Linda Sieber

Master’s thesis

VIV fatigue of rigid spool for subsea

template by a time domain model

Master’s thesis in Maritime Engineering

Supervisor: Prof. Svein Sævik (NTNU), Prof. Jonas Ringsberg
(Chalmers University of Technology) and Zhenhui Liu (Aker Solutions)
June 2020
NTNU
Norwegian University of Science and Technology
Faculty of Engineering
Department of Marine Technology
Linda Sieber

VIV fatigue of rigid spool for subsea

template by a time domain model

Master’s thesis in Maritime Engineering

Supervisor: Prof. Svein Sævik (NTNU), Prof. Jonas Ringsberg (Chalmers
University of Technology) and Zhenhui Liu (Aker Solutions)
June 2020

Norwegian University of Science and Technology

Faculty of Engineering
Department of Marine Technology
 VIV fatigue of rigid spool for
subsea template by a time domain
model

Master’s thesis

Submitted by Linda Sieber

Submission 10th June 2020

Supervisors Prof. Svein Sævik, Prof. Jonas Ringsberg

Industry contact Zhenhui Liu (Aker Solutions)

Department of Marine Technology

Faculty of Engineering
Norwegian University of Science and Technology

Department of Mechanics and Maritime Sciences

Division of Marine Technology
Chalmers University of Technology
 NTNU Trondheim
Norwegian University of Science and Technology Faculty of Engineering
Department of Marine Technology

MASTER THESIS WORK SPRING 2020

for

Stud. Tech. Linda Sieber

VIV fatigue of rigid spool for subsea template by a time domain model
Virvelindusert utmatting av stivt rørseksjon på undervannsramme etter en tidsplan modell

Vortex Induced Vibrations (VIV) can lead to fast accumulation of fatigue damage in offshore
slender structures, such as rigid spools applied in the inboard piping of subsea templates for oil
& gas production. A new empirical method for time domain (TD) calculation of VIV has been
developed by NTNU. This method is capable of accounting for structural non-linearity and
time-varying flow compared to the traditional frequency domain analyses. The objective of this
study is to validate the TD VIV prediction tool for pipeline spool applications. The work is to
be carried out as a continuation of the thesis work conducted during fall 2019 and is to include
the following items:

1. Literature study on the fundamental theory of VIV and numerical prediction tools,
relevant DnV GL rules and recommended practices related to spool design and VIV
response analysis.
2. Learn the TD VIV prediction tool SIMLA.
3. Define a case scenario including mechanical properties, geometry and environmental
conditions for an existing VIV model test
4. Perform eigenvalue analysis and explore the differences in terms of eigenvalues and
mode shapes with respect to different modelling choices, i.e. with or without bend
elements.
5. Perform TD VIV simulation for different load cases and compare the response to model
test results and the results obtained from DNVGL-RP-F105. This is to include
sensitivity analyses with respect to model parameters and fatigue estimates to explore
differences.
6. Run eigenvalue analysis in Abaqus with and without bend elements and compare with
eigenvalues from Simla.
7. Based on the above procedure, explore other relevant data published in literature and/or
alternatively a realistic full-scale scenario to be agreed upon.
8. Conclusions and recommendations for further work

The work scope may prove to be larger than initially anticipated. Subject to approval from the
supervisors, topics may be deleted from the list above or reduced in extent.

1
 NTNU
Faculty of Engineering
Norwegian University of Science and Technology Department of Marine Technology

In the thesis report, the candidate shall present her personal contribution to the resolution of
problems within the scope of the thesis work

Theories and conclusions should be based on mathematical derivations and/or logic reasoning
identifying the various steps in the deduction.

The candidate should utilise the existing possibilities for obtaining relevant literature.

Thesis report format

The thesis report should be organised in a rational manner to give a clear exposition of results,
assessments, and conclusions. The text should be brief and to the point, with a clear language.
Telegraphic language should be avoided.

The report shall contain the following elements: A text defining the scope, preface, list of
contents, summary, main body of thesis, conclusions with recommendations for further work, list
of symbols and acronyms, references and (optional) appendices. All figures, tables and
equations shall be numerated.

The supervisors may require that the candidate, in an early stage of the work, presents a written
plan for the completion of the work.

The original contribution of the candidate and material taken from other sources shall be clearly
defined. Work from other sources shall be properly referenced using an acknowledged
referencing system.

The report shall be submitted in electronic format (.pdf):

- Signed by the candidate
- The text defining the scope shall be included (this document)
- Drawings and/or computer models that are not suited to be part of the report in terms of
appendices shall be provided on separate (.zip) files.

Ownership
NTNU has according to the present rules the ownership of the thesis reports. Any use of the report
has to be approved by NTNU (or external partner when this applies). The department has the right to
use the report as if the work was carried out by a NTNU employee, if nothing else has been agreed in
advance.

Thesis supervisors:

Prof. Svein Sævik, NTNU, Zhenhui Liu, Aker Solutions and Prof. Jonas Ringsberg, Chalmers
University of Technology

Deadline: June 10th, 2020

2
Abstract
Subsea spools and jumpers are commonly placed close to the seabed where they are
exposed to low velocity currents. These currents lead to vortex induced vibrations
(VIV) in in-line (IL) and cross-flow (CF) direction. VIV can lead to fast accumula-
tion of fatigue damage and the prediction of VIV is therefore important in fatigue
assessment of those structures.

In this thesis work, a typical ’M’-shaped jumper model was investigated. Due to
the three-dimensional geometry, VIV predictions for spools and jumpers are more
complicated compared to straight pipes, both from hydrodynamic and structural-
dynamic perspective. The tri-axiality of flow makes it difficult to separate into IL
and CF response, but it is crucial to understand which modes are active in VIV
response. On a structural level, the stiffness is affected by bent pipe elements due to
deformation of the cross-section and additional strains. The special considerations
that must be made in analysis of spools and jumpers are pointed out in the literature
study.

Further, two models were investigated, which differed only in the element type
used to model the 90-degree pipe bends. One of these was a simple linear-elastic
pipe element, while the other one was an elbow element, capable of accounting for
deformation of the cross-section and additional strain terms due to bending. This
allowed for evaluation of the importance of change in stiffness on modal analysis.
Two software packages were used to verify the eigenvalue results and the use of
elbow elements: Simla, developed at NTNU Trondheim, and Abaqus, developed by
Dessault Systèmes R . It was shown that the reduced stiffness of the model including
elbow elements improved the results of modal analysis, especially with regard to
in-line bending, and caused a shift in the mode shape order.
Flexibility factors for a full-scale bend model including internal and external pres-
sure were evaluated.

TD VIV simulations for different load cases were carried out in Simla. With this TD
tool it was possible to account for large deflections and time-varying flow. A param-
eter study was conducted on the input parameters in Simla, especially for the TD
VIV prediction tool. The results from VIV motion were compared to the empirical
response model approach from DNVGL-RP-F105 (2019) and to experimental data.
With varied parameters of the TD VIV tool it was possible to obtain results that are
in good correlation with model test data. The DNV GL response model approach
for many cases gave very conservative motion amplitudes, especially for CF VIV,

4
compared to experimental data.

Stresses and fatigue damage from VIV were calculated by two different methods, of
which one used the moment signals obtained from the VIV tool in Simla directly
and applied rainflow counting to find stress ranges. The other method followed the
response model approach from the recommended practice DNVGL-RP-F105 (2019).
For this latter method, modal analysis results were used together with response am-
plitudes obtained from the guidelines. For both methods the maximum principal
stress criterion was used for equivalent stresses. Fatigue damage was calculated us-
ing the SN-data from DNVGL-RP-C203 (2005).
For most cases the fatigue damage calculated using the DNV GL response model pro-
cedure was higher than obtained from the TD tool, which was expected since motion
amplitudes obtained from the guidelines were already very conservative. However,
for several load cases with low-velocity currents up to 0.3 m/s the Simla tool gave
higher fatigue estimates. These cases are due to IL VIV which occurs at low current
velocities and is especially relevant for structures placed close to the seabed. It was
also found that torsional stresses are relevant for observed flow directions and, thus,
cannot be neglected. Sensitivity of fatigue damage to changes of the drag coefficient
and model parameters of the VIV tool was discussed. Based on the findings of this
thesis work, recommendations for future work were made.

5
List of Abbreviations
VIV - vortex induced vibration
FD - frequency domain
TD - time domain
IL - in-line
CF - cross-flow
KC - Keulegan-Carpenter number
CFD - computational fluid dynamics
SIF - stress intensification factors
FF - flexibility factors
FEM - finite element method
FEA - fine element analysis
DOF - degrees of freedom
FTT - Fast Fourier Transformation

List of Symbols

a SN-curve parameter
A motion amplitude
A/D normalized VIV response amplitude
(Ay /D)j normalized IL VIV amplitude for the j-th mode in the DNVGL proce-
dure
(Az /D)j normalized CF VIV amplitude for the j-th mode in the DNVGL proce-
dure
Ca added mass coefficient
Ca,CF −RES added mass coefficient due to CF response
CD ,CDrag Drag coefficient
CL Lift coefficient
CM Inertia coefficient
CV CF lift force coefficient
CV I1 IL force coefficient for region one
D hydrodynamic diameter (outer pipe diameter)
E Young’s modulus
Fd damping force

7
fCF −RES,j response frequency for j-th CF VIV mode
fcyc,IL , fcyc,CF cycle counting frequency for IL or CF stress cycles
fn,j natural still-water frequency for the j-th mode
fIL,j , fCF,j natural frequency in IL or CF direction for the j-th mode
fratio ratio of two consecutive CF modal frequencies
fvs vortex shedding frequency
fˆosc non-dimensional oscillating frequency
fˆ1I1 minimum non-dimensional oscillating frequency that gives energy input
for IL region one
Iy , Iz area moment of inertia with respect to y or z-axis
Ix polar moment of inertia
k stiffness
L characteristic length, for a cylinder usually equal to D
m SN-curve parameter, depending on context
m mass, depending on context
ma added mass
µ dynamic fluid viscosity
Ni number of constant-amplitude cycles to failure
ni number of cycles for one block in stress histogram
Φv instantaneous phase of the excitation force during VIV
Re Reynolds number
ρw fluid density
∆Si stress range for one block in stress histogram
∆Seq equivalent stress range for fatigue analysis
σ1 principal stress
σxx flexural stress for beam/pipe element
St Strouhal number
t wall-thickness
τxy torsional stress for beam/pipe element
θ1 time derivative of instantaneous force phase for IL region one
θ2 time derivative of instantaneous force phase for IL region two

8
θσ1 direction of principal stress
U ,Uc undisturbed flow velocity, here equal to current velocity
Vr ,Vred reduced velocity
ω circular frequency

9
Contents

1 Introduction 21
1.1 Background and motivation . . . . . . . . . . . . . . . . . . . . . . . 21
1.2 Research questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.3 Organization of thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

2 Background and Literature review 24

2.1 Fundamental theory of VIV . . . . . . . . . . . . . . . . . . . . . . . 24
2.1.1 Governing parameters . . . . . . . . . . . . . . . . . . . . . . 24
2.1.2 Formation of VIV . . . . . . . . . . . . . . . . . . . . . . . . 27
2.1.3 In-line vibrations of a circular cylinder . . . . . . . . . . . . . 29
2.2 Numerical prediction tools . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2.1 Computational fluid dynamics . . . . . . . . . . . . . . . . . . 32
2.2.2 Wake oscillator models . . . . . . . . . . . . . . . . . . . . . . 32
2.2.3 Semi-empirical models . . . . . . . . . . . . . . . . . . . . . . 32
2.3 Time domain prediction tool . . . . . . . . . . . . . . . . . . . . . . . 33
2.4 DNV GL rules and recommended practice . . . . . . . . . . . . . . . 36
2.4.1 Response models . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.5 Fatigue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.6 Special case of spools and jumpers . . . . . . . . . . . . . . . . . . . 41
2.6.1 Pipe bend elements . . . . . . . . . . . . . . . . . . . . . . . . 41
2.7 Previous work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

3 Methodology 44
3.1 Simla analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.1.1 Pipe formulation . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.1.2 Eigenvalue analysis . . . . . . . . . . . . . . . . . . . . . . . . 48
3.1.3 Modal stresses . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.1.4 TD VIV simulation . . . . . . . . . . . . . . . . . . . . . . . . 49
3.1.5 Stresses from VIV analysis . . . . . . . . . . . . . . . . . . . . 52
3.1.6 Fatigue damage . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.2 Abaqus analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

10
 3.2.1 Element formulation . . . . . . . . . . . . . . . . . . . . . . . 54
3.2.2 Eigenvalue analysis . . . . . . . . . . . . . . . . . . . . . . . . 55
3.2.3 Modal stresses . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.3 DNV GL resonse model procedure . . . . . . . . . . . . . . . . . . . . 56
3.3.1 Response models . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.3.2 Stresses from VIV response models . . . . . . . . . . . . . . . 59
3.3.3 Fatigue damage . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.4 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

4 Input for case-study 64

4.1 Case scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.1.1 Jumper models and element types . . . . . . . . . . . . . . . . 64
4.1.2 Model setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.2 VIV simulation in Simla . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.2.1 Parameter study . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.3 DNV GL response model approach . . . . . . . . . . . . . . . . . . . 67
4.4 Fatigue life . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.5 Diameter-over-thickness study with internal and external pressure . . 70

5 Results 72
5.1 Eigenvalue analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.1.1 Pipe31 model in Simla and Abaqus . . . . . . . . . . . . . . . 72
5.1.2 Pipe34 model in Simla and Elbow31B model in Abaqus . . . . 74
5.2 Modal stresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.3 Diameter-over-thickness study with internal and external pressure . . 85
5.4 VIV analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.4.1 Parameter study in Simla . . . . . . . . . . . . . . . . . . . . 87
5.4.2 Comparison of VIV motion . . . . . . . . . . . . . . . . . . . 89
5.4.3 Summary of the section about VIV motion . . . . . . . . . . . 108
5.5 Stresses and fatigue damage for VIV . . . . . . . . . . . . . . . . . . 109
5.5.1 Sensitivity to modeling choices and model parameters . . . . . 125
5.5.2 Summary of the section about VIV stresses and fatigue damage132

6 Conclusion and recommendations 133

6.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
6.2 Recommendations for further work . . . . . . . . . . . . . . . . . . . 134

Appendices 137

11
List of Figures

2.1 Reynolds number dependency of the vortex shedding process. Taken

from Techet [1] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.2 Relationship between Reynolds number and Strouhal number for a
circular cylinder. Taken from Techet [1] . . . . . . . . . . . . . . . . . 27
2.3 Vortex shedding process. Taken from Sumer and Fredsoe [2] . . . . . 28
2.4 First and second instability region of IL VIV. Taken from Sumer and
Fredsoe [2] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.5 First instability region. Taken from Sumer and Fredsoe [2] . . . . . . 31
2.6 SN-curve extrapolation for cumulative damage. Taken from Ås and
Berge [3] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.7 Principle of rainflow counting method. Taken from Lee, Barkey and
Kang [4]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.8 Jumper model with accelerometer positions. Taken from Liu et al. [5] 43

3.1 Dof of pipe element. Taken from Sævik [6]. . . . . . . . . . . . . . . . 47

3.2 Local coordinate system and bending moment definition for pipe el-
ement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.3 Stress state of a 2D element. Taken from Liu et al. [5]. . . . . . . . . 50
3.4 Example of a stress histogram obtained as output from rainflow count-
ing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.5 Generation principle for CF response curve. Taken from DNVGL-
RP-F105 [7] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.6 Generation principle for IL response curve. Taken from DNVGL-RP-
F105 [7] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.7 Correction of added mass coefficient. Taken from DNVGL-RP-F105 [7] 60
3.8 Flowchart of work steps. . . . . . . . . . . . . . . . . . . . . . . . . . 63

4.1 Safety factors for the DNV GL response model procedure. Taken
from DNVGL-RP-F105 [7]. . . . . . . . . . . . . . . . . . . . . . . . . 68
4.2 SN-curves in seawater with cathodic protection. Taken from DNVGL-
RP-C203 [8]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

12
4.3 Parameters of SN-curves in seawater with cathodic protection. Taken
from DNVGL-RP-C203 [8]. . . . . . . . . . . . . . . . . . . . . . . . 70
4.4 Half of the 90◦ bend model. Left hand side un-deformed, right hand
side after rotation was applied on the right end. . . . . . . . . . . . . 71

5.1 Comparison of natural frequencies in Simla and Abaqus . . . . . . . . 73

5.2 Relative error of natural frequencies obtained for the two Pipe31 mod-
els in Simla and in Abaqus . . . . . . . . . . . . . . . . . . . . . . . . 73
5.3 Relative error of eigenvalue results obtained for the two elbow element
models in Simla and in Abaqus . . . . . . . . . . . . . . . . . . . . . 75
5.4 Mode one of the Pipe34 model. Out-of-plane bending of the top
horizontal in y-direction. . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.5 Mode two of the Pipe34 model. In-plane bending in x-direction. . . . 76
5.6 Mode three of the Pipe34 model. Out-of-plane twist. . . . . . . . . . 77
5.7 Mode four of the Pipe34 model. In-plane bending in z-direction. . . . 77
5.8 Mode five of the Pipe34 model. Symmetric out-of-plane bending of
the two bottom segments in y-direction. . . . . . . . . . . . . . . . . 77
5.9 Mode six of the Pipe34 model. Diagonal distortion in both x and
z-direction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
5.10 Mode seven of the Pipe34 model. Asymmetric out-of-plane bending
of the two bottom segments in y-direction. . . . . . . . . . . . . . . . 78
5.11 Mode eight of the Pipe34 model. Out-of-plane bending in y-direction.
Largest displacement occurs as buckling of the top horizontal. . . . . 78
5.12 Mode nine of the Pipe34 model. In-plane bending in z-direction.
Largest displacement occurs as buckling of the top horizontal. . . . . 78
5.13 Maximum principal unit stress obtained from the Simla Pipe31 model 80
5.14 Maximum principal unit stress obtained from the Simla Pipe34 model 80
5.15 Flexural unit stress obtained from the Simla Pipe31 model . . . . . . 81
5.16 Flexural unit stress obtained from the Simla Pipe34 model . . . . . . 81
5.17 Torsional unit stress obtained from the Simla Pipe31 model . . . . . . 82
5.18 Torsional unit stress obtained from the Simla Pipe34 model . . . . . . 82
5.19 Maximum principal unit stress of mode three for the Pipe31 model
and mode two for the Pipe34 model . . . . . . . . . . . . . . . . . . . 83
5.20 Maximum principal unit stress of mode four for the two Simla models 83
5.21 Maximum principal unit stress of mode six for the two Simla models . 84
5.22 Maximum principal unit stress of mode nine for the two Simla models 84
5.23 Flexibility factors obtained for Pipe34 elements in Simla with pressure
from Table 4.6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

13
5.24 Flexibility factors obtained for Elbow31B elements in Abaqus with
pressure from Table 4.6 . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.25 VIV history for 10◦ flow, Pipe31 . . . . . . . . . . . . . . . . . . . . . 92
5.26 VIV history for 10◦ flow, Pipe34 . . . . . . . . . . . . . . . . . . . . . 92
5.27 Comparison of VIV response in x-direction for 10◦ flow at accelerom-
eter five. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.28 Comparison of VIV response in y-direction for 10◦ flow at accelerom-
eter five. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.29 Comparison of VIV response in z-direction for 10◦ flow at accelerom-
eter five. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5.30 Comparison of VIV response in x-direction for 10◦ flow at accelerom-
eter three. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5.31 Comparison of VIV response in y-direction for 10◦ flow at accelerom-
eter three. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.32 Comparison of VIV response in z-direction for 10◦ flow at accelerom-
eter three. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.33 VIV history for 90◦ flow, Pipe31 . . . . . . . . . . . . . . . . . . . . . 98
5.34 VIV history for 90◦ flow, Pipe34 . . . . . . . . . . . . . . . . . . . . . 99
5.35 Displacement in y-direction for Pipe31 model at accelerometer seven
for the 90◦ flow. Current velocity Uc =0.75 m/s. . . . . . . . . . . . . 100
5.36 Comparison of VIV response in x-direction for 90◦ flow at accelerom-
eter seven. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.37 Comparison of VIV response in y-direction for 90◦ flow at accelerom-
eter seven. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.38 Comparison of VIV response in z-direction for 90◦ flow at accelerom-
eter seven. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.39 Comparison of VIV response in x-direction for 90◦ flow at accelerom-
eter three. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.40 Comparison of VIV response in y-direction for 90◦ flow at accelerom-
eter three. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.41 Comparison of VIV response in z-direction for 90◦ flow at accelerom-
eter three. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
5.42 VIV history for 45◦ flow, Pipe31 . . . . . . . . . . . . . . . . . . . . . 105
5.43 VIV history for 45◦ flow, Pipe34 . . . . . . . . . . . . . . . . . . . . . 106
5.44 Comparison of VIV response in x-direction for 45◦ flow at accelerom-
eter seven. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
5.45 Comparison of VIV response in y-direction for 45◦ flow at accelerom-
eter seven. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

14
5.46 Comparison of VIV response in z-direction for 45◦ flow at accelerom-
eter seven. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
5.47 Critical locations for stress assessment along the jumper . . . . . . . 109
5.48 Moment signal during VIV for the Pipe34 model in 10◦ flow at loca-
tion A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
5.49 VIV fatigue damage per year for the Pipe34 model in 10◦ flow at
location A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.50 VIV fatigue for the 10◦ flow at location A calculated with the DNV GL
response model procedure and modal stresses from the Simla Pipe34
model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.51 VIV fatigue for the 10◦ flow at location A calculated with the DNV GL
response model procedure and modal stresses from the Simla Pipe34
model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.52 Moment signal during VIV for the Pipe34 model in 10◦ flow at loca-
tion E . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.53 VIV fatigue damage per year for the Pipe34 model in 10◦ flow at
location E . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
5.54 VIV fatigue for the 10◦ flow at location E calculated with the DNV GL
response model procedure and modal stresses from the Simla Pipe34
model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
5.55 Moment signal during VIV for the Pipe34 model in 90◦ flow at loca-
tion B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.56 VIV fatigue damage per year for the Pipe34 model in 90◦ flow at
location B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.57 VIV fatigue for the 90◦ flow at location B calculated with the DNV GL
response model procedure and modal stresses from the Simla Pipe34
model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.58 Stress details for Uc = 0.45 m/s over ten seconds in 90◦ flow at
location B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
5.59 VIV fatigue damage per year for the Pipe34 model in 90◦ flow at
location A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
5.60 VIV fatigue for the 90◦ flow at location A calculated with the DNV GL
response model procedure and modal stresses from the Simla Pipe34
model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
5.61 Moment signal during VIV for the Pipe34 model in 45◦ flow at loca-
tion B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
5.62 VIV fatigue damage per year for the Pipe34 model in 45◦ flow at
location B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121

15
5.63 VIV fatigue for the 45◦ flow at location B calculated with the DNV GL
response model procedure and modal stresses from the Simla Pipe34
model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
5.64 VIV fatigue damage per year for the Pipe34 model in 45◦ flow at
location A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
5.65 VIV fatigue for the 45◦ flow at location A calculated with the DNV GL
response model procedure and modal stresses from the Simla Pipe34
model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
5.66 VIV fatigue damage per year for the Pipe34 model in 45◦ flow at
location E . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
5.67 VIV fatigue for the 45◦ flow at location E calculated with the DNV GL
response model procedure and modal stresses from the Simla Pipe34
model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
5.68 Sensitivity of fatigue damage with respect to modeling choices (with
or without bend elements) for the 10◦ flow. . . . . . . . . . . . . . . . 126
5.69 Sensitivity of fatigue damage with respect to modeling choices (with
or without bend elements) for the 90◦ flow. . . . . . . . . . . . . . . . 126
5.70 Sensitivity of fatigue damage with respect to the drag coefficient for
the 10◦ flow. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
5.71 Sensitivity of fatigue damage with respect to the drag coefficient for
the 90◦ flow. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
5.72 Sensitivity of fatigue damage with respect to the CF force coefficient
for the 10◦ flow. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
5.73 Sensitivity of fatigue damage with respect to the CF force coefficient
for the 90◦ flow. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
5.74 Sensitivity of fatigue damage with respect to the IL force coefficient
for the 10◦ flow. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
5.75 Sensitivity of fatigue damage with respect to the IL force coefficient
for the 90◦ flow. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
5.76 Sensitivity of fatigue damage with respect to the minimum vibration
frequency that gives energy input for IL region one for the 10◦ flow. . 131
5.77 Sensitivity of fatigue damage with respect to the minimum vibration
frequency that gives energy input for IL region one for the 90◦ flow. . 131

1 Maximum principal unit stress of mode one for the two Simla models 138
2 Maximum principal unit stress of mode two for the Pipe31 model and
mode three for the Pipe34 model . . . . . . . . . . . . . . . . . . . . 139
3 Maximum principal unit stress of mode five for the two Simla models 139
4 Maximum principal unit stress of mode seven for the two Simla models140

16
5 Maximum principal unit stress of mode eight for the two Simla models140
6 Maximum principal unit stress of mode one for the two Pipe31 models
in Simla and Abaqus . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
7 Maximum principal unit stress of mode two for the two Pipe31 models
in Simla and Abaqus . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
8 Maximum principal unit stress of mode three for the two Pipe31 mod-
els in Simla and Abaqus . . . . . . . . . . . . . . . . . . . . . . . . . 142
9 Maximum principal unit stress of mode four for the two Pipe31 models
in Simla and Abaqus . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
10 Maximum principal unit stress of mode five for the two Pipe31 models
in Simla and Abaqus . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
11 Maximum principal unit stress of mode six for the two Pipe31 models
in Simla and Abaqus . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
12 Maximum principal unit stress of mode seven for the two Pipe31
models in Simla and Abaqus . . . . . . . . . . . . . . . . . . . . . . . 144
13 Maximum principal unit stress of mode eight for the two Pipe31 mod-
els in Simla and Abaqus . . . . . . . . . . . . . . . . . . . . . . . . . 144
14 Maximum principal unit stress of mode nine for the two Pipe31 mod-
els in Simla and Abaqus . . . . . . . . . . . . . . . . . . . . . . . . . 145
15 Maximum principal unit stress of mode one for the two Elbow models
in Simla and Abaqus . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
16 Maximum principal unit stress of mode two for the two Elbow models
in Simla and Abaqus . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
17 Maximum principal unit stress of mode three for the two Elbow mod-
els in Simla and Abaqus . . . . . . . . . . . . . . . . . . . . . . . . . 146
18 Maximum principal unit stress of mode four for the two Elbow models
in Simla and Abaqus . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
19 Maximum principal unit stress of mode five for the two Elbow models
in Simla and Abaqus . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
20 Maximum principal unit stress of mode six for the two Elbow models
in Simla and Abaqus . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
21 Maximum principal unit stress of mode seven for the two Elbow mod-
els in Simla and Abaqus . . . . . . . . . . . . . . . . . . . . . . . . . 148
22 Maximum principal unit stress of mode eight for the two Elbow mod-
els in Simla and Abaqus . . . . . . . . . . . . . . . . . . . . . . . . . 149
23 Maximum principal unit stress of mode nine for the two Elbow models
in Simla and Abaqus . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
24 IL response model for the 10◦ flow direction . . . . . . . . . . . . . . 174
25 CF response model for the 10◦ flow direction . . . . . . . . . . . . . . 174

