A. (Daniel) Okret


2 Introduction

Figure 1-1: Drilling set-up [1]

Because of the increase in drilling time and loads acting on the well system, it becomes
more vulnerable to fatigue damage. Although this is an issue with severe consequences,
there are currently no codes or international standards on how to calculate the fatigue of
these systems. Companies use different methods to assess this fatigue behaviour, with results
varying significantly.
Moreover, are many knowledge gaps which need to be filled to fully understand its dynamic be-
haviour and the main parameters which influence it. The currently applied methods probably
do not give a good enough representation of reality and could underestimate or overestimate
the fatigue damage on the system because crucial aspects are not accounted for.
Furthermore, well systems are located at the bottom of the sea and extend until the reservoir,
which makes it difficult to access them. This combined with the low interest in this topic in
the past are the reasons for the small number of measurements campaigns that were conducted
to acquire data on the dynamic behaviour of wells. Results of these campaigns have not been
shared broadly within the industry such that clear conclusions can be derived. There is a
benefit in following up with monitoring campaign to compare the predictions. Validation of
any fatigue estimation method with measurement data is therefore extremely challenging.
After the Deepwater Horizon incident, the industry started to pay more attention to this topic,
and in 2010 a Joint Industry Project was created to develop a starting point for recommended
practice for the well system fatigue calculation. It published its first report in 2011, and a

D.A. (Daniel) Okret Master of Science Thesis


1-2 Problem Definition 3

follow up report has been published in 2015, but only general guidelines for the assessment
have been developed so far. These guidelines are considered as being over conservative for
the included aspects, but not all important aspects are taken into account [2]. Some of the
highlighted parameters that could play a significant role, but have not been fully incorporated
into the current guidelines, are the effect of soil damping, VIV and time variation of the top
tension.

1-2 Problem Definition

The currently applied methods for calculating well system fatigue do not provide a reliable
outcome. Small variations on important input parameters result in large outcome varia-
tions and may exceed the required safety margin. It is therefore necessary to verify existing
calculation methods and determine which input parameters influence the outcome the most.

1-3 Research Questions

From the described problem definition, it is clear that the industry would benefit from ob-
taining better guidance on which parameters are mostly influencing the behaviour of well
systems. The models used to calculate resulting fatigue are maybe not detailed or specific
enough to simplify the problem and guide the industry. The first question for this research
project is therefore defined as:

1. How can the well system be modelled, accounting for all the loads acting on the structure
and its dynamic behaviour, in an accurate way for fatigue damage calculation purposes?

Once a suitable model has been developed, it will become possible to determine the fatigue
lifetime of a well system. Another identified gap is the uncertainty in the parameters which
need to be considered during the calculation of the fatigue behaviour of a well system. The
second research question is posed as follows:

2. What are the key parameters that influence the fatigue damage of the well system
and how can these be properly included in an model developed for reliable fatigue
calculations?

Previous research has shown that parameters such as time variation of top tension and soil
damping are generally not included in fatigue calculations. The influence of such parameters
will be tested with in the model.
Answering these questions will provide a better understanding of the dynamic behaviour of
well systems and lead to the opportunity in optimizing the well system. Moreover the results
will be presented with adequate safety margins and allow insight in the real boundary condi-
tions for well drilling operations. Ultimately, this insight will also enable well fatigue analysis
by using more straightforward models to describe the dynamic behaviour. Understanding the
principles of dynamic behaviour of wells systems with adequate calculations tools, will allow
implementation of adequate equipment and modifications that overall reduce cost by as well
as building more robust wells.

Master of Science Thesis D.A. (Daniel) Okret


4 Introduction

1-4 Thesis Structure

This thesis is organised as follows. Chapter 1 is an introduction to the project in which the
background, the problem and objectives are discussed.
Chapter 2 gives a description of the system that will be assessed in this thesis. The loads
acting on this system are presented in Chapter 3. Chapter 4 describes the dynamics of the
system, and how they can be modelled. Chapter 5 presents the frequency domain model, and
the results obtained with this model. Chapter 6 introduces the time domain model, and how
time varying tension affects the fatigue behaviour of the system.
Finally, in Chapter 7 the conclusions and recommendations obtained from this project are
presented.

D.A. (Daniel) Okret Master of Science Thesis


Chapter 2

System Description

2-1 Overall Description

The well system is part of a larger system. During its lifetime it is attached to a MODU
through a marine riser multiple times. The activities involved can be divided into drilling,
completion and workover operations. Each activity has a particular set of configurations. The
largest fatigue damage on the well system occurs during drilling operations [1]. Therefore,
this configuration will be used to assess the fatigue behaviour of the system. A representation
of the drilling configuration with all of its components is displayed in Figure 2-1.
To analyse this complex system it is necessary to first define the elements that compose it.
From Figure 2-1 the following elements can be identified:

• Mobile Offshore Drilling Unit


• Tensioner
• Upper Flex Joint
• Drilling riser
• Lower Flex Joint
• Lower Marine Riser Package + Blowout Preventer
• Wellhead
• Well System

There are elements that are not part of the structural system but have to be included in the
model. These are the soil surrounding the conductor and the fluid between the seabed and
water surface.

Master of Science Thesis D.A. (Daniel) Okret


6 System Description

Drill Deck (RKB)

Upper Flex Joint

Tensioners
MWL
Slip Joint

Riser Buoyancy Joints

Bare Riser Joints

Lower Flex Joint


LMRP
BOP
Mudline

Conductor Casing

Figure 2-1: Drilling riser configuration and components

D.A. (Daniel) Okret Master of Science Thesis


2-2 Upper systems 7

2-2 Upper systems

2-2-1 Mobile Offshore Drilling Unit

There are several types of MODU’s, including jack-ups, drill ships and semi-submersibles.
Drilling operations in water depths surpassing 150 metres water depth are usually done with
the last two types. Aboard these vessels, equipment needed for the drilling operations is
present. Moreover, it serves as the upper attachment point for the drilling riser.

2-2-2 Tensioning System

The main function of the tensioning system is to maintain the top tension of the riser constant.
They provide stiffness to the riser and ensure there is no compression above the BOP. There
are two types of riser tensioning systems, namely Direct Acting Tensioner (DAT) and wireline
tensioners. Both systems use hydraulic cylinders to develop riser tension. In DAT systems, the
cylinders form a direct connection between the MODU and the tensioner ring. The cylinders
of a wireline system are placed on deck and the connection between the cylinder and the
drilling ring occur through wires and blocks. Tensioning systems are further elaborated in
Appendix D.

2-2-3 Drilling fluid

The function of the drilling fluid is to maintain pressure in the well bore to prevent fluid from
entering or collapse, lubricate the drill bit and transport drill cuttings up the riser to the the
MODU. The necessary flow velocity to enable this transport is usually given in ft/min and
varies from 50 ft/min and 200 ft/min [3]. The density of the drilling fluid is dependent of its
application. It is usually given in ppg (pound per gallon) and can vary from 8 ppg to 20 ppg,
which correspond to 960 kg/m3 and 2400 kg/m3 respectively. However, for usual applications
the mud density is somewhere between 9 and 15 ppg.
The effect of drilling fluid on the dynamic behaviour of the system will be elaborated in
Section 5-5-2.

2-2-4 Drilling Riser

The drilling riser is a complex system formed by slick joints, kill lines and auxiliary lines.
The slick joints are large diameter pipes through which the drill string and fluid are inserted,
which serve as the main riser pipe. Kill lines and auxiliary lines are attached to the main
riser pipe and are used to control the well and BOP from the MODU. Additional buoyancy
elements may be attached to the main riser pipe to optimise the tension of the system.
Only the stiffness of the main riser pipe needs to be accounted in the calculations. However,
its increased hydraulic drag diameter will be larger than the diameter of the slick joints. It
will be calculated as described in [4], where the maximum projected diameter of the drilling
riser is considered. This can be obtained by applying Equation 2-1.

Master of Science Thesis D.A. (Daniel) Okret


8 System Description

Ddrag = Dmain pipe + 2Dkill (2-1)

The used parameters are summarized in Table 2-1.

Table 2-1: Riser Properties

Parameter Value Unit


Diameter 0.533 metres
Wall Thickness 0.022 metres
Drag diameter 0.875 metres

2-2-5 Flex Joints

There are usually two flex joints in the system, namely the Upper Flex Joint and Lower Flex
Joint. Their function is to reduce the transmission of bending stresses from the vessel to the
riser, and from the riser to the well system respectively. Their importance gets larger when
excursions of the MODU are present.

In this project, it is assumed that the UFJ acts as a pinned connection, and therefore has no
rotational stiffness. Furthermore, it is located in the RAO centre of the vessel, which implies
that pitch and roll motions will have no effect on the riser system.

The LFJ has been assumed to have a linear rotational stiffness of 38 kNm/degree [5].

2-3 Lower systems

2-3-1 Lower Marine Riser Package

The LMRP is usually formed by an annular BOP, a control pod and a connector to the LFJ.
It seals the space between the drill pipe and the wellbore, or annulus. During an emergency
disconnect procedure, the LMRP is disconnected from the BOP.

2-3-2 Blowout Preventer

The BOP is used to prevent well fluids from escaping in case of a blowout. This means that
the pressure of the well fluids is higher that the drilling fluid on top of it, and therefore go up
the well bore. A blowout can lead to severe complications including an oil spill and explosions.
The size and dimensions of BOP’s can vary significantly depending on the application and
used MODU. The used parameters for this thesis are given in Tables 2-2.

D.A. (Daniel) Okret Master of Science Thesis


2-3 Lower systems 9

Table 2-2: BOP size and weight [6]

BOP Stack Height BOP Stack Weight


Vessel
(m) (Tonnes)
3rd Generation Vessel 10.1 153.6
4th Generation Vessel 14.1 186.7
5th Generation Vessel 16.3 290.1

Table 2-3: BOP and LMRP Hydrodynamic Properties [7][8]

Parameter Value Unit


Hydrodynamic Diameter 5-8 metres
Drag Coefficient 2.12 -
Inertia Coefficient 2.19 -

The BOP is assumed as being infinitely stiff. Although this is a conservative approach, it has
been shown in [9] that the difference in fatigue lifetime between an elastic and stiff modelling
method is smaller than 2%. Moreover, this makes modelling the system significantly easier.

2-3-3 Wellhead

The wellhead is the structural element on which the BOP is supported during drilling opera-
tions. Furthermore, all casings and pipes of the well system, with exception of the conductor,
are hung of from the wellhead. It is a complex system where multiple sealing mechanisms are
applied.

2-3-4 Well System

Components

The well system is composed of a series concentric steel pipes, as displayed in Figure 2-2.
The drilling of a subsea well is done in phases, in which steel pipes of decreasing diameter are
lowered into the well and cemented in place.

Master of Science Thesis D.A. (Daniel) Okret


10 System Description

Figure 2-2: Schematic of typical well construction [4]

The outer pipe is called the conductor and its size is usually 30 or 36 inch . Its function
is providing structural support for the wellhead and other pipes that will be lowered. It
is usually 100 metres long and is not formed by a single segment, but a series of segments
connected through connectors and welds. These connectors are extremely fatigue sensitive
and need to be considered in the fatigue life calculations. There are many types of connectors,
and in Figure 2-3 an example is given. The remaining pipes which do not reach the reservoir
are called casings, and the final pipe through which well fluid reach the mudline is called
production tubing.

Figure 2-3: Example of connector (Leopard SD by Oilstate)

D.A. (Daniel) Okret Master of Science Thesis


2-3 Lower systems 11

It is assumed that the conductor is has the governing fatigue lifetime. Therefore this element
will be used in the fatigue calculations. A standard conductor is chosen, with a diameter of
30 inches and wall thickness of 1.5 inches.

Drilling Phases

Each phase has its own drilling configuration. Initially, a large diameter hole is drilled for the
installation of the conductor. After the conductor is cemented in place, the surface casing is
drilled and installed. Usually no riser is attached up to this point. Once the surface casing is
successfully cemented in place, the BOP and riser are connected to the wellhead. From here
on, drilling is done through the riser. An overview of the stages and configurations is given
in Figure 2-4.

Figure 2-4: Drilling riser configuration and components

Master of Science Thesis D.A. (Daniel) Okret


12 System Description

Types of failure

Well systems are designed to endure the pressure of the reser-


voir and avoid unwanted escapes of hydrocarbons. To achieve
this goal, several entities such as the Petroleum Safety Au-
thority in Norway, have imposed a dual barrier construction
of wells [10]. This requirement is enforced to decrease the
probability of hydrocarbon leaks. As an example, the barri-
ers during drilling operations are shown in Figure 2-5. The
primary well barrier is the hydrostatic pressure of the drilling
fluid column and the second barrier is composed by the cas-
ing, wellhead, drill sting and BOP.

The two most probable well system failure causes are acci-
dental loading, for example a drilling rig drift off, and accu-
mulated fatigue damage [1]. The consequences of structural
failure is dependent on the operation which is being under-
taken. Fatigue structural failure occurs when the crack gets
larger than a certain size, called critical crack size.

In a fatigue failure, the loss of pressure containment capa-


bility may occur before the actual structural fatigue failure.
This failure mode can be referred to as "leak before failure".
This is a situation when a fatigue through-wall crack has not
reached the critical crack size, but leaks hydrocarbons.

The opposite scenario is also possible. It occurs when


the crack reaches the critical crack size without reaching
a though-wall size through which hydrocarbons can escape.
This failure mode is referred to as "failure before leak".

Because both scenarios can occur below the BOP, it is ex-


tremely hard to control a leakage in such an event. Not only
is the area hard to gain access to (might be several metres
Figure 2-5: Sec-
below mudline and at great water depths), but the pressures ondary well barrier
might also be very large. Therefore it is critical that such a [10]
failure is avoided.

D.A. (Daniel) Okret Master of Science Thesis


Chapter 3

Loads acting on the Well System

In Chapter 2, the well system and its components has been described. The loads acting on
this system will be presented in this Chapter. In Section 3-1, the sources of loading and their
representation will be introduced. Section 3-2 elaborated the effect these cyclic loads have on
structures.

3-1 Sources of Cyclic Loading

The forces that influence the vibrations of the system can be divided into external forces
and those which are a response to the dynamics of the system itself. The fluid surrounding
the system can result in hydrodynamic damping and added mass, which are a reaction to
its vibration. The same applies to the restoring forces of the soil on the conductor. These
dynamic forces are dependent on the deflection, velocity and acceleration of the entire system
and have a large influence on its response.
By nature these forces are not the source of vibrations. The external excitation mechanisms
of the well system that will be approached are:

• Waves
• Vessel motions
• Currents

3-1-1 Waves

Ocean waves influence vibrations on the system in two manners. The first one are the direct
wave forces acting on the riser itself. The second way is the excitation of the MODU, which
will be described in the following section.
The waves acting on the system are irregular waves. Such waves can be described with a
power density spectrum. There are various spectrum types which have been developed for

Master of Science Thesis D.A. (Daniel) Okret


14 Loads acting on the Well System

different purposes. A JONSWAP spectrum will used to describe the waves. This spectrum
was developed for fetch-limited wind generated seas, but is able to describe fully developed
seas by changing the peakedness factor. Because many oil and gas fields are located relatively
near the coast line, this spectrum is an ideal choice. This spectrum is expected to be a
reasonable model for for the limit of Equation 3-1 [11].

Tp
3.6 < √ <5 (3-1)
Hs

where: Hs = significant wave height [m]


Tp = peak period [s]

An example JONSWAP is depicted in Figure 3-1, for a Hs = 4m and Tp = 10s. In Appendix


A-1 there is a more into depth explanation on the JONSWAP.

JONSWAP Spectrum
5

4.5

4
Spectral Density [m2 rad/s]

3.5

2.5

1.5

0.5

0
0 0.5 1 1.5 2 2.5 3 3.5
Frequency [rad/s]

Figure 3-1: JONSWAP Spectrum

The wave forces are proportional to the velocity of the water particles due to waves. So to
obtain the wave forces, first the velocity field due to waves needs to be defined. This can easily
be done from a wave spectrum such as a JONSWAP. These forces are largest at the surface
and gradually become smaller with with the water depth. This can clearly be observed in
Equation 3-2, in which the velocity field for deep water waves is presented.

u = ζa ω exp{kz} cos(ωt) (3-2)

D.A. (Daniel) Okret Master of Science Thesis


3-1 Sources of Cyclic Loading 15

where: ζa = wave amplitude [m]


ω = frequency [rad/s]
k = wave number [rad/m]
z = water depth [m]

Time series of waves elevations and velocities can be created from their respective spectra,
by assuming a random wave phase. This is further explained in Appendix A-3.

3-1-2 Vessel motions

The upper part of the drilling riser is attached to the MODU. As a result, any motions of
the vessel are partially transmitted to the riser itself. Therefore, the vessel acts as a upper
boundary condition for the drilling system.
The motions of a MODU are caused by environmental loads. These environmental loads come
from wind, current and waves. Due to the station keeping ability of most drilling vessels,
current and wind forces can be counteracted by either DP or mooring systems. Therefore,
the vessel motions due to wave loads will be taken into account.
Combining the wave spectrum with the drilling vessel’s RAO, the response can easily be
obtained. The amplitude of the vessel motion is calculated by applying Equation 3-3 and the
phase of the vessel motion by summing the spectrum phase with the RAO’s phase.

za (ω) = ζa (ω) · RAO (3-3)

The complete procedure to calculate the vessel response in time and frequency domain is
explained in Appendix A-4.

3-1-3 Currents

The water flowing around the drilling riser causes various phenomena which can result in
loads. The load forms that influence the dynamic response of the drilling riser are drag,
inertia and the lift force. Drag and inertia act parallel to the flow direction and can be
calculated with the Morison Equation, which is presented in Equation 3-4.

π 1
F = ρ Cm D2 u̇ + ρ Cd Du|u| (3-4)
4 2
The lift force acts perpendicular to the flow direction and can cause VIV. This phenomenon
occurs when the frequency of vortex shedding gets close to the natural frequency of the riser
system. The excitation frequency of these vortices can be calculated with the Strouhal number
with Equation 3-5.

fv · D
St = (3-5)
V

Master of Science Thesis D.A. (Daniel) Okret


16 Loads acting on the Well System

where: St = Strouhal number [-]


fv = vortex excitation frequency [Hz]
D = hydraulic diameter [m]
V = flow velocity [m/s]

The value of the Strouhal number is dependent on the Reynolds number, which is given in
Equation 3-6.

V ·D
Rn = (3-6)
ν

where: Rn = Reynolds number [-]


V = flow velocity [m/s]
D = hydraulic diameter [m]
ν = kinematic viscosity of fluid [m2 /s]

The dependency of the Srouhal and Reynolds Number can be seen in Figure 3-2. It is evident
that the Strouhal Number is approximately 0.2 in the region 102 ≤ Rn ≤ 2 · 105 . If the
Reynolds Number becomes larger, it becomes harder to define the exact value of the Strouhal
Number.