17
26 IL response model for the 90◦ flow direction . . . . . . . . . . . . . . 178
27 CF response model for the 90◦ flow direction . . . . . . . . . . . . . . 178
28 IL response model for the 45◦ flow direction . . . . . . . . . . . . . . 182
29 CF response model for the 45◦ flow direction . . . . . . . . . . . . . . 182

18
List of Tables

4.1 Segment lengths of the jumper model . . . . . . . . . . . . . . . . . . 65

4.2 Jumper model properties . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.3 Parameters for the parameter study on VIV prediction . . . . . . . . 66
4.4 Unit mass components of the jumper model . . . . . . . . . . . . . . 68
4.5 Properties of full-scale bend model . . . . . . . . . . . . . . . . . . . 71
4.6 Load cases for full-scale bend model . . . . . . . . . . . . . . . . . . . 71

5.1 Eigenfrequencies and mode shapes of Pipe31 models . . . . . . . . . . 74

5.2 Eigenfrequencies and mode shapes of Pipe34 model (Simla) and El-
bow31B model (Abaqus) . . . . . . . . . . . . . . . . . . . . . . . . . 76

1 VIV output of Pipe34 model in 10◦ flow at accelerometer five, CDrag =

1.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
2 VIV output of Pipe34 model in 10◦ flow at accelerometer three,
CDrag = 1.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
3 VIV output of Pipe34 model in 10◦ flow at accelerometer five, CDrag =
1.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
4 VIV output of Pipe34 model in 10◦ flow at accelerometer three,
CDrag = 1.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
5 VIV output of Pipe31 model in 10◦ flow at accelerometer five, CDrag =
1.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
6 VIV output of Pipe31 model in 10◦ flow at accelerometer three,
CDrag = 1.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
7 VIV output of Pipe31 model in 10◦ flow at accelerometer five, CDrag =
1.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
8 VIV output of Pipe31 model in 10◦ flow at accelerometer five, CDrag =
1.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
9 VIV output of Pipe34 model in 45◦ flow at accelerometer seven,
CDrag = 1.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
10 VIV output of Pipe34 model in 45◦ flow at accelerometer three,
CDrag = 1.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159

19
11 VIV output of Pipe34 model in 45◦ flow at accelerometer seven,
CDrag = 1.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
12 VIV output of Pipe34 model in 45◦ flow at accelerometer three,
CDrag = 1.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
13 VIV output of Pipe31 model in 45◦ flow at accelerometer seven,
CDrag = 1.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
14 VIV output of Pipe31 model in 45◦ flow at accelerometer three,
CDrag = 1.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
15 VIV output of Pipe31 model in 45◦ flow at accelerometer seven,
CDrag = 1.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
16 VIV output of Pipe31 model in 45◦ flow at accelerometer three,
CDrag = 1.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
17 VIV output of Pipe34 model in 90◦ flow at accelerometer seven,
CDrag = 1.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
18 VIV output of Pipe34 model in 90◦ flow at accelerometer three,
CDrag = 1.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
19 VIV output of Pipe34 model in 90◦ flow at accelerometer seven,
CDrag = 1.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
20 VIV output of Pipe34 model in 90◦ flow at accelerometer three,
CDrag = 1.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
21 VIV output of Pipe31 model in 90◦ flow at accelerometer seven,
CDrag = 1.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
22 VIV output of Pipe31 model in 90◦ flow at accelerometer three,
CDrag = 1.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
23 VIV output of Pipe31 model in 90◦ flow at accelerometer seven,
CDrag = 1.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
24 VIV output of Pipe31 model in 90◦ flow at accelerometer three,
CDrag = 1.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
25 IL unit response amplitudes for 10◦ flow with Pipe34 model . . . . . 175
26 CF unit response amplitudes for 10◦ flow with Pipe34 model . . . . . 176
27 Ca -Correction and CF-induced IL modes for 10◦ flow with Pipe34 model177
28 IL unit response amplitudes for 90◦ flow with Pipe34 model . . . . . 179
29 CF unit response amplitudes for 90◦ flow with Pipe34 model . . . . . 180
30 Ca -Correction and CF-induced IL modes for 90◦ flow with Pipe34 model181
31 IL unit response amplitudes for 45◦ flow with Pipe34 model . . . . . 183
32 CF unit response amplitudes for 45◦ flow with Pipe34 model . . . . . 184
33 Ca -Correction and CF-induced IL modes for 45◦ flow with Pipe34 model185
34 Ca -Correction and CF-induced IL modes for 45◦ flow with Pipe34 model186

20
Chapter 1

Introduction

1.1 Background and motivation

In the offshore industry, rigid spools and jumpers are often placed close to the seabed
in the environment of low-velocity currents. These currents can induce vortex in-
duced vibration (VIV) of the structures as a result of oscillating forces from the fluid.
VIV can lead to a fast accumulation of load cycles, which significantly decreases the
life of the structures. For environmental and economical reasons, the prediction of
VIV is therefore of high importance in fatigue life assessment.
Semi-empirical models in frequency domain (FD) have been widely used in industry
applications to predict cross-flow (CF) VIV. A problem is that these models are not
capable of accounting for structural non-linearity and time-varying flow. A time
domain (TD) prediction tool for VIV has been developed at NTNU to overcome
these drawbacks.
Furthermore, a response model approach, published in DNVGL-RP-F105 [7], is com-
monly used for the prediction of in-line (IL) VIV. The approach was developed for
free spanning pipelines, thus, special considerations apply for non-straight pipes.
For the latter one, it can be difficult to differentiate in damage contribution from IL
VIV and CF VIV, since they no longer occur in each other’s neutral plane. However,
the response model approach can be used to compare and evaluate the results from
the TD tool.
A new element type has been implemented into the TD prediction tool at NTNU
that can account for reduced stiffness and cross-section deformation of bent pipe
elements. The two main objectives of this project work are to show how this new
element type affects the results in modal analysis and to demonstrate that the TD
tool is applicable for pipeline spool and jumper applications with non-straight pipe
parts.

21
1.2 Research questions
1. Is the TD VIV prediction tool - with adjustment of certain control parameters -
applicable to non-straight pipes?

2. Which significance does the modeling choice between regular pipe elements and
elbow elements have with respect to modal analysis and stresses for a jumper model?

1.3 Organization of thesis

The thesis consists of the following chapters.

Chapter 1 presents the motivation, objectives and research questions. It provides

the outline of the thesis.

Chapter 2 provides a literature review on fundamental theory of VIV and numerical

prediction tools, fatigue assessment and relevant DNV GL rules and recommended
practices. This chapter was part of a previous study during fall semester 2019 and
has been adopted with some modifications. Section 2.5 has been added.

Chapter 3 includes the methodology applied in this study. It gives detailed de-
scriptions about element formulations and solvers of the utilized software, Simla and
Abaqus. It explains the routines used for fatigue assessment. Further, the response
models from DNVGL-RP-F105 are described in more detail.

Chapter 4 summarizes the input data for this study. This includes mechanical
and geometrical model properties, environmental conditions, the choice of elements
as well as input parameters for the VIV prediction tool in Simla and the DNV GL
response models. It also includes a section about a single bend element with real-
istic full-scale dimensions and internal and external pressure to further explore the
modeling choice between elbow elements and regular pipe elements.

Chapter 5 presents and discusses obtained results. It compares eigenfrequencies,

mode shapes and modal stresses of the models with and without elbow elements in
the bends from both Simla and Abaqus. It evaluates results from the TD VIV tool
in Simla against experimental data and the DNV GL response models.
The response model approach and the TD tool in Simla are compared with respect
to stresses and fatigue arising from VIV. Sensitivity of fatigue estimates is evaluated
for the different modeling choices and VIV tool parameters.

22
As approved by the supervisors task seven from the initial description of the work
scope was replaced due to the lack of other relevant publications. Instead, the ef-
fect of internal and external pressure on flexibility factors for elbow elements was
observed for a realistic full-scale scenario.

Chapter 6 summarizes key results and main conclusions from chapter five. It
also gives recommendations for future work.

Appendix A includes additional plots of modal stresses from Simla and Abaqus.
Oscillating frequencies, motion amplitudes and identified modes from the VIV out-
put in Simla can be found in Appendix B.
Appendix C contains calculation steps and intermediate results of the DNV GL
response model approach.
The Matlab routine used to calculate stresses and fatigue from moment signals is
presented in Appendix D. Due to the large amount of data, time signals obtained
from Simla were submitted in a separate folder in addition to this report.

23
Chapter 2

Background and Literature review

2.1 Fundamental theory of VIV

This chapter provides the fundamental theory of occurrence and formation of VIV.
The main parameters that are used in VIV analysis will be discussed here. The main
objectives are to give a better understanding of the methodology to be used later
and to summarize selected parts of the research that has been conducted previously.

2.1.1 Governing parameters

Reynolds number
The flow regime around a structure is governed by the Reynolds number. It is
defined as the ratio between inertia forces and viscous forces as

UL
Re = (2.1)
µ

where
U is the undisturbed flow velocity
L is the characteristic length, often also denoted as cylinder diameter D
µ is the dynamic viscosity of the fluid.

Generally, the flow regimes can be classified as laminar flow, turbulent flow, and
the transition region. At low Reynolds’s numbers (Re < 200), the viscous forces
dominate and the streamlines in the wake field of the structure are attached. Ve-
locity fluctuations are small and the flow is orderly. This is defined as laminar flow.
At higher Reynolds numbers (300 < Re), when a critical velocity of the flow is
exceeded and the inertia forces are large compared to the viscous forces, the flow
becomes turbulent. Characteristics of a turbulent flow are high velocity fluctuations
and chaotic motions [9]. The regime in between (200 < Re < 300) is denoted as

24
transition region. The flow regime also defines the vortex shedding pattern.

Formation of VIV occurs for any cylinder with a Reynolds number higher than
40, but in a laminar regime the vortex-street is two-dimensional. In a fully turbu-
lent wake field (300 < Re) the vortex shedding becomes three-dimensional. The
boundary layer of a circular cylinder changes with further increase of the Reynolds
number.
In the sub-critical regime (300 < Re < 3 × 105 ) the boundary layer remains laminar.
In the critical regime (3 × 105 < Re < 3.5 × 105 ) the boundary layer becomes tur-
bulent at the separation point of one side of the cross-section, but remains laminar
at the separation point of the other side. This phenomenon causes an asymmetric
flow and non-zero mean lift force [10]. For further increased Reynolds numbers the
boundary layer becomes turbulent at both separation points, but not yet in be-
tween the stagnation point and the separation points. In the supercritical regime
(3.5×105 < Re < 1.5×106 ), the transition from laminar to turbulent is shifted closer
to the stagnation point. For higher Reynolds numbers (1.5 × 106 < Re < 4.5 × 106 )
the flow is asymmetric again, since first it becomes fully turbulent only on one
side. This is denoted as upper-transitional regime. Beyond this, in the transcritical
regime, the flow is turbulent everywhere [10].

The vortex shedding process as a function of the Reynolds number is also shown in
Figure 2.1. To maintain kinematic similarity in experiments, is is important that
the Reynolds number in model tests is the same as in full-scale. Experiments on
the occurrence of VIV are often conducted in the sub-critical regime. Applying the
results to full-scale cases, which are often in the critical and supercritical regime, is
assumed to be conservative [11]. Another useful quantity in VIV analysis of cylinders
with different dimensions and in flow of different velocities is the Strouhal number,
which is related to the Reynolds number.

Strouhal number
The Strouhal number is a dimensionless parameter defined as

fs L
St = (2.2)
U

where fs is the vortex shedding frequency.

For Reynolds numbers in the sub-critical regime, the Strouhal number is almost
constant St ≈ 0.2 [1]. This means that the vortex shedding process remains almost
unchanged in this regime. In VIV analysis, the Strouhal number is often set to this

25
Figure 2.1: Reynolds number dependency of the vortex shedding process. Taken
from Techet [1].

value. The relationship between Strouhal number and Reynolds number for cylin-
ders with circular cross-section is shown in Figure 2.2.

Reduced velocity
In VIV analysis, the motion amplitude is usually plotted as non-dimensional vibra-
tion amplitude A/D against the reduced velocity [7]:

Uc + Uw
Vr = (2.3)
fn,j D

where
fn,j is the natural still-water eigenfrequency for the j-th mode
Uc is the mean current velocity, normal to the pipe
Uw is the significant wave-induced flow velocity.

Keulegan-Carpenter number
When a cylinder is exposed to oscillating forces from the fluid, also the Keulegan-
Carpenter (KC) number is an important dimensionless parameter to describe the

26
Figure 2.2: Relationship between Reynolds number and Strouhal number for a cir-
cular cylinder. Taken from Techet [1].

flow properties [7]:

Uw
KC = (2.4)
fw D
Where
Uw is the Significant wave-induced flow velocity
fw is the significant wave/oscillating frequency fw = 1/Tw .

2.1.2 Formation of VIV

Sections 2.1.2 and 2.1.3 are based on the book by Sumer and Fredsoe [2].
When bluff bodies are exposed to motions of a fluid such as wave and current,
a so-called Karman vortex street can arise on the downstream side of the body.
In this section a smooth cylinder in stationary flow is assumed. As mentioned
before, vortex shedding starts at a Reynolds number of 40, but vortices arise even
earlier from Reynolds numbers of five. First, a fixed pair of small vortices arises
simultaneously on both sides of the cross-section. The size of the vortices increases
with the Reynolds number. For Reynolds numbers higher than 40, the vortex growth
becomes unstable and the wake is sensitive to small perturbations. In addition, the
pressure gradient of a circular cross-section changes sign when moving from the
stagnation point towards the wake field. As a result, the boundary layer that has

27
formed over the cylinder surface starts to separate. These two phenomena lead
to one vortex growing bigger than the other. This is visualized in Figure 2.3 and

Figure 2.3: Vortex shedding process. Taken from Sumer and Fredsoe [2].

can be described as follows [2]. When vortex A, here rotating clockwise, becomes
significantly larger than vortex B, vortex A will be able to draw vortex B across the
wake. The counteracting rotation of vortex B will then cause the velocity supply of
vortex A to stop. The same process repeats now with vortex B and C, where vortex
B is growing bigger and becomes able to draw vortex C across the wake. Thus,
the vortex shedding alters from one side to the other. These alternating vortices
cause oscillatory forces, which can then induce structural vibrations [1]. The force
component in the direction of the flow is called drag, while the one perpendicular
to the flow direction is called lift. In dynamic analysis, the forces are commonly
expressed as dimensionless coefficients.

Fx Fy
CD = 1 CL = 1 (2.5)
2
ρDU 2 2
ρDU 2
Here,
CD is the drag coefficient
CL is the lift coefficient
Fx the in-line force per unit length
Fy the cross-flow force per unit length.
The terms cross-flow and in-line refer to the oscillation direction of the structure
with respect to the flow angle. The lift force oscillates at the vortex shedding

28
frequency, while the drag force oscillates at a frequency twice as large. The frequency
of structural vibration is a compromise between the vortex shedding frequency and
the eigenfrequency of the relevant mode. When the frequency of structural vibration
is equal to the vortex shedding frequency, the structure is said to be ’locked-in’. At
this point the largest amplitudes of oscillation occur [1].

StU
Vortex shedding frequency ωvs = 2πfvs = 2π (2.6)
D

s
k
Natural frequency ωn = (2.7)
m + ma
Here,
k is the stiffness
m is the mass and
ma is the added mass.
Since the added mass depends on the non-dimensional oscillating frequency fˆosc =
fosc D/U and the amplitude ratio A/D, the natural frequency of an oscillating body
keeps changing [10]. This is important, because resonance can occur over a whole
range of frequencies.

2.1.3 In-line vibrations of a circular cylinder

While research has focused on CF VIV for decades, IL VIV and a combination of the
two phenomena have lately come more into focus. The vortex shedding frequency
of IL VIV is approximately twice as high compared to CF VIV. Thus, IL VIV gen-
erally occurs at lower flow velocities. It also indicates that fatigue damage from
IL VIV accumulates much faster than fatigue damage from CF VIV. Furthermore,
long slender marine structures such as free-spanning pipelines are designed with the
aim to avoid CF VIV. Accordingly, flexibly-mounted pipelines mainly experience
oscillating drag force arising from IL vortex shedding as a cause for fatigue damage.
IL vibrations can be classified into three categories, of which two occur for low-
velocity current, while the third one is observed at higher velocities in connection
with CF VIV.

First instability region

The so-called first instability region occurs for reduced velocities between one and
2.5. It is the result of two coupled processes. First, normal vortex shedding occurs,
which induces two oscillations per shedding. Secondarily, the IL motion of the cylin-
der relative to the fluid leads to symmetric vortex shedding, as is shown in Figure
2.5. The formation of secondary, symmetric vortex shedding indicates that vibra-

29
Figure 2.4: First and second instability region of IL VIV for a cylinder with Re =
6 × 104 . Taken from Sumer and Fredsoe [2].

tions from normal vortex shedding must already exist. In a flow without turbulence
or perturbation, the secondary system will not be found. However, if excited, the
IL force of the flow occurs to oscillate with a frequency that is approximately three
times the Strouhal frequency [2]:

fx D
= 3St (2.8)
U

where
fx is the frequency of the oscillating IL force.
Large amplitudes arise if this frequency is close to a natural frequency, where the
so-called first lock-in occurs. From Equation 2.8 and with fx = fn it can be derived
that this occurs at a reduced velocity of 1.7 [2]:

1 1
Vr = = ≈ 1.7 (2.9)
3St 3 × 0.2

Second instability region

When the reduced velocity is further increased, the IL-force frequency becomes
larger and so does the difference between fx and fn . Consequently, the symmetric
vortex shedding stops. When only normal vortex shedding is active, the IL-force
frequency is approximately twice the Strouhal frequency, which is a lower value than
in the first instability region. For further increase of the reduced velocity of the flow,
fx approaches fn again (second lock-in) which yields large amplitudes. The second
instability region is specified for reduces velocities between 2.5 and four due to the

30
Figure 2.5: Combination of normal vortex shedding and secondary, symmetric vortex
shedding, leading to the first instability region. Taken from Sumer and Fredsoe [2].

relation [2]:
1 1
Vr = = = 2.5 (2.10)
2St 2 × 0.2
It must be noted that until here, a constant Strouhal number of 0.2 was assumed,
since this is the case for the sub-critical regime, see section 2.1.1. For larger Strouhal
numbers, the first and second instability regions occur at lower reduced velocity val-
ues, and the graphs should be shifted to the left.
Generally, IL VIV amplitudes in these two regions are significantly smaller than CF
VIV amplitudes. This is because IL VIV can be initiated from smaller forces and
thus at smaller current velocities. Also, the drag coefficient CD during IL VIV is
smaller than the lift coefficient CL during CF VIV [2].

Third kind of in-line vibrations

A third kind of IL VIV occurs in a reduced velocity region where CF VIV is usually
initiated and far away from a lock-in point. However, much larger IL-motion and
IL-force amplitudes compared to the first and second instability regions can be ob-
served here. The IL-force frequency acting on the cylinder is still twice the Strouhal
frequency. To initiate IL VIV in the higher reduced velocity region, the following
phenomena must arise simultaneously. The cylinder experiences much higher veloc-
ities than in the IL region and larger vibration amplitudes in CF direction arise.
Together with stronger and more orderly vortex shedding this leads to an increase
in the drag force coefficient, which then causes IL VIV. A strong growth of the
fluctuating drag coefficient for a cylinder oscillating in CF direction is observed for
A/D-ratios greater than 0.2 to 0.3 [2].

31
2.2 Numerical prediction tools
Numerical prediction tools for VIV have been under development for many years,
thus, a large amount of tools is available. They have been divided into the following
categories by The Resistance Committee et al. [12].

2.2.1 Computational fluid dynamics

The term CFD covers different discretization methods to simulate the fluid flow,
which are usually applied in combination with a finite element model to simulate
the response of the structure. The working principle of these methods is to find the
fluid forces acting on a body by integration of the fluid pressure and shear stresses,
which are coupled with the velocity field formulation. CFD methods solve the
Navier-Stokes equation by taking viscosity into account. However, not all available
methods account for turbulence. Two popular methods for turbulence modeling
are Large Eddy Simulation (LES), which solves a filtered Navier-Stokes equation,
and Reynolds-averaged Navier-Stokes (RANS), which solves a time-averaged Navier-
Stokes equation.
CFD methods are generally promising for VIV prediction when using 3D models,
but they are very time-consuming. Several approaches using 2D models have been
developed. They solve the 2D flow at many planes along the riser with the aim of
saving computational time. However, 2D models are not commonly applied, since
they give less accurate results than semi-empirical models [10].

2.2.2 Wake oscillator models

Models that describe the oscillating forces acting during IL VIV and CF VIV by
using distributed Van der Pol wake oscillators are denoted as wake oscillator models
[12]. Srinil et al [13] have succeeded in including the effect of initial curvature and
time-varying lift and drag forces in a reduced-order model. From comparison with
CFD, the results of wake oscillator models are in good agreement. A major problem
is to find a set of parameters to solve for both free and forced vibration [12].

2.2.3 Semi-empirical models

Semi-empirical models are commonly applied to predict CF VIV. Two popular tools
are VIVANA, developed by Sintef Ocean, and SHEAR7, developed by the MIT.
They are based on Morison’s equation and utilize databases to determine coeffi-
cients for hydrodynamic forces such as added mass, damping and lift forces [14].
The structure is typically modeled with finite element methods. Semi-empirical
prediction tools operate in FD, assuming stationary flow and harmonic response

32
patterns. When using these tools, IL VIV must be treated separately from CF VIV.
A commonly applied procedure is to utilize the response model approach from
DNVGL-RP-F105 [7] for IL VIV and riser analysis methodology such as above
mentioned tools for CF VIV assessment. However, this approach does not hold for
flow-variations and structural non-linearity.

2.3 Time domain prediction tool

Since 2012 a TD prediction tool for VIV has been under development at NTNU. The
main aim is to achieve precise predictions for non-linear structural-dynamic behav-
ior and time-varying flow. In 2014, the tool was developed for CF VIV of cylindrical
slender structures, for which it works with high accuracy in stationary flow [10]. In
2017 an extended version for IL VIV was published. This section is based on the
doctoral thesis of Mats J. Thorsen [10], on the paper by Thorsen, Sævik and Larsen
[15] and on the paper by Ulveseter, Sævik and Larsen [16].
The most important feature of the TD tool is the ability to synchronize forces with
cylinder motion. Synchronization provides the possibility to transfer energy to the
cylinder at specific frequencies and motion amplitudes [16]. This also allows the
vortex-shedding frequency to lock on to the structure’s frequency of oscillation.

Cross-flow model
In the pure CF model, the lift force acting on the cylinder is a function of the CF
amplitude-over-diameter (A/D) ratio. A phase angle (0 < φ < 2π) is introduced to
account for fluctuations of the lift force. Furthermore, it is assumed that the time
derivative of the lift force phase is a function of the phase difference between the
lift force itself and the induced cylinder velocity [10]. The time derivative of the lift
force phase is the lift force frequency.
A new hydrodynamic damping formulation for CF VIV was published by Thorsen,
Sævik and Larsen [15]. The damping force is formulated as

1
Fd,y = − ρDCd,y |ẏ|ẏ (2.11)
2

where
Fd,y is the CF damping force
ρ is the fluid density.
The coefficient Cd,y is described with a linear function so that the energy loss per
cycle is approximately the same as for the model proposed by Vikestad et al. [15].

33
The total CF fluid force is written as [15]

1 1 πD2
Fy = ρDU 2 Cv cos φexc,y − ρDCd,y |ẏ|ẏ − ρ ÿ (2.12)
2 2 4

where the first term represents the CF excitation force, the second term represents
the damping force and the third term represents the added mass force. However,
the total added mass is also partly represented in the first term. The dimensionless
coefficient Cv determines the magnitude of the excitation force.

In-line model
In the pure IL model, it is assumed that vortex shedding gives rise to an IL fluc-
tuating component and that the frequency is approximately twice the CF vortex
shedding frequency. The vortex shedding frequency is described as a function of the
phase difference between cylinder motion and the force itself. Non-zero mean drag
due to low pressure in the wake region is not part of the excitation force model [16].
The damping term for IL VIV is formulated with respect to the x-axis where the
coefficients C1 and C2 are determined so that the dissipated energy per oscillation
cycle is approximately the same as for the model proposed by Venugopal [16].

1 1
Fd,x (t) = − ρDC1 U ẋ − ρAx C2 |ẋ|ẋ (2.13)
2 2

here,
Fd,x (t) is the IL damping force and
Ax is the motion amplitude in IL direction.
As discussed in section 2.1.3 two instability regions occur for pure IL VIV in low-
velocity current. Within these two low-velocity regions, the excitation force oscillates
with the same frequency as the cylinder velocity, but slightly out of phase with
displacement and acceleration. The excitation force acting in the second instability
region is given by the equation

1
Fexc,2 = ρDU 2 Cv2 cos (Φv2 ) (2.14)
2

where
Cv2 is a function of the in-line A/D-ratio and
Φv2 is the instantaneous phase of the force [16].
The time derivative of the instantaneous force phase is a function of the instanta-
neous phase of the velocity minus the instantaneous phase of the force itself [16].

dΦv2
= g2 (θ2 ) (2.15)
dt

34
 θ2 = Φẋ − Φv2 (2.16)

The frequency range of synchronization for the second instability region with a peak
value of fb = 0.35 is derived after Aronsen [17] who calculated the excitation force
coefficient as a function of the non-dimensional frequency and A/D-ratio.