Figure 3-2: Strouhal Number as a function of Reynolds Number [12]

D.A. (Daniel) Okret Master of Science Thesis


3-2 Fatigue 17

The criterion that is usually used to access the possibility of VIV occurring is the reduced
flow velocity, shown in Equation 3-7. When the reduced velocity is equal to 5, there is an
alignment between the vortex shedding frequency and the natural frequency. However, it has
been shown that for reduced velocities ranging from 4 to 11 frequency locking may occur.

V
Vv = (3-7)
Dfv

An important feature of VIV is that the amplitudes do not grow indefinitely, but are limited
to about one diameter [13]. Therefore, these vibrations do not produce very large amplitudes,
but do have an important effect on the fatigue lifetime of the riser.
Although VIV may cause severe fatigue damage to riser systems, it can be relatively easily
mitigated by application of strakes or fairings. Therefore this phenomenon will not be included
in the fatigue assessment of the well system. A method to investigate this effect has been
developed in [14], which could also be applied to the well system.

3-2 Fatigue

Fatigue is the weakening of the material due to cyclic loading. A single load cycle is not large
enough to cause structural failure, but the cumulative damage caused by each cycle may cause
fatigue failure. The fatigue mechanism will be further elaborated in C-1.
To estimate the total fatigue damage, Miner’s Rule will be applied. This theory assumes
that the sequence of the cycles does not influence the fatigue damage. In other words, the
combined damage caused by a low stress cycle after a high stress cycle is equal to the combined
damage caused by a high stress cycle after a low stress cycle. In reality, this is not always
true. Miner’s Rule is displayed in 3-8.

X ni
D= (3-8)
Ni

where: D = accumulated fatigue damage [-]


ni = number of stress cycles ∆σi [cycles]
Ni = fatigue endurance to load ∆σi [cycles]

If the total accumulated damage is smaller than 1, the element will not fail.
The number of stress cycles (ni in Equation 3-8) can be calculated in the time domain and
frequency domain. The method that will be applied for the time domain calculation is the
Rainflow Counting. The algorithm reduces the time history of stresses to cycles of stresses.
These cycles can directly be inserted into Equation 3-8 [15]. An example of how this algorithm
works is given in Appendix C-2.
For the frequency domain approach, Dirlik’s Method will be applied. This an empirical
method that approximates the cycle-amplitude distribution by using a combination of one

Master of Science Thesis D.A. (Daniel) Okret


18 Loads acting on the Well System

exponential and two Rayleigh probability densities. The method has been shown to accurately
describe the fatigue damage when compared to time domain Rainflow Counting [16]. The
method is further elaborated in Appendix C-3.
The fatigue endurance to a certain load (Ni in Equation 3-8) is described by the SN Curves.
These curves give the mean minus two standard deviations from relevant experimental data.
This gives a 99.7% probability of survival. They have the following form:

Ni (∆σi )m = a (3-9)

where: Ni = fatigue endurance to load


∆σi = stress amplitude
a = experimental constant
m = experimental constant

Using a log-log scale, the SN curve is linear and Equation 3-9 can be rewritten as:

log Ni = log a − m log ∆σi (3-10)

With this formulation, m becomes the negative inverse slope of SN curve and log a becomes
the intercept of the log N axis by the SN curve. The SN curves for offshore steel structures,
such as jackets and risers, are defined in [17]. The experiments used for the development of
these curves were done under a zero mean stress. Therefore effects of a non zero mean stress
is not accounted for. This effect will be considered by application of the Goodman relation,
as elaborated in Appendix C-5.
The stress amplitudes used for the fatigue damage calculations need to take local geometries
into account. This is done by applying local SCF’s to the nominal stresses before inserting
them in Equation 3-10. Stress Concentration Factors are applied as follows:

∆σ = SCF · ∆σnominal (3-11)

After the fatigue damage has been calculated with Miner’s rule, an additional safety factor
needs to be applied to obtain the final fatigue lifetime. The applied SCF’s, SN curves and
safety factors are further elaborated in Appendix C-4.

D.A. (Daniel) Okret Master of Science Thesis


Chapter 4

Riser and Well System Dynamics

4-1 Structural Dynamics

When a load is applied to a structure, internal stresses and displacements occur. In (quasi)
static analysis the influence of inertia through time can be neglected in the response of the
structure. In this case, the following equilibrium equations hold [18]:

X X
F = 0, Tr = 0 (4-1)

where: F = force [N]


Tr = moment [Nm]

When the variation of the of the movements and stresses through time is not negligible, inertia
must be taken into account. In this case a dynamic analysis must be performed, and the time
dependence of the forces and moments must to be considered.

4-1-1 Discrete Systems

Discrete systems have a response that can be described with a finite number of degrees of
freedom. The most simple example of such a dynamic system is the SDOF mass-spring
system. It is displayed in Figure 4-1 and its motion can be described by Equation 4-2:

∂2x ∂x
m 2
+c + kx = F (t) (4-2)
∂t ∂t

Master of Science Thesis D.A. (Daniel) Okret


20 Riser and Well System Dynamics

F(t) x(t)

k c

Figure 4-1: Mechanical scheme of SDOF

where: m = mass [kg]


c = viscous damping [Ns/m]
kspr = spring stiffness [N/m]
x = displacement [m]
t = time [s]

Due to the simplicity of the SDOF system, several important concepts of structural dynamics
can be explained with it. One of them is the natural frequency (or eigenfrequency). This is
the frequency at which the system would vibrate given an arbitrary initial condition. For a
undamped SDOF, the natural frequency can be determined with Equation 4-3.

s
1 kspr
fn = (4-3)
2π m

If a forcing frequency is equal or near the natural frequency of the system, resonance occurs.
The response amplitude becomes relatively large in resonance, especially if little damping is
present. Large response amplitudes may cause damage to the structure. Therefore, structures
are designed such that its natural frequencies and force excitation frequencies do not coincide.
This system has been extensively researched and has many applications. A relevant example
of such an application for this research is a mass on top of beam. If the mass of the beam is
very small relative to the mass on top of it, and the rotation of the mass is very small, the
beam can be described as a spring. An analogy between this model and a BOP on top of
a wellhead can be made, even though the rotation of the BOP may not be neglected. This
idealisation is displayed in Figure 4-2.
The stiffness of the spring representing the beam fixed rigidly to the ground is given in
Equation 4-4.
EI
kspr = 3 3 (4-4)
l

D.A. (Daniel) Okret Master of Science Thesis


4-1 Structural Dynamics 21

F(t)
m
m

k F(t)
m

EI,l

Figure 4-2: Analyse beam as a spring

Which leads to a natural frequency of the system being equal to:

s s
kspr 3EI
fn = = (4-5)
m l3 m

In reality, one degree of freedom is not sufficient to describe the response of most systems.
These systems can be described as a continuous system, which will be described in Section
4-1-2, or as a Multiple Degree of Freedom System. The equation of a MDOF is similar to the
one of a SDOF as can be seen in Equation 4-6. Here the mass, damping and spring term are
matrices and the displacement and force terms are vectors. The number of eigenmodes and
natural frequencies is equal to the number of degrees of freedom of the system. The lowest
natural frequency corresponds to the first eigenmode, and so on.

Mẍ + Cẋ + Kx = F (t) (4-6)

Numerical software that use Finite Element Model, such as FREERISE and USFOS (see
Chapter 6), use discrete systems to represent continuous systems. If the amount of degrees
of freedom is large enough, these programs are able to correctly describe the behaviour of the
system.

4-1-2 Continuous Systems

A continuous system is an infinite degree of freedom system. Therefore it has an infinite


number of eigenmodes and natural frequencies. The dynamics of such a system are usually
described by means of partial differential equations. An example of a continuous system is
the bending beam. The mechanical representation of a beam is given in Figure 4-3.

Master of Science Thesis D.A. (Daniel) Okret


22 Riser and Well System Dynamics

Figure 4-3: Simply supported beam [18]

where: w = vertical displacement [m]


E = Young’s Modulus [N/m2 ]
I = area moment of inertia [m4 ]
ρ = density [kg/m3 ]
A = area [m2 ]

The most commonly used model to describe this system is the Euler-Bernoulli beam theory.
Its assumption is that the plane cross-section initially perpendicular to the axis of the beam
remain plane and perpendicular to the neutral axis during bending [19]. This implies that
shear deformations are not taken into account. The bending of an Euler-Bernoulli beam is
given by Equation 4-7:

∂ 4 w(z, t) ∂ 2 w(z, t)
EI + ρA = F (z, t) (4-7)
∂z 4 ∂t2
In addition to this equation, the boundary conditions on both extremities of the beam need
to be defined. If these is a system of multiple coupled beams, interface conditions between
the beams need to be defined as well.
If the shear deformation does play a role, this model is not sufficient. In this case the
Timoshenko beam model needs to be used [18]. This model adds a term to Equation 4-7
which accounts for this extra deformation. The effect of including the shear deformation can
be seen in Figure 4-4.

Figure 4-4: Timoshenko vs Euler-Bernoulli beam models

D.A. (Daniel) Okret Master of Science Thesis


4-2 Riser Dynamics 23

Shear deformations need to be included if the length of the beam is smaller than 10-20 times
its height [20]. Considering a main drilling riser with 21 inches (0.533 metres) diameter, its
length should be greater than ten metres for shear deformations to be neglected. Since this
will always be the case, the Euler-Bernoulli theory will be applied for the fatigue assessment
of the well system.
The dynamics of a beam are further explained in Appendix B-1.

4-2 Riser Dynamics

The vibration of a riser can be well described as the bending of a beam. However, a simple
beam equation as Equation 4-7 will not be able to account all the interactions occurring in
this system. Extra terms need to be included into the equation of motion [21]. These terms
will be introduced in the following sections.

4-2-1 Pipe Conveying Fluid

The dynamics of a pipe conveying fluid have been studied by several researchers. The first
publication of the correct equation for the transverse motion of such a system has been done
by Mr. François Joseph Bourrières in 1939 [21]. This equation has been widely studied, and
the can be written as shown Equation in 4-8 [22].

∂4w   2
2 ∂ w ∂2w ∂2w
EI − T − m u
f f + 2mp uf + (mf + m p ) =0 (4-8)
∂x4 ∂x2 ∂x∂t ∂t2

∂2w
where: uf 2 = Centripetal acceleration
∂x2
∂2w
2uf = Coriolis acceleration
∂x∂t
∂2w
= Inertia acceleration
∂x2

Note that the a tension term T is already included in this equation for completeness. This
term will be further elaborated in Section 4-2-2.

4-2-2 Riser Tension

The tension on a riser has two important functions. The first one is the stabilisation and
reduction of stresses and displacements. Furthermore, the tension decreases the possibility
of a collision between the LMRP and the BOP if a emergency disconnection is performed.
Therefore it is imperative that during all drilling operations, the tension just below the LMRP
is always positive (no compression).
Not only the true wall tension needs to be accounted for, but also the effect of the internal
and external pressures along the beam. The combination of true wall tension and pressures

Master of Science Thesis D.A. (Daniel) Okret


24 Riser and Well System Dynamics

results in the effective tension. The true tension on a riser system is given by Equation 4-9
[23].

Te (z, t) = T (z, t) + Ae (z)pe (z) − Ai (z)pi (z) (4-9)

where: Te = effective tension [N]


T = true wall tension [N]
Ae = external area [m2 ]
pe = external pressure [Pa]
Ai = internal area [m2 ]
pi = internal pressure [Pa]

Due to the vessel motions, mainly the heave motion, there is also a time varying component of
the tension. This is a result of the limitation of tensioners to compensate the heave motion,
which will be further discussed in Appendix D-1. The true wall tension can therefore be
described with a mean value and a time varying value, as displayed in Equation 4-10:

T (z, t) = Tmean (z) + Tvarying (t) (4-10)

Here it is assumed that the axial wave propagation frequency is much higher than the trans-
verse vibration frequencies. As a result, the time varying term of the tension for transverse
vibrations is not spatially dependent.
The equation of motion of a tensioned beam is given by Equation 4-11.

∂ 4 w(z, t) ∂ ∂w ∂ 2 w(z, t)
 
EI 4
− Te (z, t) + ρA = F (z, t) (4-11)
∂z ∂z ∂z ∂t2

Combining Equation 4-11 and Equation 4-9 the following is obtained:

∂ 4 w(z, t) ∂ ∂w ∂ 2 w(z, t) ∂ ∂w
   
EI 4
− T (z, t) + ρA 2
− (Ae pe (z) − Ai pi (z)) = F (z, t)
∂z ∂z ∂z ∂t ∂z ∂z
(4-12)

4-2-3 Material Damping

When a material is subjected to motions, part of mechanical vibration energy is converted to


heat, and the motion is damped. This phenomenon is called material damping and is usually
included into the equation of motion by adding a damping term which is proportional to the
curvature. In the frequency domain, this will be done by adding a complex term to the
by adding an complex stiffness term, as shown in Equation 4-13.

D.A. (Daniel) Okret Master of Science Thesis


4-3 Well System Dynamics 25

∂ 4 w̃ ∂ 2 w̃
EI(1 + Cdamp i) + ρA = F̃ (Ω, t) (4-13)
∂z 4 ∂t2

The material damping coefficient is usually chosen as the critical damping for steel, which is
0.3%. In [24] an experimental study was performed to determine the critical damping value
of a plain steel risers. The obtained values for material damping are considerably lower than
what is usually used, reaching 0.13%. However, the damping values for a complex drilling
riser are expected to be higher than a plain riser, and therefore the critical value for steel will
be used.

4-2-4 Riser equation of motion

Combining the effective tension and conveying fluid with Equation 4-7, the equation of motion
of a tensioned riser is obtained. The behaviour of this system has been thoroughly investigated
[25]. The resulting equation of motion is given by Equation 4-14.

!
∂4w ∂ ∂w ∂2w ∂2w ∂2w
 
EI 4 − T (z, t) + ρf Ai uf 2 2 − 2uf + 2 −
∂z ∂z ∂z ∂z ∂z∂t ∂t
∂ ∂w ∂2w
 
(Ae pe (z) − Ai pi (z)) + ρA = F (z, t) (4-14)
∂z ∂z ∂t2

4-3 Well System Dynamics

The well system can also be modelled with an Euler-Bernoulli beam. The interaction between
soil and the well system is included into the equation of motion by means of an elastic Kelvin
foundation [19], which varies along the depth. A mechanical representation of a beam on
elastic foundation is given in Figure 4-5.

EI A
z

ks(z),cs(z)

w
Figure 4-5: Beam on viscoelastic foundation

This visco-elastic foundation can easily be modelled by introducing a distributed spring (ks )
and dashopt (cs ) elements into Equation 4-7. The resulting equation of motion is given by
Equation 4-15.

∂ 4 w(z, t) ∂ 2 w(z, t) ∂w(z, t)


EI 4
+ ρA 2
+ cs (z) + ks (z) · w(z, t) = F (z, t) (4-15)
∂z ∂t ∂t

Master of Science Thesis D.A. (Daniel) Okret


26 Riser and Well System Dynamics

4-3-1 Soil Stiffness

A model has been adopted in which the stiffness of this spring is related to the ultimate
lateral capacity of the soil and the lateral displacement. Because the stiffness is dependent
on the lateral displacement, its profile is nonlinear. The soil stiffness can be represented by
p-y curves. An simplified example of a p-y curve is displayed in Figure 4-6.

Figure 4-6: Example of p-y curve [26]

where: p = pressure [Pa]


pu = ultimate resistance [Pa]
y = displacement [m]
yc = measure of ultimate displacement [m]

Extensive research has been done on p-y curves. The current guidelines are provided by
API [27] and are based on a paper by Matlock [28]. There are some papers that argue the
proposed curves may be conservative [29], however none of them has presented sufficient data
and validation to be considered reliable. Therefore the API p-y curves will be applied.
To perform analyses in the frequency domain the p-y curves need to be linearised. DNV has
recommended a piece-wise linearisation method in [30]. In this method the initial stiffness is
calculated as the slope of the first discrete point of the p-y curve. If the displacement of the
system exceeds the displacement described by the first point, the second point is chosen, and
so on. This linearised stiffness is also referred to as the secant stiffness.
The p-y curves are further elaborated in Appendix B-3.

D.A. (Daniel) Okret Master of Science Thesis


4-4 Challenges of Analytical Modelling 27

4-3-2 Soil Damping

There are two damping mechanisms resulting from the interaction between soil and well
system. These are radiation damping and hysteretic damping. Radiation damping comes
from the energy transmitted from structure to soil due to wave propagation. This type of
damping is frequency dependent and is related to the inertial effects. Hysteretic damping
is causes by the shear deformations in the material and is amplitude dependent. For low
frequencies, which act on the conductor system, radiation damping will have a small effect
compared to hysteretic damping [31][2].
Damping effects will be approximated in a linear analysis. This is achieved by applying the
procedure described in [32]. A viscous damper with energy dissipation equal to the dissipated
energy within the hysteresis (load-unload) loop for given frequencies and amplitudes.
The derivation of the damping ratio is further elaborated in Appendix B-4.

4-4 Challenges of Analytical Modelling

As elaborated in this chapter, a simple Euler-Bernoulli beam model is not able to capture
all the dynamic interactions occurring in the system. Multiple terms need to be added to
Equation 4-7 to account for this effects. Due to the addition to these extra terms, there
are complications to solve the system in a purely analytical manner. These complications
come from the spacial varying tension and the tune varying boundary conditions. In this
section approaches to account for these challenges are presented, in addition to alternatives
to a purely analytical approach.

4-4-1 Coupled modes due to spatial varying tension

The vibration of a beam can be described as the sum of uncoupled sinusoidal and hyperbolic
eigenmodes. In these cases the equation of motion can be solved by applying the classical
separation of variable method:

w(z, t) = W (z)eiωt (4-16)

where: w(x, t) = displacement


W (z) = eigenmodes
ω = natural frequency

When the tension varies along the length of the riser, it is not possible to find uncoupled
sinusoidal modes. However, uncoupled modes can still be calculated, either numerically or
analytically. An analytical approach to obtain the eigenmodes was developed in [33]. Eigen-
modes are calculated by assuming a power series expansion, as in Equation 4-17:

Master of Science Thesis D.A. (Daniel) Okret


28 Riser and Well System Dynamics


X ∞
X ∞
X ∞
X
W (z) = a0 Fn z n + a1 Gn z n + a2 Hn z n + a3 In z n (4-17)
n=0 n=0 n=0 n=0

This approach has been further developed and validated in [25].

4-4-2 Time varying boundary conditions

Because of the time varying boundary condition at the top of the system, imposed by the
vessel motions, it is not possible to solve it with classical separation of variables method. At
each moment in time, the upper boundary is different and a different set of modes define the
system. In [34] an analytical method to deal with such situations is developed. An extra
term that accounts for the varying boundary condition is added to the general solution. The
proposed solution is displayed in Equation 4-18.


X
v(x, y) = w(x, t) + fi (t) gi (x) (4-18)
i=1

where: w(x, t) = classical separation of variables method


fi (t) = time varying boundary condition
gi (x) = shape function that allows the boundary condition to be satisfied

Applying this analytical solution to a simple Euler-Bernoulli beam without complicated terms
is relatively straightforward. However, when applied to more complex systems, this method
becomes less adequate. Moreover, the classical separation of variables needs to be possible
and initial conditions need to be included.

4-4-3 Time vs. Frequency Domain

From a mathematical point of view, a beam equation can be classified as a partial differential
equation dependent of space and time. Solving such an equation analytically is only possible in
very specific situations. For other situations, it can be solved numerically. There are several
ways in which this can be done, either in the time or frequency domain. Both numerical
methods can handle the time varying boundary condition and spatial varying tension. This
makes a numerical methods extremely efficient for solving more elaborate beams.
Each calculation type has its advantages and disadvantages. The frequency domain has as its
main advantage that the computational time is much lower when compared to time domain
calculations. However, the frequency domain can only describe linear features. Therefore, any
non-linearities need to either be linearised or ignored while working in the frequency domain.
This is not necessary for frequency domain calculations.
By applying the Fourier Transforms, as explained in Section 5-2, the equation of motion
can be transformed to the frequency domain. Because the computation time of this method

D.A. (Daniel) Okret Master of Science Thesis


4-4 Challenges of Analytical Modelling 29

is much lower than the time domain analysis, it is better suited for the assessment of the
influence of parameters. This analysis will be performed in Chapter 5.
Another method to solve the equation of motion of a beam is to discretise space, by for ex-
ample through finite elements or finite differences. By applying these techniques the partial
differential equations become ordinary differential equations dependent on time. This ap-
proach is more accurate if non-linearities play a significant role in the motions of the riser. In
Chapter 6 a FEM will be developed with the USFOS program.

Master of Science Thesis D.A. (Daniel) Okret


30 Riser and Well System Dynamics

D.A. (Daniel) Okret Master of Science Thesis


Chapter 5

Frequency Domain Analysis

In this chapter, a frequency domain model is developed to investigate the fatigue behaviour of
the well system. In Section 5-1 the model is described, which consists of a system of coupled
beams representing the various elements of the system. In Section 5-2 and 5-3 the governing
equations of motion and boundary conditions will be elaborated.

In Section 5-4 the developed model is validated numerically and dynamically with two different
FEM programs. One performs frequency domain simulations and the other performs time
domain simulations. In this way the algorithm and the validity of the algorithm can be
assessed.

5-1 Model description

A semi-analytical frequency domain model is developed to describe the dynamics of the well
system and wellhead. The applied programming technique is explained in Appendix E. To
accurately describe the dynamics of the well system, the components on top of it need to be
included in the model. To achieve this, the system is divided in three separate subsystems.
These correspond to the riser, wellhead and well system. Each subsystem is modelled as
a specific beam with its own material and geometrical properties. The surrounding fluid,
internal drilling fluid and soil interaction are also considered.

A schematic overview of the model is displayed in Figure 5-1. Note that this scheme is not
in scale. Usually the length of the riser and well systems are much larger compared to the
wellhead.

Master of Science Thesis D.A. (Daniel) Okret


32 Frequency Domain Analysis

Vessel Motions

Water

Hydrodynamic Beam (riser)


Forces

Rotational spring (LFJ)


m,J (BOP)

Beam (wellhead)

Soil

Beam (well system)

Figure 5-1: Scheme of Riser and Well System model

All the segments are subjected to drilling mud flow, even though its effect is negligible,
as elaborated in Section 5-5-2. The first beam represents the riser and is subjected to a
variable tension and hydrodynamic forces (added mass and drag). The middle beam models
the wellhead. The third one stands for the well system and is subjected to soil-structure
interaction. The equations of motion for each segment will be further discussed in Section
5-2.
Several simplifications are considered:

• There is no tension variation in time. This implies that only fatigue damage due to
bending stresses variations will be taken into account, and not due to axial stress vari-
ations. However, the time-independent component of tension is included in this model.
• Soil structure interaction is described with a Kelvin foundation.
• Complex riser will be described as a simple pipe. It is assumed that flowlines and
umbilicals do not affect the stiffness, but only the hydrodynamic parameters.
• The BOP is assumed as infinitely stiff, and its effect, together with the LFJ, are ac-
counted for through proper interface conditions. This will be further elaborated in
5-3.
• The vibrations caused by the drill sting and further drill equipment are not taken into
account.
• The MODU is located directly on top of the well will be considered. Possible excursions
could be accounted for by varying the top tension.

D.A. (Daniel) Okret Master of Science Thesis


5-2 Governing Equation of Motion 33

• Only the conductor of the well system is modelled. It is assumed that because of its
larger diameter, it will have the governing fatigue lifetime. Furthermore, its bending
stiffness it very large when compared to the next pipe of the well system (e.g factor
close to 10).

5-2 Governing Equation of Motion

The equations of motion that describe each subsystem are derived in Chapter 4, and are
displayed in Equations 5-1 to 5-3 for convenience. Subscript r stands for the riser, subscript
wh for the wellhead and ws for the well system.

!
∂4w ∂ ∂w 2∂
2w ∂2w ∂2w
 
EIr 4 − Tr (z) + ρf Ai uf − 2uf +
∂z ∂z ∂z ∂z 2 ∂z∂t ∂t2
∂ ∂w ∂2w
 
− (Ae pe (z) − Ai pi (z)) + ρr Ar = F (z, t) (5-1)
∂z ∂z ∂t2

!
∂4w ∂2w ∂2w ∂2w ∂2w
EIwh 4 − Twh 2 + ρf Ai uf 2 2 − 2uf + 2
∂z ∂z ∂z ∂z∂t ∂t
∂ ∂w ∂2w
 
− (Ae pe (z) − Ai pi (z)) + ρwh Awh = F (z, t) (5-2)
∂z ∂z ∂t2

!
∂4w ∂2w ∂2w ∂2w ∂2w
EIwh 4 − Tws 2 + ρf Ai uf 2 2 − 2uf + 2
∂z ∂z ∂z ∂z∂t ∂t
∂ ∂w ∂2w ∂w
 
− (Ae pe (z) − Ai pi (z)) + ρwh Awh 2
+ cs + ks w = Fw s(z, t) (5-3)
∂z ∂z ∂t ∂t

The forcing terms at the riser and wellhead are given by the fluid surrounding it. These
hydrodynamic forces can be separated in drag and inertia and are displayed in Equation 5-4
and together form the hydrodynamic forces. Because there is no fluid surrounding the well
system, there is no forcing term acting on the well system, so Fws (z, t) = 0.

1 π π
F (z, t) = ρ Cd D(uc + uw − ẇ)|uc + uw − ẇ| + ρ Cm D2 u̇ − ρ (1 − Cm )D2 ẅ (5-4)
2 4 4

where: uc = current velocity [m/s]


uw = horizontal water particle velocity [m/s]
ẇ = riser velocity [m/s]

Master of Science Thesis D.A. (Daniel) Okret


34 Frequency Domain Analysis

The riser tension is dependent on the drilling fluid velocity, the riser weigh and buoyancy [25].
It is calculated as stated in Equation 5-5.

Tr (z) = ρr Ar g(L − z) − ρf Ar gL (5-5)

The aforementioned system of equations can be solved in the frequency domain by applying
the Fourier Transforms as described in Appendix B-5. This requires the linearisation of the
soil stiffness and the non-linear forcing terms on the right hand side of the system of equations
above. Furthermore, the material damping as described in Section 4-2-3 will be added. After
transformation, Equations 5-6 to 5-8 are obtained:

!
∂ 4 w̃ ∂ ∂ w̃ ∂ 2 w̃ ∂ w̃
 
EIr (1 + Cdamp i) 4 − Tr (z) + ρf Ai uf 2 2 − iΩ2uf − Ω2 w̃
∂z ∂z ∂z ∂z ∂z
∂ ∂ w̃
 
− (Ae pe (z) − Ai pi (z)) − Ω2 ρr Ar w̃ = F̃ (Ω, z) (5-6)
∂z ∂z

!
∂ 4 w̃ ∂ 2 w̃ ∂ 2 w̃ ∂ w̃
EIw h(1 + Cdamp i) 4 − Twh 2 + ρf Ai uf 2 2 − iΩ2uf − Ω2 w̃
∂z ∂z ∂z ∂z
∂ ∂ w̃
 
− (Ae pe (z) − Ai pi (z)) − Ω2 ρwh Awh w̃ = F̃ (Ω, z) (5-7)
∂z ∂z

!
∂ 4 w̃ ∂ 2 w̃ ∂ 2 w̃ ∂ w̃
EIw s(1 + Cdamp i) 4 − Tws 2 + ρf Ai uf 2 2 − iΩ2uf − Ω2 w̃
∂z ∂z ∂z ∂z
∂ ∂ w̃
 
− (Ae pe (z) − Ai pi (z)) − Ω2 ρws Aws w̃ + cs iΩw̃ + ks w̃ = F̃ (Ω, z) (5-8)
∂z ∂z

The linearisation procedure applied on the drag force and soil stiffness are elaborated in
Appendix A-5 and Appendix B-3 respectively.

5-3 Interface and Boundary Conditions

In addition to the equations of motion of the system, boundary and interface conditions are
needed to represent the system’s behaviour. Each subsystem has two extremities, which will
be denoted by top (nearest to the MSL) and end.

Top Boundary Conditions The top boundary condition is given by the vessel. The surge
motions of the vessel are directly transferred to the riser. Therefore the displacement at the
top of the riser is equal to the surge motion of the MODU. For simplicity, the moment at the
top of the vessel is assumed to be zero. This assumption is based on the fact that the UFJ,
to which the top of the riser is attached, will not transfer significant moments to it.

D.A. (Daniel) Okret Master of Science Thesis


5-3 Interface and Boundary Conditions 35


w(z, t) = imposed displacement

z=rtop
(5-9)
∂ 2 w(z, t)
−EIr =0
∂z 2

z=rtop

Interface conditions Riser-Wellhead The BOP and LFJ are present at the interface con-
ditions between the riser and the wellhead. These elements will be taken into account by
including them into this interface condition. There is a continuity of displacement between
the two elements. The slope is not continuous due to the LFJ, so nothing can be said about
the slopes. The LFJ causes a discontinuity in the bending moments, together with the mo-
ment of inertia of the BOP. The mass of the BOP causes a discontinuity in the shear force
between the riser and wellhead. This resulting interface condition is displayed in Equation
5-10.


w(z, t) = w(z, t)

z=rend z=whtop
∂ 2 w(z, t) ∂w(z, t) ∂w(z, t)
 
(EI)r = −kθ −

∂z 2

z=rend ∂z z=rend ∂z z=whtop
∂ 2 w(z, t) ∂ 3 w(z, t) ∂w(z, t) ∂w(z, t)
 
(EI)wh −J = −kθ −

∂z 2 ∂t2 ∂z z=whtop

z=whtop ∂z z=whtop ∂z z=rend
∂ 3 w(z, t) ∂ 3 w(z, t) 2
∂ w(z, t)
(EI)r − (EI)r = mBOP + F̃ (Ω, z)
∂z 3 ∂z 3 ∂t2

z=rend z=whtop z=rend
(5-10)

Interface conditions Wellhead-Well System There is no discontinuity between the wellhead


and well system. Therefore the interface between these elements can be assumed continuous
for all the derivatives of the beam equation.


w(z, t) = w(z, t)

z=whend z=wstop
∂w(z, t) ∂w(z, t)
=
∂z z=whend ∂z z=wstop
(5-11)
∂ 2 w(z, t) 2
∂ w(z, t)
=

∂z 2 ∂z 2

z=whend z=wstop
∂ 3 w(z, t) 3
∂ w(z, t)
=

∂z 3 ∂z 3

z=whend z=wstop

Bottom Boundary Conditions The bottom boundary condition is assumed to be pinned.


This is based on the assumption that waves travelling down the well system will be completely
damped out once they reach the end, and no reflection will take place. The length of the well
system is chosen to be sufficiently large so that this assumption is true.

Master of Science Thesis D.A. (Daniel) Okret


36 Frequency Domain Analysis


w(z, t) =0

z=wsend
(5-12)
∂ 2 w(z, t)
EIws =0
∂z 2

z=wsend

5-4 Validation

The ideal validation method would be through measurement data of a real offshore structure.
This measurements could be done in several ways. The benefits and possible configurations
are have been the topic of various papers [35, 36, 37]. However, not many measurement
campaigns have been performed and unfortunately the existing data was not available for
this thesis.

Three alternative validation methods were used to validate the semi-analytical model. The
first one is a numerical validation which will be elaborated in Section 5-4-1. The second
method is a validation with the finite element software USFOS, and will be addressed in
Section 5-4-2. The last method is a comparison with a case study performed by an offshore
contractor called 2H Offshore [7].

5-4-1 Numerical validation

A numerical validation is performed to make sure the calculations being performed by the
model are done correctly. It does not give an indication whether the obtained results give an
accurate description of what happens in reality. The real validation will be elaborated further
in this section.

The semi-analytical model was numerically validated with an existing FEM software called
FREERISE. This software was developed in-house at Shell by Ir. J. W. van de Graaf [38].
It is able to perform simple riser analysis in the frequency domain. One aspect that makes
FREERISE particularly suitable for the numerical validation is that the drag forces are lin-
earised in the same way as in the semi-analytical model. If both programs are run with the
same or as similar as possible input parameters, the outputs should be comparable.

Because the Finite Element software gives a standard output, it is not possible to compare
the intermediate calculation steps. Two outputs from the FREERISE were used to perform
the numerical validation. The first one are the standard deviation of the displacements,
bending stresses and shear forces along the length of the riser. The second one are the
natural frequencies of the system.

An example of such a validation is given for a conductor of 36 inch diameter in 143 m water
depth. This conductor is driven 18 m into the mudline, which gives a total length of 161 m.
A representation of this conductor can be seen in Figure 5-2.

D.A. (Daniel) Okret Master of Science Thesis


5-4 Validation 37

Water

143m

36''

Soil
18m

Figure 5-2: Scheme of conductor in 143 m water depth

The result of this validation is displayed in Figure 5-3 and Table 5-1. It is clear that the
results are very similar. The differences are probably due to the method used for modelling
the soil and the discrete versus continuous system approximation.

Table 5-1: Comparison of first three natural frequencies

Frequency FREERISE Frequency Semi-Analytical Difference


Natural Frequency
[s] [s] [%]
1 9.955 9.990 0.3
2 2.779 2.829 1.4
3 1.298 1.330 2.4

The frequency domain model does not provide the natural frequencies directly. These need
to be extracted from the Frequency Response Function of the system, as described in Section
B-2. This FRF gives the response of the system for each frequency for an unit force. The
peaks of this function are related to the natural frequencies of the system. An example of an
FRF for the case described is shown in 5-4.

For fatigue purposes, the most important parameter is the standard deviation of the bending
stresses. This is because the stress cycles are the source of fatigue for the well and riser
system.

Master of Science Thesis D.A. (Daniel) Okret


38 Frequency Domain Analysis

Standard Deviation Displacement Standard Deviation Bending Stress


0 0

-20 -20

-40 -40

-60 -60

Depth [m]
Depth [m]

-80 -80

FREERISE FREERISE
-100 -100
MATLAB MATLAB

-120 -120

-140 -140

-160 -160

-180 -180
0 0.1 0.2 0.3 0.4 0.00E+00 1.00E+07 2.00E+07 3.00E+07
Displacement [m] Stress [kN/m2]

Standard Deviation Shear Force


0

-20

-40

-60
Depth [m]

-80

FREERISE
-100
MATLAB

-120

-140

-160

-180
0 20000 40000 60000 80000
Shear Force [kN/m]

Figure 5-3: Comparison of Semi-Analytical model and FREERISE results

D.A. (Daniel) Okret Master of Science Thesis


5-4 Validation 39

FFR of Displacement at 10m below MSL


10 -2

10 -3

10 -4

10 -5
Magnitude

10 -6

10 -7

-8
10

10 -9

10 -10
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Frequency [Hz]

Figure 5-4: Frequency Response Function of semi-analytical model

5-4-2 Validation with FEM

To ensure that the results obtained with the semi-analytical model give an accurate description
of reality, it is validated with the FEM USFOS. The output of the finite element software
which was used to compare the models is a time series of bending moments for each element.
From these bending moment, the stresses can be easily calculated with Equation 5-13.

My
σ= (5-13)
I

where: σ = bending stress [N/m2 ]


M = moment around neutral axis [Nm]
y = perpendicular distance to neutral axis [m]
I = area moment of inertia [m4 ]

With the bending stress time series, the standard deviation can be calculated for each element.
Moreover, the fatigue lifetime for each element can be calculated using rainflow counting
method.
The same example set-up displayed in Figure 5-2 was used for an initial validation. Figure
5-5 gives the result of this validation. The fatigue lifetime is almost identical. There is one
big gap in the fatigue lifetime comparison at about -113 m depth. This is due to the absence
of an element in the FEM at this particular depth. Apart from this larger gap, the maximum
difference in the standard deviation of bending moments is less than 3 %.

Master of Science Thesis D.A. (Daniel) Okret


40 Frequency Domain Analysis

Fatigue Lifetime of Conductor Standard Deviation of Bending Moment


0 0
USFOS USFOS
-20 MATLAB -20 MATLAB

-40 -40

-60 -60
Depth [m]

Depth [m]
-80 -80

-100 -100

-120 -120

-140 -140

-160 -160
10 0 10 5 10 10 0 1 2 3 4 5 6 7
Lifetime [years] Bending Moment [N] ×10 5

Figure 5-5: Comparison of Semi-Analytical model and USFOS results

Both simulations were done for a duration of twenty minute. That is why a there is no
significant difference in the fatigue lifetime. The differences in bending stresses are probably
due to the linear vs nonlinear drag forces included in both models.

5-4-3 Comparison with Case Study

Because of the lack of measurement data, alternative methods to validate the model were
investigated. One chosen method which will be used as a comparison is the fatigue lifetime
assessment performed by 2H Offshore for a shallow water well in New Zealand [7]. Please
note that not all input parameters required for the frequency domain model were available in
the report, and some parameters had to be linearised. Therefore this will just serve as a high
level comparison of the order of magnitude of the fatigue lifetime.

As can be observed in Figure 5-6, the trend of the fatigue lifetime of both assessments have
a pretty good match. The largest differences occur near the bottom of the riser. This is
probably due to the linearisation of the BOP applied on the frequency domain model.

D.A. (Daniel) Okret Master of Science Thesis


5-5 Sensitivities 41

Figure 5-6: Unfactored fatigue lifetime of riser system for well in New Zeeland calculated by 2H
Offshore (red) and frequency domain model (blue). Adapted from [7]

5-5 Sensitivities

In this section, the sensitivity of the most important parameters will be investigated. Special
attention will be given to the soil damping, as this has been pointed out as a knowledge gap
in the literature study, and could significantly contribute to the fatigue lifetime of the well
system. For many of the sensitivities, a riser configuration for 270 metres water depth is dis-
played as a reference. This has been chosen because its critical modes for fatigue calculations
(modes with large displacements at the mudline) are in the region where wave excitation
occurs. Therefore, it is ideal to access the effect of parameters on the fatigue lifetime of the
well system.