 0.35 + 0.055 sin (θ2 ) if − π < θ2 < 0
fb (θ2 ) = (2.17)
0.35 + 0.035 sin (θ2 ) if π > θ2 > 0



Ulveseter et al. reformulated this as a sinusoidal function for the synchronization

range, so that the instantaneous frequency of the force can vary with the difference
of the instantaneous phase of the velocity and the instantaneous phase of the force
itself [16]:

U
dΦv2

2π D (0.35 + 0.055 sin (θ2 )) if − π < θ2 < 0
(θ2 ) = (2.18)
dt  2π U (0.35 + 0.035 sin (θ ))

if π > θ2 > 0
D 2

The first instability region is described in a similar way to the second one [16]:

1
Fexc,1 = ρDU 2 Cv1 cos (Φv1 ) (2.19)
2

dΦv1
= g1 (θ1 ) (2.20)
dt

θ1 = Φẋ − Φv1 (2.21)

The range of positive excitation is formulated for a peak value of fb = 0.45 [16].

U
dΦv1

2π D (0.45 + 0.055 sin (θ1 )) if − π < θ1 < 0
(θ1 ) =  (2.22)
dt  2π U (0.45 + 0.275 sin (θ )) if π > θ1 > 0
D 1

The total pure in-line force for a single degree of freedom system can then be written
as [16]:

1 1 1 1 πD2
Fx (t) = − ρDC1 U ẋ− ρAx C2 |ẋ|ẋ+ ρDU 2 Cv1 cos Φv1 + ρDU 2 Cv2 cos Φv2 −ρ ẍ
2 2 2 2 4
(2.23)
where the first two terms represent the linear and the quadratic damping term, the
third and fourth term represent the excitation forces of the first and second insta-
bility region and the last term represent the force arising from hydrodynamic added

35
mass. The formulation for only one degree of freedom is applied for low velocity
current. The magnitude of the excitation force is determined by the dimensionless
coefficients Cv1 and Cv2 . Equations 2.12 and 2.23 can be combined with a simple
FE model for the structure.

2.4 DNV GL rules and recommended practice

In DNVGL-RP-F105 (2019) [7], design criteria for VIV avoidance and fatigue are
proposed together with a response model and a force modal approach for VIV.
At present, the empirical response model approach is more detailed compared to
the force model approach and therefore is the suggested choice. The recommended
practice refers to free spanning pipelines. However, it can also be applied to subsea
spools and jumpers under some special considerations.

Free spanning pipelines

For free spanning pipelines, assessment of IL VIV and CF VIV are conducted sep-
arately. Eigenvalues and mode shapes are generally obtained from FE-analysis.
Modes are divided into dominating, weak and irrelevant modes by so-called mode
competition. The dominating mode is the one with largest A/D-ratio or the one
that leads to the largest stress, respectively, for CF VIV and IL VIV. Modes that
contribute with ten to 100 percent of the magnitude of the dominating mode are
identified as weak modes and must be taken into account in the analysis. Modes
that contribute with less than ten percent of the magnitude of the dominating mode
are identified as irrelevant modes and can be disregarded in the further procedure.

Jumpers, spools, flexible loops and subsea piping

Several exceptions apply when VIV fatigue is calculated for non-straight pipes by
the DNVGL-RP-F105 procedure. They are described in Appendix A of DNVGL-
RP-F105 (2019) [7]. Modes can no longer be strictly classified as IL or CF VIV
modes. They can change their classifications depending on the incoming flow angle
and the position of the non-straight parts. If a mode can be both IL and CF mode
at the same time, the according flow velocity shall not be reduced with respect to
the angle of attack. Mode reductions due to mode competition do not apply. Thus,
each contributing mode is handled as the dominating mode and applied with the
full stress range. Damage contribution from IL and CF VIV can act on the same
hot spots. The damage contribution is thus added for both IL and CF rather than
treated separately. Damage must either be calculated conservatively at one critical
location, or be calculated for at least 16 points along the circumference of each girth
weld.

36
2.4.1 Response models
The response models in DNVGL-RP-F105 are empirical models to determine the
maximum steady-state VIV response. Input parameters are basic hydrodynamic
and structural parameters. The models are valid for the following conditions [7]:
- IL VIV in steady current and current dominated conditions
- CF-induced IL motion, which is relevant in all regions where CF VIV occurs
- CF VIV in steady current and combined wave and current conditions
- CF VIV in wave-dominated Keulegan-Carpenter regimes.

The response models depend on the following hydrodynamic parameters [7]:

- reduced velocity, Vr
- Keulegan-Carpenter number, KC
- current flow velocity ratio, α
- turbulence intensity ratio, Ic
- flow angle, relative to the pipe, θ
- stability parameter, Ks .

The current flow velocity ratio is defined as [7]:

Uc
α= (2.24)
Uc + Uw

If no other information is available, the turbulence intensity factor is set to 0.05

according to the guidelines. The stability parameter is calculated from the equation
[7]
4πme ζT
Ks = (2.25)
ρw D2
where
ρw is the water density
ζT is the total modal damping ratio
me is the effective mass, defined as unit mass for the relevant mode.

2.5 Fatigue
If a structure is exposed to cyclic loading over time, it can experience fatigue dam-
age. The load that causes fatigue failure is often far below the yield stress of the
material [3]. This is because damage is accumulated cycle by cycle. For marine
structures which are generally exposed to dynamic loading, fatigue failure is a com-
mon issue. Fatigue damage is categorized into two types [3]:
-high-cycle fatigue with 105 to 107 cycles where the stress is essentially elastic

37
-low-cycle fatigue with two to 105 cycles where plastic behavior occurs.
Fatigue data is obtained from empirical testing and usually presented in SN-diagrams,
where the stress range is plotted over the number of cycles that would lead to fail-
ure if constant amplitude loading was applied alone. Such diagrams are available
for different environments, geometries and load directions. Also, SN-diagrams are
specified for different probabilities of failure. For engineering purposes it is common
to use a mean-minus-two-standard-deviations curve. The SN curve is found from
regression analysis of test data and lies two standard deviations below the mean re-
gression line. The remaining probability of failure for this design curve is 2.3 percent.

High-cycle fatigue for marine applications

Marine structures are generally in the range of high-cycle fatigue. In this range, the
SN-data tends to follow a log-linear relationship, which is described by

log N = log a − m × log S (2.26)

where a and m are constants of the mean regression curve. When two standard
deviations σ are included in the SN-curve this becomes [3]:

log N = log ā − m × log S (2.27)

log ā = log a − 2σlogN (2.28)

SN-curves for seawater with cathodic protection as given in the guidelines mostly
change slope at 1E + 06 cycles. This is because fatigue of welded joints, which are
commonly applied in marine engineering, mainly is due to crack growth of initially
small cracks. For cyclic loading with constant amplitude a threshold value exists
for fatigue crack growth. For stress ranges below this threshold the crack will not
grow. For environmental loads, some stress cycles are above this limit and some
below, where the ones above the limit contribute to crack growth. As the fatigue
limit is gradually lowered, more stress cycles will exceed the limit. The effect of
crack-growth on the fatigue limit is represented by an extrapolation of the SN-curve
with a slope of (2m − 1)−1 , where m−1 is the initial slope of the curve, as shown in
Figure 2.6. This model is valid for stationary load histories [3].

Palmgren-Miner hypothesis
To assess the cumulative damage over a certain time, the Palmgren-Miner approach

38
Figure 2.6: SN design curve with Haibach extrapolation beyond the fatigue limit to
take into account small cycles. Taken from Ås and Berge [3].

is typically used. Damage D is calculated as the sum

X ni
D= (2.29)
Ni

where ni is the number of cycles of a given stress range experienced during a defined
period and Ni is the number of constant-amplitude cycles which would cause failure
for this stress range as found from the SN-curve. Summation is applied over all load
cases i. Equation 2.29 is also known as ’Miner sum’. Failure is reached when D is
equal to one.

Rainflow counting
In many engineering applications vibrations are assumed to have a time history of
stress with Gaussian distribution and zero mean stress. A difference when consider-
ing real vibrations is that stress ranges of all magnitudes between zero and a fixed
maximum can occur. For broad banded load histories, rainflow counting is a suit-
able method to calculate stress ranges. It is possible to identify small cycles within
a group of larger cycles, independent of variations in the mean stress.
This method is implemented by counting the number of half-cycles in the stress
diagram. This is done by identifying the first peak or valley as seen in Figure 2.7.
When it reaches the edge, it drips down. The flow is terminated either when it
merges with a previous flow or when the absolute value of a subsequent peak or
valley is larger. Thus, small variations in the stress diagram are neglected. Finally,
the difference between start and end point of each half cycle yields its magnitude.
Half cycles of equal or similar magnitude, but opposite sign, are grouped to create
a full cycle.

Uni-axial and multi-axial fatigue

Fatigue can be classified based on its stress state as uni-axial or multi-axial fatigue.

39
Figure 2.7: Principle of rainflow counting method. Taken from Lee, Barkey and
Kang [4].

Stress can be described by six components. Three normal stress components act
along the x, y or z-axis, while three shear stress components act in the yz, xz or
xy-plane.
When cyclic stress acts in only one principal direction or plane, the stress state and
the resulting fatigue damage are uni-axial.
On the contrary, when cyclic stress acts in more than one principal direction or

40
plane, the stress and its resulting fatigue damage are classified as multi-axial.

2.6 Special case of spools and jumpers

Modeling and investigation of spools and jumpers yields additional challenges com-
pared to straight slender structures. Due to the three-dimensional geometry, the
response under VIV is much more complicated to predict. The mode shapes are
complex, including in-plane modes, out-of-plane modes and twist modes. For struc-
tures with complex geometry, each active mode can act as if it were dominating.
This means all modes can apply their full stress range. With traditional methods,
an accurate prediction of amplitudes may not be possible [18]. Also, semi-empirical
methods that are often developed for risers or straight pipes may not be applicable
to jumpers due to the complex flow around the structure.

2.6.1 Pipe bend elements

A typical jumper consists of several vertical and horizontal pipes that are connected
by curved elements. Bends lead to decreased stiffness and hence increased stress
compared to straight pipes. Unless the in-pipe pressure is very high, these param-
eters must be taken into account. This can be challenging when generally applying
the straight beam theory. Shell models can be used, but the analysis costs can be
very high.
For linear analysis, correction factors are commonly applied to estimate the flexi-
bility and stress intensification. Such factors are more complicated to predict for
non-linear problems. Another option to deal with the modified stiffness is to im-
plement another pipe element type that allows for ovalization of the cross-section.
A simple elbow element has been developed by Klaus-Jürgen Bathe [19], on whose
article this section is based. Bathe’s elbow elements hold for kinematic non-linearity
and elastic-plastic material behaviour and are applicable in static and dynamic anal-
ysis. The elements are assumed to distort only in the cross-section plane when being
bent, thus, warping is excluded. The ovalization displacement is described by the
following equations [19]:
 

Nd
N
4 X

c 
hk ckm sin 2mφ hk dkm cos 2mφ
X X
wξ (r, φ) = + (2.30)
 

 
k=1 m=1 m=1
| {z } | {z }
in-plane bending out-of-plane bending

dwξ
wζ = − (2.31)
dφ

41
 where
hk are isoparametric interpolation functions
wζ and wξ are the pipe-skin displacements in the ζ and ξ direction, respectively,
and c and d are the unknown generalized ovalization displacements.
Strain terms are applied to the initial configuration of the pipe elements to counte-
nance ovalization. The following strain terms represent the common beam normal
and shear strains and the additional ovalization strains [19].
 !2 
dwz
wξ sin φ + dφ
cos φ 1 d2 w ζ 
ηη = − ζ (2.32)
R − a cos φ R − a cos φ dθ2

" #
1 d2 wζ
ξξ = − 2 wζ + ζ (2.33)
a dφ2
!
1 dwξ
γηξ = (2.34)
R − a cos φ dθ

These formulations are used to determine the strain-displacement matrix. To cover

all ovalization modes of these elements 24 integration points are required, while
for pure in-plane bending twelve points are suggested. For integration through the
thickness of the pipe-skin, three points are usually considered appropriate. Total
Lagrangian formulation is applied to make the analysis efficient for large displace-
ments.
From comparison to curved shell elements it is evaluated that the predicted stiffness
of elbow elements is slightly larger. However, the results are still in good accor-
dance. The results for longitudinal stresses are similar, but circumference stresses
show larger deviations [19]. Based on this element type, a new pipe element was
implemented in Simla at NTNU to modify the flexibility of bends for the M-shaped
jumper.

2.7 Previous work

Model tests with an M-shaped jumper model were carried out by ExxonMobil with
the aim to assess validity of potential VIV prediction approaches [18]. Uniform bot-
tom currents were simulated to achieve reduced velocities that excite the vibration
modes of interest. With a scale ratio of 4.525 the model had a total arch length of
13.96m. The lengths of each segment and the model properties are listed in Table
4.1 and 4.2, respectively. In the first phase of the experiment the main question
was if VIV occurs as expected at very low velocities under realistic flow conditions.

42
Tests with a straight cylinder section of one meter were also performed to measure
drag forces and evaluate the sensitivity to changes in the Reynolds number. In the
second phase static pull tests were performed to excite and identify the first modes.
Thirteen tri-axial accelerometers and three strain gauges were installed along the
jumper model. The test setting is shown in Figure 2.8. For flow angles of ten and 45
degrees current velocities between 0.05 and 0.98 m/s were tested. For a flow angle
of 90 degrees data was measured for current velocities up to 0.79 m/s due to the
capacity of the end dynamometers.
The following observations were made when the model was towed with an orienta-
tion of 10 degrees to the model plane [18]:
A typical cross-flow VIV response occurred at a towing speed of 0.412m/s, where
out-of-plane vibration was observed at a frequency close to the model’s out-of-plane
natural frequency. This refers to the first mode. In-plane vibration (second mode)
occurred at a towing speed of 0.206m/s and close to the model’s in-line natural fre-
quency. Out-of-plane twist vibration (third mode) was observed at a towing speed
of 0.527m/s and at a frequency close to the jumper’s out-of-plane twist vibration
natural frequency.
For these three cases the response amplitudes were given for all accelerometers. Fur-
ther information about VIV responses for a wider range of reduced velocities were
available in plots for accelerometers three, five and seven.

Figure 2.8: Jumper model with locations of accelerometers, strain gauges and pipe
segments and the current directions. Taken from Liu et al. [5].

43
Chapter 3

Methodology

3.1 Simla analysis

Simla is a finite element method (FEM) based program for non-linear static and dy-
namic analysis developed at NTNU. It is specialized in simulations of pipe and cable
structures and operations. Step-by-step time integration is used to describe load his-
tories and analysis sequences. This section is based on the Simla Theory Manual [6].

Static analysis
Static analysis is performed based on the Newton-Raphson method. Equilibrium is
found at each load step. More generally, the Newton-Raphson algorithm solves x
for the problem f(x)=0 with the equation

f (xn )
xn+1 = xn − (3.1)
f 0 (xn )

where f 0 (xn ) is the first derivative of f(x) with respect to x at x = xn . This algorithm
can be used to find the relation between stress and strain increments. The iterative
equation to find the incremental stress-strain relationship can be written as [6]

∆rin+1 = K−1i i
T,n+1 ∆Rn+1 (3.2)

∆rn+1 = rn+1 − rn (3.3)

i
∆Rn+1 = R − Rint (3.4)

where
∆r is the incremental displacement vector
r is the displacement vector

44
 ∆R is the incremental load vector
R is the total load vector
Rint is the internal load vector
K−1
T is the inverse of the tangential stiffness matrix.
Subscript n+1 refers to the next load step and subscript n to the current one. Su-
perscript i denotes the iteration step.
The load increment is given from equilibrium at step one to equilibrium at step two
and results in a displacement increment. Stiffness matrix and internal load vector
must be updated and iterations repeated until the convergence requirement is met.

Dynamic analysis
Nonlinear problems in dynamic analysis cannot be solved by using modal superposi-
tion. Instead, direct time-integration of the equation of motion is applied. This can
be done by using either explicit or implicit methods. Explicit methods solve for the
displacement at the next time step based on information from the current and pre-
vious steps. Implicit methods depend on quantities at the next time step together
with information from the current step. Since they solve for the displacement at the
next step by using information at this allocated step, implicit methods have higher
numerical stability. The dynamic solution in Simla is found using the implicit HHT-
α method in a time integration scheme. It solves the following modified equilibrium
equation [6]:

I
Mr̈k+1 + (1 + α)Cṙk+1 − αCṙk + (1 + α)Rk+1 − αRkI = (1 + α)Rk+1
E
− αRkE (3.5)

M is the mass matrix

C is the damping matrix
RI is the internal load vector
RE is the external load vector
r, ṙ, r̈ are the displacement vector, the velocity vector and the acceleration vector,
respectively. Subscript k+1 refers to the next time step and k to the current one.
The velocity and acceleration at the next time step are found by using the same
equations as for the Newmark-β method:

1 1 1
∆r̈k+1 = r̈k+1 − r̈k = 2
∆rk+1 − ṙk − r̈k (3.6)
∆t β ∆tβ 2β

!
γ γ γ
∆ṙk+1 = ṙk+1 − ṙk = ∆rk+1 − ṙk − ∆t − 1 r̈k (3.7)
∆tβ β 2β
∆t is the increase in time and α, β and γ are parameters of the HHT-α method. If
α is set equal to zero, the method coincides with the Newmark-β family.

45
Since time integration does not generally lead to fulfillment of the equilibrium equa-
tion (Equation 3.5), iterations are necessary before the next time step. The iteration
procedure can be formulated by a Newton-Raphson scheme [6] as:
h i  
I,i
K̂ik δri+1 E i i E I i
k+1 = (1 + α) Rk+1 − Rk+1 − Cṙk+1 − Mr̈k+1 − α Rk − Rk − Cṙk (3.8)

K̂ is the effective stiffness matrix, calculated from below equation.

K̂k = a0 M + c0 C + b0 KT,k (3.9)

KT is the tangential stiffness matrix that should be updated at each iteration cycle
to improve the convergence rate. The parameters a0 , b0 and c0 are determined as
follows [6]:
1 α1 γ
a0 = 2
+ (1 + α)
∆t β ∆tβ
γ
c0 = (1 + α) (3.10)
∆tβ
α2 γ
b0 = (1 + α)
∆tβ
The right-hand side of Equation 3.8 represents the unbalance in inertia, damping
and internal forces and vanishes when equilibrium is obtained.
In dynamic analysis, high-frequency modes are not usually of interest, while lower
modes are described with high accuracy. An advantage of the HHT-α method is
that it damps out high frequencies without losing second order accuracy [6].

Element loads, mass and damping

Mass matrices can be lumped or consistent in Simla. Both lumped damping and
Raleigh damping are applied on element level. Hydrodynamic loads are established
by the following form of the Morison equation [6]:

π 2 1 q
ρ D Cm,t üy = ρCd,t (ẇx − u̇x )2 (ẇx − u̇x ) (3.11)
4 2

π π 1 q
ρ D2 (Cm,n − 1) üy = ρ D2 ẅy + ρCd,n (ẇy − u̇y )2 + (ẇz − u̇z )2 (ẇy − u̇y )
4 4 2
(3.12)

π π 1 q
ρ D2 (Cm,n − 1) üz = ρ D2 ẅz + ρCd,n (ẇy − u̇y )2 + (ẇz − u̇z )2 (ẇz − u̇z )
4 4 2
(3.13)
where

46
 Cm,t is the tangential added mass coefficient
Cm,n is the normal added mass coefficient
Cd,t is the tangential drag coefficient
Cd,n is the normal drag coefficient
ux − uz describe the body displacements
wx − wz describe the fluid particle displacement in the respective degree of free-
dom.

3.1.1 Pipe formulation

This section is based on the Simla Theory Manual [6] and the paper about elbow-
elements by Bathe [19]. Pipe elements in Simla are formulated as two-node beam
elements with twelve regular degrees of freedom (DOF). The orientation and motion
is referred to a global coordinate system. For torsion rotation and axial displace-

Figure 3.1: Dof of pipe element. Taken from Sævik [6].

ment linear interpolation is applied, while cubic interpolation is used for transverse
directions. The same interpolations are used to establish the mass matrix.
Pipe31 is an elastic pipe element, for which plane stress is assumed. All DOF are
uncoupled from hoop stresses and hoop strains, since they are known from shell
theory.
Pipe34 is a bend element type implemented with reference to Bathe [19] and as de-
scribed in section 2.6.1. It is able to account for deformation of the cross-section and
kinematic non-linearity. For the jumper model, deformation of the cross-section is a
result from bending an initially straight pipe during manufacturing. In contrast to
Bathe’s bend elements, Pipe34 is only linear-elastic. The ovalization of an initially
circular cross-section is included in the strain terms in Equations 2.32 to 2.34. Six

47
ovalization DOF are introduced in addition to the twelve regular DOF. For thin-
walled pipes, 48 additional integration points are used around the circumference,
while no additional integration points through the thickness are required for D/t-
ratios between 15 and 45. The model which includes Pipe34 elements is expected
to have lower stiffness in the bend parts.

3.1.2 Eigenvalue analysis

Eigenvalues and their associated frequencies are found from solving the general
eigenvalue problem [20]
 
M0 − ωi2 K0 Φi = 0 (3.14)

where Φi is the eigenvector of mode i and ωi is the associated circular frequency.

The mass matrix M0 includes added mass, for which the added mass coefficient for
still-water conditions is specified in the input file. The eigenvalue λ of mode i has
the following relation to the circular frequency:

λi = ωi2 (3.15)

In Simla, the required number of eigenvalues, starting with the smallest one, is found
from a Lanczos solver. The natural frequencies are obtained from the eigenvalue
calculations with the following relation.
√
λi
fn,i = (3.16)
2π

The number of relevant modes for this study was based on the maximum oscillation
frequency that occurred during VIV simulations.

3.1.3 Modal stresses

Information about axial strain, torsion and the curvature in both transverse direc-
tions could be extracted from the output file from eigenvalue analysis for each mode
shape. Data was given for each element in its local coordinate system and scaled
with respect to the maximum modal displacement equal to one. To obtain unit
stresses, the information had to be multiplied with the outer diameter. Accord-
ing to the guidelines, total flexural stress was calculated at 16 points around the
circumference with the equation

σxx = −κy ER cos θi + κz ER sin θi (3.17)

48
where the curvature κ was scaled with the outer diameter before and E is the
Young’s modulus. In this and all following equations the radius R denotes the outer
radius. The angle θi was specified for 16 points around the circumference. The local
coordinate system is also shown in Figure 3.2. The maximum principal stress was
then found with the equation:
s
2
σxx σxx

σ1 = + 2
+ τxy (3.18)
2 2

It is defined as the stress at a direction where only normal stress is acting and the
shear stress is zero. The stress state of a 2D element is illustrated in Figure 3.3.
The shear stress about the longitudinal pipe axis was calculated as

τxy = ϕ,x GR (3.19)

where the torsion ϕ,x was given in radians per meter and scaled with the outer
diameter before and G taken as the shear modulus.

Figure 3.2: Local coordinate system and bending moment definition for pipe element

3.1.4 TD VIV simulation

Both CF VIV and IL VIV can be analyzed in the same tool in Simla. The program
applies additional vortex-shedding terms to the Morison equation together with the
synchronization model as was described in section 2.3.
For CF VIV hydrodynamic forces normal to the cylinder are calculated with Mori-
son’s equation and the additional vortex shedding term acts perpendicular to the
normal component of the relative velocity [21].

πD2 πD2 1 1
Fn = C M ρ u̇n −(CM − 1) ρ ẍn + ρDCD |vn | vn + ρDCvc |vn | (j3 × vn ) cos φexc
4 4 2 2
(3.20)
CM is the inertia coefficient from the Morison equation

49
 Figure 3.3: Stress state of a 2D element. Taken from Liu et al. [5].

CD is the drag coefficient from the Morison equation

Cvc is a coefficient representing the magnitude of the CF vortex shedding force
u̇n is the fluid particle acceleration normal to the cylinder
ẍn is the cylinder acceleration normal to the cylinder
vn is the relative flow velocity normal to the cylinder, found as v = u̇ − ẋ.
The instantaneous phase φexc is linked to the dimensionless excitation frequency
[21],
dφexc 2π |vn | ˆ
= 2πfexc = fexc (3.21)
dt D
where the synchronization model comes into place.
For pure IL VIV hydrodynamic forces are calculated from the Morison equation,
including two potential vortex shedding force terms [21].

πD2 πD2 1
Fn =CM ρ u̇n − (CM − 1) ρ ẍn + ρDCD |vn | vn +
4 4 2 (3.22)
1 1
ρDCvi1 |vn | vn cos φexi,1 + ρDCvi2 |vn | vn cos φexi,2
2 2

Cvi1 is a dimensionless coefficient representing the magnitude of the IL vortex

shedding force for the first instability region
Cvi2 is a dimensionless coefficient representing the magnitude of the IL vortex
shedding force for the second instability region
φexi,1 is the instantaneous phase of the first vortex shedding force (referred to as
region one in literature)

50
 φexi,2 is the instantaneous phase of the second vortex shedding force (referred to
as region two).
Other parameters have the same definition as in Equation 3.20. Synchronization
works in a similar way to CF VIV. The first time derivative of the instantaneous
phase is linked to the phase difference between structural velocity in relative IL
direction and the force itself [21].

dφexi,j 2π |vn | ˆ
= 2πfexi,j = fexi,j (3.23)
dt D

The decision to switch between CF or IL forces is made based on comparison of

the CF and IL amplitudes. A synchronization model for CF-induced IL VIV is also
included. The applied synchronization model follows the equation [21]

dφexi dφexc
=2 [1 + α sin (φẋrel − φexi )] (3.24)
dt dt

where α is a CF-induced IL force synchronization parameter. The procedure is de-

scribed in more detail in the Simla user manual.
Furthermore, several parameters can be manipulated to control which VIV models
are activated for certain reduced velocity ranges.
To obtain the best fit with the model test data, a parameter study was conducted
on the Pipe31 model. Several parameters that can influence the VIV tool were
varied and the results plotted over reduced velocities. Later, the parameters that
gave the best fit were also applied on the Pipe34 model. The main reason to use
the Pipe31 model for the study was time-efficiency, since the Pipe34 element type
required much more computational time. The parameters are described in the input
section 4.2.1.

Mode identification
To compare results from the VIV prediction tool with experimental data, non-
dimensional response amplitudes were plotted against reduced velocities. Reduced
velocities were defined in Equation 2.3 and depend on the flow velocity, the outer
diameter and the natural frequency of the active mode. To find the active mode for
each load case, Fast Fourier Transformation (FTT) was applied to transfer the TD
signal into FD. From the location of the peak value, the oscillating frequency was
found. The active mode was then identified as the one with its natural frequency
closest to the structure’s oscillating frequency.