5-5-1 Water Depth

The water depth defines the length of the drilling riser. Larger water depths have longer
drilling risers. With increasing riser length, the natural frequencies get higher. This can
clearly be observed in Figure 5-7, where the FRF’s for five water depths (100, 200, 500, 1000
and 2000 metres) are displayed.

Master of Science Thesis D.A. (Daniel) Okret


42 Frequency Domain Analysis

100 m 500 m
200 m 1000 m

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Frequency [Hz] Frequency [Hz]

2000 m

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1


Frequency [Hz]

Figure 5-7: Frequency Response Function for different water depths

There are only three natural frequencies below one Hertz for a water depth of 100 metres,
while there are over thirty eigenfrequencies in the same bandwidth for 2000 metres water
depth. Because of this, the lower natural frequencies of longer risers move away from the
excitation bandwidth by waves.

The excitation by wave forces usually occurs at the top 100 metres below the MSL, and the
vessel motions act on the top of the riser. An increased riser length results in a larger damping
path for vibrations along the riser before they reach the well system. The combination
of increased damping path and lower frequencies moving away from excitation bandwidth
result in an increased fatigue lifetime for larger water depth, when only wave excitations are
considered. This can clearly be observed in Figure 5-8, where the risers are subjected to a
sea state with peak period of six seconds and wave height equal to three metres.

D.A. (Daniel) Okret Master of Science Thesis


5-5 Sensitivities 43

−100
100m
200m
500m
1000m
−101 2000m
Depth [metres]

−102

−103

100 102 104 106 108 1010


Lifetime [years]

Figure 5-8: Fatigue lifetime for different water depths

This is just a representation for a specific sea state. However, a relatively high period sea
state was chosen because it should excite lower modes of large riser lengths. Even for these
sea states lower water depths still suffer more fatigue damage. Note that the time variation
of tension is not taken into account. If this is done, as in Section 6-3, larger water depths can
still be prone to fatigue.

5-5-2 Effects of drilling fluid

The drilling fluid is taken into account in four separate terms of the equation of motion, as
can be seen in Equations 5-6. Two of these terms are related to flow velocity (centripetal and
Coriolis effect) and the two terms are related to its mass (inertia and internal pressure). Both
these properties can vary depending of the applications, and will be optimised to transport
drilled particles to the surface.
The mass of the drilling fluid gives a mass term of the same order of magnitude of the steel
riser itself. Increasing this mass will therefore result in a significant reduction of the natural
frequencies of the system. Furthermore, an increase in density will decrease the effective
tension in the lower part of the riser, due to a larger difference between the inner and outer
pressure of the riser (see Equation 4-9). A decrease in tension of the riser will affect the
system in two ways. The first one is a reduction of the natural frequencies of the system.
The second effect is a reduction of the tension dependent restoring force towards the mudline.
This increases the amplitude of the curvature occurring towards the mudline. These effects
can be observed in Figure 5-9a. The natural frequencies can be adjusted by increasing the
top tension of the riser. However, it is not possible to fully compensate the increase of
curvature amplitudes towards the mudline, as displayed in Figure 5-9b. The ratio between
the curvature amplitudes between 9 and 15 ppg at 0.15 Hertz (second natural frequency)
remains unchanged, at 5.

Master of Science Thesis D.A. (Daniel) Okret


44 Frequency Domain Analysis

9 ppg 9 ppg
11 ppg 11 ppg
-50 13 ppg -50 13 ppg
15 ppg 15 ppg
Muldine Muldine
-100 -100
Depth [metres]

Depth [metres]
-150 -150

-200 -200

-250 -250

-300 -300

100 105 1010 1015 100 105 1010 1015


Lifetime [years] Lifetime [years]

(a) Equal top tension (b) Larger top tension for heavier mud

Figure 5-10: Effect of drilling mud density on fatigue lifetime

9 ppg 9 ppg
11 ppg 11 ppg
13 ppg 13 ppg
15 ppg 15 ppg
Amplitude

Amplitude

0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
Frequency [Hz] Frequency [Hz]

(a) Equal top tension (b) Larger top tension for heavier mud

Figure 5-9: Effect of drilling mud density on curvature Frequency Response Function 5 metres
below the mudline

The larger curvature amplitudes at critical modes can therefore not be compensated by in-
creasing the top tension. However, shifting the natural frequencies away from the excitation
bandwidth can still significantly improve the fatigue lifetime of the system. Figure 5-10 and
Table 5-2 show the effect of mud weight on the fatigue lifetime for a sea state with T p = 6s
and Hs = 3m. Note that the peak period of this sea state is near the second natural frequency
of the system (critical mode). Increasing the top tension can increase the fatigue lifetime by
a factor 24 for a mud density of 15 ppg.
The mud flow is assumed to occur in the through the entire inner section of the riser. However,
it has a small effect on the dynamic properties of the well system is very small. As a result,
its effect on the fatigue lifetime of the well system is also small, and can be neglected. The
effect of mud flow for a 15 ppg mud in 270 metres water depth is displayed in Figure 5-11.
Note that there is an increase smaller than 0.1% in fatigue lifetime when including the mud

D.A. (Daniel) Okret Master of Science Thesis


5-5 Sensitivities 45

Table 5-2: Effect of drilling mud density on unfactored fatigue lifetime at 5 metres below mudline

Mud Density Lifetime equal tension Lifetime adjusted tension Ratio


[ppg] [years] [years] [-]
15 4.00E+06 9.70E+07 24.25
13 2.50E+08 1.00E+09 4
11 4.41E+09 5.40E+09 1.22
9 1.10E+10 1.10E+10 1

flow.

Fatigue lifetime different mud weights

0 ft/min
-50 100 ft/min
200 ft/min
Muldine
-100
Depth [metres]

-150

-200

-250

-300

100 105 1010 1015


Lifetime [years]

Figure 5-11: Effect of mud flow velocity for T p = 6s and Hs = 3m

5-5-3 Effect of BOP

The size and mass off the BOP has a significant influence on the dynamic parameters of the
well system. Three BOP sizes were used in the assessment, and its parameters are displayed
in Table 2-2. Larger BOP’s move the natural frequencies with large displacements at the
mudline down. This shift can be larger than 20% when comparing the 3rd and 5th generation
BOP’s. This can clearly be observed in Figure 5-12.

Master of Science Thesis D.A. (Daniel) Okret


46 Frequency Domain Analysis

3rd Gen 3rd Gen


4th Gen 4th Gen
5th Gen 5th Gen
Amplitude

Amplitude
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
Frequency [Hz] Frequency [Hz]

(a) 120 metres water depth (b) 270 metres water depth

Figure 5-12: Effect of BOP size on curvature Frequency Response Function 5 metres below the
mudline

The influence of the shift in natural frequency on the fatigue lifetime of the system depends
on the sea state to which the system is exposed. For a high peak period, large BOP’s perform
better, whereas in low peak periods small BOP’s will suffer less fatigue damage. This effect
can be seen in Figure 5-13, where the 270 metre riser configuration is exposed to a sea state
with a Tp of 6 and 12.

-250 -250
3rd Gen 3rd Gen
4th Gen 4th Gen
-260 6th Gen -260 6th Gen
Mudline Mudline
Depth [metres]

Depth [metres]

-270 -270

-280 -280

-290 -290

-300 -300
100 105 1010 1015 100 105 1010 1015
Lifetime [years] Lifetime [years]

(a) Tp = 6 and Hs = 3 (b) Tp = 12 and Hs = 3

Figure 5-13: Effect of sea state on fatigue lifetime for 270 metres water depth

5-5-4 Soil Stiffness

The soil stiffness has a two main effects on the system. The first one is a change in natural
frequencies corresponding to eigenmodes with large displacements at and below the mudline.
Larger stiffness results in an increase of these natural frequencies. Furthermore, there is a
shift of the location where the minimum fatigue lifetime occurs according to the soi stiffness.

D.A. (Daniel) Okret Master of Science Thesis


5-5 Sensitivities 47

Larger soil stiffness will have the critical fatigue spot closer to the mudline. Figure 5-14
shows this effect for a well and riser system at 270 metres water depth. The undrained shear
strength (su) was set at 24 kPa (soft clay) and varied according to the legend.

0
24%
mode 1
50%
mode 2
-50 100%
mode 3
200%
mode 4
400%
mode 5
-100 Mudline

Amplitude
Depth [m]

-150

-200

-250

-300
0 0.1 0.2 0.3 0.4 0.5
-1 -0.5 0 0.5 1 Frequency [Hz]

(a) Modes for su = 24 kPa (b) FRF of well and riser system

Figure 5-14: Effect of soil stiffness on curvature Frequency Response Function 5 metres below
the mudline

It can clearly be seen in Figure 5-14a that the largest deflections at and below the mudline
occur at the second and third modes. Therefore it is expected that these modes suffer a shift
in natural frequency depending on the soil stiffness. From Figure 5-14b it is shown that this
indeed is true. Table 5-3 gives the exact natural frequencies and shift from the base case
for the modes of interest. Moreover, the amplitude of the second natural mode below the
mudline reduces significantly with an increase of stiffness.

Table 5-3: Effect soil stiffness on the critical modes 5 metres below mudline

Soil Stiffness Mode 2 Frequency shift Mode 3 Frequency shift


[% from 24 kPa su] [Hz] [% from 24 kPa su] [Hz] [% from 24 kPa su]
24 0.147 -6.4 0.168 -7.7
50 0.153 -2.5 0.173 -5.0
100 0.157 0.0 0.182 0.0
200 0.158 0.6 0.191 4.9
400 0.159 1.3 0.201 10.4

To access the influence of the frequency shift on the fatigue lifetime of the system, it was
simulated with a sea state with T p = 6s and Hs = 3m. The peak period of this sea state is
located between the second and third natural frequencies for all soil stiffness cases. Figure
5-15 shows the fatigue lifetime of this system, focussing on the region near the mudline.

Master of Science Thesis D.A. (Daniel) Okret


48 Frequency Domain Analysis

-250
24%
50%
-260 100%
200%
500%
Mudline
Depth [metres]

-270

-280

-290

-300
100 105 1010 1015
Lifetime [years]

Figure 5-15: Fatigue lifetime for different soil stiffness parameters

It can be seen that the worse fatigue lifetimes occur at 50% and 100% of the soil stiffness.
This is the case because for these soil parameters the second mode, which has the highest
curvature amplitude, is in the bandwidth of high excitation. Higher soil stiffness parameters
have the second natural frequency closer to the peak period, but due to their relatively small
curvature amplitude for this mode, their fatigue lifetime is larger. Moreover, the third natural
frequency for larger soil stiffness moves away from the applied Tp .

5-5-5 Soil Damping

Soil damping is currently not used for the assessment of the fatigue analysis of well systems.
Because of this, the amount of research conducted on this topic is small. In 2015 a model
to describe the soil damping of a conductor has been published in [2]. It is developed by
considering a soil element as a SDOF system. However, the validity of this assumption is
questionable. In view of the absence of a better description, this model will be used for
further investigation. The method is further elaborated in Appendix B-4.

The inclusion of soil damping has an influence in the damping ratio of the modes that have
large deflections at and below the mudline. These are the modes which contribute the most
to the fatigue of the system when excited. Therefore increasing the damping ratio for these
modes can significantly contribute to improve the fatigue performance of the well system.
Figure 5-16 illustrates the effect of damping for a well configuration in 270 metres water
depth.

D.A. (Daniel) Okret Master of Science Thesis


5-5 Sensitivities 49

0
0%
mode 1
25%
mode 2
-50 50%
mode 3
75%
mode 4
100%
mode 5
-100 150%
Mudline
200%

Amplitude
500%
Depth [m]

-150

-200

-250

-300
0 0.1 0.2 0.3 0.4 0.5
-1 -0.5 0 0.5 1 Frequency [Hz]

(a) Modes for damping = 0% (b) FRF of well and riser system

Figure 5-16: Effect of soil damping on curvature Frequency Response Function 5 metres below
the mudline

From Figure 5-16a it is possible to see that the second and third mode have a large displace-
ment. The modal damping of these modes is significantly increased when the soil damping is
included, as shown in Figure 5-16b. The exact modal damping ratios corresponding to these
modes are displayed in Table 5-4

Table 5-4: Effect of soil damping on modal damping of critical modes 5 metres below mudline

% Soil damping ratio Modal damping mode 2 Modal damping mode 3


[-] [-] [-]
0 0.038 0.033
25 0.055 0.036
50 0.072 0.045
75 0.078 0.053
100 0.098 0.059
150 0.134 0.077
200 0.189 0.092
500 - -

It was not possible to calculate the modal damping ratios for the case where the soil damping
ratio was increased by 500% due to the absence of clear peaks. However, an increase in modal
damping is clearly observed when including soil damping. The modal damping ratios of the
second and third modes are increased by a factor 2.5 and 2 respectively when comparing no
soil damping with 100% soil damping ratio.
To illustrate this effect, the riser configuration in 270 metres water depth with a sea state
with Tp = 6s and Hs = 3m was simulated with varying soil damping ratios. The resulting
fatigue lifetime near the mudline is displayed in Figure 5-17.
The fatigue lifetime improves with a factor 2 when comparing no damping with 100% damping
ratio. This factor becomes even larger when the damping ratio is further increased, getting

Master of Science Thesis D.A. (Daniel) Okret


50 Frequency Domain Analysis

-250
0%
25 %
-260 50 %
75 %
100 %
Depth [metres] 150 %
-270 200 %
500 %
Muldine
-280

-290

-300
100 105 1010 1015
Lifetime [years]

Figure 5-17: Fatigue lifetime for varying soil damping ratios

to 8 at 500% damping ratio. It should be noted that there is a large uncertainly in the
damping ratio, and only one source of parameters was available. Therefore it is possible that
the damping ratio is indeed much higher than what was used. Table 5-5 gives the modal
damping ratios for the first two modes. It is possible to observe that increasing the damping
coefficient from 1 to 2 almost doubles the modal damping experienced by the system.

5-5-6 Drag Coefficient and Diameter

The drag coefficient can be assessed interchangeably, because they appear linearly in the
hydrodynamic force term. Increasing the drag parameter increases the modal damping ratio.
This occurs for all modes.

Cd = 0.5
Cd = 0.75
Cd = 1
Cd = 1.25
Cd = 1.5
Cd = 2
Amplitude

0 0.1 0.2 0.3 0.4 0.5


Frequency [Hz]

Figure 5-18: Fatigue lifetime for varying damping ratios

D.A. (Daniel) Okret Master of Science Thesis


5-5 Sensitivities 51

Table 5-5: Effect of soil damping on modal damping of critical modes 5 metres below mudline

Damping Coefficient Modal damping mode 1 Modal damping mode 2


[-] [-] [-]
0.5 0.011 0.053
0.75 0.013 0.056
1 0.016 0.060
1.25 0.018 0.068
1.5 0.022 0.078
2 0.028 0.089

To assess the effect of the increased drag coefficient on the fatigue lifetime of the well system,
it was subjected to a sea state with Tp = 6 and Hs = 3. The results are displayed in Figure
5-19b. The increase from Cd from 1 to 2 results in an increase in fatigue lifetime of 5 times.
-250
Cd = 0.5
Cd = 0.5
Cd = 0.75
-50 Cd = 0.75
Cd = 1
-260 Cd = 1
Cd = 1.25
Cd = 1.25
Cd = 1.5
-100 Cd = 1.5
Cd = 2
Cd = 2
Depth [metres]

Depth [metres]

Muldine
-270 Muldine
-150

-200 -280

-250
-290

-300
-300
100 105 1010 1015 100 105 1010 1015
Lifetime [years] Lifetime [years]

(a) Fatigue lifetime for entire system (b) Fatigue lifetime around mudline

Figure 5-19: Effect of riser drag coefficient on fatigue lifetime

It is relatively easy to influence the drag parameters of the drilling riser. This can be done
by applying buoyancy neutral elements to it, with large diameters. Another possibility is to
attach stakes such as the ones used in the mitigation of VIV. One big pro of applying one of
these methods is that they can be launched from the moonpool with the equipment which is
already on board. Therefore, no large modifications need to be performed on the MODU to
apply one of these techniques.

5-5-7 Placement of the connector

The placement of the connector has a very large influence in the fatigue lifetime of the well
system. This effect is due to the SCF applied during the fatigue lifetime calculations for
this component, and not due changes in dynamic properties of the system. Therefore this
has a deterministic effect on the fatigue lifetime. For a SCF of 5, and applying the SN
curves defined in Table C-3, the connector decreases the fatigue lifetime with a factor 3000.

Master of Science Thesis D.A. (Daniel) Okret


52 Frequency Domain Analysis

Placing the connector away from areas where large fatigue damage occurs partially mitigates
its negative effect on the fatigue lifetime. This effect can be observed in Figure 5-20.
-195
4m
6m
-200 8m
10m
12m
14m
Depth [metres]

-205

-210

-215

-220
100 105 1010 1015
Lifetime [years]

Figure 5-20: Fatigue lifetime for different connector placements

When different connector are used, the fatigue lifetime can significantly change. Table 5-6
illustrates the effect a change in connector type has on the fatigue lifetime of the well system.
Therefore using a connector with a lower SCF should be used if it will be placed in a location
where large stress amplitudes occur.

Table 5-6: Effect of soil damping on modal damping of critical modes 5 metres below mudline

SCF Type of connector Fatigue reduction ratio


[-] [-] [-]
5 QUIK-STAB HD90 3124.8
2.7 Merlin 36x2 HDIF 143.5
1.3 Riser connector 3.7
1 DNV ’11 B1 1.0

D.A. (Daniel) Okret Master of Science Thesis


Chapter 6

Time Domain Analysis

In this chapter, a time domain model is developed to investigate the fatigue behaviour of the
well system. The model, which was developed in USFOS, is described in 6-1. In the following
sections, the obtained findings and results are elaborated.

6-1 Model description

In addition to the frequency domain model, a time domain model is developed to describe
the fatigue behaviour of the well system. The USFOS software, a FEM program that can
accurately perform dynamic analysis of offshore structures, was chosen to develop this model.
USFOS has extensively been used for the assessment of (slender) bottom founded structures,
such as jackets and mono towers.
The simulations of the program are performed in the time domain. Therefore nonlinear terms
such as the drag force and soil structure interaction can be modelled without linearisation.
An new feature has recently been developed, which makes it possible to incorporate vessel
motions into the analysis. This makes it ideal tool to describe the behaviour of the well system
during drilling operations. Furthermore, all the analysis can be done within one program,
which makes keeps the modelling simple and avoids complicated couplings.
Furthermore, the USFOS software has an extremely user friendly Graphical User Interface
interface called Xact. This makes visualising the model and high level verification easy. An
example of an USFOS model visualisation in Xact is shown in Figure 6-1.
The same subsystems developed in the frequency domain method will be applied in this
model. The drilling riser is modelled as a series of continuously connected beams. Each beam
element is defined as a pipe with structural and material properties equal to the main riser
pipe. The same is done with the wellhead and well system. The increased drag diameter of
the riser is included by applying a factor on the drag coefficient for these elements. In this
way all the dynamic effects are correctly accounted for.

Master of Science Thesis D.A. (Daniel) Okret


54 Time Domain Analysis

Figure 6-1: USFOS representation of model

Beam elements in USFOS are two node elements. Each node has three degrees of freedom,
giving a total of six degrees of freedom for each beam element. The definition of degrees of
freedom is displayed in Figure 6-2. Therefore both axial and transverse vibrations of the well
and riser system are included in this model. The element formulation for beam elements is
based on the exact solution of the 4th order differential equation for a beam subjected to end
forces [39].

Mz

Qz
z My x
Qy Mx
θz N

w j
y
θy
v
θx i
u

Figure 6-2: USFOS beam

The BOP is also modelled with beam elements with pipe geometry. The sum of the length of
the beams is equal to the BOP height. The mass of these elements also needs to be equal to
the mass of the BOP. To achieve this, the diameter of the pipes is fixed as its drag diameter,
and the wall thickness is optimised in such a way that the mass of the beams becomes equal
to the BOP mass. By applying this procedure, both the mass and hydrodynamic properties

D.A. (Daniel) Okret Master of Science Thesis


6-2 Post processing 55

are correct. Furthermore, because of the large diameter relative to the riser and wellhead
diameter, this structure is extremely stiff.
The vessel motions of the USFOS are created by providing the RAOs of the vessel and the sea
state parameters. As in the frequency domain model, a JONSWAP spectrum will be assumed,
that requires the peak period, significant wave height and direction of the waves as input.
Only head waves are analysed, as it is assumed that the MODU’s motions are optimised for
this wave direction. Only surge motions will be created by USFOS. Heave motions will be
included in the model as a time variation of tension. This is calculated outside of the USFOS
program and included as a node load on the upper node of the drilling riser. In Appendix D
the applied procedure to calculate the tension variation signal is elaborated.
Soil stiffness is modelled with nonlinear py-curves. The damping of the soil as elaborated
is Appendix B-4 cannot be included in this model because of limitations of the USFOS
software, and no suitable alternative was identified. Therefore no soil damping was included
in this model. The hydrodynamic forces are calculated with the Morison Equation, without
linearisation of the relative velocity.

6-2 Post processing

For the calculation of fatigue lifetime, a time signal of stresses is needed. USFOS only provides
time signals of moments and forces. By applying Equation 6-1, the stress time series can be
calculated.

M (t) · y F (t)
σ(t)f ront = +
I A (6-1)
M (t) · y F (t)
σ(t)back = −
I A

Note that the axial stresses need to be added and subtracted from the bending stresses. This
is the case because the bending stresses on opposing sides of the riser have inverse signs. The
front of the riser is defined as the side where the waves come from, and back of the riser is
defined as the side where the waves go to.
Once the time series of stresses is obtained, the rainflow counting algorithm is applied to
calculate the stress cycles for fatigue calculation. The programming procedure and routines
are further elaborated in Appendix E-2.

6-3 Effect of Tension variation across time

The time variation of tension can significantly reduce the fatigue lifetime of the well system.
This effect is particularly accentuated in steep sea states. This effect can clearly be observed
in Figure 6-3. For the both water depth, the fatigue lifetime is reduced when the time varying
tension is included. In 270 metres water depth, the reduction is by a factor 120, while a the
fatigue lifetime decreases by a factor 4.

Master of Science Thesis D.A. (Daniel) Okret


56 Time Domain Analysis

0 0
No Tension No Tension
With Tension With Tension
-50 Mudline
-100
Mudline

-100
Depth [metres]

Depth [metres]
-200

-150
-300

-200
-400
-250

-500
-300
100 105 1010 1015 100 105 1010 1015
Lifetime [years] Lifetime [years]

(a) 270 metres water depth (b) 500 metres water depth

Figure 6-3: Comparison of fatigue lifetime including and excluding tension variation for a sea
state with Hs = 5m and T p = 12s

Both reductions occur due to the same reason. However, because the fatigue lifetime is
decreased more severely in the 270 metres water depth, this case will be further elaborated
upon in the following section.

6-3-1 Source of fatigue damage

The main reason for the reduction of fatigue lifetime at the well system are not the axial
stresses resulting from the variation of tension across time. This can be seen in Figure 6-4,
where the fatigue lifetimes including and excluding the axial stresses are plotted. Along the
riser, there is a significant reduction of fatigue lifetime. However, below the LFJ the fatigue
lifetimes are almost the same.

0
Front of riser
Back of riser
-50
Only bending

-100
Depth [metres]

-150

-200

-250

-300
100 105 1010 1015
Lifetime [years]

Figure 6-4: Comparison of fatigue lifetime including and excluding the fatigue lifetime reduction
caused by axial stresses

D.A. (Daniel) Okret Master of Science Thesis


6-3 Effect of Tension variation across time 57

It can therefore be concluded that bending stresses cause the reduction in fatigue lifetime.
For this to be the case, the moments at the mudline should be larger when including the
time variation of tension. Figure 6-5 clearly shows that larger moments occur once tension
variations are included in the model.

×105
4
No tension variation
Tension variation
2
Moment [Nm]

-2

-4
0 100 200 300 400 500 600 700 800 900 1000
Period [s]

Figure 6-5: Moments as a function of time at mudline

Not only the amplitude of the moments increase, but there is also an increase in frequency.
The natural periods and corresponding modes of this configuration are shown in Table 6-1
and Figure 6-6. The largest displacements at the mudline occur at the third mode, that has
a natural period of 5.3 seconds.

Table 6-1: Natural periods of riser configuration

Mode Number Natural Period


[-] [s]
1 17.6422
2 7.8574
3 5.017
4 4.406
5 2.929

To get a better understanding of which vibrations are occurring, the time series are trans-
formed to the frequency domain. Figure 6-7 shows how the transformed time series of Figure
6-5 in the frequency domain. A large increase in moment amplitude can be observed when
the time varying tension is included in the model. This increase is most accentuated at the
at the second and third natural periods. It is plausible that the time variation of tension is
responsible for the excitation of this mode.

Master of Science Thesis D.A. (Daniel) Okret


58 Time Domain Analysis

0
mode 1
mode 2
-50 mode 3
mode 4
mode 5
-100 Mudline
Depth [m]

-150

-200

-250

-300
-1 -0.5 0 0.5 1

Figure 6-6: Eigen modes of riser configuration

105
No tension variation
Tension variation
104
Moment [Nm]

103

102

101

100
0 5 10 15 20
Period [s]

Figure 6-7: Moments as a function of period at the mudline

The tension variation in the frequency domain is displayed in 6-8. Note that a moving average
filter has been applied to give a better visualisation of the peaks. A clear peak is observed
at the peak period of the sea state, at 12 second. Two peaks at 2.3 and 4.1 second are also
identified, which are the result of the damping of the tensioning system. Although there is
no peak at the third natural period of the well system, at 5.0 seconds, the tension variation
amplitude is sufficient to cause resonance of this mode.

D.A. (Daniel) Okret Master of Science Thesis


6-4 Verification of tensioning model 59

102 102

101
101
Tension variation [kN]

Tension variation [kN]


100

100

10−1

10−1
10−2

10−3 10−2
0 5 10 15 20 0 5 10 15 20
Period [s] Period [s]

(a) Unfiltered tension (b) Filtered tension

Figure 6-8: Effect of riser drag coefficient on fatigue lifetime

The excitations caused by the time varying tension suffers little damping along the riser due
to its axial character. Therefore it can excite (critical) higher modes of the well system,
which are less excited by wave and vessel motions because of the large damping in transverse
direction. These higher modes can cause large bending stresses on the system, depending on
the riser configuration. The coupling between axial and transverse vibrations is possible in
this model because of the beam element formulation.

6-4 Verification of tensioning model

The horizontal vibrations of the riser cause a vertical displacement of the tip of the riser,
where it is attached to the tensioning system. If this displacement is large enough it could
contribute to the variation of tension in time. The vertical displacement of the top node of
the riser for a sea state with T p = 12 seconds and Hs = 5 metres is displayed in Figure 6-9.
0.04

0.02
Displacement [m]

-0.02

-0.04
0 100 200 300 400 500 600 700 800 900 1000
Time [s]

Figure 6-9: Vertical displacement of top node

As can be seen, the displacement of the top node is very small when compared to the heave
motion shown in Figure 6-10. Therefore its effect on the time variation of tension can be
neglected.

Master of Science Thesis D.A. (Daniel) Okret


60 Time Domain Analysis

1
Heave motion [m]

-1

-2
0 100 200 300 400 500 600 700 800 900 1000
Time [s]

Figure 6-10: Heave motion

6-4-1 Mitigation

Time varying tension can be counteracted by an active compensation system. Such a com-
pensation system should be developed to only compensate the variation of tension, and not
providing the static tension to the riser. The compensation system could be formed by elec-
tric cylinders, attached to the tensioning ring (the same place where the tensioning system is
attached) that produce the inverse the tension variation. It would be convenient if each ten-
sioner cylinder had its own compensating electric cylinder, to avoid eccentricities and simplify
the control system.

The control system should be set as feedback loop, where the piston movements are measured,
and transformed to variation of tension. This control system is very simple because only the
piston movements and transfer functions need to be known. However, the feedback loop
duration (time between measurement and reaction of the electric cylinders). If the duration
of the feedback loop is to long, the tension compensation system might actually increase the
time variation of tension. This has a negative effect on the fatigue lifetime of the system. This
effect can be seen in Figure 6-11 for a sea state with Tp = 12 and Hs = 5. The uncompensated
time variation of tension is shown in Figure 6-12 for comparison.

800
0.1 s
0.2 s
600
0.5 s
1s
400 2s
3s
∆ Tension [kN]

200

-200

-400

-600
0 100 200 300 400 500 600 700 800 900 1000
Time [s]

Figure 6-11: Effect of feedback loop duration on time varying tension

D.A. (Daniel) Okret Master of Science Thesis


6-4 Verification of tensioning model 61

500
∆ Tension [kN]

-500
0 100 200 300 400 500 600 700 800 900 1000
Time [s]

Figure 6-12: Uncompensated time varying tension

In this particular case, it can be seen that a feedback loop larger than 1 second starts to have
a negative effect on the time varying tension. Furthermore, the time variation of tension is
almost completely compensated if the feedback loop duration is equal to 0.1 seconds.

Master of Science Thesis D.A. (Daniel) Okret


62 Time Domain Analysis

D.A. (Daniel) Okret Master of Science Thesis


Chapter 7

Conclusions and Recommendations

In the beginning of the thesis, two goals were set. The first was to create a model which can
describe the dynamic behaviour of the well system for fatigue purposes. The second was to
identify key parameters needed for this calculation, and assess their sensitivity on the fatigue.
To achieve this goal, a frequency domain and a time domain model of the system were
developed. In this Chapter, a reflection on the goal will be done by providing the conclusions
in Sections 7-1 and the recommendations in Section 7-2.

7-1 Conclusions

The conclusions are presented separately for the frequency domain and time domain models.
Both models assume the riser and well system as being beam elements with vessel motions
and wave forces acting on it. However, the time domain model has more sophisticated beam
elements which can take axial deformations into account. The main difference between these
two models is that the frequency domain model is linear. This limits the application of this
model, because forces such as the time varying tension cannot be included in the model.
Furthermore, nonlinear features such as soil stiffness and drag forces need to be linearised for
the frequency domain model.

7-1-1 Frequency domain model

The frequency domain model is very fast and easy to set up. A single run can take anywhere
from a couple of seconds to a few minutes depending on the length of the riser and the
magnitude of the excitation. The magnitude of the excitation influences the run time of this
model, because it performs iteration on its mesh until a predefined residual error is achieved.
This is called a residual control based adaptive mesh. Larger magnitudes result in large
residuals, and therefore need more iterations. The short run time makes it ideal for running
multiple consecutive simulations to perform sensitivity analysis.
The following limitations have been identified for this model:

Master of Science Thesis D.A. (Daniel) Okret


64 Conclusions and Recommendations

1. Because only wave loading and vessel motions are taken into account, this model only
will predict significant fatigue damage until approximately 500 metres water depth.
This occurs because for larger water depths the vibrations caused by wave loading and
surge vessel motions are almost completely damped once they reach the mudline.

2. The applied soil linearisation method gives similar results when compared to the time
domain analysis near for the first few meters of soil. After that, the stiffness of the fre-
quency domain model gets to low due to the small displacements that occur in deeper
soil layers. This occurs because the secant linearisation method is only applied for
displacements near the mudline, because that is the region where the governing fa-
tigue lifetimes occur. Therefore, the secant method will give accurate estimates for the
governing conditions.

3. The linearisation method used for the hydrodynamic drag gives a accurate representa-
tion of the forces acting on the riser, when compared to the time domain model.

4. Only the case with MODU located directly on top of the well system was considered.
However, possible horizontal plane excursions could easily be included in this system
by adjusting the top tension accordingly.

A sensitivity analysis of several important parameters was performed. The main findings are
presented below:

1. Drilling fluid flow The fluid flow velocity of the drilling fluid through the riser was
found to have a very small effect on the fatigue of the well system and does not need
to be included in the fatigue calculations.

2. Drilling fluid density The density of the chosen drilling fluid has a significant effect
on the dynamic behaviour of the riser and well system, and could worsen its fatigue
lifetime with a factor larger than 100. Larger densities cause the natural frequencies of
the system to get lower (e.g. 20% increase in density caused 6% reduction in natural
period). Furthermore, larger densities reduce the effective tension of the riser, which
results in larger amplitudes at the mudline. If the shift in natural frequencies is a
problem, it can be partially mitigated by increasing the top tension of the system.

3. BOP size The BOP mass and dimension have a large influence in the dynamic be-
haviour of the system. Increasing the size and mass of the BOP will shift the natural
frequencies of modes with a large displacement at the mudline down. The effect of this
shift on the fatigue lifetime can be positive depending on the sea states to which the
system will be exposed, but is significant. The drag coefficient of the BOP has small
effect on the modal damping of the overall system. This is because of the small velocities
of the BOP.

4. Riser drag parameters The drag coefficient/diameter of the riser has a large effect
on the modal damping of all modes. Doubling one of these drag properties of the can
increase the modal damping by a factor five. This will result in a improvement of fatigue
lifetime, independently of the sea state to which the system is exposed.

D.A. (Daniel) Okret Master of Science Thesis


7-1 Conclusions 65

Special attention was given to the soil structure interaction. This has been pointed out as an
uncertain area by the literature. Current fatigue estimation methods use p-y curves to model
soil-structure interaction, neglecting any type of soil damping. The following was found from
the sensitivity analysis of soil properties.

1. Soil stiffness Lower soil stiffness will decrease the natural frequencies of modes with
a large amplitude near the mudline. The effect on the fatigue lifetime is dependent of
the sea state to which the system is exposed.

2. Soil damping The applied soil damping model has a significant effect on the modal
damping of modes which have large displacements near the mudline. Including the soil
damping can increase the modal damping by a factor larger than 5. However, there is
still uncertainty on the applied soil damping representation and parameters. The fatigue
lifetime can be improved by a factor close to 10 depending on the applied parameters.

7-1-2 Time domain model

The time domain model was developed in USFOS software. It is robust and can include
nonlinear features that have a significant effect on the fatigue lifetime of the well system.
However, it is significantly slower than the frequency domain model (e.g. 15 times longer
runtime than the frequency domain model).
The most important conclusion that can be drawn from the analysis with this model is that
the time variation of tension plays a very important role in the fatigue lifetime of the system.
This is the variation of tension of tension in the riser due to heave motions of the MODU. It
can reduce the lifetime of the well by a factor larger than 100. Therefore it is imperative that
this effect is considered for the calculation of the lifetime of the well system. This is generally
not included during assessments performed by the industry.
The method for calculating the time variation of tension is also important. Not only the
stiffness, but also the damping of the system need to be modelled. Damping has a significant
role for steep sea states, where the velocity of the heave motions is large. Neglecting this
effect can result in
The main limitation of USFOS software is that only built-in functions can be used. There is
no way to include external software to it. Because it has no built in VIV function, this cannot
be accessed with this software. However, the vendor is happy to assist in the development
this functionality. Another limitation is that soil can only be modelled through springs, and
not as an elastic solid, which gives a more accurate physical description. The applied soil
damping model can also not be included in USFOS because it is an equation based method.

7-1-3 Main identified improvements

Summarizing, following improvements to the fatigue lifetime calculations and fatigue lifetime
of the system have been identified:

1. Fatigue lifetime calculations

Master of Science Thesis D.A. (Daniel) Okret


66 Conclusions and Recommendations

(a) Including the time varying tension is imperative for the calculations
(b) Soil damping should be included

2. Fatigue lifetime of the system

(a) Increase the drag diameter of the riser. This can be done by applying buoyancy
neutral elements to the riser. These could be based on exiting buoyancy elements,
as they can be launched from the existing systems at the MODU.
(b) Actively compensating the time variation of tension. This can be achieved by a
feedback control system that measures the piston movements in combination with
electric cylinders.
(c) Placing the connectors and fatigue sensitive elements away from regions where
large cyclic stresses occur. Moreover, the connector type also has a large effect on
the fatigue performance of the system.

7-2 Recommendations

Although this project gives more insight into the fatigue behaviour of the well system, fur-
ther work could increase its understanding. There were several limitations and boundary
conditions which were set to this project. The following topics are perceived as promising for
further investigation.

Application of developed models The frequency domain model was developed to assess the
sensitivities of the fatigue behaviour well system. It is suitable for this purpose. However,
it has many limitations, such as the inclusion of the time varying tension. It is therefore
recommended to use the time domain model when assessing the fatigue behaviour of a specific
riser configuration. This will give a more complete, and therefore reliable, estimate of the
fatigue lifetime of the system.

Obtain measurement data Measurements would enable a better understanding of the dy-
namic properties of the system. Also, it would enable a more realistic validation of the
developed fatigue estimation models. Preferably measurements would be done at multiple
locations of the system, either with strain gauges or accelerometers. A minimum of three lo-
cations is recommended, somewhere along the riser, just above the LFJ and below the BOP.
Furthermore, the environmental conditions acting during the measurement campaign should
also be captured. This would enable a validation of the entire well and riser system model.

Validation with measurement data It was not possible to obtain relevant measurement
data to validate the developed models. Even though the models were validated against each
other and with FREERISE, it would still be extremely useful to perform a validation with
actual measurements. This would enable a more reliable validation of the developed models,
which in turn would decrease uncertainties on the calculated fatigue lifetimes.

D.A. (Daniel) Okret Master of Science Thesis


7-2 Recommendations 67

Soil modelling In this project, soil was modelled with a visco-elastic Kelvin foundation in
the frequency domain model, and as nonlinear springs (py-curves) in the time domain model.
Both approaches only account for the local reaction of the soil. Modelling the soil as a elastic
solid, and therefore accounting for its global behaviour, gives more realistic representation.
This should not result in a large increase in computation time. However, such a representation
cannot be included in either one of the developed models due to programming limitations.