51
3.1.5 Stresses from VIV analysis
In the TD VIV prediction tool, forces and moments in all DOF can be obtained
directly. This has the advantage that no mode identification and combination of
modal stresses is needed and uncertainties are limited. Moments around the x, y
and z-axis were output for each sea state. They are given in the local coordinate
system for each element. To reduce computational time, the output was specified
for elements at critical locations. These are the jumper ends, the center point of the
top horizontal, and the bends. Since welds are not placed in the bends, but close
to it, the elements next to the bend at the longer of two adjacent straight parts
were chosen. As for modal stresses, the flexural stresses were evaluated at 16 points
around the jumper circumference.
Especially for larger current velocities, the stresses were fluctuating around a non-
zero mean. This was due to a constant displacement of certain jumper parts, usually
in flow-direction. Since constant stresses are not considered in fatigue calculations
and they were not induced by VIV, the stresses were scaled so that zero mean
stresses were obtained for all load cases.
Flexural stress was found from the moments around the y and z-axis.

My Mz
σxx = − R sin θi + R cos θi (3.25)
Iy Iz

The area moments of inertia, Iy and Iz , are calculated with the equation for thin-
walled circular cross-sections.

Iy = Iz = πR3 t (3.26)

The shear stress is constant around the circumference of each pipe section and
calculated with the equation
Mx
τxy = R (3.27)
Ix
where the polar moment of inertia Ix equals 2πR3 t. From the two stress signals, the
maximum principal stress was calculated with Equation 3.18 at each time step. To
the best of knowledge, the first principal stress criterion gives conservative results,
since the maximum stress is assumed to contribute to crack-opening regardless of
the real stress direction. When calculated at each time step, shifts between flexural
and torsional stress cycles should also be taken care of. The direction of principal
stress is a measure for how much torsional stress contributes to the total stress state.
According to the coordinate transformation shown in Figure 3.3, the principal stress

52
direction was found from the following equation.

1 2τxy
 
θσ1 = × arctan (3.28)
2 σxx

However, both signals were also observed individually for fatigue assessment.

3.1.6 Fatigue damage

Figure 3.4: Example of a stress histogram obtained as output from rainflow counting.

To transform the signal of oscillating stresses into a set of stress reversals with
constant amplitudes a rainflow counting algorithm was used. The basic principle
was explained in section 2.5.
For each load case, a set of stress ranges ∆Si with the corresponding number of cycles
ni during an observed time period were output. Obtained data can be illustrated
in a stress histogram as shown in Figure 3.4. To enter the SN-curve and calculate
damage from the Miner sum, an equivalent stress range was required. The concept
is to find a constant amplitude stress range ∆Seq that represents the combined effect
of all stress ranges in a real load history. This was found by the direct summation
method with the equation [3]
#1
Σi ni (∆Si )m
"
m
∆Seq = (3.29)
Σi ni

53
 where
ni is the number of cycles for one block in the histogram
∆Si is the stress range of the corresponding block
m is a constant of applied SN-curve.
Fatigue was calculated and compared for each load case under the assumption that
it is present over one year with a probability of occurrence of 100 percent. Con-
sequently, the number of cycles was adjusted. The SN-curve was entered with the
equivalent stress range and the number of cycles that would lead to failure was
obtained. By comparing this number with the amount of cycles per year obtained
from rainflow counting, fatigue damage could be calculated with the Miner sum, as
was explained in section 2.5. First, the procedure was applied on the flexural stress
and the torsional stress signals separately to observe the significance of torsional
stress. Later, the first principal stress was calculated at each time step and rainflow
counting was applied to the new signal. It is assumed that phase shifts between
torsional and flexural stress are taken care of in the rainflow algorithm, since they
are represented in the combined stress signal.

3.2 Abaqus analysis

Abaqus/CAE is a set of finite element analysis (FEA) applications from SIMULIATM by
Dessault Systèmes R . In this study, Abaqus/Standard was used, where several com-
mands from Abaqus/Aqua were implemented in the input file. Abaqus/Standard can
be used for linear and non-linear static and dynamic analysis. With Abaqus/Aqua
sea loads, waves and winds can be added. This was used to include coefficients for
added mass and drag in the eigenvalue analysis.

3.2.1 Element formulation

This section is based on the Abaqus documentation version 6.13 [22]. Pipe31 el-
ements in Abaqus are linear two-node beam element with six DOF at each node.
They are Timoshenko beams which allow for transverse shear deformation. Lin-
ear interpolation functions are used for displacements and rotations. Plane stress
theory is applied for both pipe elements and elbow elements. Hoop stresses and
hoop strains are not considered for standard pipe elements. They can, however, be
considered when elbow elements are used.
The purpose of elbow elements is to obtain accurate results for initially circular
pipes with distorted cross-section from bending during manufacture. They are ap-
plied for non-linear problems, when it is not sufficient to use flexibility factors to

54
correct results obtained from beam theory. Elbow elements in Abaqus appear like
beam elements, but are actually shell type elements, which allows for complex defor-
mation patterns. Because of the shell formulation, the number of possible DOF per
element is large. Accordingly, the number of integration points through the thick-
ness, the number of integration points around the pipe and the number of Fourier
modes should be specified in the input file. For an Elbow31 element this means that
linear interpolation is applied along its length, together with Fourier interpolation
around the pipe and shell theory to model the behavior. Without a given number
of Fourier modes, the elements become simple pipe elements with hoop stress and
hoop strain included. In the Abaqus documentation 18 integration points around
the pipe and six Fourier modes are suggested for thin-walled pipes [22]. Different
elbow element types can be selected in Abaqus. For two-node elements, Elbow31 is
the most complete one. In these elements, ovalization of the cross-section is contin-
uous from one element to another and warping is included.
Elbow31B are simplified elements without warping and for which ovalization is dis-
continuous between elements. Since warping is excluded, they are most similar to
Pipe34 elements in Simla.

3.2.2 Eigenvalue analysis

In this study, Abaqus was used to verify the results from modal analysis, in especially
with respect to the bend element types. Thus, the analysis procedure was kept
similar to that in Simla. Abaqus/Standard uses a set of implicit solvers for static
problems. For eigenvalue analysis a Lanczos solver was chosen.

3.2.3 Modal stresses

Various stress components can be specified for the Abaqus output. Since they are
commonly applied for fatigue analysis, maximum principal stresses were also output
here. They were defined in Equation 3.18 in the section about Simla analysis. A
more general form for first principal unit stress is
s
2
σflexural + σhoop σflexural − σhoop

2
σ1 = + + τshear (3.30)
2 2

where σflexural is bending stress and τshear is torsional stress. Hoop stresses, denoted
as σhoop were neglected in this study. This also led to a zero-mean stress level.
To obtain unit stresses, the output stresses were multiplied with

Dout
(3.31)
max(wi )

55
where wi is the displacement or rotation given for each of the six DOF.

3.3 DNV GL resonse model procedure

Empirical response models as proposed in DNVGL-RP-F105 (2019) [7] were used
to assess the response amplitudes for IL VIV and CF VIV. Based on the response
amplitude and the modal stresses obtained from FEA, combined stress ranges for
CF VIV and IL VIV including CF-induced IL VIV could be calculated.
This section is based on the recommended practice DNVGL-RP-F105.

3.3.1 Response models

For each flow angle the amplitude response curves were developed by a set of para-
metric equations. They are plotted as non-dimensional response over reduced veloc-
ities. The basic generation principle for the CF and IL response curves are shown
in Figure 3.5 and 3.6, respectively. It should be noted that the coordinate system
is different from the one used in Simla. Especially, the guideline uses index y for IL
direction and index z for CF direction.

CF VIV response
For CF VIV the amplitude response Az /D was calculated as a function of the
wave-current flow ratio α and the KC number. Ψproxi,onset and Ψtrench,onset are
reduction factors for onset CF due to seabed proximity and due to the effect of a
trench, respectively. fratio is the minimum ratio between natural frequencies of two
consecutive CF modes.

CF 3 · Ψproxi,onset · Ψtrench,onset
VR, onset = (3.32)
γon ,CF

 
CF
CF
7 − vR, onset

Az,1

VR,1 =7− · 1.3 − (3.33)
1.15 D

7 Az,1
   
CF CF
VR,2 = VR, end − · (3.34)
1.3 D

CF
VR, end = 16 (3.35)

56
 



 0.9 α > 0.8 fratio < 1.5



0.9 + 0.5 · (fratio − 1.5) α > 0.8 1.5 ≤ fratio ≤ 2.3









AZ,1
 
 1.3 α > 0.8 fratio < 2.3
= (3.36)
D 
 0.9 α ≤ 0.8 30 ≤ KC ≤ 40




0.7 + 0.01(KC − 10) α ≤ 0.8 10 ≤ KC ≤ 30








0.7 α ≤ 0.8 KC < 10



AZ,2 AZ,1
   
= (3.37)
D D
It should be pointed out that for pure current conditions the maximum amplitude
of the CF response curve depends on the ratio of natural frequencies for the first
three CF modes. These are taken from FEA.

IL VIV response
In current dominated conditions, IL VIV is associated with either alternating or
symmetric vortex shedding. The IL response model includes contributions from the
first and the second instability region. Parametric equations for the response curve
are given below.

1
for Ksd < 0.4


γon,IL




IL 0.6+Ksd
VR, onset =  γon,IL
for 0.4 ≤ Ksd < 1.6 (3.38)
 
2.2



γon,IL
for Ksd ≥ 1.6

IL Ay,1 IL
VR,1 = 10 · + VR,onset (3.39)
D

IL IL Ay,2
VR,2 = VR,end −2· (3.40)
D

IL

 4.5 − 0.8Ksd for Ksd < 1.0
VR,end = (3.41)
 3.7 for Ksd ≥ 1.0

Ay,1 Ksd Ay,2

   
= max 0.18 1 − · RIθ,1 ; (3.42)
D 1.2 D

Ay,2 Ksd
 
= 0.13 1 − · RIθ,2 (3.43)
D 1.8
Here, the amplitude response Ay /D is a function of the design stability parameter
Ksd and of the reduction factors RIθ,1 and RIθ,2 that account for the effect of turbu-

57
lence intensity and for the angle of attack. The design stability parameter includes
the stability parameter, as was defined in equation 2.25, and a safety factor. The
reduction factors are calculated as

π √
 
RIθ,1 = 1 − π 2 − 2 · θrel (Ic − 0.03) 0 ≤ RIθ,1 ≤ 1 (3.44)
2

(Ic − 0.03)
RIθ,2 = 1.0 − 0 ≤ RIθ,2 ≤ 1 (3.45)
0.17
with the turbulence intensity factor Ic and the relative flow angle θrel .

Figure 3.5: Generation principle for CF response curve. Taken from DNVGL-RP-
F105 [7]

Figure 3.6: Generation principle for IL response curve. Taken from DNVGL-RP-
F105 [7]

58
3.3.2 Stresses from VIV response models
In the DNV GL response model procedure, stresses arising from VIV are calcu-
lated from modal stresses, found from FEA, and the response model curves. Modal
stresses were calculated for all relevant modes. Whether a mode is active and ap-
plies its stress range depends on the reduced velocity of each mode at a given current
velocity and the associated response model curve. As mentioned above, all active
modes are dominating modes for non-straight pipelines. Thus, only the procedure
for dominating modes is covered in this section.

Stresses from CF VIV

The CF VIV-induced stress range is found from the equation

Az
 
SCF,i (x) = 2 × ACF,i (x) × × Rk × γs (3.46)
D i

where
ACF,i (x) is the unit diameter amplitude stress of CF-mode i at location x along
the jumper
 
Az
D i
is the normalized response amplitude according to the CF response model
Rk is a reduction factor to cover the effect of damping
γs is the safety factor on stress amplitudes.
For a given sea state, the combined stress range from all active modes is calculated
from the equation v
um
uX
Scomb,CF (x) = t (SCF,j (x))2 (3.47)
j=1

where m is the number of active CF modes.

To find the correct cycle-counting frequency for CF VIV, a correction of the added
mass coefficient is required. This is due to variations of added mass when the
structure oscillates. The corrected added mass coefficient Ca,CF −RES is taken from
Figure 3.7 from the guidelines and applied to correct the CF response frequency
fCF −RES,j with the equation
s
sg + C a
fCF −RES,j = fCF,j (3.48)
sg + Ca,CF −RES

where
fCF,j is the CF response frequency without correction
sg is the specific gravity of the pipe, calculated as (q + b)/b, where q is the
submerged weight and b is the buoyancy
Ca is the still-water added mass coefficient and

59
Figure 3.7: Correction of added mass coefficient to account for the difference of VIV
added mass and still water added mass coefficient. Taken from DNVGL-RP-F105
[7].

Ca,CF −RES is the added mass coefficient due to CF response.

It should be noted that the correction curve has the effect of still-water added mass
already built into it.
The CF cycle-counting frequency is given by the following equation.
v
u
m
!2
uX SCF,j (x)
fcyc,CF (x) = fCF −RES,j · (3.49)
u
t
j=1 Scomb,CF (x)

Stresses from IL VIV and CF-induced IL VIV

For IL VIV, a preliminary stress range for pure IL VIV is calculated using the
following equation, before comparison with CF-induced IL VIV is applied.

Ay
 
P
SIL,j (x) = 2 × AIL,j (x) × × ψα,IL × γs (3.50)
D j

AIL,j (x) is the unit diameter amplitude stress of IL-mode j at location x along
the jumper
 
Ay
D j
is the normalized response amplitude according to the IL response model
ψα,IL is a correction factor for the current flow ratio α.
CF-induced IL VIV can arise from the dominating CF-mode. Accordingly, for non-
straight pipelines all acting CF-modes must be considered. The candidate for CF-
induced IL VIV is the IL mode with its eigenfrequency closest to twice the oscillating

60
frequency of the CF-mode with the largest response amplitude. This candidate is
found from the minimum of the expression

part
|fIL,k − 2 · fCF −RES,i | (3.51)

part
where fIL,k is the natural frequency of the k-th participating IL-mode with k ∈
{1, 2, . . . , n}. The stresses due to CF-induced IL VIV are calculated from the equa-
tion
Az
 
SIL−CF (x) = 0.8 × AIL,k (x) × × Rk × γs (3.52)
D max
where
AIL,k (x) is the unit diameter amplitude stress of the selected candidate
 
Az
D max
is the maximum normalized response amplitude from the CF response
model
Rk is the reduction factor to cover the effect of damping as applied in
Equation 3.46.
If the candidate mode for CF-induced IL VIV is an active IL mode, the stress from
IL VIV is taken as the maximum of stresses due to CF-induced IL VIV and pure
IL VIV. If the candidate is not among the active modes, it is added as a new active
mode with its stress range. The combined IL stress range is then calculated as [7]
v
umaug
(SIL,j (x))2
uX
Scomb,IL (x) = t (3.53)
j=1

where
SIL,j is the response stress range due to the j-th contributing IL mode
maug is equal to the number of contributing modes m if the CF-induced IL mode
is among these. Otherwise, it is equal to m + 1.
The IL cycle-counting frequency is given by the equation [7]
v
um
aug
!2
uX SIL,j (x)
fcyc,IL (x) = fIL,j · (3.54)
u
t
j=1 Scomb,IL (x)

where fIL,j is the natural frequency of the j-th contributing IL mode. For the CF-
induced IL mode this is set to 2 · fCF −RES,i .

3.3.3 Fatigue damage

From the DNV GL procedure two combined stress ranges, Scomb,CF and Scomb,IL , were
obtained that represent all stresses for a specific load case. The number of cycles
per year was obtained from the corresponding cycle-counting frequency. With the

61
combined stress ranges the SN-curve could be entered and the number of cycles to
failure was found. The damage from CF VIV and IL VIV could simply be added to
a total damage.

3.4 Limitations
The work-flow of this thesis work is shown in the figure below. Simplifications were
made when a case scenario was developed from a realistic full-scale scenario. Es-
pecially was the model initialized in a stress-free condition. Further, no internal
pressure and only small external pressure from the hydrostatic condition one meter
below surface were applied. Positive internal pressure would have a stiffening effect
on the structure, while larger external pressure would soften it. These effects were
not taken into account for the jumper model to be in agreement with the experi-
mental setup. However, the stiffening and softening effects, also with respect to the
modeling choice between elbow elements and regular pipe elements for the bends,
were observed from a single bend model with realistic full-scale dimensions.
In modal analysis special focus was put on modeling flexibility of the bends cor-
rectly by utilizing the Pipe34 and Elbow31B element types. However, also regular
pipe elements have different stiffness properties in Simla and Abaqus with respect
to transverse shear. Thus, some deviation in results was accepted.
For VIV analysis and the resulting fatigue damage two procedures were applied.
In Simla, VIV motion and moments were obtained directly as a time signal. For
VIV motion, experimental data at two locations along the jumper could be used for
comparison to ensure that the Simla results are in a conservative range. For stresses,
no experimental data was available. To evaluate the fit of simulation results and
adjust parameters in the VIV tool in Simla, the experimental data was considered
precise. For the ten degree flow, information was given about active modes and
shifts in vibration patterns. Such information was not made available for the other
two flow directions.
Since both flexural and torsional stresses can have an important contribution to fa-
tigue damage, the total damage was calculated from the first principal stress. This
was calculated at each time step of the signal before rainflow counting was applied.
It is assumed that the maximum principal stress is conservative for tensile crack-
growth, since stresses contribute to crack-growth with their full range regardless
of their direction. However, this method can be less precise than other criteria for
multi-axial fatigue. Especially when torsional stresses are very large, the main driver
for fatigue is shear crack growth, which behaves different from crack growth due to
tensile load. In this study, the first principal stress approach is used, but limitations
are discussed in the results.

62
 Figure 3.8: Flowchart of work steps.

The second procedure for VIV assessment is the DNV GL response model approach,
which utilizes the maximum principal stress from modal analysis and scales it with
response amplitudes from the guidelines itself. Modal analysis assumes small dis-
placements, thus, limitations arise when it is applied for non-linear structural be-
havior. Further, the DNV GL procedure was adopted from the approach for straight
pipes with several modifications. Thus, less precise, but conservative results must
be expected here. Nonetheless, it was used to compare the Simla results for fatigue.

63
Chapter 4

Input for case-study

4.1 Case scenario

The case scenario was defined according to the experimental setup of the ExxonMobil
model test to obtain comparable results. With a scale ratio of 4.525 an arch length
of 13.96m is obtained. The jumper geometry was modeled using the dimensions
given in Table 4.1 and Table 4.2. The model consisted of aluminum. As in the
test, the highest point of the model was placed one meter below sea surface and the
seawater density was 1000kg/m3 .

4.1.1 Jumper models and element types

The bends were modeled with a bending radius defined as three times the outer
diameter of the pipe, which resulted in a bending radius of 0.1815 meters and an
arch length of 0.2851 meters for each bend. The number of elements was chosen
in such way that the element length in bent pipe parts did not exceed the outer
diameter size. Five elements per bend were used in the simulation, which led to an
element length of 0.057 meters. For the straight parts the element length was varied
between once and twice the outer diameter. During modal analysis in Simla it was
found that an element length of twice the outer diameter and accordingly twice the
element length from the bent parts was sufficient.
In Simla, two models were used and compared, of which one consisted completely
of Pipe31 elements, also referred to as ’Pipe31 model’. For the other model Pipe31
elements were used for the straight parts and Pipe34 elements for the pipe bends.
This model will later be referred to as ’Pipe34 model’.
In Abaqus, two models were used to compare eigenvalue results from Simla and
obtain modal principal stresses. One model consisted of Pipe31 elements, while the
other one was modeled with Elbow31B elements for the bends. As mentioned in
section 3.2.1, Elbow31B elements do not include warping and have most similar

64
properties compared to Pipe34 elements. As recommended in the Abaqus docu-
mentation, 18 additional integration points around the circumference and 6 Fourier
modes were specified for ovalization.
Linear elastic material definition was applied for all models. The drag coefficient
applied in the Morison equation was initially chosen according to Wang et al. [18],
who had reported a normal drag coefficient of 1.1 from the model test. Since in
Simla a value of 1.2 is recommended for the drag coefficient, simulations for VIV
were later carried out for both values. The normal added mass coefficient was set
to 1.0, since it is a typical value for structures with circular cross-section. Both the
tangential drag and the tangential mass coefficients were set to zero.

Table 4.1: Segment lengths of the jumper model

Segment name Length [m]

Vertical 1 1.495
Horizontal 1 1.000
Vertical 2 2.323
Horizontal 2 4.327
Vertical 3 2.326
Horizontal 3 1.000
Vertical 4 1.495

Table 4.2: Jumper model properties

Parameter Value Unit

Pipe density 2700 kg/m3
Outer diameter 0.0605 m
Inner diameter 0.055 m
Wall thickness 2.77 × 10−3 m
Elastic modulus (E) 6.90 × 1010 N/m2
Bending stiffness (EI) 1.44 × 104 N/m2
Shear modulus (G) 2.61 × 1010 N/m2
Torsion stiffness (GJ) 1.08 × 104 N/m2

65
4.1.2 Model setup
In Simla, the model was set up in a stress-free configuration by placing a straight
pipe in a vertical position in the sea and applying initial curvature on the parts to
form 90-degree bends. In a second step, gravity and external hydrostatic pressure
were added. Fixed boundary conditions were added on both ends of the jumper.
In Abaqus, the geometry could be modeled as in its final state after bending during
manufacture. Gravity was added and the model was placed in water to add the
added mass coefficient and the drag coefficient.
In both tools the weight of the internal fluid was accounted for in the total unit
mass. By this means, the dry mass consisted of two components, the structural
weight and the weight of content. Both are considered constant along the pipe. No
internal pressure or free-surface effects inside the pipe were considered.

4.2 VIV simulation in Simla

VIV simulations were run for at least 100 seconds per load case. During five seconds
after initiation of the model setup the current velocity was smoothly increased from
zero to the required value. This was done by ramping on the load in form of a half
cosine function, starting at zero slope [21]. At the end of a simulation, the load was
ramped off again during five seconds.
Simulations were run for flow angles of ten and 45 degrees with current velocities
of 0.05 to 0.98 m/s, where the velocity was increased in steps of 0.05 m/s. For a
flow angle of 90 degrees simulations were run up to a current velocity of 0.79 m/s,
according to available model test data. Non-dimensional responses were plotted over
reduced velocities. With these data the VIV response from Simla could be compared
with model test results and with the DNV GL response models.

4.2.1 Parameter study

Table 4.3: Parameters for the parameter study on VIV prediction

Parameter Default value Observed values Definition

CDrag 1.2 1.1, 1.2 Drag coefficient
CV 1.0 1.0, 1.1, 1.2 CF lift force coefficient
CV I1 0.8 0.8, 0.9, 1.0, 1.1, 1.2 IL force coefficient for region one
fˆ1I1 0.3 0.1, 0.13, 0.3 Minimum non-dimensional vibra-
tion frequency that gives energy
input for IL region one

66
 In a parameter study the influence on VIV prediction of the drag coefficient and
several control parameters of the VIV tool were observed. In the first part of this
study, the results were compared with the model test data for an incoming flow
angle of ten degrees. After a first selection of parameters was decided on, these were
tested for 45 and 90 degree flow angles and adjusted. The parameters are listed and
explained in Table 4.2.1. They were observed with the goal to obtain parameters
that give conservative results for all flow directions.
In the literature study Section 2.3 it was mentioned that another force coefficient
can be defined for the second IL instability region. This coefficient was by default
set to zero, since in the present model the first IL instability region can be excited
over a wider range of reduced velocities.

4.3 DNV GL response model approach

The response models were calculated as described in section 3.3. For the purpose of
comparability with the Simla results, all applied safety factors were here set to 1.0.
Usually, the values are chosen according to the guidelines and as shown in Figure
4.1. They are defined as
γk - safety factor on stability parameter
γs - safety factor on stress amplitude
γon - safety factor on onset value for CF or IL reduced velocity
γf - safety factor on natural frequency.
Further, seabed proximity and the effect of a trench were neglected. The turbulence
intensity factor Ic was set to 0.05 as recommended when no other information is
available.
For pure current conditions, the flow ratio α is equal to one. The vortex-shedding
frequency is steady and the KC-terms of the CF response model can be neglected.
For the calculation of the stability parameter, the effective mass per unit length was
required, which consists of structural mass, added mass and mass of internal fluid.
The mass components are listed in Table 4.4. Since the effect of added mass is taken
into account with the added mass coefficient of 1.0 and the other two components are
constant along the jumper length, the effective mass is constant, too. A correction
for the deviation of the added mass during VIV and the still water added mass had
to be applied on the CF response model, when the cycle-counting frequencies were
calculated.

67
Figure 4.1: Safety factors for the DNV GL response model procedure. Taken from
DNVGL-RP-F105 [7].

Table 4.4: Unit mass components of the jumper model

Parameter Value Explanation

mpipe 1.347 kg/m
mcontent 5.532 kg/m
mdry 6.879 kg/m mpipe +mcontent
madded 2.875 kg/m with Ca = 1.0
meff 9.754 kg/m mpipe +mcontent + madded

68
4.4 Fatigue life
The SN-curve was chosen according to the guideline DNVGL-RP-C203 (2005). For
pipelines with single side or double side welds, curve D is recommended. SN-curves
for structures in seawater with cathodic protection are shown in Figure 4.2 with as-
sociated parameters shown in Figure 4.3. Curve D changes slope at 1.00E+06 cycles,
where the equivalent stress is approximately 83 mega-pascal. If the equivalent stress
of a load case lies below this value, it is over-conservative to apply the parameters
from the left side of the curve with a higher slope. Instead, the parameters from
the right side of the curve should be chosen, which are still conservative for larger
stresses, but not over-conservative for smaller equivalent stresses.
Thus, the parameters m and log ā were chosen with 5.0 and 15.606, respectively.
Correction with the thickness exponent was not applied, since the wall-thickness
of the jumper model is smaller than the reference thickness and it is not allowed
to take advantage from thickness correction. For fatigue of welds, more detailed
analysis with respect to hot-spot stresses would be required. However, in this study
the SN data is mainly used to compare fatigue results and no hot-spot correction
was applied.

Figure 4.2: SN-curves in seawater with cathodic protection. Taken from DNVGL-
RP-C203 [8].

69
Figure 4.3: Parameters of SN-curves in seawater with cathodic protection. Taken
from DNVGL-RP-C203 [8].

4.5 Diameter-over-thickness study with internal

and external pressure
It was assumed that the use of elbow elements in bends will have large effects for very
thin-walled pipes, while the effect should vanish for relatively thick-walled pipes. A
common definition is that pipes with a diameter-over-thickness (D/t) ratio of ten or
larger are thin-walled, while with a ratio lower than ten they are thick-walled.
To further explore the effect of the modeling choice between regular pipe elements
and elbow elements, a single bend section was modeled with constant outer diameter
and varied wall-thickness, so that D/t-ratios between five and 45 were obtained.
A typical scenario with realistic pipe dimensions and steel material was used. The
properties are given in Table 4.5 and the model is shown in Figure 4.4. Due to
symmetry, only half of the 90-degree bend section was modeled with a radius of
2.25 meters and adjacent straight parts of 2.25 meters length. The element size
was set equal to the outer diameter. On the left end translation in x-direction and
rotation around the y-axis were restricted. The model was free to move in vertical
direction and to deform its cross-section. On the right tip translation in both y and
z-direction and rotation around the x- and z-axis were restricted, while translation
in x-direction was possible. Here, a rotation of five degrees around the y-axis was
applied which led to a constant bending moment.
Flexibility factors were measured at the intersection between bend and straight part
using the following definition.

(My /θy )Pipe31

FF = (4.1)
(My /θy )elbow

First, the behavior without internal or external pressure was observed. In a second
and third step, internal and external pressure were included separately. The mag-
nitudes are shown in Table 4.6 and were chosen corresponding to 80 percent of the
pipe capacity with respect to elastic buckling pressure in hoop direction and yield
stress. The simulations were carried out in both Simla and Abaqus.

70
 Table 4.5: Properties of full-scale bend model

Pipe density [kg/m3 ] 7850

Outer diameter [m] 0.45
Elastic modulus [N/m2 ] 2.1 × 1011
Shear modulus [N/m2 ] 8.08 × 1010

D/t=15 D/t=30 D/t=45

Inner diameter [m] 0.39 0.42 0.43
Wall thickness [m] 0.03 0.015 0.01
Bending stiffness [N/m2 ] 1.83 × 108 1.02 × 108 7.02 × 107
Torsion stiffness [N/m2 ] 1.41 × 108 7.83 × 107 5.40 × 107

Table 4.6: Load cases for full-scale bend model

D/t=15 D/t=30 D/t=45

Internal pressure [N/m2 ] 51.4 × 106 24.82 × 106 16.36 × 106
External pressure [N/m2 ] 39.2 × 106 10.55 × 106 2.65 × 106

Figure 4.4: Half of the 90◦ bend model. Left hand side un-deformed, right hand
side after rotation was applied on the right end.

71
Chapter 5

Results

5.1 Eigenvalue analysis

In a previous project work it was observed that the eigenfrequencies of the Pipe31
model in Simla are in very good correlation with the results from other numerical
studies. The Pipe34 model in Simla, on the other hand, shows some differences,
but is in better agreement with the mode shape order from the ExxonMobil model
test reported by Wang et al. [18]. To verify the use of elbow elements and the
observations from eigenvalue analysis in Simla, the two Simla models were compared
to the two Abaqus models.

5.1.1 Pipe31 model in Simla and Abaqus

The eigenvalue results for the two models consisting completely of Pipe31 elements
are in good correlation with each other, as shown in Figure 5.1. Relative errors of
the eigenvalue results between the two Pipe31 models are plotted in Figure 5.2.
The maximum error occurs for mode one with about nine percent which is compa-
rably high. For all other modes the error lies below 3.5 percent. The differences
in results can be due to the fact that Pipe31 elements in Abaqus are Timoshenko
beams which allow for transverse shear deformation. In Simla, on the contrary, the
cross-section of Pipe31 elements remains plane and does not distort due to trans-
verse shear. An overview of the first nine modes with mode shapes is also given
in Table 5.1. Mode shapes are identical for the two Pipe31 models. Modes three,
four, six and nine are in-plane bending modes, of which modes three and six have
their largest displacement in x-direction and modes four and nine in z-direction. For
these four modes the relative error is very small. Modes one, five, seven and eight
are out-of-plane bending modes and mode two is an out-of-plane twist mode.
The eigenfrequencies of modes two and three are very close to each other. This gives
a reason why they appear in switched order in the simulation results to how they

72
 Figure 5.1: Comparison of natural frequencies in Simla and Abaqus

Figure 5.2: Relative error of natural frequencies obtained for the two Pipe31 models
in Simla and in Abaqus

were reported from the ExxonMobil model test. Also, the eigenfrequencies of modes
six and seven are very close to each other. However, information about the shapes
reported from the model test were available only for the first three modes.

73
 Table 5.1: Eigenfrequencies and mode shapes of Pipe31 models

Mode fn in Simla [Hz] fn in Abaqus [Hz] Mode shape

1 0.8377 0.8744 out-of-plane bending in y-direction
with largest displacement at top hor-
izontal
2 2.1991 2.2343 out-of-plane twist around center
point of top horizontal
3 2.2859 2.2844 in-plane bending in x-direction
4 2.6285 2.6341 in-plane bending in z-direction
5 3.3800 3.4094 symmetric out-of-plane bending in
y-direction with largest displace-
ment at the two bottom horizontals
6 3.7261 3.7294 diagonal distortion in both x and z-
direction
7 3.7646 3.7949 asymmetric out-of-plane bending in
y-direction with largest displace-
ment at the two bottom horizontals
8 6.3413 6.3851 buckling of the top horizontal in y-
direction
9 7.1614 7.1678 buckling of the top horizontal in z-
direction

5.1.2 Pipe34 model in Simla and Elbow31B model in Abaqus

Eigenfrequencies for the Pipe34 model in Simla and the Elbow31B model in Abaqus
are generally lower than the ones from the Pipe31 models, also shown in Figure
5.1. Moreover, mode two and three are switched in order for the elbow models
as shown in Table 5.2, so that in-plane bending occurs at a lower frequency than
out-of-plane twist. This is in correlation with the observations from the model test.
Mode shapes of the two elbow element models are identical. The maximum relative
error of eigenvalue results again occurs for mode one with 6.6 percent, as plotted in
Figure 5.3. For the other modes, the error is below three percent which is a good
correlation. Both Pipe34 and Elbow31B elements are beam elements for which shell
theory is applied to model the deformation of the cross-section. The classical von
Kármán approach is implemented to account for ovalization from pure bending.
Since the two elements are very similar, the derogation of results, especially for the
first mode, can also be attributed to the differences of Pipe31 elements to which the
elbow elements are connected.
It should be pointed out that the shift in mode shapes could not be obtained with

74
models consisting only of Pipe31 elements, even with a refined mesh. This gives
a relevant reason to use elbow elements in the analysis of subsea jumpers. When
following the DNV GL response model procedure, FEA modal results are utilized
for fatigue analysis, too. Thus, it is of great importance to obtain the correct order
of mode shapes and correct modal frequencies. The mode shapes are also shown in
Figures 5.4 to 5.8.

Figure 5.3: Relative error of eigenvalue results obtained for the two elbow element
models in Simla and in Abaqus

Figure 5.4: Mode one of the Pipe34 model. Out-of-plane bending of the top hori-
zontal in y-direction.

75
Table 5.2: Eigenfrequencies and mode shapes of Pipe34 model (Simla) and El-
bow31B model (Abaqus)

Mode fn in Simla [Hz] fn in Abaqus [Hz] Mode shape

1 0.8076 0.8336 out-of-plane bending in y-direction
with largest displacement at top hor-
izontal
2 1.9131 1.9406 in-plane bending in x-direction
3 2.1047 2.1344 out-of-plane twist around center
point of top horizontal
4 2.4522 2.4689 in-plane bending in z-direction
5 3.3357 3.3568 symmetric out-of-plane bending in
y-direction with largest displace-
ment at the two bottom horizontals
6 3.4739 3.4949 diagonal distortion in both x and z-
direction
7 3.7181 3.7389 asymmetric out-of-plane bending in
y-direction with largest displace-
ment at the two bottom horizontals
8 6.2740 6.2998 buckling of the top horizontal in y-
direction
9 6.6148 6.6539 buckling of the top horizontal in z-
direction

Figure 5.5: Mode two of the Pipe34 model. In-plane bending in x-direction.

76
 Figure 5.6: Mode three of the Pipe34 model. Out-of-plane twist.

Figure 5.7: Mode four of the Pipe34 model. In-plane bending in z-direction.

Figure 5.8: Mode five of the Pipe34 model. Symmetric out-of-plane bending of the
two bottom segments in y-direction.

Figure 5.9: Mode six of the Pipe34 model. Diagonal distortion in both x and z-
direction.

77
Figure 5.10: Mode seven of the Pipe34 model. Asymmetric out-of-plane bending of
the two bottom segments in y-direction.

Figure 5.11: Mode eight of the Pipe34 model. Out-of-plane bending in y-direction.
Largest displacement occurs as buckling of the top horizontal.

Figure 5.12: Mode nine of the Pipe34 model. In-plane bending in z-direction.
Largest displacement occurs as buckling of the top horizontal.

78
5.2 Modal stresses
An overview of modal stresses for the first nine modes is given in Figures 5.13 to
5.18 where the maximum principal stress, the flexural stress and the shear stress is
shown for the Pipe31 and Pipe34 models in Simla. For the Pipe34 model resulting
stress in the bends is increased up to factor two compared with the Pipe31 model.
In the adjacent straight parts, on the other hand, the stresses are decreased by up to
35 percent. The coupled effect of decreased stresses in straight parts and increased
stresses in bends is most significant for the in-plane bending modes, which are shown
in Figures 5.19, 5.20, 5.21 and 5.22. Generally, largest principal stresses occurred at
the boundaries or in the midpoint of the top horizontal for the Pipe31 model, while
they occurred in bends for the Pipe34 model. The difference is due to the flexural
stress component. Torsional stresses have the same magnitude for both models.
The remaining plots that compare modal stresses, also with Abaqus results, can be
found in Appendix A due to the amount of data. When comparing the results for
the Pipe31 model in Simla and Abaqus as plotted in Figures 6 to 14 (Appendix
A), minor differences are observed for the first four modes. They might be due to
discontinuous curvature in Abaqus, while Pipe31 elements in Simla have continuous
curvature. Although the relative difference for the first four modes is between 2.4
and 7.5 percent, the absolute difference is no more than 2.5 mega-pascal, since these
modes generally have smaller stresses. For mode five and seven with large bending
of the shorter segments, larger deviations are observed with a relative difference of
about ten percent. The deviation for mode eight is most significant. Here, the long
top horizontal bends outwards with large displacement in y-direction. The relative
difference in results is 38.1 percent at the center point of the top horizontal. This
might be attributed to higher stiffness of the Timoshenko beam elements in Abaqus.
For all nine modes, Simla gives larger stresses than Abaqus. Thus, the Simla results
are assumed to be conservative.
For the Pipe34 and Elbow31B model larger difference in peak stress occurs in the
bends, as shown in Figures 15 to 23 (Appendix A). With deviations of 13 to 30
percent the derogation is quite large. An additional shell model should be used to
review which of the models is more reliable. Even though elbow elements theoreti-
cally behave like shells, they are connected to straight pipe elements which in turn
can cause imprecision. However, the Simla results are again larger than the Abaqus
results, hence, they are used for further analysis without correction.

79
Figure 5.13: Maximum principal unit stress of the first nine modes obtained from
the Simla Pipe31 model. Plotted over accumulated arch length and scaled with
respect to the outer diameter and maximum modal displacement.

Figure 5.14: Maximum principal unit stress of the first nine modes obtained from
the Simla Pipe34 model. Plotted over accumulated arch length and scaled with
respect to the outer diameter and maximum modal displacement.

80
Figure 5.15: Flexural unit stress of the first nine modes obtained from the Simla
Pipe31 model. Plotted over accumulated arch length and scaled with respect to the
outer diameter and maximum modal displacement.

Figure 5.16: Flexural unit stress of the first nine modes obtained from the Simla
Pipe34 model. Plotted over accumulated arch length and scaled with respect to the
outer diameter and maximum modal displacement.

81
Figure 5.17: Torsional unit stress of out-of-plane twist or bending modes obtained
from the Simla Pipe31 model. Plotted over accumulated arch length and scaled with
respect to the outer diameter and maximum modal displacement.

Figure 5.18: Torsional unit stress of out-of-plane twist or bending modes obtained
from the Simla Pipe34 model. Plotted over accumulated arch length and scaled with
respect to the outer diameter and maximum modal displacement.

82
Figure 5.19: Maximum principal unit stress of mode three for the Pipe31 model
and mode two for the Pipe34 model, which describe the same mode shape. Plotted
over accumulated arch length and scaled with respect to the outer diameter and
maximum modal displacement.

Figure 5.20: Maximum principal unit stress of mode four for the two Simla models.
Plotted over accumulated arch length and scaled with respect to the outer diameter
and maximum modal displacement.

83
Figure 5.21: Maximum principal unit stress of mode six for the two Simla models.
Plotted over accumulated arch length and scaled with respect to the outer diameter
and maximum modal displacement.

Figure 5.22: Maximum principal unit stress of mode nine for the two Simla models.
Plotted over accumulated arch length and scaled with respect to the outer diameter
and maximum modal displacement.

84
5.3 Diameter-over-thickness study with internal
and external pressure
The effect of elbow elements in the bends was further explored for different D/t-
ratios and including internal or external pressure. Flexibility factors (FF) at the
intersection between bend and straight part from simulations with full-scale dimen-
sions and steel material properties are shown in Figures 5.23 and 5.24.
The results from Simla and Abaqus are in good correlation with each other for the
zero pressure case, but the Elbow31B elements in Abaqus are less affected by inter-
nal and external pressure. The curves for the different load cases are only slightly
shifted for Elbow31B elements, while Pipe34 elements show increasing impact of
pressure for thinner models as was expected.
The jumper model has a D/t-ratio of 21.8 which corresponds to a FF of 1.8 without
pressure. This explains the stress amplification of modal stresses. With external
pressure included the FF obtained from Pipe34 elements would raise up to 2.25 for
the same D/t-ratio. With internal pressure the bends become stiffer and the FF for
the jumper model would decrease to 1.45.
The chosen values for internal pressure are realistic scenarios for pipes in operat-
ing condition. The chosen external pressure values, on the other hand, will not be
reached in a typical north-sea application due to comparably shallow water. Here,
internal pressure is usually much higher than external pressure. With a FF of 1.45
the effect of bend elements is still relevant, but lower compared to the case scenario
of the jumper model.

85
Figure 5.23: Flexibility factors obtained for Pipe34 elements in Simla with pressure
from Table 4.6

Figure 5.24: Flexibility factors obtained for Elbow31B elements in Abaqus with
pressure from Table 4.6

86
5.4 VIV analysis
5.4.1 Parameter study in Simla
A parameter study was made by comparing the simulation results to the Exxon-
Mobil model test results. In experimental data, peak values mainly occurred for
a very narrow range, e.g. for reduced velocities between six and eight for the ten
degree flow. In Simla, significant responses were usually spread over a wider range
of reduced velocities. With the parameters of the TD tool, especially with the non-
dimensional vibration frequencies that give energy input for specific regions, it is
possible to narrow down the response to a certain reduced velocity range. However,
this is problematic for complex geometries where the response is entirely different if
the angle of attack is changed.
Accordingly, all three flow directions for which experimental results were available
had to be observed in the parameter study. The parameters explained in section
4.2.1 were varied with the aim to apply the same values for each load case and obtain
conservative results. Sensitivity of fatigue estimates to mentioned parameters is also
discussed in Section 5.5.1.

CVI1
First, the IL force coefficient for IL region one, CVI1, was decided on. It mainly
affects the non-dimensional response amplitude of IL VIV and to a smaller extent
also the response amplitude in CF direction. The latter one might be due to cou-
pling between IL and CF VIV.
The biggest effect was seen for the ten degree flow, since amplitudes in x-direction,
which is associated with IL VIV here, are generally smaller than in y-direction for
the model geometry. When this coefficient was changed from its default value 0.8
to 1.2, the response amplitude in x-direction was increased by up to 53.3 percent
at its peak value. The response in y-direction was increased by up to 34.7 percent
in the reduced velocity range where IL modes were dominating. In the higher re-
duced velocity range, where CF modes were dominating, the response amplitude in
y-direction was increased by maximum 8.2 percent due to the change in CVI1.
The default value 0.8 of the IL force coefficient for region one in Simla was suggested
for free spanning pipelines. It was observed that for the jumper model this value was
not sufficient to capture the maximum IL response amplitudes for all flow directions.
For the ten degree flow direction and a drag coefficient of 1.1, conservative results
were obtained when CVI1 was set to 1.1 or 1.2. For a drag coefficient of 1.2 and the
same flow direction, conservative results were obtained only for CVI1 equal to 1.2.
Accordingly, it was set to 1.2 for all further analyses.

87
 fˆ1I1
The minimum non-dimensional vibration frequency that gives energy input for IL
region one had to be modified in such way that IL VIV could be activated in a
wider range of reduced velocity. Since it is a non-dimensional quantity, the reduced
velocity range where IL VIV can be activated is approximately the inverse of fˆ1I1 . It
is not exactly equal to this value, because reduced velocity was calculated with the
natural frequency that was found closest to the vibration frequency and not with
the vibration frequency itself. However, the default value in Simla referred to free
spans. For the jumper model, IL VIV is not limited to the low velocity range.
From experimental data, it was observed that large displacement in IL direction
could occur for reduced velocities of up to eight. This is in agreement with the
paper by Wang, Ji, Chi and Wu [23] who investigated the coupling of IL and CF
motion and reported that IL motion contributed to the path described by cylinder
motion for Vred = 8, while at Vred = 10 the IL motion was negligible.
Therefore, fˆ1I1 was set to 0.13. With this value conservative results were obtained
for all flow directions when the drag coefficient was kept at 1.1.

CDrag
The drag coefficient is represented in the excitation force in Simla. A strength of
this TD prediction tool is that it operates in local systems and automatically ob-
tains the drag amplification during VIV. The drag force oscillates with twice the
vortex shedding frequency. However, it is crucial to choose an appropriate value for
the still-water drag coefficient, since amplification is applied on this. A lower drag
coefficient led to larger VIV response amplitudes for IL VIV and often also for CF
VIV due to coupling. Therefore, the effect of changes in the drag coefficient is in the
same range for all load cases. The value of 1.1 was chosen after Wang et al. [18] who
had reported this from the model test. Further, conservative response amplitudes
were obtained for all flow directions with this value. An increased drag coefficient
can be obtained e.g. by strakes for VIV mitigation.

CV
The default value of the CF lift force coefficient was 1.0, referring to straight pipes.
For the ten and 45 degree flow, conservative results were obtained with the default
value. For the 90 degree flow, motion in x-direction is associated with CF VIV. Due
to the jumper geometry it is easier to excite large amplitudes in y-direction than in
x-direction. Since in experimental data large responses were reported for x-direction
in the 90 degree flow, the CF lift force coefficient had to be increased to 1.2.

88
5.4.2 Comparison of VIV motion
Experimental data was available for accelerometer positions three and five for the
ten degree flow and accelerometers three and seven for the 45 and 90 degree flow
directions. Accelerometer five is located on the top horizontal H2 next to the bend,
while number seven is located at the center of H2. Accelerometer three is placed
at the lower part of the jumper, next to the bend at vertical part V2, as shown in
Figure 2.8.
For structures with similar geometry to the jumper model the displacement close to
bends typically is smaller compared with the straight parts. Thus, the positions of
accelerometers three and five are predestined if significant out-of-plane bending dis-
placements are expected rather than rotational deformations or in-plane distortions.
The position of accelerometer seven, on the contrary, is predestined to capture all
kinds of deformation including rotation and in-plane bending.
In this section test data is plotted together with results from the DNV GL response
model procedure and from the TD VIV tool in Simla. The CF response model
depends on modal frequencies, thus, it is plotted for both the Pipe31 and Pipe34
model. Also, Simla results are plotted for both models. Comparison is made using
the following parameters for VIV analysis.
CV I1 = 1.2
CV = 1.2
CDrag = 1.1
fˆ1I1 = 0.13
Sensitivity to an increased drag coefficient CDrag = 1.2 is discussed. Relative differ-
ence in percent is always calculated with respect to the Pipe34 model.

10◦ flow
In Figures 5.25 and 5.26 the displacement during VIV in x, y and z-direction is
shown for the Pipe31 and Pipe34 model, respectively. From left to right each block
represents a period of constant current velocity, starting at 0.05 m/s and increased
in steps of 0.05 m/s up to 0.98 m/s. For numerical stability, each load case started
from zero current speed and was increased to the required speed. Current velocity
was also reduced to zero before a new load case started. Since a steady state was
of greatest interest, the blocks in these Figures each represent a period of stable
vibrations.
At the position of accelerometer five, response in x-direction dominates for both
the Pipe31 and Pipe34 model for low current velocities of 0.1 to 0.2 m/s. For this
flow direction, this is linked with IL VIV. For current velocities 0.25 to 0.98 m/s
the maximum response occurs in y-direction, which is associated with CF VIV. For

89
Uc =0.05 m/s no VIV was observed. From oscillating frequencies it was concluded
that the in-plane bending mode (mode three for Pipe31 and mode two for Pipe34
model) is active for low current velocities up to 0.2 m/s. From 0.25 m/s the out-of-
plane bending mode (mode one) is active. Transition to the twist mode (mode two
for Pipe31 and mode three for Pipe34 model) occurs between 0.50 and 0.55 m/s for
the Pipe34 model and between 0.6 and 0.65 m/s for the Pipe31 model. The earlier
transition of the Pipe34 model is in better agreement with observations from the
model test reported by Wang et al. [18]. This transition also causes a cut in the
moment and stress diagrams for the ten degree flow, as will be shown later. This is
due to the change of oscillating frequency, direction of maximum VIV motion and
amplitude of vibrations.
Figures 5.27 to 5.29 show the ten degree amplitude response plotted over reduced
velocities measured at accelerometer five. Both the Pipe31 and the Pipe34 model
results are in very good agreement with experimental data.
In x-direction significant response from the simulations and test data is observed
around Vred = 1.7, which is linked to the first IL instability region and first lock-in
region. The largest results here come from the simulations at 0.15 and 0.2 m/s. No
response was recorded in the second IL instability region, which would occur around
Vred = 2.5. However, about twice as large IL response compared to the first insta-
bility region occurs at reduced velocities between six and eight, presumably due to
CF-induced IL VIV. Larger amplitudes compared to the first and second instability
region without lock-in are characteristic for the CF-induced IL region.
Peak values for pure CF VIV in y-direction are observed as in the model test at
reduced velocities from six to eight. As mentioned previously, model test results
show quite narrow peaks, while they are spread wider in Simla results. However,
the Simla models are in a conservative range.
For motion in z-direction, the Simla models reach their maximum response at higher
reduced velocities, between ten and twelve, while it occurs between six and eight in
experimental data. However, the results in z-direction are very small for this flow
direction. Generally, there is some uncertainty about reduced velocities, whether
they are calculated from the modal frequency associated with motion in each direc-
tion individually or from the modal frequency found for the direction of maximum
response.
For motion in x and z-direction, the Pipe34 model gives larger response than the
Pipe31 model due to its decreased stiffness in the bends. The largest relative dif-
ference measures 17.4 percent and appears at Vred = 6 for the x-direction. Also, in
z-direction the Pipe34 results are up to 17 percent larger around the same reduced
velocity. In y-direction the Pipe31 model gives up to four percent larger responses
when mode one is active. On the other hand, when the twist mode is dominant the

90
Pipe34 model shows larger response amplitudes in y-direction, too, with up to 10.5
percent difference.
When the drag coefficient is increased from 1.1 to 1.2, all VIV response amplitudes
turn out smaller, but the Pipe31 results are more affected than the Pipe34 results.
Since drag affects the IL force, the biggest difference for the Pipe31 model occurs for
x and z-directions with an average decrease of seven and 6.3 percent, respectively.
For the Pipe34 model the average difference in IL direction is only 0.7 percent.
Regarding the DNV GL response models, it is observed that they are very conser-
vative for CF VIV. The maximum amplitude in particular is predicted over a very
wide range of reduced velocities. For IL VIV the response model fits the maximum
amplitudes for reduced velocities up to four. A drawback of the response model is
that CF-induced IL VIV is only considered when stress ranges are calculated, but
no amplitude response estimates are given for it. However, the formulation for stress
ranges of CF-induced IL VIV is similar to the one for pure CF VIV, but scaled with
factor 0.8 instead of factor two. With this knowledge it can be assumed that the
response model also gives conservative results for CF-induced IL VIV over a wide
span of reduced velocities.
Amplitude responses at accelerometer three are shown in Figures 5.30 to 5.32. Since
this is located at the shorter part V2, smaller responses were expected compared
to the ones from accelerometer five. In contrast to this, model test data showed
significantly larger response in y-direction than for accelerometer five. Therefore,
the data was considered less credible for comparison.
However, responses of the Simla models could be compared. Oscillating frequencies
are in agreement with the active modes identified from the data at accelerome-
ter position five. Also here, the Pipe34 model showed larger amplitudes in x and
z-direction, while the Pipe31 model showed larger response in y-direction. For
CDrag = 1.2 the results in all three directions are affected to a very similar ex-
tent. A decrease in results between 4.1 (for x-direction) and 4.9 (in y-direction)
percent on the average is observed.
Altogether, the simulation results from Simla are in very good agreement with exper-
imental data. Wang et al. [18] reported a strong IL VIV response at 0.206 m/s with
an oscillating frequency close to the second modal frequency and an out-of-plane
twist vibration at 0.527 m/s associated with the third modal frequency. Strong CF
response with a frequency close to the first modal frequency was reported for 0.412
m/s. The Simla results show a promising correlation with test data.

91
Figure 5.25: Displacement of Pipe31 model at accelerometer five for the 10◦ flow.
Load cases from left to right are the current velocities from 0.05 to 0.98 m/s increased
in steps of 0.05 m/s. For each load case, the current speed started at zero and was
reduced to zero again after a time of stable vibrations.

Figure 5.26: Displacement of Pipe34 model at accelerometer five for the 10◦ flow.
Load cases from left to right are the current velocities from 0.05 to 0.98 m/s increased
in steps of 0.05 m/s. For each load case, the current speed started at zero and was
reduced to zero again after a time of stable vibrations.

92
Figure 5.27: Comparison of VIV response in x-direction for 10◦ flow at accelerometer
five.

Figure 5.28: Comparison of VIV response in y-direction for 10◦ flow at accelerometer
five.

93
Figure 5.29: Comparison of VIV response in z-direction for 10◦ flow at accelerometer
five.

Figure 5.30: Comparison of VIV response in x-direction for 10◦ flow at accelerometer
three.

94
Figure 5.31: Comparison of VIV response in y-direction for 10◦ flow at accelerometer
three.

Figure 5.32: Comparison of VIV response in z-direction for 10◦ flow at accelerometer
three.