Soil damping The applied soil damping method had a significant impact on the dynamic
behaviour of the well system, and therefore on its fatigue lifetime. Including this parameter
can significantly improve the fatigue lifetime of the well system. However, it has been pointed
out that the assumptions made in the development of this model are questionable. Therefore,
its validity is uncertain and so are the parameters that were used for its calculation. Identi-
fying alternative representations for this soil property could improve the level confidence in
this method.

Vortex Induced Vibrations The developed models only take vessel motions and wave loading
into account. This limits the riser configuration length that are prone to fatigue to about
500 metres water depth. Including the time variation of tension may cause longer riser
configuration to be prone to fatigue damage as well. However, if VIV is included, well systems
in larger water depths could become sensitive to fatigue damage as well. A possible model to
account for VIV is a phenomenological tuned wake oscillator.

Time varying tension Time varying tension was included based on the tensioning system
data from the Transocean Spitsbergen vessel. This is a harsh environment semi-submersible,
that is optimised to minimise the effect of environmental conditions. It is therefore plausible
that other tensioning systems would result in a larger variation of tension, that could have
a negative effect on the fatigue lifetime of the well system. It is recommended that other
tensioning systems be assessed.

Next step in modelling It is recommended to develop a model that can include additional
features such as the effect of VIV and soil as solid. It is not possible to develop such a model
in USFOS do to the limitations of the software. A possible program in which such a model
can be developed is Orcaflex, where additional features can be included due to its Python
and Matlab coupling capabilities.

Master of Science Thesis D.A. (Daniel) Okret


68 Conclusions and Recommendations

D.A. (Daniel) Okret Master of Science Thesis


Bibliography

[1] L. Reinås, Wellhead Fatigue Analysis : Surface casing cement boundary condition for
subsea wellhead fatigue analytical models. 2012.

[2] A. Zakeri, E. Clukey, B. Kebadze, P. Jeanjean, D. Walker, and G. Piercey, “Recent


Advances in Soil Response Modeling for Well Conductor Fatigue,” 2015.

[3] A. Kristensen, Flow properties of water-based drilling fluids. PhD thesis, 2013.

[4] DNV, “Wellhead Fatigue Analysis Method,” Proceedings of the ASME 2011 30th Inter-
national Conference on Ocean, Offshore and Arctic Engineering, 2011.

[5] NORSOK, “NORSOK U-001: Well System loads,” tech. rep., 2014.

[6] P. Ward, A. Rimmer, and H. Howells, “Evaluation of Wellhead Fatigue Using In-Service
Structural Monitoring Data,” OTC, no. May, pp. 6–9, 2013.

[7] 2H Offshore, “Shell Todd Oil Services Limited - Drilling Riser and Conductor Analysis,”
no. March, pp. 1–67, 2014.

[8] Shell, “EASTERN GULF OF MEXICO - DSME Rig Drilling Riser Analysis,” no. May,
2016.

[9] A. Haukanes and E. Harildstad, “Effects of BOP Stack Modelling on Estimated Wellhead
Fatigue Damage,” 2013.

[10] NORSOK, “NORSOK D-010: Well Integrity in Drilling and Well Operations,” Tech.
Rep. June, 2004.

[11] Det Norske Veritas, “DNV-RP-C205: Eviromental conditions and environmental loads,”
tech. rep., 2010.

[12] J. H. Lienhard, “Synopsis of lift, drag, and vortex frequency data for rigid circular
cylinders,” 1966.

Master of Science Thesis D.A. (Daniel) Okret


70 Bibliography

[13] R. Gopalkrishnan, Vortex-induced forces on oscillating bluff cylinders. PhD thesis, Mas-
sachusetts Institute of Technology, 1993.

[14] A. Mina, “Modeling the vortex-induced vibrations of a multi-span free standing riser,”
2013.

[15] I. Rychlik, “A new definition of the rainflow cycle counting method,” International Jour-
nal of Fatigue, vol. 9, no. 2, pp. 119–121, 1987.

[16] A. Halfpenny, “A frequency domain approach for fatigue life estimation from Finite Ele-
ment Analysis 2 . Review of S-N analysis in the time domain,” International Conference
on Damage Assessment of Structures, no. Damas 99, pp. 401–410, 1999.

[17] Det Norske Veritas, “DNV-RP-C203: Fatigue Design of Offshore Steel Structures,” tech.
rep., 2011.

[18] J. Spijkers, A. Vrouwenvelder, and E. Klaver, “Structural Dynamics CT 4140: Part 1


Structural Vibrations,” no. January, pp. 1–204, 2005.

[19] A. Metrikine, Dynamics, Slender Structures and an Introduction to Continuum Mechan-


ics. Delft University of Technology.

[20] A. Öchsner, Elasto-Plasticity of Frame Structure Elements. Springer, 1st ed., 2014.

[21] M. P. Paidoussis, “Pipes Conveying Fluid: Linear Dynamics I,” in Fluid-Structure In-
teractions, vol. 1, ch. Chapter 3, pp. 63–233, 2 ed., 2014.

[22] M. P. Paidoussis, “Concepts, Definitions and Methods,” in Fluid-Structure Interactions,


ch. Chapter 2, pp. 6–58, 1998.

[23] C. P. Sparks, “Mechanical Behavior of Marine Risers Mode of Influence of Principal


Parameters,” Energy Resources, vol. 102, pp. 214–222, 1980.

[24] F. Lim, C. Bridge, S. Hatten, L. Robinson, and P. Beynet, “Structural Damping Test
for the BP Shah Deniz Risers,” Offshore Technology Conference, pp. 3–8, 2005.

[25] G. L. Kuiper, Stability of Offshore Risers Conveying Fluid. 2008.

[26] J. B. Stevens and J. M. E. Audibert, “Re-Examination of P-Y Curve Formulations,”


Offshore Technology Conference, 1979.

[27] American Petroleum Institute, “API-RP-2GEO: Geotechnical and Foundation Design


Considerations,” tech. rep., 2014.

[28] H. Matlock, “Correlations for design of laterally loaded piles in soft clay,” 1970.

[29] P. Jeanjean, “Re-Assessment of P-Y Curves for Soft Clays from Centrifuge Testing and
Finite Element Modeling,” 2009 Offshore Technology Conference, no. Vm, pp. 1–23, 2009.

[30] Det Norske Veritas, “DNV-OS-J101: Design of Offshore Wind Turbine Structures,” Tech.
Rep. May, 2014.

[31] J. M. Hegseth, “Assessment of Uncertainties in Estimated Wellhead Fatigue,” Msc The-


sis, 2014.

D.A. (Daniel) Okret Master of Science Thesis


71

[32] S. L. Kramer, Geotechnical Earthquake Engineering. Pearson, 1996.

[33] D. W. Dareing and T. Huang, “Natural Frequencies of Marine Drilling Risers,” pp. 813–
818, 1976.

[34] J. Aranda-Ruiz and J. Fernández-Sáez, “On the use of variable-separation method for
the analysis of vibration problems with time-dependent boundary conditions,” vol. 226,
pp. 2912–2924, 2012.

[35] E. Myhre, L. E. Eilertsen, M. Russo, S. Asa, F. Johansen, and C. Hart, “Successful Real
Time Instrumentation Of The Conductor And Surface Casing Of An Exploration Subsea
Well In The North Sea To Measure The,” 2015.

[36] H. Howells, R. Baker, A. Rimmer, and O. E. Limited, “Measurement of Wellhead Fa-


tigue,” no. May, pp. 4–7, 2015.

[37] H. B. Lindstad, G. Grytøyr, M. Russo, and S. Asa, “Direct And Indirect Measurement
Of Well Head Bending Moments,” no. May, pp. 4–7, 2015.

[38] J. W. van de Graaf, “FREERISE: Part 1 - Theory,” tech. rep., 1983.

[39] T. Holmås, USFOS User’s Manual: Modelling. 1999.

[40] J. Journée and W. Massie, Offshore Hydromechanics. Delft University of Technology,


2001.

[41] L. P. Krolikowski and T. A. Gay, “An Improved Linearization Technique For Frequency
Domain Riser Analysis,” in Offshore Technology Conference, pp. 341–354, 1980.

[42] J. Ruigrok, “Laterally Loaded Piles,” p. 360, 2010.

[43] J. Schijve, “Fatigue of Structures and Materials in the 20 th Century and the State of
the Art,” Materials Science, vol. 39, no. 3, pp. 1–52, 2003.

[44] T. Irvine, “Rainflow Cycle Counting in Fatigue Analysis,” pp. 1–5, 2011.

[45] A. Ringeval and Y. Huang, “Random Vibration Fatigue Analysis with LS-DYNA,” 12th
International LS-DYNAÂő Users Conference, no. 2, pp. 1–16, 2012.

[46] T. Dirlik, “Application of Computers in Fatigue Analysis,” p. 234, 1985.

[47] NORSOK, “NORSOK N-004: Design of steel structures,” Tech. Rep. October, 2004.

[48] N. E. Dowling, “Mean Stress Effects in Stress-Life and Strain-Life Fatigue,” 2004.

[49] S. Cycles, “Elements of Metallurgy and Engineering Alloys,” ch. Chapter 14, 2008.

[50] G. Grytoyr, P. Sharma, S. Vishnubotla, and D. N. V. U. S. A, “Marine drilling riser


disconnect and recoil analysis,” American Association of Drilling Engineers, 2011.

[51] C. Gallagher, D. Williams, and D. Lang, “Modelling of marine riser tensioner load vari-
ations and implications for,” in International Conference on Ocean, Offshore and Arctic
Engineering, pp. 1–9, 2012.

Master of Science Thesis D.A. (Daniel) Okret


72 Bibliography

[52] G. L. Kuiper, J. Brugmans, and A. V. Metrikine, “Destabilization of deep-water risers


by a heaving platform,” Journal of Sound and Vibration, vol. 310, pp. 541–557, 2008.

[53] F. S. Gökhan, “Effect of the Guess Function & Continuation Method on the Run Time
of MATLAB BVP Solvers,” Edited by Clara M. Ionescu, p. 1, 2011.

[54] Infield, “Infield Rigs,” 2016.

D.A. (Daniel) Okret Master of Science Thesis


Glossary

List of Acronyms

APV Air Pressure Vessel

API American Petroleum Institute

BOP Blowout Preventer

DAT Direct Acting Tensioner

DP Dynamic Positioning

DNV Det Norske Veritas

FEM Finite Element Model

FRF Frequency Response Function

GUI Graphical User Interface

HP High Pressure

JIP Joint Industry Project

JONSWAP Joint North Sea Wave Project

LFJ Lower Flex Joint

LMRP Lower Marine Riser Package

LP Low Pressure

MDOF Multiple Degree of Freedom

MODU Mobile Offshore Drilling Unit

Master of Science Thesis D.A. (Daniel) Okret


74 Bibliography

MSL Mean Sea Level

NPV Nitrogen Pressure Vessel

PLC Programmable Logic Controller

RAO Response Amplitude Operator

SDOF Single Degree of Freedom

SCF Stress Concentration Factor

UFJ Upper Flex Joint

VIV Vortex Induced Vibrations

List of Symbols

Latin
A [m2 ] Area
Ae [m2 ] External area
Ai [m2 ] Internal area
c [Ns/m] Damping coefficient
cs [Ns/m] Soil damping coefficient
C [Ns/m] Damping matrix
Cd [-] Drag coefficient
Cm [-] Inertia coefficient
D [m] Diameter
Ddrag [m] Drag diameter
E [Pa] Young’s Modulus
F [N] Force
fn [Hz] Natural frequency
fv [Hz] Vortex excitation frequency
Hs [m] Significant wave height
I [m4 ] Area moment of inertia
J [kg[m2 ]] Moment of inertia
k [rad/m] Wave number
K [N/m] Stiffness matrix
ks [N/m] Soil stiffness
kspr [N/m] Translation spring stiffness
kθ [Nm/rad] Rotation spring stiffness
m [kg] Mass
M [kg] Mass Matrix
mf [kg] Mass fluid
l [m] Length
p [Pa] Pressure
pe [Pa] External pressure
pi [Pa] Internal pressure

D.A. (Daniel) Okret Master of Science Thesis


75

pu [Pa] Ultimate resistance


Rn [-] Reynolds number
St [-] Strouhal number
t [s] Time
T [N] Tension
Te [N] Effective tension
Tp [s] Peak period
Tr [Nm] Moment
u [m/s] Water particle velocity
uf [m/s] Fluid velocity
V [m/s] Flow velocity
Vv [-] Reduced velocity
w [m] Transverse deflection
W [-] Eigenmode
x [m] Axial direction
y [m] Displacement
yc [m] Ultimate displacement
z [m] Water depth
za [m] Heave amplitude

Greek
ω [rad/s] Frequency
ζa [m] Wave amplitude
ν [m2 /s] Kinematic viscosity
ξ [-] Damping ratio
ρ [kg/m3 ] Density
σ [Pa] Stress
ω [rad/s] Frequency
ωn [rad/s] Natural frequency
Ω [rad/s] Forcing frequency

Master of Science Thesis D.A. (Daniel) Okret


76 Bibliography

D.A. (Daniel) Okret Master of Science Thesis


Appendix A

Hydrodynamics

A-1 Wave Spectrum, Amplitudes and Velocity

The wave amplitudes and velocities are required to calculate the vessel motions and hydro-
dynamic forces acting on the riser. These wave properties can be calculated from the wave
spectrum. The applied procedure is elaborated in this section.
A JONSWAP spectrum was chosen to describe the wave amplitudes. It developed in the late
1960’s to describe the spectral formulation for fetch-limited wind generated seas. This is an
augmented version of the Pierson-Moskowitz spectrum, that applies for fully developed sea
states. The JONSWAP spectrum is given in Equation A-1.
" 4 #
320 H 2 5 ωp

Sζ (ω) = 4 5s exp − γA (A-1)
Tp ω 4 ω

Where:
" #
(ω − ωp )2
A = exp −
2σ 2 ωp2

ωp = (circular peak frequency)
Tp
(
3.3 for North Sea
γ = peakedness factor
1 for Pierson-Moskowitz
(
0.07 ω ≤ ωp
σ=
0.09 ω ≥ ωp

The JONSWAP spectrum is actually the probability density function of the wave amplitude.
To obtain the actual wave amplitudes, which are needed to calculate the vessel motions and
wave forces, Equation A-2 is applied.

Master of Science Thesis D.A. (Daniel) Okret


78 Hydrodynamics

q
ζa (ω) = 2 Sζ (ω) · ∆ω (A-2)

Once the wave amplitude spectrum is determined, and assuming linear wave theory, the
horizontal water particle velocity profile due to waves can be calculated with Equation A-3.
Note that this is also a spectrum.

cosh k(h + z)
u(z) = ζa ω cos(ωt) (A-3)
sinh kh
The wave number is obtained from the dispersion relation, given in Equation A-4. Because k
appears in a nonlinear way, it needs to be solved iteratively for each frequency.

ω 2 = kg tanh kh (A-4)

In deep water, kh becomes large, resulting in tanh kh = 1. In this case the dispersion relation
can be simplified to Equation A-5, which does not require iteration to be solved.

ω 2 = kg (A-5)

The horizontal water particle velocity obtained with Equation A-3 can be directly used in
Equation 5-4 to calculate the hydrodynamic forces acting on the riser.

A-2 Wave Steepness

The wave steepness is defined as the ratio between the wave amplitude and its length, as
displayed in Equation A-6.

H
steepness = (A-6)
λ
The wave number and wavelength are related to each other as shown in Equation A-7.


k= (A-7)
λ
For deep water waves, the wavelength can be rewritten to Equation A-8 by combining Equa-
tions A-5 and A-7, and applying the relation between period and frequency.

g
λ= · T2 (A-8)

Therefore, the steepness in deep waters is given by Equation A-9.

H · 2π
steepness = (A-9)
T2 · g

D.A. (Daniel) Okret Master of Science Thesis


A-3 Spectra and Time Series 79

The steepness is proportional to the water particle velocity. In other words, if the wave
height is maintained constant, smaller wave periods will cause higher water particle velocities.
However, individual waves in deep water break if the steepness exceeds 1/7 [40].
The steepness of a sea state can be calculated by inserting its significant wave heigh and zero
up-crossing period in Equation A-9. Because it is more convenient to work with the peak
period in stead of the zero up-crossing period, a conversion needs to be applied. The resulting
steepness, as e function of the significant wave height and peak period is given in Equation
A-10.

Hs · 2π
steepness = (A-10)
(1.4Tp )2 · g

A-3 Spectra and Time Series

Obtaining a time series of wave elevation directly from the JONSWAP spectrum is not possi-
ble, because it only contains information on the elevation of the waves, and not on the phase.
Fortunately it is not necessary to reproduce the time series exactly. By adding a random
phase to the spectrum, a statistically undistinguishable signal is created [40]. By applying
the Fourier Transform, which will be further explained in Section B-5, a time series of the
wave amplitude is obtained.
A scheme of the procedure of going from the frequency domain to the time domain and back
can be seen in Figure A-1.

Figure A-1: Wave Record Analysis and Recreation [40]

Master of Science Thesis D.A. (Daniel) Okret


80 Hydrodynamics

A-4 Vessel Motion

The vessel motions can be described as a function of its amplitude and phase. Te heave
motion is displayed in Equation A-11 as an example.

z = za cos ωt + εzζ (A-11)

Assuming the ship motions are linear, a set of Response Amplitude Operators has been
developed. This is a transfer function between the wave and the vessel motions for every
wave frequency. It gives the amplitude of the vessel’s motion za as a function of the wave
amplitude ζa and its phase. The RAO for the heave motion of e vessel is given in Equation
A-12 and Figure A-2.

za
(ω) = amplitude characteristics
ζa (A-12)
εzζ (ω) = phase characteristics

Figure A-2: Heave Motions of a Vertical Cylinder

By multiplying the magnitudes of wave amplitude spectrum and the RAO, the magnitude of
the vessel motion spectrum is defined. The phase of the vessel motion spectrum is calculated
by adding the phases of the wave amplitude spectrum and the RAO. A time series of the vessel
motion can be obtained by applying the Fourier Transform on the vessel motion spectrum.

D.A. (Daniel) Okret Master of Science Thesis


A-5 Linearisation Of Hydrodynamic Forces 81

A-5 Linearisation Of Hydrodynamic Forces

To perform analysis in the frequency domain it is necessary to linearise any nonlinear terms.
The drag force is given in Equation A-13.

1
FD (z, t) = ρ CD D(uc + uw − ẇ)|uc + uw − ẇ| (A-13)
2
Because the drag force is dependent on the squared velocity it is nonlinear. The chosen
linearisation method was developed by Krolikowski [41]. It minimises the mean squared error
between the output of the original nonlinearity and the linear displacement. Equation A-14
gives the linearisation procedure.