95
 90◦ flow
Figures 5.33 and 5.34 show the displacement in x, y and z-direction of respectively
Pipe31 and Pipe34 model for 90 degree flow. For this flow angle, several current
velocities show peaks in the beginning or end of the load case, where the current
speed was increased or decreased. For current velocities between 0.7 and 0.79 m/s
the response amplitudes in y-direction were not stable, even in the region where
the current speed was constant. To find oscillating frequencies and response am-
plitudes, narrow peaks were excluded from the analysis. Nonetheless, for this flow
direction it takes more time to reach a state of stable vibrations compared with
the ten degree flow. What happens in y-direction for the highest current velocities
could be observed during a longer simulation. As an example, the displacement
in y-direction for 0.75 m/s is plotted from a simulation over 200 seconds in Figure
5.35. The displacement here is not scaled around a zero mean like it was done for
the cases with stable vibrations. Unlike other load cases, the oscillating frequencies
from rainflow counting and from FFT do not agree for this case. In the beginning,
the oscillating frequency from rainflow counting is close to the natural frequency
of mode nine, while from FFT it is close to mode one. This indicates that smaller
and bigger cycles are present in the signal. After a period of irregular response, it
approaches the same frequencies, but with smaller response amplitude. Finally, the
signal approaches a frequency close to the first modal frequency, both when rainflow
counting and FFT is applied, but amplitudes are very large. When the current
velocity was not ramped out at this point, but the simulation prolonged, numerical
problems occurred.
From the model test it was reported that end dynamometers had reached their
load capacity and, thus, simulations were not continued beyond 0.79 m/s [18]. The
correlation of results is interesting and shows that these load cases are critical. In
Simla, quite a stable response was obtained for x and z-direction, also for the largest
current velocities, which may be used for comparison.

For this flow angle the response in y-direction, which is associated with IL VIV,
dominates for both models up to 0.3 m/s. The oscillating frequency is closest to
the first modal frequency. For further increased current velocities, from 0.4 to 0.65
m/s the in-plane bending mode (mode three and two for Pipe31 and Pipe34, re-
spectively) is active with largest non-dimensional amplitudes in x-direction. For
the 90 degree flow, this is associated with CF VIV. Due to decreased stiffness, the
Pipe34 model shows significantly larger responses in x-direction for above mentioned
flow velocities. This is represented in Figure 5.36 for reduced velocities around five.
Nonetheless, the peak values which occur at reduced velocities around six are in
good agreement for both models and also in good correlation with experimental

96
results.
From 0.7 to 0.79 m/s the maximum response occurs in z-direction. For these load
cases both models again show very similar response. The oscillating frequency is
closest to the natural frequency of mode four, which is the up-and-down bending
mode.
In Figures 5.36 to 5.38 the Simla results for both models at accelerometer seven
are compared with model test data and the DNV GL response models, plotted over
reduced velocities. Since the ratio of consecutive CF modes is similar for the Pipe31
and Pipe34 model for this flow direction, there is only one CF response model. For
all three directions, the Simla results are in correlation with experimental data. For
z-direction the Simla response might be over-conservative, but the plots for x and
y-direction justify to keep both force coefficients in the VIV tool at 1.2.
When the drag coefficient is increased to 1.2, the Pipe31 results are more affected
for higher current velocities than the Pipe34 results. For x and z-direction Pipe31
results are decreased by about 16 percent due to increased drag.
Regarding the DNV GL response models, the conclusion is similar as for the ten
degree flow. The CF response model has a far wider range where large amplitudes
are expected compared with the Simla and model test data .
At the position of accelerometer three stable vibrations were observed for all cur-
rent velocities in Simla. The response in y-direction dominates up to Uc = 0.6 m/s.
The oscillating frequency here is associated with the frequency of mode one for the
lowest current velocities up to 0.2 m/s, and later with mode five whose mode shape
shows symmetric out-of-plane bending of the two shorter horizontals. For all higher
current velocities the oscillating frequency is closest to the fourth modal frequency
for both models. This is in agreement with the findings from accelerometer position
seven. In Figures 5.39 to 5.41 the simulation results are compared with experimental
data and the DNV GL response models. The response in Simla is a bit offset from
model test results with respect to reduced velocities. In especially occur peak values
from the Simla results at lower reduced velocities than from experimental data. The
maximum responses in x and z-direction come from the simulation at 0.7 to 0.79
m/s. It is not known which oscillating frequency was measured during the exper-
iment for the 90 degree flow and, thus, from which modal frequency the reduced
velocities were calculated. However, the large response in z-direction supports the
assumption that the oscillating frequency was close to the fourth modal frequency.
Further, the responses from the two Simla models overestimate amplitudes for x and
z-direction at this location, while at the position of accelerometer seven they just
met the experimental data. Thus, VIV motion from the TD tool gives conservative
results also for this flow angle. When the drag coefficient was increased to 1.2, a
small reduction in peak values was observed of 4.7 to 6.4 percent.

97
For the 90 degree flow, less detailed description was available from the model test.
Especially, it was not reported for which flow velocities largest responses or mode
shifts occurred. However, from A/D-comparison plotted over reduced velocities, the
simulation results are in good correlation with experimental data.

Figure 5.33: Displacement of Pipe31 model at accelerometer seven for the 90◦ flow.
Load cases from left to right are the current velocities from 0.05 to 0.79 m/s increased
in steps of 0.05 m/s. For each load case, the current speed started at zero and was
reduced to zero again after a time of stable vibrations.

98
Figure 5.34: Displacement of Pipe34 model at accelerometer seven for the 90◦ flow.
Load cases from left to right are the current velocities from 0.05 to 0.79 m/s increased
in steps of 0.05 m/s. For each load case, the current speed started at zero and was
reduced to zero again after a time of stable vibrations.

99
Figure 5.35: Displacement in y-direction for Pipe31 model at accelerometer seven
for the 90◦ flow. Current velocity Uc =0.75 m/s.

Figure 5.36: Comparison of VIV response in x-direction for 90◦ flow at accelerometer
seven.

100
Figure 5.37: Comparison of VIV response in y-direction for 90◦ flow at accelerometer
seven.

Figure 5.38: Comparison of VIV response in z-direction for 90◦ flow at accelerometer
seven.

101
Figure 5.39: Comparison of VIV response in x-direction for 90◦ flow at accelerometer
three.

Figure 5.40: Comparison of VIV response in y-direction for 90◦ flow at accelerometer
three.

102
Figure 5.41: Comparison of VIV response in z-direction for 90◦ flow at accelerometer
three.

103
 45◦ flow
For the 45 degree flow direction it is more complex to identify IL and CF VIV.
For the DNV GL response model approach it was considered that all modes with
maximum deformation in x and y-direction could be both IL and CF modes, while
the ones with maximum displacement in z-direction can solely be CF modes. In the
Simla VIV tool it is not necessary to decide which direction is associated with which
kind of VIV.
The time signals at location of accelerometer seven are plotted in Figures 5.42 and
5.43. For low velocity currents the response in y-direction shows maximum ampli-
tudes and its oscillating frequency is associated with mode one. For the Pipe31
model the response in y-direction remains dominating up to 0.8 m/s, where the
response in x-direction becomes larger for higher load cases. For the Pipe34 model,
the response in x-direction is significant also for lower current velocities (0.25, 0.3
and 0.4 m/s) and is dominating for current velocities of 0.65 to 0.9 m/s. For these
current velocities the mode with its frequency closest to the oscillating frequency is
mode two for both models. Since both x and y-direction show large amplitudes for
low velocity currents, it supports the theory that modes with main displacement in
either of the directions must be considered for IL VIV, when the response model
approach is used. The response in z-direction only becomes large for the highest
current velocities and oscillates with a frequency close to the fourth modal frequency.
For many current velocities the signal is not steady at this flow angle and amplitudes
are decreased or increased during a period of steady flow. Especially for some lower
current velocities the response starts with large amplitudes in y-direction, which are
canceled out soon and are followed by VIV motions in x or z-direction. This indi-
cates that different types of oscillations are overlaying or cancel out each other. It
stands out that changes of vibration amplitudes during a single load case are more
significant for the Pipe31 model than for the softer Pipe34 model, especially for low
current velocities.
In the scope of this thesis it was not possible to explore all irregularities, e.g. tran-
sition between modal frequencies for each current velocities. However, these cases
are very interesting, since rapid chances in structural dynamics also affect the flow
and much energy is transferred.
It is also interesting that the Pipe31 signal is more volatile than the Pipe34 model.
For the Pipe31 model, longer simulations were run to observe the development of
VIV motion. Results over a region with stable vibrations were used for comparison,
as shown in Figures 5.44 to 5.46. The Simla results are in a conservative range
and describe the peak response at reduced velocities between six and eight where
they occur also in test data. For both models very large response in z-direction was
observed for the highest current velocities, which is represented at Vred =6. For the

104
other current velocities, on the other hand, the results are in good correlation.
Since the trend for all three flow directions is that responses at largest current ve-
locities are over-conservative, it can be considered to decrease the TD VIV tool
parameters CV and CVI1 for these cases in future work.
The guideline’s response model approach is quite over-conservative for CF VIV and
CF-induced IL VIV.

Figure 5.42: Displacement of Pipe31 model at accelerometer seven for the 45◦ flow.
Load cases from left to right are the current velocities from 0.05 to 0.98 m/s increased
in steps of 0.05 m/s. For each load case, the current speed started at zero and was
reduced to zero again after a time of stable vibrations.

105
Figure 5.43: Displacement of Pipe34 model at accelerometer seven for the 45◦ flow.
Load cases from left to right are the current velocities from 0.05 to 0.98 m/s increased
in steps of 0.05 m/s. For each load case, the current speed started at zero and was
reduced to zero again after a time of stable vibrations.

Figure 5.44: Comparison of VIV response in x-direction for 45◦ flow at accelerometer
seven.

106
Figure 5.45: Comparison of VIV response in y-direction for 45◦ flow at accelerometer
seven.

Figure 5.46: Comparison of VIV response in z-direction for 45◦ flow at accelerometer
seven.

107
5.4.3 Summary of the section about VIV motion
• The dominating VIV response with respect to direction of motion and location
along the jumper could be associated with one of the first four modes for all
three flow directions. Thus, these four modes require highest standards of
accuracy. In the direction where smaller VIV motion occurred and at the
shorter jumper segments, also higher modal frequencies were identified.

• The Simla results are in a conservative range and in good correlation with test
data. Thus, the VIV tool is promising for non-straight geometries in steady
flow. The DNV GL response models are very conservative for the jumper
model, especially for CF VIV. Both from experimental data and from the
Simla VIV tool peak values occurred for reduced velocities of six to eight and
in quite a narrow range for most load cases.

• For the ten degree flow, VIV histories have very stable amplitudes. For the
90 degree flow, irregularities are observed when the current velocity changes
and for the highest load cases, while for the 45 degree flow irregularities are
present for most load cases.
Generally, the 45 degree flow is very interesting, since VIV motions in the
different directions tend to cancel out each other and are followed by a different
vibration pattern during a steady flow condition. These rapid chances in
structural dynamics also affect the flow.

• The Pipe31 model tends to have more volatile VIV response than the Pipe34
model. This makes sense, since stiffer structures generally show higher-frequent
response, while softer ones oscillate with lower frequencies. Presumably, the
Pipe34 model is less sensitive to higher-frequent disturbances. This can be in
favor to observe a steady state of vibrations.

• The shift from one oscillation pattern to another one (with respect to oscillat-
ing frequency and direction of largest amplitudes) for consecutive load cases
happened at lower current velocities for the Pipe34 model than for the Pipe31
model.

• When the drag coefficient is increased, the IL VIV results of the Pipe31 are
more affected than of the Pipe34 model. This indicates that the higher drag
coefficient has a softening effect on the structure in flow direction.

108
 Figure 5.47: Critical locations for stress assessment along the jumper

5.5 Stresses and fatigue damage for VIV

The stress ranges for fatigue calculations were assessed by two different methods.
One of them was to follow the procedure proposed in DNVGL-RP-F105 and using
modal stresses obtained from Simla. The other one was to use the moment output
from TD analysis in Simla directly and apply rainflow counting on obtained stress
signals.
To get comparable results all safety factors in the DNV GL response model were set
to one.
Stresses were compared at five locations (A to E) as shown in Figure 5.47, which
are considered as critical locations. As discussed earlier, maximum modal stresses
usually occurred in the ends for the Pipe31 models and in the bends for the Elbow
element models. Accordingly, the jumper end was chosen as one of the critical
locations. Since welds are usually placed next to the bend parts, but with a small
offset, the elements next to the bends and at the longer of two adjacent straight parts
were chosen as locations B to D. Location E is the center of the top horizontal, which
is the part with longest unsupported length.
Moment signals were extracted from Simla and the observations fit into one of the
the following three patterns:
(1) Steady oscillations over the observed time. When ramped on, the oscillating
moments increased smoothly. When ramped off, they decreased smoothly.
(2) Steady oscillations at the beginning of the simulation, later transition to another
steady pattern.
(3) Steady oscillations over most of the time. When current velocity was ramped
out, large peaks occurred in the moment signal.

109
 10◦ flow
For the ten degree flow, very stable oscillations were observed for the lower current
velocities, while the signal for higher current velocities became more volatile. Large
peaks were observed for Uc =0.98 m/s. At location A next to the support a transition
from one oscillation pattern with smaller amplitudes to one with larger amplitudes
was observed at 0.4 m/s. This is shown in Figure 5.48. During longer simulations
for this case it was observed that before the transition the oscillation frequency is
close to the third modal frequency, while after it the oscillation frequency is close
to the seventh modal frequency. These events can be interesting for future model
test series. Since both the equivalent stress range and frequency are larger after
the transition and it remains stable, only the latter case was taken into account for
fatigue calculation. Torsional moments are close to zero in the IL VIV region up to
0.2 m/s. For locations A and B, which are closer to the support, torsional moments
are present for higher current velocities, but still comparatively small for this flow
angle. Therefore, fatigue estimates from the first principal stress are dominated by
flexural stresses. The fatigue damage obtained from Simla is shown in Figure 5.49
for location A. For the Pipe34 model, a change in the moment signal is also seen
between Uc =0.5 m/s and 0.55 m/s, since the out-of-plane twist mode becomes active
here.
For comparison the fatigue damage obtained from the DNV GL response model
procedure is shown in Figure 5.50. Also here, the main contribution to the maximum
principal stress comes from flexural stresses. The damage estimates for current
velocities 0.5 to 0.6 m/s stand out from the other current velocities. The contribution
from IL VIV and CF VIV can be observed separately in Figure 5.51. Here it is seen
that the drop of fatigue estimates is due to jumps in IL VIV fatigue damage. This is
because of small contribution from the pure IL response model for current velocities
of 0.5 m/s and higher, while at 0.65 m/s the contribution from CF-induced IL VIV
becomes larger. At 0.9 m/s also CF-induced IL VIV is smaller.
For both methods the overall fatigue damage has its maximum at location A. For
current velocities 0.05 and 0.1 m/s Simla gives larger fatigue estimates, while for
the higher current velocities, the DNV GL response model procedure calculates
significantly higher fatigue damage.
At locations C, D and E torsional moments become more significant for larger current
velocities. As an example the moment history from Simla at location E is shown in
Figure 5.52 and associated fatigue damage in Figure 5.53. For current velocities of
0.6 m/s and higher, torsional stresses have a relevant contribution to fatigue damage.
The DNV GL response model fatigue estimates at the same location are shown in
Figure 5.54. Here, torsional stresses slowly increase within the plotted range starting
at 0.3 m/s. For lower current velocities from 0.05 to 0.2 m/s the fatigue damage

110
obtained with Simla is larger than from the response model approach. For higher
current velocities the damage from flexural stress and thus, from the first principal
stress is estimated higher with the DNV GL procedure than with the Simla output.
Damage from torsional stresses is similar from both methods. In the DNV GL
procedure all active modes apply their full stress range, which can lead to over-
estimation of fatigue damage. Torsional stresses do not occur for all modes, thus,
this effect is less significant here. When excluding the load case with 0.55 m/s, the
fatigue damage at location E differs by up to factor 20 and the largest difference
occurs for the highest current velocities. However, the difference at location A is
much larger. This indicates, that the jumper ends experience far less VIV stress in
a TD simulation than obtained from the guidelines by scaling modal stresses.

Figure 5.48: Moment signal during VIV for the Pipe34 model in 10◦ flow at location
A

111
Figure 5.49: VIV fatigue damage per year for the Pipe34 model in 10◦ flow at
location A

Figure 5.50: VIV fatigue for the 10◦ flow at location A calculated with the DNV
GL response model procedure and modal stresses from the Simla Pipe34 model

112
Figure 5.51: VIV fatigue for the 10◦ flow at location A calculated with the DNV
GL response model procedure and modal stresses from the Simla Pipe34 model

Figure 5.52: Moment signal during VIV for the Pipe34 model in 10◦ flow at location
E

113
Figure 5.53: VIV fatigue damage per year for the Pipe34 model in 10◦ flow at
location E

Figure 5.54: VIV fatigue for the 10◦ flow at location E calculated with the DNV GL
response model procedure and modal stresses from the Simla Pipe34 model

114
 90◦ flow
For the 90 degree flow the in-plane flexural moment My is small for lower current
velocities up to 0.45 m/s. At locations B and C next to the bends at the lower
part, there are several load cases for which the torsion moment is larger than the
flexural moments. Further, the flexural moments are more unsteady in the same
region. At location B this can be seen at 0.3 to 0.5 m/s in Figure 5.55. In Figure
5.56 corresponding fatigue estimates are shown. In the above mentioned region the
fatigue damage calculated from torsional stresses becomes very large and at some
load cases even exceeds the damage calculated from the first principal stress.
Here, the torsion stress has a lower number of cycles, but a higher equivalent stress
range compared with the first principal stress. It seems that in these cases it is not
suitable to calculate the principal stress at each time step and determine the stress
ranges afterwards from rainflow counting how it was done in this method. This is
because amplitudes can be decreased when two signals are combined. Details of
the two stress components are shown in Figure 5.58. Stress ranges should rather
be calculated for both signals and be combined within the rainflow algorithm. This
can be done by defining a master and slave for each load case and find the stress
ranges for the master. The stress ranges of the slave component should be found
and the principal stress ranges be calculated based on the time where the master
has its peaks and troughs. Since not all moment signals were steady, especially for
described load cases, a major intervention in the algorithm would be necessary to
capture the stress cycles correctly. These issues are also known from flexible pipe
analyses.
It was not possible in the scope of this work to amend the rainflow algorithm, but
it is recommended to investigate the possibilities further.
At locations A, D and E the torsion stress is much smaller and such irregularities
as mentioned above are not present. The maximum fatigue damage occurs at the
supports. Figures 5.59 and 5.60 show the fatigue estimates at location A from Simla
and from the DNV GL procedure, respectively. For current velocities up to 0.3 m/s
the Simla model gives larger fatigue damage, while for all higher load cases the DNV
GL response model gives higher fatigue estimates. The largest difference occurs at
0.5 m/s, because in the response model the flexural stress contribution to fatigue
damage is increased steadily, but in Simla strong flexural stress response only arises
beyond that.
As was observed for the ten degree flow direction, the smallest difference in results
between the two methods occurs at the center of the top horizontal, while the
difference at locations A, B and C is larger. This is reasonable, since the DNV GL
procedure is adopted from straight pipes and the response at location E is least
influenced by deviations of the jumper compared to a straight pipe. Nevertheless,

115
deviations at location E are significant, too. It highlights the differences in the two
approaches.
When fatigue damage is calculated, the SN-curve coefficient m is applied as an
exponent on the stress range. For chosen SN-curve this means that the difference in
stress levels is represented with the power of five in the fatigue estimates. Further,
when following the recommended practice for non-straight pipes all modes that are
active based on their reduced velocity apply their full stress range. Already from
the VIV motion amplitudes it was seen that the DNV GL procedure gives higher
response estimates for many reduced velocities compared to the Simla results and
model test data. In Simla, several modes can be active at the same time, too, but
obviously not all apply their full stress range. From the oscillating frequencies in
Simla the dominating response could always be associated with one of the first four
modes, which have lower unit stresses compared to the higher modes. At the shorter
segments also higher modes were identified during VIV, but the response amplitudes
are significantly lower than from the guidelines. Following the DNV GL procedure
also higher modes were active, which contribute with higher unit stresses.
It is interesting that for some cases with low velocity currents Simla gives larger
fatigue estimates, however, fatigue damage for these cases is small in general.

Figure 5.55: Moment signal during VIV for the Pipe34 model in 90◦ flow at location
B

116
Figure 5.56: VIV fatigue damage per year for the Pipe34 model in 90◦ flow at
location B

Figure 5.57: VIV fatigue for the 90◦ flow at location B calculated with the DNV
GL response model procedure and modal stresses from the Simla Pipe34 model

117
Figure 5.58: Stress details for Uc = 0.45 m/s over ten seconds in 90◦ flow at location
B

Figure 5.59: VIV fatigue damage per year for the Pipe34 model in 90◦ flow at
location A

118
Figure 5.60: VIV fatigue for the 90◦ flow at location A calculated with the DNV
GL response model procedure and modal stresses from the Simla Pipe34 model

119
 45◦ flow
Also for the 45 degree flow several load cases have large torsion moments at locations
B and C close to the bends in the lower part. The moment signal at location B is
shown in Figure 5.61 and associated fatigue damage from Simla in Figure 5.62.
The same conclusion as for the 90 degree flow is drawn here, that in regions with
large torsion moments it is not sufficient to calculate the first principal stress from
stress amplitudes of the two stress components and instead it should be calculated
using stress ranges after specifying master and slave for each case. From the fatigue
estimates it stands out that they do not increase as steadily and show more jumps
than for the other flow directions. Also from VIV motion it was observed that
the vibration patterns of consecutive load cases with respect to main direction and
oscillation frequency differ much more for this flow angle. It can be concluded that
the change in vibration patterns is also evident in fatigue damage calculations.
Next to the bends the DNV GL response model procedure gives much higher fatigue
estimates which also steadily rise for increasing current velocities. For location B
this is shown in Figure 5.63.
Largest fatigue damage occurs at the support as for the other flow directions. From
VIV motion it was shown that large response arises in z-direction for 0.95 and
0.98 m/s. This causes an increase in fatigue damage at locations A, B and C. At
locations D and E it has the opposite effect. Nonetheless, at location E the Simla
results exceed the fatigue estimates from the response model approach for current
velocities of 0.7 to 0.9 m/s. As mentioned before it can be considered to reduce the
load parameters in the VIV tool for the highest load cases to avoid over-conservative
results.

120
Figure 5.61: Moment signal during VIV for the Pipe34 model in 45◦ flow at location
B

Figure 5.62: VIV fatigue damage per year for the Pipe34 model in 45◦ flow at
location B

121
Figure 5.63: VIV fatigue for the 45◦ flow at location B calculated with the DNV
GL response model procedure and modal stresses from the Simla Pipe34 model

Figure 5.64: VIV fatigue damage per year for the Pipe34 model in 45◦ flow at
location A

122
Figure 5.65: VIV fatigue for the 45◦ flow at location A calculated with the DNV
GL response model procedure and modal stresses from the Simla Pipe34 model

Figure 5.66: VIV fatigue damage per year for the Pipe34 model in 45◦ flow at
location E

123
Figure 5.67: VIV fatigue for the 45◦ flow at location E calculated with the DNV GL
response model procedure and modal stresses from the Simla Pipe34 model

124
5.5.1 Sensitivity to modeling choices and model parameters
In this section sensitivity of fatigue damage results to modeling choices, especially
with or without bend elements, and to the model parameters of the TD VIV tool is
discussed. The results were obtained from simulations in Simla, for which only one
parameter was changed at a time. They are reported for the pure flow angles of ten
and 90 degrees, since these cases are most suitable to draw conclusions.

Pipe31 and Pipe34 model

From modal analysis the main difference in fatigue damage between the two Simla
models is expected to occur in the bends and at the transition to the straight parts.
In the bends the Pipe34 model experiences higher stresses, whereas the Pipe31 model
shows higher stresses right next to the bends and, thus, at possible weld locations.
In correlation with that gives the Pipe31 model higher fatigue damage at locations
B, C and D for the ten degree flow and current velocities higher than 0.2 m/s. For
the lowest current velocities mainly IL VIV is active, which is associated with vi-
brations in x-direction at this flow angle. Thus, for these cases the Pipe34 model
experiences higher fatigue damage. In Figures 5.68 and 5.69 the relative difference
in fatigue damage is shown. At 0.55 m/s some results are not within the plotted
range, but larger. This is because the twist mode becomes active and causes smaller
vibration amplitudes for the Pipe34 model. From 0.7 m/s on the trend is that the
Pipe31 model gives increasingly larger fatigue damage.
For the 90 degree flow, where motion in x-direction is associated with CF VIV, the
Pipe31 model gives significantly lower fatigue damage for current velocities 0.5 to
0.65 m/s, not only close to the bends, but also at location E. Here, the maximum
displacement due to VIV was in x-direction. For 0.7 to 0.79 m/s, where the largest
displacement was in z-direction, the Pipe31 model again gives larger fatigue esti-
mates.

CDrag
When the drag coefficient is changed to 1.2 and all other parameters are kept the
same, fatigue damage for the ten degree flow is decreased by 15 to 35 percent for
almost all load cases. Larger deviations occur for the 0.05, 0.1 and 0.55 m/s load
cases. The influence on fatigue damage is shown in Figures 5.70 and 5.71 for the ten
and 90 degree flow, respectively. For the 90 degree flow angle the relative difference
in fatigue damage per year is decreased between 25 to 45 percent for most load cases.
Here, exemptions occur at current velocities 0.05 and 0.5 m/s.
For the lowest current velocities the results are less affected when the drag coefficient
is changed, but damage here is generally low. Also at 0.55 and 0.5 m/s respectively

125
Figure 5.68: Sensitivity of fatigue damage with respect to modeling choices (with
or without bend elements) for the 10◦ flow.

Figure 5.69: Sensitivity of fatigue damage with respect to modeling choices (with
or without bend elements) for the 90◦ flow.

for the two flow angles, major changes in vibration patterns occurred, which led to
smaller stress amplitudes. Besides that the change in drag coefficient has a similar
effect at all locations along the jumper and all flow velocities. This is reasonable,
since the drag coefficient is defined for still-water conditions and amplification during
VIV is applied to this value.

126
Figure 5.70: Sensitivity of fatigue damage with respect to the drag coefficient for
the 10◦ flow.

Figure 5.71: Sensitivity of fatigue damage with respect to the drag coefficient for
the 90◦ flow.