(uc + uw − ẇ)|uc + uw − ẇ| =⇒ B2 uc + B1 (uw − ẇ) (A-14)

The coefficients B1 and B2 are defined by Equation A-15. In this equation, ur stands for the
relative velocity and is equal to uw − ẇ.

uc uc
     
B1 = 4 P F + 2uc 2 P I −1
σur σur
2 !  (A-15)
uc uc uc
     
B2 = 2 σur P F + 1+ 2PI − 1 uc
σur σur σur

With P F , P I and σur being the probability function, probability integral and the standard
deviation of the relative velocity respectively. They are defines as:
!
1 x2
P F (x) = √ exp −
2π 2
! (A-16)
y2
Z x
1
P F (x) = √ exp − dx
2π −∞ 2

The linearised drag force is given by Equation A-17.

1 1 ˙
FD (z, t) = ρ CD DB2 Uc + ρ CD DB1 (uw − (w)) (A-17)
2 2

Master of Science Thesis D.A. (Daniel) Okret


82 Hydrodynamics

D.A. (Daniel) Okret Master of Science Thesis


Appendix B

Dynamics

B-1 Beam Dynamics

In section 4-1-2 the dynamics of a bending beam were introduced. In addition to its equation
of motion, boundary conditions are needed to correctly describe its behaviour. The boundary
conditions are formulated relative to the derivatives of the beam to space, that are given in
Equation B-1.

w(z, t) = displacement
∂w(z, t)
= slope
∂z
∂ 2 w(z, t) (B-1)
EI = bending moment
∂z 2
∂ 3 w(z, t) ∂w(z, t)
EI 3
−T = shear force
∂z ∂z

There are two two types of boundary conditions. These are kinematic conditions, that are
related to the displacement and slope, and dynamic conditions, that are related to the bending
moment and shear force. A selection of possible boundary conditions is given in FIgure B-1.

Master of Science Thesis D.A. (Daniel) Okret


84 Dynamics

Figure B-1: Examples of boundary conditions of a beam [19]

B-2 Experimental Modal Analysis

To obtain the dynamic properties of the system, experimental modal analysis techniques
in the frequency domain are applied. These properties can be extracted by combining the
Frequency Response Functions of several locations along the well system. Mode identification
is performed by examining the FRF of one point of the well system. Peaks in the FRF
correspond to natural frequencies of the system, and can be determined by peak picking. The
corresponding eigenmodes can be found with the magnitude of the imaginary part of FRF’s
at natural frequencies for different points along the well system.

The modal damping of the dynamic system is calculated with the half-power bandwidth
method. This frequency domain method used the peak sharpness at a natural frequency to
estimate the modal damping. To calculate the modal damping, two half power point are
defined, as per Figure B-2. The frequency bandwidth between these points is defined as the
half-power bandwidth.

D.A. (Daniel) Okret Master of Science Thesis


B-3 Soil Stiffness 85

Figure B-2: Half-power bandwidth method

Once the half power points have been identified, the modal damping can be calculated with
Equation B-2.

ω2 − ω1
ζ= (B-2)
2ωn

B-3 Soil Stiffness

The stiffness of the soil is described by p-y curves. The p-y curves which are currently adopted
in the API guidelines [27] were developed by Matlock [28] . To calculate the values of these
curves, first the ultimate unit lateral bearing capacity of the soil. This is given by Equation
B-3.
( 0
3su D + γ zD + Jsu z forz ≤ zr
pu D = (B-3)
9su D forz ≥ zr

where: pu = ultimate resistance [Pa]


D = diameter [m]
su = undrained shear strength of soil [Pa]
0
γ = submerged soil unit weight [N/m3 ]
J = dimensionless empirical constant with values ranging from 0.25 to 0.5
z = depth below mudline [m]
zr = depth below soil surface to bottom of reduced resistance zone [m]

The transition depth zr is the depth at which the ultimate resistance would be greater than
the limiting 9su D. For a constant strength along the depth, this can be calculated with
Equation B-4.

Master of Science Thesis D.A. (Daniel) Okret


86 Dynamics

6D
zr = 0 (B-4)
γ D
su +J

The cyclic p-y curve for z > zr is given by Equation B-5.

  1
y 3
0.5p fory ≤ 3yc

u yc



p= 0.72pu fory > 3yc (B-5)


forz ≥ zr

9s D
u

The cyclic p-y curve for z ≤ zr is given by Equation Equation B-6.

  1
y 3
0.5p fory ≤ 3yc

u yc


    
p = 0.72pu 1 − 1 − z y−3yc for3yc < y ≤ 15yc (B-6)

 zr 12yc
 z

0.72pu zr fory > 15yc

where: y = is the local pile lateral displacement [m]


yc = 2.5εc D
εc = strain at one-half the maximum deviator stress in laboratory undrained
compression tests of undisturbed soil samples [-]

For purpose of completeness, the static and dynamic curves are shown in Figure B-3 and
Figure B-4.

Figure B-3: Static P-Y curve given by API [27]

D.A. (Daniel) Okret Master of Science Thesis


B-3 Soil Stiffness 87

Figure B-4: Dynamic P-Y curve given by API [27]

Linearisation needs to be applied before using these curves in the frequency domain. The
secant procedure will be used, which is described in [30]. Initially the first point of Table B-1
is chosen as stiffness. If the lateral pile displacement exceeds y/yc , the next point is taken,
and so on. This approach is conservative for stresses underneath the mudline, because of
the underestimation of the stiffness, which results larger displacements and a lower fatigue
lifetime.
Table B-1: Point-wise cyclic p-y curve

p/pu y/yc
0 0
0.23 0.1
0.33 0.3
0.5 1
0.72 3

Some typical values of soil properties for normally and overly consolidated clays are given in
Table B-2 and Table B-3 respectively.
Table B-2: Representative properties for normally consolidated clay [42]

Consistency of Clay Average Undrained Shear Strength [kPa] εc


Soft <48 0.020
Medium 48-96 0.010
Stiff 96-192 0.005

For overconsolidated clays, the initial stiffness for static and cyclic loading is given in Table
B-4.

Master of Science Thesis D.A. (Daniel) Okret


88 Dynamics

Table B-3: Representative properties for overconsolidated clay [42]

Consistency of Clay Average Undrained Shear Strength [kPa] εc


Soft 50-100 0.007
Medium 100-200 0.005
Stiff 200-400 0.004

Table B-4: Representative values of initial stiffness for overconsolidated clay [42]

Average Undrained Shear Strength kpy (static) kpy (cyclic)


[kPa] [MN/m3 ] [MN/m3 ]
50-100 135 55
100-200 270 110
200-400 540 540

B-4 Soil Damping

The used approach to describe the soil damping is developed in [32]. This method claims to
use a viscous damper to approximate the hysteretic damping. The hysteresis loop is given in
Figure B-5. The energy dissipated by hysteresis in one load-unload cycle (Wd ) is equal to the
area of the ellipse.

WS
u
Wd

u0
Figure B-5: Stress-strain behaviour implied by viscous damping, adapted from [32]

When a soil element, which is modelled as a SDOF system shown in Figure 4-1, is subjected
to a harmonic displacement

u(t) = u0 sin Ωt (B-7)

the force exerted on that element is equal to:

F (t) = ku(t) + cu̇(t) = ku0 sin Ωt + cΩu0 cos Ωt (B-8)

D.A. (Daniel) Okret Master of Science Thesis


B-4 Soil Damping 89


The total energy dissipated from t0 to t0 + ω (one cycle of the hysteresis loop) can be
calculated with Equation B-9.

Z t0 + 2π
ω du
Wd = F dt = πc Ωu20 (B-9)
t0 dt

At the maximum displacement, there is no velocity and the strain energy stored in the system
(Ws ) is given by Equation B-10.

1
Ws = k u20 (B-10)
2

From Equations B-9 and B-10, the stiffness k and viscous damping factor c can be calculated.
Assuming the natural frequency is equal to the natural frequency of the system, the damping
ratio (ζ) is given by Equation B-11.

Wd
ξ= (B-11)
4πWs

Once the damping ratio has been defined, the viscous damping factor can be calculated as:

2kξ
c= (B-12)
ωn

A SDOF system only has one natural frequency and calculating the viscous damping is
straightforward. However, continuous systems as a well systems have infinite natural fre-
quencies, and obtaining the viscous damping coefficient is not that simple. Moreover, it is
not straightforward which stiffness should be used. The applied method assumes equivalence
between the natural frequency and forcing frequency to handle this complication. Because
the response of dynamic systems is largest at resonance, where damping is the dominant
term, this assumption should give reasonable results when applied to a SDOF. However, the
validity of this assumption for a continuous system is questionable. For the stiffness term,
the method assumes the secant soil stiffness, as explained in Section B-3.

Moreover, by making this assumption, the damping becomes frequency independent. This
can clearly be shown in by transforming the damping term of the equation of motion to the
frequency domain. This is done in Equation B-13.

2kξ
ciΩw̃ = iΩw̃ = 2kξiw̃ (B-13)

The damping ratios for kaolin clay, for three motion amplitudes, have been experimentally
obtained in [2], and are displayed in B-6.

Master of Science Thesis D.A. (Daniel) Okret


90 Dynamics

Figure B-6: Damping ratio profiles for several displacement amplitudes [2]

These damping ratios will be considered as representative for other soil types.

B-5 Fourier Transforms

The transformation of signals from time to frequency domain and the other way around is
done by performing Fourier transformations. This algorithm decomposes time signals into
the frequencies and corresponding amplitudes that make it up. The Fourier transform and
Inverse Fourier Transform are displayed in Equation B-14 and Equation B-15.

Z +∞
1
w(z, t) = w̃(z, Ω) · eiΩt dΩ (B-14)
2π −∞

Z +∞
w̃(z, Ω) = w(z, t) · e−iΩt dt (B-15)
−∞

Fourier transforms cannot account for the transient response unless initial conditions are
specified. Because including initial conditions makes the problem significantly more com-
plicated, only the steady state response is investigated using this technique. Therefore, the
system should reach the steady state response before starting the analysis, i.e. transient
effects should be damped out.

D.A. (Daniel) Okret Master of Science Thesis


Appendix C

Fatigue

C-1 Fatigue mechanism

The weakening of the material when cyclic loading is applied occurs due to the formation of
microscopic cracks. The fatigue life can be divided in three phases, namely crack initiation,
crack growth and final failure. Structural failure occurs after the crack reaches a certain size.
These stages are displayed in Figure C-1. For unwelded components, most of the fatigue life
is spent in the initiation period, while welded components spend most of its fatigue life in the
crack growth period.

cyclic crack micro crack macro crack final


slip nucleation growth growth failure

initiation period crack growth period

Figure C-1: Phases of fatigue life [43]

Due to the high amount of loads cycles acting on the riser and well system, these components
are sensitive to fatigue. There are multiple effects that need to be taken into account to
determine the fatigue damage of these elements. The following procedure will be used for the
fatigue lifetime calculation:

1. Calculate stresses and amount of cycles

2. Apply SCF’s to account for local geometry

3. Apply the Goodman relation to account for nonzero mean stress

4. Calculate fatigue lifetime with SN curves

5. Apply safety factors

Master of Science Thesis D.A. (Daniel) Okret


92 Fatigue

C-2 Fatigue Analysis in Time Domain

The fatigue analysis in time domain will be done with rainflow counting. This method is
extensively been applied throughout the industry and is able to convert a stress time series
into a stress cycles with a certain amplitude. The rainflow counting algorithm works as follows
[44]:

1. The stress history plot is rotated 90◦ clockwise (see Figure C-2)

2. A line is drawn downward from every stress peak

3. The number of half cycles is determined by counting the amount of terminations in the
flow that either:

(a) Reach the end of the time history


(b) Merge with the flow of another tensile peak
(c) Encounters a trough of greater magnitude

4. A magnitude is assigned to each half cycle, and half cycles with identical magnitude are
coupled forming a complete cycle

An example of rainflow counting is depicted in Figure C-2 and Table C-1.

Figure C-2: Rainflow plot [44]

This stress cycles can be inserted in Miner’s rule (Equation 3-8), after application of the ap-
porpriate factors, to determine the fatigue damage. The applied rainflow counting algorithm
was obtained from the Mathworks website1 .
1
https://nl.mathworks.com/matlabcentral/fileexchange/38834-simple-rain-flow-counting-
algorithm/content/myRainFlow.m

D.A. (Daniel) Okret Master of Science Thesis


C-3 Fatigue Analysis in Frequency Domain 93

Table C-1: Cycle Counting

Path Cycles Stress Range


A-B 0.5 3
B-C 0.5 4
C-D 0.5 8
D-G 0.5 9
E-F 1 4
G-H 0.5 8
H-I 0.5 6

C-3 Fatigue Analysis in Frequency Domain

The fatigue calculations in the frequency domain use the power spectral density of stress.
Most of the frequency domain methods, for example Bendat’s method [45], assume that
every stress peak is followed by a through of equal magnitude. This assumption is valid when
applied to narrow band signals, but is conservative when applied to wide band signals. This
can be seen in Figure C-3.

Narrow band
Stress

time

Wide band
Stress

time

Cycle counting using Bendat’s method


Stress

time

Figure C-3: Why Bendat’s Method is Conservative [16]

Master of Science Thesis D.A. (Daniel) Okret


94 Fatigue

An alternative method which can describe wide band signals is Dirlik’s methos. It is the result
of extensive computer simulations using the Monte Carlo technique. It used four moments of
area of the power spectral density to calculate the fatigue damage. This empirical method
was introduced by Dirlik [46]. The total fatigue damage is given by Equation C-1.

Z ∞
T 1
Damage = · σr m p(σr )dσr (C-1)
Tc A 0

where: T = total time [s]


m2
r
Tc = mean time between peaks [s] =
m4
mx = xth moment of area
A = coefficient of SN curve [N/m2 ]
σr = stress amplitude
m = coefficient of SN curve
p(σr ) = damage factor

The damage factor is calculated with Equation C-2.

D1 −Z D2 · Z −Z22 −Z 2
·e Q + 2
· e 2R + D3 Ze 2
Q R
p(σr ) = √ (C-2)
2 m0

Where:

2 · xm − γ 2

D1 =
1 + γ2
1 − γ − D1 + D12
D2 =
1−R
D3 = 1 − D1 − D2
σr
Z= √
2 · m0
1.25 · (γ − D3 − D2 · R)
Q=
D1
γ − xm − D12
R=
1 − γ − D1 + D12
m2
γ=√
m0 · m4
m1 m2
r
xm = ·
m0 m4

D.A. (Daniel) Okret Master of Science Thesis


C-4 SCF’s, SN curves and Safety Factors 95

C-4 SCF’s, SN curves and Safety Factors

Once the stress cycles have been defined, it it necessary to consider local geometric disconti-
nuity effects. In the well system, such discontinuities mainly occur at welds and connectors.
These areas are more prone to fatigue, and are considered fatigue hotspots. By applying
SCF’s, the additional fatigue damage of these hotspots is taken into account. The SCF’s used
in this thesis are based on [7].

For the calculation of the fatigue damage, the fatigue endurance will be described with SN
curves. The SN curves as prescribed by DNV are bilinear, having two different profiles
according to the numbers of cycles the material is subjected to. These curves are displayed
in Figure C-4.

1000
Stress range (MPa)

B1
B2

C
100 C1
C2
D
E
F
F1
F3
G
W1
W2
W3

10
1.00E+04 1.00E+05 1.00E+06 1.00E+07 1.00E+08
Number of cycles

Figure C-4: SN curves in sea water with cathodic protection [17]

Risers are hollow sections without welded sections. This would result in it being classified as
a C curve. However, due to its complexity and connection with surrounding auxiliary lines
a T curve will be used. The conductor has several locations where welds occur, as at the
connector and wellhead. This structure will also be evaluated as a tubular joint with SN
curve T.

To ensure that the structure endure the loads it is subjected to, and fulfil its intended function,
NORSOK imposed an additional safety factor. This factor is dependent on the consequence
of failure and accessibility, as can be seen in Table C-2. For elements below 150m water
depth, it may be assumed that access is not possible.

Master of Science Thesis D.A. (Daniel) Okret


96 Fatigue

Classification of Access for inspection and repair


structural components No access or Accessible
based on damage in the splash Below splash Above
consequence zone zone splash zone

Substantial consequences 10 3 2
Without substantial 3 2 1
consequences
“Substantial consequences” in this context means that failure of the joint will entail
danger of loss of human life;
significant pollution;
major financial consequences.

“Without substantial consequences” is understood failure where it can be demonstrated that the structure satisfy the requirement to
damaged condition according to the ALSs with failure in the actual joint as the defined damage.

Table C-2: Fatigue Safety Factor [47]

The following SCF’s and SN curves will be applied for the elements of the riser and well
system:

Table C-3: SCF’s, SN curves and safety factors

Component SN curve SCF Safety Factor


Riser DNV ’11 B1 1 10
Wellhead DNV ’11 B1 1 10
Conductor DNV ’11 B1 1 10
First Conductor Connector DNV ’11 B1 5 10

The base case for the connector is a QUIK-STAB HD90 Connector.


The application of of SCF’s and safety factors can have a significant effect on the fatigue
lifetime of an element. This effect will be clarified by means of the following example. Suppose
a conductor has a unfactored fatigue lifetime of 10 thousand years, a SCF of 5 and a safety
factor of 10. Taking the high cycle fatigue values for the T curve (m = 5 and log a = 15.606),
the factored fatigue lifetime can be calculated as:

10000 years
F actored f atigue lif e = SCF m
= 5 years (C-3)
log a · SF

Therefore, at hotspots, a relatively large unfactored fatigue lifetime can still be become critical
once the appropriate factors are applied.

C-5 Nonzero mean stress

A nonzero mean stress has an effect on the fatigue damage of materials. Tensional stresses
have a negative effect, while compression has a positive effect on the fatigue damage. Several

D.A. (Daniel) Okret Master of Science Thesis


C-5 Nonzero mean stress 97

empirical curves were developed to estimate the effect of mean stresses on fatigue. The most
relevant relations were developed by Goodman, Gerber and Soderberg, and are displayed in
Figure C-5.

Δσ i
Alternating stress Δσ a

Gerber

Goodman

Soderberg

σy σu
Mean stress σ m

Figure C-5: Effect of nonzero mean stress on fatigue

where: ∆σa = alternating stress


σm = mean stress
σy = yield strength
σu = ultimate tensile strength
∆σi = effective alternating stress

The empirical relations are given in the equations below:

σa σm
Solderberg : + =1 (C-4)
∆σi σy
σa σm
Goodman : + =1 (C-5)
∆σi σu
σa σm 2
 
Gerber : + =1
∆σi σu
(C-6)

Several experiments have shown that for steel, the effect of mean stress lays somewhere be-
tween the Goodman and Gerber curve [48][49]. Because the Goodman curve is more conser-
vative, never underestimating the fatigue damage, this will be used for the fatigue assessment.