CVC
When the CF lift force coefficient is changed back to its default value of 1.0, the
impact varies more for the different locations at the jumper and different current
velocities. This is shown in Figures 5.72 and 5.73. Generally, smaller VIV amplitudes

127
and, thus, lower fatigue damage is expected for the default value when CF VIV is
active. For the ten degree flow, this can be seen for current velocities of 0.6 to 0.98
m/s and for the 90 degree flow for current velocities of 0.5 to 0.79 m/s. Location
A at the support is an exception for the ten degree flow, since fatigue damage here
is between 20 and 40 percent higher at most load cases for the lower CF force
coefficient. Furthermore, in the ten degree flow an increase in fatigue damage is
observed for small flow velocities up to 0.5 m/s, while for the 90 degree flow it had
only small impact for the low current velocities.

Figure 5.72: Sensitivity of fatigue damage with respect to the CF force coefficient
for the 10◦ flow.

CVI1
Also for the IL force coefficient the effect on fatigue damage varies at the different
locations. In contrary to the CF force coefficient, the biggest impact is expected for
low-velocity currents, for which IL VIV dominates. The impact is clearer for the 90
degree flow compared to the ten degree flow as shown in Figures 5.74 and 5.75. For
the 90 degree flow, fatigue damage for current velocities up to 0.35 m/s is decreased
by 85 to 90 percent when the IL force coefficient of 0.8 is used. This is observed
at all five locations. For the ten degree flow, the same effect is observed for Uc =
0.15 and 0.2 m/s. Besides that, the deviations are spread wider for this flow angle.
For current velocities of 0.6 to 0.98 m/s fatigue damage at the top horizontal is
even increased. This might be due to the coupling between IL and CF VIV. When
IL VIV and therefore also CF-induced IL VIV are decreased more energy can be

128
Figure 5.73: Sensitivity of fatigue damage with respect to the CF force coefficient
for the 90◦ flow.

utilized for the pure CF response at higher reduced velocities. This may be a reason
why fatigue damage is increased for the lower IL force coefficient in a region, where
CF response is typically dominating.
The observations of the two force coefficients again show the complexity of the flow
around such structure and of the structural-dynamic response.

f̂1I1
With the default value of the minimum vibration frequency that gives energy input
to IL region one of 0.3, IL VIV could only be activated over a smaller range of reduced
velocity than with the chosen value of 0.13. However, even for low-velocity currents,
where IL VIV was active for both values of fˆ1I1 , fatigue damage is significantly lower
when using the default value. For the 90 degree flow, fatigue damage is lowered for
almost all cases, whereas for the ten degree flow fatigue damage is increased for the
higher current velocities, where CF VIV is dominating. As for the force coefficients,
it is assumed that this is due to coupling between the two kinds of VIV.

129
Figure 5.74: Sensitivity of fatigue damage with respect to the IL force coefficient for
the 10◦ flow.

Figure 5.75: Sensitivity of fatigue damage with respect to the IL force coefficient for
the 90◦ flow.

130
Figure 5.76: Sensitivity of fatigue damage with respect to the minimum vibration
frequency that gives energy input for IL region one for the 10◦ flow.

Figure 5.77: Sensitivity of fatigue damage with respect to the minimum vibration
frequency that gives energy input for IL region one for the 90◦ flow.

131
5.5.2 Summary of the section about VIV stresses and fa-
tigue damage
• Largest fatigue damage is observed at the jumper supports for all three flow di-
rections. The largest deviations of fatigue estimates from the two methods also
occur at this location. Except for the lowest current velocities, fatigue damage
from the DNV GL response models is significantly higher than obtained from
Simla. This is because the response model procedure assumes that all active
modes contribute with their full stress range. In the TD tool several modes
are active in the same load case, too, but they have the possibility to compete
with each other and contribute with less than the full amount. The small-
est difference between the two methods is observed at the center of the top
horizontal. This is reasonable, since the DNV GL procedure is adopted from
straight pipes and at this location the response is presumably least affected
by the non-straight jumper geometry.

• Torsional stresses are relevant for all three flow directions. Especially for the
45 and 90 degree flow and at the lower jumper segment, the contribution to
fatigue damage is significant. A weakness in the way that principal stresses
were calculated here was discovered for these cases. When the first principal
stress is calculated at each time step from the maximum flexural stress around
the circumference and the torsional stress, resulting stress ranges can turn out
smaller than when each component is considered separately. It is recommended
to include a master and slave condition in the rainflow algorithm, and calculate
the principal stress from stress ranges instead of stress amplitudes.

• Largest difference in fatigue estimates between the Pipe31 and Pipe34 model
occurs next to the bends.
When the value of the drag coefficient is changed, the impact on fatigue dam-
age is very similar for all locations and load cases, whereas, when the parame-
ters of the VIV tool are changed the affect on fatigue estimates is much more
complex.
For the lower default values of the load coefficients, the damage as expected is
smaller in the corresponding velocity region, but apart from that coupling ef-
fects occur for other current velocities, too. For the higher default value of the
minimum vibration frequency that gives energy input for IL region one, the
impact on fatigue estimates at higher current velocities varies for the different
flow angles and locations along the jumper, too. It highlights the complex-
ity of VIV for non-straight pipes and emphasizes the importance of precise
prediction tools.

132
Chapter 6

Conclusion and recommendations

6.1 Conclusion
Modal analysis has been carried out for two models in Simla and Abaqus. It was
shown that the models using elbow elements in the bends are in better agreement
with results from the ExxonMobil model test, since the first in-plane bending mode
has a lower natural frequency compared to the out-of-plane twist mode for these
models. The results from modal analyses have shown the relevance of accounting
for additional strains and deformation of the cross-section due to bending during
manufacture.
For the model with Pipe34 elements in the bends the modal stress in bends is
increased by up to factor two compared with the Pipe31 model. At the transition to
straight parts the stresses are decreased by up to 35 percent. These coupled effects
are most significant for in-plane bending modes.
The parameters of the TD VIV tool in Simla have been adjusted, since default
values referred to straight pipes. In this study the IL and CF force coefficients were
increased from 0.8 to 1.2 and 1.0 to 1.2, respectively. The minimum non-dimensional
vibration frequency that gives energy input to IL region one was changed from 0.3
to 0.13, so that IL VIV can be activated over a wider range of reduced velocities.
With these parameters conservative results for VIV motion were obtained for all
observed load cases. From comparison with experimental data, it was shown that
Simla results are in very good correlation. The DNV GL response model over-
predicted the response for many cases, especially for CF VIV.
With respect to fatigue damage, the relevance of torsional stresses was pointed out.
Especially close to the bends in the lower jumper segment, torsional stresses became
the largest stress components for several flow velocities in the 90 and 45 degree flow.
Maximum fatigue damage occurred at the supports for all flow angles. At the same
location, the results from the DNV GL procedure and from Simla differed most.

133
Smaller derogation between the two methods was observed at the center of the
long top horizontal part, but also here the response model approach gave far higher
fatigue estimates compared with the TD tool. When the DNV GL procedure is
used for non-straight pipes, all active modes apply their full stress range, whereas
in Simla, several modes can be active, too, but they can compete with each other.
Sensitivity of fatigue estimates to the control parameters of the TD VIV tool was
discussed. For the lower default values of the force coefficients fatigue damage as
expected was smaller in the corresponding velocity region. Also for the higher default
value of the minimum non-dimensional oscillating frequency that gives energy input
to IL region one, fatigue damage was smaller for low current velocities. Apart from
that, coupling effects occurred in the other velocity regions, too. It once more
highlighted the complexity of VIV for non-straight pipes and the importance of
accurate prediction tools.

6.2 Recommendations for further work

Some recommendations for future work are as follows:

• Based on the large difference in modal stresses between the two models with
and without elbow elements, relevance of thinning in bends can be explored

• Include internal and external pressure in the analysis to observe stiffening and
softening effects for the full jumper model

• Initialize the jumper model in a pre-stressed condition

• Observe VIV for the jumper model when placed horizontally and close to the
seabed, like it is often done for shallow water

• Investigate multi-axial stress. Furthermore, principal stresses should be cal-

culated from stress ranges of contributing components within the rainflow
algorithm instead of stress amplitudes at each time step. Therefore a master-
slave relationship should be implemented. Different stress criteria for bend
segments and straight parts can be used.

134
Bibliography

[1] A.H. Techet. 13.42 lecture: Vortex induced vibrations. Massachusetts Institute
of Technology, Open Courseware, 21, 2005.

[2] B. Mutlu Sumer et al. Hydrodynamics around cylindrical strucures, volume 26.
World scientific, 2006.

[3] Sigmund Kyrre Ås and Stig Berge. Compendium fatigue and fracture design
of marine structures, Revised as of 2017.

[4] Yung-Li Lee, Mark E Barkey, and Hong-Tae Kang. Metal fatigue analysis
handbook: practical problem-solving techniques for computer-aided engineering.
Elsevier, 2011.

[5] Zhenhui Liu, Laila Aarstad Igeh, Jie Wu, and Muk Chen Ong. Fatigue dam-
age assessment to a rigid planar jumper on model scale. Journal of Offshore
Mechanics and Arctic Engineering, 142(1), 2020.

[6] Svein Sævik. Simla-theory manual. Revised as of 2017.

[7] DNVGL-RP-F105. Recommended practice. free spanning pipelines. Revised as

of 2019.

[8] DNVGL-RP-C203. Recommended practice. fatigue design of offshore steel

structures. 2005.

[9] Scott Post. Applied and computational fluid mechanics. Jones & Bartlett Pub-
lishers, 2010.

[10] Mats Jørgen Thorsen. Time domain analysis of vortex-induced vibrations. 2016.

[11] Laila Aarstad Igeh. Viv fatigue investigation for subsea planar rigid spools and
jumpers. Master’s thesis, University of Stavanger, Norway, 2017.

[12] Resistance Committee et al. Final report and recommendations to the 26th ittc.
In Proceedings of 26th International Towing Tank Conference, Rio de Janeiro,
Brazil, volume 38, 2011.

135
[13] Narakorn Srinil, Marian Wiercigroch, and Patrick O Brien. Reduced-order
modelling of vortex-induced vibration of catenary riser. Ocean Engineering,
36(17-18):1404–1414, 2009.

[14] Mats Jørgen Thorsen, Svein Sævik, and Carl Martin Larsen. A simplified
method for time domain simulation of cross-flow vortex-induced vibrations.
Journal of Fluids and Structures, 49:135–148, 2014.

[15] Mats Jørgen Thorsen, Svein Sævik, and Carl Martin Larsen. Time domain sim-
ulation of vortex-induced vibrations in stationary and oscillating flows. Journal
of Fluids and Structures, 61:1–19, 2016.

[16] Jan Vidar Ulveseter, Svein Sævik, and Carl Martin Larsen. Time domain model
for calculation of pure in-line vortex-induced vibrations. Journal of Fluids and
Structures, 68:158–173, 2017.

[17] Kristoffer Høyem Aronsen. An experimental investigation of in-line and com-

bined in-line and cross-flow vortex induced vibrations. 2007.

[18] Howard Wang, Jerry Huang, Sungho Lee, Paulo Gioielli, Wan Kan, Don
Spencer, and Mohammed Islam. Viv response of a subsea jumper in uniform
current. In ASME 2013 32nd International Conference on Ocean, Offshore and
Arctic Engineering. American Society of Mechanical Engineers Digital Collec-
tion, 2013.

[19] Klaus-Jürgen Bathe, Carlos A Almeida, and Lee W Ho. A simple and effective
pipe elbow element - some nonlinear capabilities. Computers & Structures,
17(5-6):659–667, 1983.

[20] Ivar Langen and Ragnar Sigbjörnsson. Dynamisk analyse av konstruksjoner:

Dynamic analysis of structures. Tapir, 1979.

[21] Svein Sævik, Gro S Baarholm, Ole D Økland, and Janne KØ Gjøsteen. Simla
version 3.16.0 user manual, 2019.

[22] AAUs Manual. Abaqus documentation version 6.13. Dassault Systems SIMU-
LIA Corp., Providence, RI, USA, 2013.

[23] Kunpeng Wang, Chunyan Ji, Qinghai Chi, and Hairong Wu. Hydrodynamic
force investigation of a rigid cylinder under the coupling cf and il motion. Jour-
nal of Fluids and Structures, 81:598–616, 2018.

136
Appendices

137
Appendix A: Modal stress output from Simla and
Abaqus

Figure 1: Maximum principal unit stress of mode one for the two Simla models.
Plotted over accumulated arch length and scaled with respect to the outer diameter
and maximum modal displacement.

138
Figure 2: Maximum principal unit stress of mode two for the Pipe31 model and
mode three for the Pipe34 model, which describe the same mode shape. Plotted
over accumulated arch length and scaled with respect to the outer diameter and
maximum modal displacement.

Figure 3: Maximum principal unit stress of mode five for the two Simla models.
Plotted over accumulated arch length and scaled with respect to the outer diameter
and maximum modal displacement.

139
Figure 4: Maximum principal unit stress of mode seven for the two Simla models.
Plotted over accumulated arch length and scaled with respect to the outer diameter
and maximum modal displacement.

Figure 5: Maximum principal unit stress of mode eight for the two Simla models.
Plotted over accumulated arch length and scaled with respect to the outer diameter
and maximum modal displacement.

140
Figure 6: Maximum principal unit stress of mode one for the two Pipe31 models in
Simla and Abaqus. Plotted over accumulated arch length and scaled with respect
to the outer diameter and maximum modal displacement.

Figure 7: Maximum principal unit stress of mode two for the two Pipe31 models in
Simla and Abaqus. Plotted over accumulated arch length and scaled with respect
to the outer diameter and maximum modal displacement.

141
Figure 8: Maximum principal unit stress of mode three for the two Pipe31 models
in Simla and Abaqus. Plotted over accumulated arch length and scaled with respect
to the outer diameter and maximum modal displacement.

Figure 9: Maximum principal unit stress of mode four for the two Pipe31 models in
Simla and Abaqus. Plotted over accumulated arch length and scaled with respect
to the outer diameter and maximum modal displacement.

142
Figure 10: Maximum principal unit stress of mode five for the two Pipe31 models in
Simla and Abaqus. Plotted over accumulated arch length and scaled with respect
to the outer diameter and maximum modal displacement.

Figure 11: Maximum principal unit stress of mode six for the two Pipe31 models in
Simla and Abaqus. Plotted over accumulated arch length and scaled with respect
to the outer diameter and maximum modal displacement.

143
Figure 12: Maximum principal unit stress of mode seven for the two Pipe31 models
in Simla and Abaqus. Plotted over accumulated arch length and scaled with respect
to the outer diameter and maximum modal displacement.

Figure 13: Maximum principal unit stress of mode eight for the two Pipe31 models
in Simla and Abaqus. Plotted over accumulated arch length and scaled with respect
to the outer diameter and maximum modal displacement.

144
Figure 14: Maximum principal unit stress of mode nine for the two Pipe31 models
in Simla and Abaqus. Plotted over accumulated arch length and scaled with respect
to the outer diameter and maximum modal displacement.

Figure 15: Maximum principal unit stress of mode one for the two Elbow models in
Simla and Abaqus. Plotted over accumulated arch length and scaled with respect
to the outer diameter and maximum modal displacement.

145
Figure 16: Maximum principal unit stress of mode two for the two Elbow models in
Simla and Abaqus. Plotted over accumulated arch length and scaled with respect
to the outer diameter and maximum modal displacement.

Figure 17: Maximum principal unit stress of mode three for the two Elbow models
in Simla and Abaqus. Plotted over accumulated arch length and scaled with respect
to the outer diameter and maximum modal displacement.

146
Figure 18: Maximum principal unit stress of mode four for the two Elbow models in
Simla and Abaqus. Plotted over accumulated arch length and scaled with respect
to the outer diameter and maximum modal displacement.

Figure 19: Maximum principal unit stress of mode five for the two Elbow models in
Simla and Abaqus. Plotted over accumulated arch length and scaled with respect
to the outer diameter and maximum modal displacement.

147
Figure 20: Maximum principal unit stress of mode six for the two Elbow models in
Simla and Abaqus. Plotted over accumulated arch length and scaled with respect
to the outer diameter and maximum modal displacement.

Figure 21: Maximum principal unit stress of mode seven for the two Elbow models
in Simla and Abaqus. Plotted over accumulated arch length and scaled with respect
to the outer diameter and maximum modal displacement.

148
Figure 22: Maximum principal unit stress of mode eight for the two Elbow models
in Simla and Abaqus. Plotted over accumulated arch length and scaled with respect
to the outer diameter and maximum modal displacement.

Figure 23: Maximum principal unit stress of mode nine for the two Elbow models
in Simla and Abaqus. Plotted over accumulated arch length and scaled with respect
to the outer diameter and maximum modal displacement.

149
Appendix B: Simla VIV motion details
Results from the VIV tool in Simla are listed below in more detail. They were ob-
tained with the following parameters:
CV I1 = 1.2
CV = 1.2
fˆ1I1 = 0.13

Table 1: Oscillating frequencies, amplitude response and identified modes for the
Pipe34 model in 10◦ flow at position of accelerometer five with CDrag = 1.1

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.72 0.0005 1 0.72 0.0005 1 0.72 0.0005 1
0.1 1.59 0.0049 2 1.59 0.0049 2 1.59 0.0049 2
0.15 1.88 0.0882 2 0.37 0.0112 1 1.88 0.0118 2
0.2 1.91 0.160 2 0.49 0.0451 1 1.91 0.0224 2
0.25 1.93 0.200 2 0.77 0.467 1 1.93 0.0291 2
0.3 1.95 0.229 2 0.80 0.621 1 1.95 0.0377 2
0.35 1.97 0.250 2 0.83 0.669 1 1.97 0.0820 2
0.4 2.00 0.257 2 0.84 0.700 1 3.47 0.105 6
0.45 2.03 0.248 3 0.87 0.670 1 3.48 0.120 6
0.5 2.07 0.225 3 0.89 0.648 1 3.49 0.132 6
0.55 1.88 0.206 2 0.95 0.449 1 7.13 0.0432 9
0.6 3.95 0.127 7 1.97 0.495 2 7.20 0.0418 9
0.65 4.11 0.0265 7 2.05 0.679 3 7.25 0.0523 9
0.7 4.17 0.0309 7 2.08 0.778 3 7.31 0.0545 9
0.75 7.85 0.0518 9 2.12 0.836 3 7.85 0.0594 9
0.8 7.91 0.0581 9 2.15 0.861 3 7.91 0.0653 9
0.85 7.95 0.0608 9 2.17 0.899 3 7.95 0.0680 9
0.9 4.39 0.0374 7 2.20 0.948 3 7.53 0.0713 9
0.95 4.45 0.0356 7 2.23 0.951 3 7.59 0.0815 9
0.98 4.48 0.0402 7 2.24 0.979 3 7.63 0.0955 9

150
Table 2: Oscillating frequencies, amplitude response and identified modes for the
Pipe34 model in 10◦ flow at position of accelerometer three with CDrag = 1.1

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.72 0.0002 1 0.72 0.0002 1 0.72 0.0002 1
0.1 1.59 0.0012 2 1.59 0.0012 2 1.59 0.0012 2
0.15 1.88 0.0154 2 0.37 0.0016 1 1.88 0.0087 2
0.2 1.91 0.0274 2 0.51 0.0051 1 1.91 0.0170 2
0.25 1.93 0.0339 2 0.77 0.0409 1 1.93 0.0218 2
0.3 1.95 0.0372 2 0.80 0.0534 1 1.95 0.0287 2
0.35 1.97 0.0702 2 0.83 0.0610 1 1.97 0.0763 2
0.4 3.47 0.0859 6 0.84 0.0635 1 3.47 0.0984 6
0.45 3.48 0.0970 6 0.87 0.0673 1 3.48 0.114 6
0.5 3.49 0.105 6 0.89 0.0782 1 3.49 0.127 6
0.55 7.13 0.060 9 0.95 0.0525 1 7.13 0.0440 9
0.6 7.20 0.076 9 1.97 0.106 2 7.20 0.0476 9
0.65 7.25 0.080 9 2.05 0.135 3 7.25 0.0582 9
0.7 7.31 0.087 9 2.08 0.144 3 7.31 0.0636 9
0.75 7.85 0.093 9 2.12 0.154 3 7.85 0.0677 9
0.8 7.91 0.096 9 2.15 0.161 3 7.91 0.0736 9
0.85 7.95 0.105 9 2.17 0.146 3 7.95 0.0756 9
0.9 7.53 0.110 9 2.20 0.166 3 7.53 0.0745 9
0.95 7.59 0.117 9 2.23 0.188 3 7.59 0.0841 9
0.98 7.63 0.124 9 2.24 0.477 3 7.63 0.100 9

151
Table 3: Oscillating frequencies, amplitude response and identified modes for the
Pipe34 model in 10◦ flow at position of accelerometer five with CDrag = 1.2

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.72 0.0007 1 0.72 0.0007 1 0.72 0.0007 1
0.1 1.57 0.0049 2 1.57 0.0049 2 1.57 0.0049 2
0.15 1.88 0.0827 2 0.37 0.0111 1 1.88 0.0110 2
0.2 1.91 0.149 2 0.51 0.0464 1 1.91 0.0208 2
0.25 1.93 0.186 2 0.76 0.435 1 1.93 0.0271 2
0.3 1.95 0.212 2 0.80 0.578 1 1.95 0.0347 2
0.35 1.97 0.233 2 0.83 0.637 1 1.97 0.0793 2
0.4 2.00 0.242 2 0.84 0.661 1 3.47 0.100 6
0.45 2.04 0.228 3 0.87 0.635 1 3.48 0.116 6
0.5 2.08 0.222 3 0.91 0.616 1 3.49 0.126 6
0.55 1.85 0.191 2 0.93 0.526 1 7.56 0.0376 9
0.6 3.93 0.106 7 1.96 0.462 2 7.20 0.0401 9
0.65 7.75 0.0564 9 2.05 0.645 3 7.75 0.0435 9
0.7 4.16 0.0274 7 2.08 0.714 3 7.31 0.0500 9
0.75 7.85 0.0499 9 2.12 0.769 3 7.85 0.0553 9
0.8 7.91 0.0537 9 2.15 0.808 3 7.91 0.0615 9
0.85 4.33 0.0336 7 2.17 0.862 3 7.48 0.0629 9
0.9 4.39 0.0355 7 2.20 0.877 3 7.55 0.0631 9
0.95 4.45 0.0342 7 2.23 0.890 3 2.23 0.0819 3
0.98 4.49 0.0352 7 2.24 0.901 3 7.65 0.0870 9

152
Table 4: Oscillating frequencies, amplitude response and identified modes for the
Pipe34 model in 10◦ flow at position of accelerometer three with CDrag = 1.2

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.72 0.0002 1 0.72 0.0002 1 0.72 0.0002 1
0.1 1.57 0.0012 2 1.57 0.0012 2 1.57 0.0012 2
0.15 1.88 0.0145 2 0.37 0.0016 1 1.88 0.0082 2
0.2 1.91 0.0254 2 0.51 0.0052 1 1.91 0.0157 2
0.25 1.93 0.0313 2 0.76 0.0378 1 1.93 0.0204 2
0.3 1.95 0.0350 2 0.80 0.0497 1 1.95 0.0265 2
0.35 1.97 0.0718 2 0.83 0.0582 1 3.45 0.0744 6
0.4 3.47 0.0845 6 0.84 0.0616 1 3.47 0.0935 6
0.45 3.48 0.0959 6 0.87 0.0635 1 3.48 0.110 6
0.5 3.49 0.102 6 0.91 0.0755 1 3.49 0.121 6
0.55 7.56 0.0444 9 0.93 0.0557 1 7.56 0.0357 9
0.6 7.20 0.0712 9 1.96 0.0981 2 7.20 0.0448 9
0.65 7.75 0.0710 9 2.05 0.134 3 7.75 0.0458 9
0.7 7.31 0.0831 9 2.08 0.135 3 7.31 0.0581 9
0.75 7.85 0.0853 9 2.12 0.148 3 7.85 0.0608 9
0.8 7.91 0.0903 9 2.15 0.152 3 7.91 0.0673 9
0.85 7.48 0.103 9 2.17 0.145 3 7.48 0.0654 9
0.9 7.55 0.100 9 2.20 0.151 3 7.55 0.0675 9
0.95 7.61 0.116 9 2.23 0.229 3 7.61 0.0820 9
0.98 7.65 0.115 9 2.24 0.287 3 7.65 0.0881 9

153
Table 5: Oscillating frequencies, amplitude response and identified modes for the
Pipe31 model in 10◦ flow at position of accelerometer five with CDrag = 1.1

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.72 0.0004 1 0.72 0.0004 1 0.72 0.0004 1
0.1 1.93 0.0023 2 1.93 0.0023 2 1.93 0.0023 2
0.15 2.15 0.0262 2 0.39 0.0100 1 2.15 0.0027 2
0.2 2.27 0.115 3 0.51 0.0294 1 2.27 0.0147 3
0.25 2.29 0.164 3 0.79 0.302 1 2.29 0.0219 3
0.3 2.31 0.189 3 0.84 0.572 1 2.31 0.0298 3
0.35 2.33 0.210 3 0.87 0.639 1 2.33 0.0333 3
0.4 2.36 0.223 3 0.88 0.645 1 2.36 0.080 3
0.45 2.39 0.231 3 0.91 0.627 1 3.73 0.101 6
0.5 2.41 0.233 3 0.93 0.587 1 3.75 0.115 7
0.55 2.44 0.234 3 0.96 0.574 1 3.76 0.127 7
0.6 2.48 0.224 4 1.00 0.547 1 3.77 0.136 7
0.65 4.25 0.158 7 2.13 0.587 2 8.28 0.0524 9
0.7 4.35 0.0603 7 2.17 0.697 2 8.36 0.0453 9
0.75 4.40 0.0292 7 2.20 0.778 2 8.44 0.0501 9
0.8 4.45 0.0287 7 2.23 0.861 2 8.51 0.0525 9
0.85 4.49 0.0317 7 2.24 0.892 2 8.57 0.0543 9
0.9 4.53 0.0327 7 2.27 0.931 3 8.64 0.0565 9
0.95 4.59 0.0394 7 2.29 1.054 3 8.71 0.0644 9
0.98 4.63 0.0345 7 2.31 0.948 3 8.75 0.0621 9

154
Table 6: Oscillating frequencies, amplitude response and identified modes for the
Pipe31 model in 10◦ flow at position of accelerometer three with CDrag = 1.1