Master of Science Thesis D.A. (Daniel) Okret


98 Fatigue

Table C-4: Mechanical properties of steel

Steel grade Yield Strength [MPa] Ultimate Tensile Strength [MPa]


X65 448-600 531-758
X80 555-750 621-837

Riser systems are made out of steel grade X80, and most of the components are made of steel
grade X65 or X80. The material properties for these steel grades are displayed in Table C-4.
For the estimation of the decrease of fatigue resistance due a non-zero mean stress, the worse
ultimate tensile strength will be used. Because tension only occurs in the drilling riser, and
not on the well, only steel grade X80 will be used for this assessment. The effect of the non-
zero mean stress on the top of the riser is displayed in Figure C-6. A LMRP with submerged
weigh of 650kN in combination with a 22 inch riser with wall thickness of 0.875 and steel
grade X80 was used for this figure.

0.6
Ultimate yield stress usage [%]

0.5

0.4

0.3

0.2

0.1

0
0 500 1000 1500 2000 2500 3000
Water Depth [m]

Figure C-6: Fatigue resistance reduction because of non-mean stress

It is possible to conclude that non-zero mean causes a large reduction in fatigue resistance,
reaching a reduction of over 50% for water depths of 3000 metres. This effect is increased
for large water depths, where the top tension is larger. Note that tension, and therefore
the non-zero mean tension, is reduced in along the riser. Therefore its affect on the fatigue
lifetime is worst at the sea level and will become less accentuated towards the mudline. No
tension is expected below the BOP, and therefore the non-zero mean stress at that point is
compressive and will not have a negative effect on the fatigue lifetime.

D.A. (Daniel) Okret Master of Science Thesis


Appendix D

Riser Tensioners

D-1 Tensioning systems and components

The basic components of both tensioning systems, wireline and DAT, are similar. They
have a set of Air Pressure Vessels or Nitrogen Pressure Vessels, an oil/air accumulator and a
hydraulic cylinder. The largest difference between the systems is that the cylinder of a DAT
is directly connected to the tension ring, while this connection is made through sheaves and
a wire in the wireline tensioner. The similarities between the two systems can be clearly seen
in Figure D-1.

Figure D-1: Schematic overview for wireline (left) and DAT (right) systems [50]

In both systems, a set of APV’s is used to pressurise the HP oil/air accumulator that provides
hydraulic fluid to the HP side of the cylinder. The LP side of the cylinder is supplied with
hydraulic fluid by a LP oil/air accumulator.
An anti-recoil valve is usually installed in the line between the HP oil/air accumulator and HP
side of the cylinder. This valve will shut the fluid flow if an emergency situation occurs, such

Master of Science Thesis D.A. (Daniel) Okret


100 Riser Tensioners

as a tensioner wireline failure or riser disconnection at the LMRP. The valve is programmed
to shut automatically if a certain flow rate is exceeded. Usually a Programmable Logic
Controller (PLC) system is used to monitor the flow velocity through the anti-recoil valve. A
more detailed representation of the components of a DAT are shown in Figure D-2.
Shackle Connection to
PLC Vessel
Tensioner Tensioner Cylinder
Stroke - LP Side
Signal

Flexible Hose
Anti-Recoil
Valve

LP Nitrogen
HP Air/Oil Accumulator
Accumulator
Flexible Hose

Tensioner Piston and Rod


Air Pressure Vessels
(APVs)
Tensioner Cylinder
- HP Side

Shackle Connection to
Tension Ring

Figure D-2: Detailed scheme of DAT system [51]

D-2 Source of time varying tension

The static top tension of the tensioning system is applied at midstroke. Although the ten-
sioning system reduces the fluctuation of top tension, it is not able to compensate the heave
motion completely. The change of top tension of the tensioners are the result of the pressure
changes within the hydraulic cylinders.The pressure in the cylinder and tension of the ten-
sioner are directly proportional. There are two mechanisms that contribute to the change of
pressure across time, both caused by the change of the stroke length of the piston.
The stroke length determines the gas volume inside the HP accumulator. An increase in
stoke length decreases this volume, with a decrease in stroke length increases the volume in
the accumulator. The change in volume is assumed to be adiabatic, i.e. no heat transfer.
This process can be regarded as the stiffness of the tensioner. The adiabatic relation is given
in Equation D-1.

P · V γ = Constant (D-1)

where: P = pressure [Pa]


V = volume [m3 ]
γ = adiabatic gas constant

Note that the stiffness is nonlinear, due to the presence of the adiabatic gas constant. For N2
(which is the main component in air) at 15 degrees, γ is 1.404.

D.A. (Daniel) Okret Master of Science Thesis


D-3 Modelling of tensioner system 101

The change in stoke length causes a flow of hydraulic fluid from the cylinder to accumulator
or the other way around. The flow of hydraulic fluid within the system is subjected to friction,
and causes a dynamic pressure drop. This friction can be calculated with the Darcy-Weisbach
equation. This pressure drop can be regarded as the dynamic damping of the tensioner, and
is given by Equation D-2.

ρ U 2f L
∆P = (D-2)
2D

where: ρ = fluid density [kg/m3 ]


U = fluid velocity [m/s]
f = friction factor [-]
L = length of piping [m]
D = diameter of piping [m]

The friction factor can be determined using the Moody Diagram. The damping of the ten-
sioner is also nonlinear, because it is dependent of the squared velocity. Another source of
pressure drop is the anti-recoil valve, which can significantly contribute to the damping of the
tensioner system.

D-3 Modelling of tensioner system

Two approaches were developed to describe the tensioner system and obtain the time fluctu-
ation of the tension in the riser. The first approach will use the measured hydraulic stiffness
and damping data of the DAT of the MODU Transocean Spitsbergen, and behaves as ex-
plained in Appendix D-2. The second approach was based on [52] and assumes the tensioner
system as being a linear spring.
Obviously the first approach is more realistic. The second method will only be used for
comparison, and to check if the more sophisticated model is necessary.

Method 1 - Hydraulic Stiffness and Damping


The first method relies on measurements of the tensioning system of the Transocean Spits-
bergen to calculate its stiffness and damping. The stiffness of the system is dependent on
the static tension of the system. Therefore, the pressure necessary to give the correct static
tension at midstroke needs to be defined, and subsequently the stiffness can be calculated by
applying Equation D-1. The damping relation of the system is only dependent on the piston
velocity, and therefore equal for all static tension settings. To calculate it, the total damping
of the system needs to be measured. The main sources of damping are the fluid flow through
the anti-recoil valve and piping.
The tensioner is assumed to be vertical, and therefore the pistons movements will be equal
to the heave motion of the vessel. An example of a stiffness and damping transfer function

Master of Science Thesis D.A. (Daniel) Okret


102 Riser Tensioners

3500 3000

2000
3000

1000
Tension [kN]

Tension [kN]
2500

2000
-1000

1500
-2000

1000 -3000
-8 -6 -4 -2 0 2 4 6 8 -2 -1 0 1 2
Heave displacement [m] Heave velocity [m/s]

(a) Tensioner Stiffness (b) Tensioner Damping

Figure D-3: Example of tensioner transfer functions

for 500 metres water depth are given in Figures D-3a and D-3b. Note that with zero heave
velocity and pistons at midstroke (zero heave displacement), the tension is equal to the static
tension.

Method 2 - Linear Spring


In [52], the heave compensator is assumed as a linear spring. Its stiffness is chosen to com-
pensate the submerged weight of the riser if the platform heaves with a critical amplitude.
No damping is taken into account by this method. This stiffness is given by Equation D-3.

Wr + WLM RP
k= (D-3)
ac

where: k = tensioner stiffness [N/m]


Wr = submerged weight riser [N]
WLM RP = submerged weight of the LMRP [N]
ac = critical amplitude [m]

Comparison
In both methods, the stiffness of the tensioning system is dependent on the water depth,
because it defines the length and therefore the submerged weight of the riser. Deeper waters
will have stiffer tensioning systems. The first method also takes into account the damping
mechanism of the tensioners, which is independent of the water depth. Therefore, the tension
variation of the second method will increase linearly with the depth. For the second method,
the tension variation will also increase with depth, but not linearly.
Comparing the tension variation of both methods, it is possible to conclude that the fist
method will underestimate the variation in tension for shallow waters and overestimate the

D.A. (Daniel) Okret Master of Science Thesis


D-3 Modelling of tensioner system 103

variation for deep waters. To illustrate this trend, the tension variations for deep and shallow
water are displayed in Figure D-4 and Figure D-5. The sea state used in this simulation is a
significant wave height of 5 metres and a peak period of 12 seconds.

Tension Variation
500
Method 1
400 Method 2

300

200
∆ Tension [kN]

100

-100

-200

-300

-400

-500
0 100 200 300 400 500 600 700 800 900 1000
Time [s]

Figure D-4: Comparison of tension variation for both methods in 250m water depth

Tension Variation
800
Method 1
600 Method 2

400

200
∆ Tension [kN]

-200

-400

-600

-800
0 100 200 300 400 500 600 700 800 900 1000
Time [s]

Figure D-5: Comparison of tension variation for both methods in 1000m water depth

Moreover, the tension variation of the second method will also not be accurate for steep sea
states, where the heave velocity is large. This can clearly be seen in Figure D-6 and Figure
D-7. Both sea states have a peak period of 12 seconds, and a wave height of 6 and 1 metres
respectively.

Master of Science Thesis D.A. (Daniel) Okret


104 Riser Tensioners

Tension Variation
800
Method 1
600 Method 2

400

200
∆ Tension [kN]

-200

-400

-600

-800
0 100 200 300 400 500 600 700 800 900 1000
Time [s]

Figure D-6: Comparison of tension variation for both methods in steep sea states

Tension Variation
80
Method 1
60 Method 2

40

20
∆ Tension [kN]

-20

-40

-60

-80
0 100 200 300 400 500 600 700 800 900 1000
Time [s]

Figure D-7: Comparison of tension variation for both methods in mild sea states

It is possible to conclude that neglecting the damping effects of the tensioner causes a signif-
icant error in the tension variation. Therefore it is important to include it in the calculations
when modelling the tensioner system.

D.A. (Daniel) Okret Master of Science Thesis


Appendix E

Programming

This appendix describes the programming techniques applied to describe the well system. In
section

E-1 Frequency Domain Model

When the equations of motion of a beam are transformed to the frequency domain, the
system of partial differential equations becomes a set of ordinary boundary value problems.
This is a set of ordinary differential equations that have a set of constraints, called boundary
conditions. Solving such a problem in a purely analytical manner can be very challenging
in certain cases. However, there are several means to solve it numerically. In this section, a
semi-analytical method will be introduced.
Matlab has a boundary value problem solver package called bvp4c. This is an residual control
based adaptive mesh solver [53]. An adaptive mesh solver does not use a uniform mesh, but
adjusts its mesh points at each iteration stage according to what is needed. This makes the
computation much simpler and reduces the storage costs.
The bvp4c solver requires three input functions:

1. mat3ode - A differential equation function, in which the differential equation’s need to


be described as a system of first-order ordinary differential equations.

2. mat4bc - A boundary condition function, in which the boundary conditions are specified.

3. mat4init - A initial guess function, in which an initial guess of the profile of each
derivative of the differential equation needs to be provided.

The output of the bvp4c solver is the value of each derivative of the differential equation (in
the case of the beam: the displacement, slope, curvature and shear force/bending stiffness)
for a given set of points.

Master of Science Thesis D.A. (Daniel) Okret


106 Programming

The method was validated statically and dynamically. The dynamic validation is described
in Section 5-4. A statical validation is performed here for three beam configurations of which
the analytical deflection is known. These are a cantilever beam with distributed load and
load at the free end, and a pinned-pinned beam with a distributed load. The equations for
their analytical deflection is displayed in Equation E-1.

−F x2
Cantilever Point Load: u(x) = (3L − x)
6EI
−qx2 2
Cantilever Distributed Load: u(x) = (x − 4Lx + 6L2 ) (E-1)
24EI
−qx2 3
Pinned-pinned Distributed Load: u(x) = (x − 2Lx2 + L3 )
24EI
The input data in Table E-1 was used for the calculations. Note that these values are merely
used for validation purposes, and are not representative for real structures.

Table E-1: Input values for bvp4c validation

Parameter Value Unit


Young’s Modulus 30 [MPa]
Area moment of inertia 25 [m3 ]
Length 40 [m]
Distributed load 100 [N/m]
Point load 1000 [N]

The results are presented in Figures E-1 to E-3. As can be seen, the results of the program
are extremely accurate.

×10−3
0
Numerical
-0.5 Analytical

-1

-1.5
Deflection [m]

-2

-2.5

-3

-3.5

-4

-4.5
0 5 10 15 20 25 30 35 40
Length [m]

Figure E-3: Pinned-pinned beam with distributed load

D.A. (Daniel) Okret Master of Science Thesis


E-1 Frequency Domain Model 107

0
Numerical
Analytical
-0.005

-0.01
Deflection [m]

-0.015

-0.02

-0.025

-0.03
0 5 10 15 20 25 30 35 40
Length [m]

Figure E-1: Cantilever beam with point load

0
Numerical
-0.005 Analytical

-0.01

-0.015
Deflection [m]

-0.02

-0.025

-0.03

-0.035

-0.04

-0.045
0 5 10 15 20 25 30 35 40
Length [m]

Figure E-2: Cantilever beam with distributed load

Master of Science Thesis D.A. (Daniel) Okret


108 Programming

A schematic overview of the program routines, including riser model, environmental loading
and fatigue calculations is displayed in Figure E-4. Note that the hydrodynamic loads are
created by a separate routine, and together with the structural parameters serve as input for
the main program.

Figure E-4: Overview of Matlab routine

D.A. (Daniel) Okret Master of Science Thesis


E-2 Time Domain Model 109

E-2 Time Domain Model

The time domain model was developed in USFOS. Because time domain simulations can take
a long time to run (e.g. 15 times longer than the frequency domain model), it is necessary to
optimise its running time. This can be done by adjusting certain programming parameters
while maintaining the accuracy of the fatigue estimates. Furthermore, tuning the program to
only output the needed data also improves the run time (e.g. making it twice as fast).
The parameters which have been identified as having the most effect on the runtime of the
program are the amount of beam elements along the riser, time step and duration of the
simulation. Sensitivity analysis were performed on the these parameters. The element size
below the mudline (e.g. of the well system) was kept at two metres, to identify where the
maximum fatigue damage occurs.
An example of a sensitivity analysis for the simulation duration is displayed in Figure E-5
and Table E-2. This was the case where there was the largest difference in fatigue lifetime
between the 2000 and 10000 second run.

Table E-2: Fatigue lifetime 2 metres under the mudline

Duration Lifetime no tension Lifetime with tension


(seconds) (years) (years)
1000 1557 1363
2000 2579 1742
5000 3033 1957
10000 3153 2043

0
1000 sec
2000 sec
-50 5000 sec
10000 sec

-100
Depth [metres]

-150

-200

-250

-300
100 105 1010 1015
Lifetime [years]

Figure E-5: Effect of simulation duration for Hs = 1 and T p = 12

It is possible to observe that the largest difference occurs from the 1000 to 2000 second
duration. Moreover, the fatigue life increases with the duration. A 2000 second duration was

Master of Science Thesis D.A. (Daniel) Okret


110 Programming

chosen as it gives a good estimation (maximum observed difference with 10000 second run is
18.2%) and has an ideal run time. To account for the possible error caused by this truncation
a safety factor could be applied.
It is common practice to have ten points to represent one vibration cycle. Therefore, the time
step should be taken as being ten times smaller than the highest period of interest. The time
step analysis demonstrated that no significant change occurs if the time step is reduced below
0.3 seconds, which is able to correctly describe frequencies up to 0.33 Hz. This is a very high
frequency for sea states, and should be sufficient. However, decreasing the time step from
0.3 to 0.1 seconds only resulted in a minor increase in total running time (e.g. around 10%).
Therefore a time step of 0.1 seconds was used for the simulations.
The element length was analysed by assessing the natural frequencies of the system. The fewer
elements, the lower the natural frequencies. It was observed that for water depths until 500
metres, 50 elements are sufficient to accurately model the system. This amount of elements
was chosen for the simulations, and the length of the elements was adapted according to
the water depth. The running time of the program could be further optimised by assessing
exactly how many elements are needed for a particular set-up.
The time domain model uses USFOS to perform the dynamic calculations and Matlab to the
the pre and post processing. An schematic overview of the routines is given in Figure E-6.
Where:

• RAO.inp contains vessel RAO


• Model file.fem contains structural parameters
• Head file.fem contains load parameters

The first step in this method is calculating the time varying tension of the riser. Once this
has been obtained, it is inserted into the Head file.fem file. USFOS only requires two input
files for its run (which are the Head and Model files), and the RAO.inp file is included as a
vessel property in the Model file. The elements and nodes for which time signals of forces,
moments and displacements will be output are defined in the Head file.
After the simulation of USFOS has been completed, the time signals of forces, moments
and displacements can be extracted from the output file res.dyn by executing the dynres.exe
executable. It will output the time series as a text file, which can be imported into Matlab
to perform the fatigue calculations.

D.A. (Daniel) Okret Master of Science Thesis


E-2 Time Domain Model 111

Varying

Figure E-6: Overview of time domain model routines

Master of Science Thesis D.A. (Daniel) Okret


112 Programming

D.A. (Daniel) Okret Master of Science Thesis


Appendix F

Drilling Rig Information

Mobile Offshore Drilling Units can be classified in according to its generation or drilling depth.
These classifications are displayed in Figures F-1 and F-1.

Table F-1: Types of Floating Drilling Rigs - Drilling Depth [54]

Drilling Depth Drilling Depth


Type
[m] [ft]
Midwater < 1220 < 4000
Deepwater 1221 - 2270 4001 - 7499
Ultra-deepwater > 2271 > 7500

Table F-2: Types of Floating Drilling Rigs - Generations [54]

Max Drilling Depth


Vessel Type Construction Year
[m]
1st Generation 1961 - 1972 180
2nd Generation 1973 - 1980 610
3rd Generation 1980 - 1985 1524
4th Generation 1986 - 1997 1524
5th Generation 1998 - 2004 N/A
6th Generation 2005 - onward 3048
7st Generation 2015 - onward 3658

Figures F-1 and F-2 show the RAO’s of the vessels which were used in the fatigue calculations.

Master of Science Thesis D.A. (Daniel) Okret


114 Drilling Rig Information

Heave RAO
4

3.5

3
Kan Tan IV
2.5
Aker H6
2 Noble Clyde Boudreaux

1.5 Ocean Guardian


Ocean Patriot
1
Noble Globetrotter
0.5

0
0 10 20 30 40
-0.5

Figure F-1: Heave RAO’s of vessels used in analysis

Surge RAO
2.5

2
Kan Tan IV
1.5 Aker H6
Noble Clyde Boudreaux

1 Ocean Guardian
Ocean Patriot
Noble Globetrotter
0.5

0
0 10 20 30 40

Figure F-2: Surge RAO’s of vessels used in analysis

D.A. (Daniel) Okret Master of Science Thesis

You might also like