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.72 0.0001 1 0.72 0.0001 1 0.72 0.0001 1
0.1 1.93 0.0007 2 1.93 0.0007 2 1.93 0.0007 2
0.15 2.15 0.0060 2 0.39 0.0023 1 2.15 0.0018 2
0.2 2.27 0.0233 3 0.51 0.0054 1 2.27 0.0101 3
0.25 2.29 0.0326 3 0.79 0.0391 1 2.29 0.0153 3
0.3 2.31 0.0386 3 0.84 0.0707 1 2.31 0.0211 3
0.35 2.33 0.0421 3 0.87 0.0815 1 2.33 0.0237 3
0.4 2.36 0.0775 3 0.88 0.0883 1 2.36 0.0740 3
0.45 3.73 0.0959 6 0.91 0.0969 1 3.73 0.0931 6
0.5 3.75 0.105 7 3.35 0.0982 5 3.75 0.107 7
0.55 3.76 0.114 7 0.96 0.0926 1 3.76 0.120 7
0.6 3.77 0.119 7 1.00 0.0796 1 3.77 0.129 7
0.65 8.28 0.0644 9 2.13 0.143 2 8.28 0.0533 9
0.7 8.36 0.0693 9 2.17 0.165 2 8.36 0.0454 9
0.75 8.44 0.0807 9 2.20 0.175 2 8.44 0.0554 9
0.8 8.51 0.0824 9 2.23 0.185 2 8.51 0.0576 9
0.85 8.57 0.0892 9 2.24 0.188 2 8.57 0.0580 9
0.9 8.64 0.0924 9 2.27 0.171 3 8.64 0.0584 9
0.95 8.71 0.0965 9 2.29 0.209 3 8.71 0.0672 9
0.98 8.75 0.100 9 2.31 0.195 3 8.75 0.0656 9

155
Table 7: Oscillating frequencies, amplitude response and identified modes for the
Pipe31 model in 10◦ flow at position of accelerometer five with CDrag = 1.2

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.72 0.0004 1 0.72 0.0004 1 0.72 0.0004 1
0.1 1.92 0.0023 2 1.92 0.0023 2 1.92 0.0023 2
0.15 2.15 0.0257 2 0.39 0.0100 1 2.15 0.0028 2
0.2 2.27 0.108 3 0.51 0.0293 1 2.27 0.0137 3
0.25 2.29 0.152 3 0.77 0.287 1 2.29 0.0203 3
0.3 2.31 0.177 3 0.84 0.532 1 2.31 0.0269 3
0.35 2.33 0.196 3 0.87 0.610 1 2.33 0.0321 3
0.4 2.36 0.208 3 0.88 0.614 1 2.36 0.0828 3
0.45 2.39 0.216 3 0.91 0.593 1 3.73 0.0976 6
0.5 2.43 0.219 3 0.93 0.570 1 3.75 0.107 7
0.55 2.45 0.214 3 0.96 0.551 1 3.76 0.119 7
0.6 2.49 0.208 4 1.00 0.534 1 3.77 0.131 7
0.65 4.23 0.153 7 2.11 0.542 2 8.28 0.0441 9
0.7 4.32 0.0529 7 2.16 0.645 2 8.36 0.0408 9
0.75 4.39 0.0251 7 2.20 0.724 2 8.44 0.0490 9
0.8 4.44 0.0285 7 2.23 0.779 2 8.51 0.0468 9
0.85 4.49 0.0298 7 2.24 0.825 2 8.57 0.0545 9
0.9 4.53 0.0311 7 2.27 0.872 3 8.64 0.0554 9
0.95 4.59 0.0312 7 2.29 0.875 3 8.71 0.0497 9
0.98 4.67 0.0659 7 2.33 0.853 3 9.32 0.0557 9

156
Table 8: Oscillating frequencies, amplitude response and identified modes for the
Pipe31 model in 10◦ flow at position of accelerometer three with CDrag = 1.2

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.72 0.0001 1 0.72 0.0001 1 0.72 0.0001 1
0.1 1.92 0.0007 2 1.92 0.0007 2 1.92 0.0007 2
0.15 2.15 0.0059 2 0.39 0.0023 1 2.15 0.0018 2
0.2 2.27 0.0219 3 0.51 0.0055 1 2.27 0.0094 3
0.25 2.29 0.0303 3 0.77 0.0371 1 2.29 0.0143 3
0.3 2.31 0.0353 3 0.84 0.0653 1 2.31 0.0189 3
0.35 2.33 0.0396 3 0.87 0.0773 1 2.33 0.0235 3
0.4 2.36 0.0789 3 0.88 0.0840 1 3.72 0.0751 6
0.45 3.73 0.0909 6 0.91 0.0899 1 3.73 0.0900 6
0.5 3.75 0.0986 7 0.93 0.0884 1 3.75 0.101 7
0.55 3.76 0.109 7 0.96 0.0866 1 3.76 0.112 7
0.6 3.77 0.115 7 1.00 0.0688 1 3.77 0.125 7
0.65 8.28 0.0633 9 2.11 0.132 2 8.28 0.0454 9
0.7 8.36 0.0684 9 2.16 0.151 2 8.36 0.0451 9
0.75 8.44 0.0743 9 2.20 0.164 2 8.44 0.0523 9
0.8 8.51 0.0821 9 2.23 0.175 2 8.51 0.0499 9
0.85 8.57 0.0858 9 2.24 0.180 2 8.57 0.0586 9
0.9 8.64 0.0878 9 2.27 0.160 3 8.64 0.0552 9
0.95 8.71 0.0886 9 2.29 0.159 3 8.71 0.0550 9
0.98 9.32 0.0907 9 2.33 0.231 3 9.32 0.0655 9

157
Table 9: Oscillating frequencies, amplitude response and identified modes for the
Pipe34 model in 45◦ flow at position of accelerometer seven with CDrag = 1.1

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.72 0.0004 1 0.72 0.0004 1 0.72 0.0004 1
0.1 0.78 0.0013 1 0.78 0.0013 1 0.78 0.0013 1
0.15 0.82 0.0084 1 0.82 0.228 1 1.63 0.0019 2
0.2 1.90 0.0198 2 0.85 0.246 1 2.40 0.0224 4
0.25 1.92 0.1864 2 0.87 0.258 1 1.92 0.0061 2
0.3 1.93 0.1701 2 0.88 0.298 1 1.93 0.0127 2
0.35 0.88 0.0514 1 0.88 0.276 1 2.40 0.0252 4
0.4 1.97 0.1786 2 0.98 0.357 1 2.38 0.0201 4
0.45 1.13 0.0376 1 1.13 0.176 1 2.20 0.0222 3
0.5 1.33 0.0420 1 1.33 0.132 1 2.22 0.0341 3
0.55 1.50 0.0670 2 1.50 0.142 2 2.25 0.0475 3
0.6 1.77 0.192 2 1.77 0.146 2 2.25 0.0555 3
0.65 1.92 0.424 2 1.92 0.180 2 2.22 0.0590 3
0.7 1.95 0.577 2 1.95 0.237 2 1.95 0.0729 2
0.75 1.98 0.537 2 1.98 0.254 2 1.98 0.0845 2
0.8 1.97 0.664 2 1.97 0.236 2 1.57 0.130 2
0.85 2.00 0.622 2 2.00 0.266 2 2.17 0.230 3
0.9 2.05 0.565 3 6.33 0.273 8 2.27 0.432 3
0.95 2.48 0.118 4 2.48 0.112 4 2.48 1.048 4
0.98 2.50 0.0407 4 2.48 0.118 4 2.48 1.066 4

158
Table 10: Oscillating frequencies, amplitude response and identified modes for the
Pipe34 model in 45◦ flow at position of accelerometer three with CDrag = 1.1

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.71 0.0001 1 0.71 0.0001 1 0.71 0.0001 1
0.1 0.79 0.0007 1 0.79 0.0007 1 0.79 0.0007 1
0.15 0.81 0.0018 1 0.81 0.0178 1 1.63 0.0025 2
0.2 2.40 0.0128 4 0.84 0.0215 1 2.40 0.0136 4
0.25 1.92 0.0319 2 0.87 0.0546 1 1.92 0.0210 2
0.3 1.93 0.0373 2 3.28 0.0912 5 1.93 0.0317 2
0.35 3.32 0.0325 5 3.32 0.138 5 3.32 0.0363 5
0.4 1.97 0.0577 2 3.36 0.141 5 3.36 0.0624 5
0.45 3.43 0.0617 6 3.43 0.157 6 3.43 0.0811 6
0.5 3.45 0.0781 6 3.45 0.171 6 3.45 0.101 6
0.55 3.48 0.0909 6 3.48 0.187 6 3.48 0.124 6
0.6 3.51 0.105 6 3.51 0.204 6 3.51 0.125 6
0.65 1.92 0.131 2 3.55 0.237 6 3.55 0.146 6
0.7 1.95 0.164 2 3.59 0.248 6 3.59 0.213 6
0.75 1.99 0.135 2 3.61 0.232 7 3.61 0.173 7
0.8 1.96 0.190 2 3.92 0.251 7 1.96 0.189 2
0.85 1.99 0.250 2 3.99 0.249 7 1.99 0.226 2
0.9 2.27 0.291 3 2.25 0.241 3 2.27 0.331 3
0.95 2.48 0.594 4 2.48 0.119 4 2.48 0.702 4
0.98 2.49 0.590 4 2.49 0.160 4 2.49 0.721 4

159
Table 11: Oscillating frequencies, amplitude response and identified modes for the
Pipe34 model in 45◦ flow at position of accelerometer seven with CDrag = 1.2

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.70 0.0004 1 0.70 0.0004 1 0.70 0.0004 1
0.1 0.78 0.0014 1 0.78 0.0014 1 0.78 0.0014 1
0.15 0.82 0.0069 1 0.82 0.207 1 1.63 0.0023 2
0.2 1.90 0.0249 2 0.85 0.223 1 2.40 0.0219 4
0.25 1.92 0.173 2 0.87 0.233 1 1.92 0.0048 2
0.3 0.80 0.0474 1 0.80 0.199 1 2.12 0.0131 3
0.35 1.95 0.157 2 0.92 0.337 1 2.40 0.0152 4
0.4 1.97 0.140 2 0.98 0.285 1 2.17 0.0192 3
0.45 1.97 0.110 2 1.07 0.321 1 2.35 0.0152 4
0.5 1.32 0.0437 1 1.32 0.132 1 2.23 0.0358 3
0.55 1.48 0.0813 2 1.48 0.159 2 2.27 0.0642 3
0.6 1.75 0.175 2 1.75 0.142 2 2.25 0.0680 3
0.65 1.92 0.3978 2 1.92 0.184 2 1.92 0.0620 2
0.7 1.95 0.5586 2 1.95 0.234 2 1.95 0.0711 2
0.75 1.98 0.4955 2 1.98 0.250 2 1.98 0.0713 2
0.8 1.93 0.5358 2 1.93 0.272 2 1.93 0.122 2
0.85 2.00 0.5847 2 2.00 0.248 2 2.17 0.222 3
0.9 2.05 0.4928 3 6.35 0.246 8 2.25 0.400 3
0.95 2.47 0.0371 4 2.47 0.104 4 2.47 0.969 4
0.98 2.48 0.0334 4 2.48 0.110 4 2.48 0.977 4

160
Table 12: Oscillating frequencies, amplitude response and identified modes for the
Pipe34 model in 45◦ flow at position of accelerometer three with CDrag = 1.2

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.71 0.0001 1 0.71 0.0001 1 0.71 0.0001 1
0.1 0.79 0.0006 1 0.79 0.0006 1 0.79 0.0006 1
0.15 0.81 0.0015 1 0.81 0.0161 1 0.81 0.0018 1
0.2 2.41 0.0130 4 0.84 0.0191 1 2.41 0.0140 4
0.25 1.92 0.0308 2 3.21 0.0516 5 1.92 0.0209 2
0.3 3.28 0.0200 5 3.28 0.0898 5 3.28 0.0204 5
0.35 1.95 0.0422 2 3.31 0.117 5 3.31 0.0391 5
0.4 3.37 0.0537 5 3.37 0.137 5 3.37 0.0622 5
0.45 3.43 0.0642 6 3.43 0.136 6 3.43 0.0801 6
0.5 3.45 0.0760 6 3.45 0.156 6 3.45 0.0985 6
0.55 3.49 0.0944 6 3.49 0.174 6 3.49 0.123 6
0.6 3.52 0.0979 6 3.52 0.188 6 3.52 0.123 6
0.65 1.92 0.121 2 3.56 0.220 6 3.56 0.122 6
0.7 1.95 0.147 2 3.60 0.237 7 3.60 0.174 7
0.75 1.99 0.125 2 3.63 0.215 7 3.63 0.149 7
0.8 1.93 0.191 2 3.88 0.215 7 1.93 0.190 2
0.85 2.16 0.217 3 3.99 0.228 7 2.16 0.224 3
0.9 2.25 0.287 3 2.25 0.212 3 2.25 0.296 3
0.95 2.47 0.544 4 2.47 0.111 4 2.47 0.660 4
0.98 2.48 0.532 4 2.48 0.121 4 2.48 0.667 4

161
Table 13: Oscillating frequencies, amplitude response and identified modes for the
Pipe31 model in 45◦ flow at position of accelerometer seven with CDrag = 1.1

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.71 0.0003 1 0.71 0.0003 1 0.71 0.0003 1
0.1 0.83 0.0008 1 0.83 0.0008 1 0.83 0.0008 1
0.15 0.85 0.0026 1 0.85 0.223 1 2.57 0.0023 4
0.2 0.88 0.0077 1 0.88 0.248 1 1.77 0.0053 2
0.25 2.28 0.0389 3 0.63 0.200 1 2.62 0.0688 4
0.3 2.29 0.1187 3 0.79 0.188 1 2.67 0.0386 4
0.35 2.31 0.1421 3 0.93 0.351 1 2.51 0.0313 4
0.4 2.41 0.0581 3 0.99 0.273 1 2.27 0.0218 3
0.45 2.32 0.0553 3 1.16 0.136 1 2.30 0.0302 3
0.5 2.32 0.0825 3 1.17 0.137 1 2.33 0.0305 3
0.55 1.47 0.0426 1 1.47 0.142 1 2.37 0.0846 3
0.6 1.65 0.061 2 1.65 0.124 2 2.39 0.0635 3
0.65 1.62 0.067 2 6.17 0.164 8 7.23 0.0698 9
0.7 1.79 0.099 2 6.23 0.178 8 1.87 0.0955 2
0.75 2.09 0.121 2 6.28 0.132 8 2.09 0.102 2
0.8 2.23 0.547 2 2.23 0.254 2 2.33 0.0855 3
0.85 2.28 0.576 3 2.28 0.241 3 1.63 0.122 2
0.9 2.32 0.635 3 2.32 0.276 3 1.79 0.178 2
0.95 2.36 0.621 3 2.36 0.270 3 2.36 0.274 3
0.98 2.38 0.5969 3 0.01 0.629 1 2.61 0.913 4

162
Table 14: Oscillating frequencies, amplitude response and identified modes for the
Pipe31 model in 45◦ flow at position of accelerometer three with CDrag = 1.1

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.72 0.0001 1 0.72 0.0001 1 0.72 0.0001 1
0.1 0.83 0.0003 1 0.83 0.0003 1 0.83 0.0003 1
0.15 0.85 0.0012 1 0.85 0.0202 1 2.57 0.0012 4
0.2 0.88 0.0032 1 0.88 0.0246 1 2.47 0.0026 4
0.25 2.61 0.0362 4 2.19 0.0574 2 2.61 0.0405 4
0.3 2.68 0.0267 4 3.32 0.0903 5 2.68 0.0270 4
0.35 2.31 0.0410 3 3.36 0.129 5 2.31 0.0331 3
0.4 3.40 0.0293 5 3.40 0.158 5 3.40 0.0243 5
0.45 3.44 0.0281 5 3.44 0.165 5 3.44 0.0329 5
0.5 3.49 0.0474 5 3.49 0.174 5 3.49 0.0488 5
0.55 3.57 0.0701 6 3.84 0.178 7 3.57 0.0865 6
0.6 3.69 0.108 6 3.69 0.155 6 3.69 0.136 6
0.65 3.73 0.123 6 3.92 0.197 7 3.73 0.156 6
0.7 1.87 0.112 2 3.96 0.222 7 3.79 0.125 7
0.75 8.37 0.164 9 2.09 0.263 2 2.09 0.187 2
0.8 2.24 0.168 2 2.24 0.232 2 2.24 0.146 2
0.85 2.28 0.179 3 2.28 0.245 3 2.28 0.190 3
0.9 2.32 0.200 3 2.32 0.234 3 2.32 0.159 3
0.95 2.36 0.238 3 2.36 0.209 3 2.11 0.165 2
0.98 2.45 0.576 3 2.39 0.200 3 2.47 0.615 4

163
Table 15: Oscillating frequencies, amplitude response and identified modes for the
Pipe31 model in 45◦ flow at position of accelerometer seven with CDrag = 1.2

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.70 0.0002 1 0.70 0.0002 1 0.70 0.0002 1
0.1 0.83 0.0009 1 0.83 0.0009 1 0.83 0.0009 1
0.15 0.87 0.0024 1 0.87 0.202 1 2.57 0.0014 4
0.2 0.90 0.0072 1 0.90 0.225 1 2.43 0.0051 3
0.25 2.30 0.130 3 0.90 0.176 1 2.30 0.0033 3
0.3 0.80 0.015 1 0.80 0.151 1 2.67 0.0320 4
0.35 0.90 0.0242 1 0.90 0.244 1 2.23 0.0207 2
0.4 2.33 0.158 3 1.00 0.356 1 0.73 0.0170 1
0.45 2.27 0.0456 3 1.17 0.124 1 2.30 0.0188 3
0.5 2.37 0.0616 3 1.17 0.145 1 2.33 0.0264 3
0.55 1.47 0.0429 1 1.47 0.136 1 2.37 0.0366 3
0.6 1.63 0.0543 2 1.63 0.125 2 2.40 0.0472 3
0.65 1.63 0.0703 2 6.17 0.156 8 1.07 0.0595 1
0.7 1.80 0.0811 2 6.23 0.144 8 1.90 0.0822 2
0.75 2.33 0.409 3 2.33 0.143 3 2.33 0.127 3
0.8 2.23 0.456 2 2.23 0.218 2 2.23 0.0750 2
0.85 2.27 0.539 3 2.27 0.224 3 1.63 0.106 2
0.9 2.33 0.590 3 2.33 0.257 3 1.77 0.138 2
0.95 2.37 0.577 3 2.37 0.286 3 2.37 0.252 3
0.98 2.37 0.544 3 6.50 0.321 8 2.47 0.433 4

164
Table 16: Oscillating frequencies, amplitude response and identified modes for the
Pipe31 model in 45◦ flow at position of accelerometer three with CDrag = 1.2

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.72 0.0001 1 0.72 0.0001 1 0.72 0.0001 1
0.1 0.83 0.0003 1 0.83 0.0003 1 0.83 0.0003 1
0.15 0.85 0.0010 1 0.85 0.0185 1 2.57 0.0010 4
0.2 0.88 0.0032 1 0.88 0.0221 1 2.51 0.0038 4
0.25 2.28 0.0264 3 0.92 0.0515 1 2.28 0.0144 3
0.3 2.68 0.0228 4 3.32 0.0868 5 2.68 0.0232 4
0.35 3.36 0.0205 5 3.36 0.129 5 3.36 0.0173 5
0.4 2.32 0.0433 3 3.40 0.142 5 3.40 0.0298 5
0.45 3.44 0.0281 5 3.44 0.153 5 3.44 0.0342 5
0.5 3.49 0.0486 5 3.49 0.159 5 3.49 0.0507 5
0.55 3.67 0.0833 6 3.67 0.147 6 3.67 0.0962 6
0.6 3.71 0.114 6 3.71 0.150 6 3.71 0.137 6
0.65 3.75 0.115 7 3.92 0.171 7 3.75 0.125 7
0.7 1.89 0.105 2 3.97 0.215 7 3.77 0.118 7
0.75 2.32 0.193 3 3.79 0.249 7 3.79 0.182 7
0.8 2.23 0.162 2 2.23 0.218 2 2.23 0.178 2
0.85 2.28 0.151 3 2.28 0.210 3 2.28 0.180 3
0.9 2.32 0.186 3 2.32 0.233 3 2.32 0.164 3
0.95 2.36 0.217 3 2.36 0.214 3 2.11 0.152 2
0.98 2.45 0.304 3 2.39 0.214 3 2.45 0.305 3

165
Table 17: Oscillating frequencies, amplitude response and identified modes for the
Pipe34 model in 90◦ flow at position of accelerometer seven with CDrag = 1.1

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.13 0.0002 1 0.13 0.0002 1 0.13 0.0002 1
0.1 1.87 0.0046 2 1.87 0.0046 2 1.87 0.0046 2
0.15 0.43 0.0041 1 0.85 0.247 1 0.42 0.0121 1
0.2 0.55 0.0030 1 0.90 0.253 1 2.33 0.0315 4
0.25 1.53 0.0216 2 0.97 0.205 1 2.63 0.0218 4
0.3 0.82 0.0349 1 1.03 0.031 1 0.93 0.0259 1
0.35 1.05 0.0358 1 1.07 0.044 1 1.03 0.0353 1
0.4 1.73 0.0591 2 1.13 0.050 1 1.10 0.0628 1
0.45 1.72 0.159 2 5.95 0.099 8 1.07 0.0505 1
0.5 1.67 0.309 2 6.08 0.113 8 1.18 0.0796 1
0.55 1.85 0.801 2 6.10 0.153 8 2.07 0.121 3
0.6 1.88 0.985 2 6.18 0.134 8 2.05 0.197 3
0.65 1.92 1.013 2 6.23 0.167 8 2.27 0.512 3
0.7 1.95 1.023 2 2.45 0.149 4 2.45 1.012 4
0.75 1.97 1.058 2 0.02 0.185 1 2.48 1.082 4
0.79 1.97 1.023 2 0.02 0.308 1 2.50 1.107 4

166
Table 18: Oscillating frequencies, amplitude response and identified modes for the
Pipe34 model in 90◦ flow at position of accelerometer three with CDrag = 1.1

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.13 0.0001 1 0.13 0.0001 1 0.13 0.0001 1
0.1 1.09 0.0023 1 1.09 0.0023 1 1.09 0.0023 1
0.15 0.41 0.0049 1 0.85 0.0200 1 2.11 0.0059 3
0.2 2.33 0.0130 4 0.91 0.0263 1 2.33 0.0170 4
0.25 2.64 0.0116 4 3.21 0.0895 5 2.64 0.0129 4
0.3 0.93 0.0196 1 3.29 0.132 5 0.93 0.0221 1
0.35 3.24 0.0284 5 3.35 0.162 5 3.24 0.0399 5
0.4 3.37 0.0606 5 3.40 0.188 5 3.37 0.0821 5
0.45 1.72 0.0568 2 3.45 0.200 6 3.45 0.0636 6
0.5 1.67 0.0933 2 3.51 0.198 6 1.67 0.0655 2
0.55 1.84 0.158 2 3.91 0.215 7 1.84 0.125 2
0.6 1.89 0.249 2 3.95 0.220 7 1.89 0.200 2
0.65 1.92 0.401 2 3.84 0.233 7 2.28 0.414 4
0.7 2.45 0.654 4 3.89 0.273 7 2.45 0.753 4
0.75 2.48 0.712 4 2.48 0.234 4 2.48 0.811 4
0.79 2.51 0.734 4 2.51 0.321 4 2.51 0.840 4

167
Table 19: Oscillating frequencies, amplitude response and identified modes for the
Pipe34 model in 90◦ flow at position of accelerometer seven with CDrag = 1.2

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.15 0.0002 1 0.15 0.0002 1 0.15 0.0002 1
0.1 1.87 0.0044 2 1.87 0.0044 2 1.87 0.0044 2
0.15 0.42 0.0033 1 0.85 0.226 1 0.42 0.0121 1
0.2 1.48 0.0061 2 0.92 0.228 1 2.33 0.0295 4
0.25 1.53 0.0229 2 1.00 0.203 1 2.65 0.0210 4
0.3 0.90 0.0349 1 1.05 0.030 1 0.97 0.0192 1
0.35 1.02 0.034 1 1.08 0.041 1 1.03 0.0382 1
0.4 1.73 0.055 2 1.17 0.049 1 1.10 0.0546 1
0.45 1.73 0.141 2 5.95 0.086 8 1.38 0.0555 2
0.5 1.67 0.289 2 6.07 0.105 8 1.18 0.0646 1
0.55 1.83 0.735 2 6.10 0.140 8 2.08 0.125 3
0.6 1.88 0.997 2 6.18 0.159 8 2.08 0.185 3
0.65 1.92 0.944 2 6.23 0.162 8 2.25 0.478 3
0.7 1.93 0.939 2 2.43 0.146 4 2.43 0.930 4
0.75 1.95 0.957 2 2.47 0.140 4 2.47 1.006 4
0.79 1.97 0.973 2 0.02 0.213 1 2.50 1.053 4

168
Table 20: Oscillating frequencies, amplitude response and identified modes for the
Pipe34 model in 90◦ flow at position of accelerometer three with CDrag = 1.2

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.13 0.0001 1 0.13 0.0001 1 0.13 0.0001 1
0.1 0.28 0.0022 1 0.28 0.0022 1 0.28 0.0022 1
0.15 0.41 0.0053 1 0.85 0.0183 1 0.41 0.0061 1
0.2 2.35 0.0122 4 2.85 0.0264 4 2.35 0.0160 4
0.25 2.64 0.0120 4 3.20 0.0803 5 2.64 0.0120 4
0.3 0.96 0.0193 1 3.29 0.129 5 0.96 0.0179 1
0.35 3.25 0.0287 5 3.35 0.149 5 3.25 0.0415 5
0.4 3.37 0.0564 5 3.40 0.169 5 3.37 0.0737 5
0.45 1.37 0.0501 2 3.45 0.182 6 1.43 0.0512 2
0.5 1.67 0.0885 2 3.52 0.188 6 1.67 0.0705 2
0.55 1.84 0.150 2 3.92 0.189 7 1.84 0.119 2
0.6 1.89 0.231 2 3.79 0.201 7 1.89 0.193 2
0.65 2.25 0.367 3 3.83 0.236 7 2.25 0.386 3
0.7 2.44 0.605 4 3.88 0.272 7 2.44 0.694 4
0.75 2.47 0.660 4 4.43 0.236 7 2.47 0.757 4
0.79 2.49 0.685 4 4.47 0.301 7 2.49 0.786 4

169
Table 21: Oscillating frequencies, amplitude response and identified modes for the
Pipe31 model in 90◦ flow at position of accelerometer seven with CDrag = 1.1

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.13 0.0002 1 0.13 0.0002 1 0.13 0.0002 1
0.1 0.27 0.0015 1 0.27 0.0015 1 0.27 0.0015 1
0.15 2.16 0.0061 2 0.89 0.246 1 0.41 0.0103 1
0.2 0.56 0.0043 1 0.95 0.256 1 2.41 0.0227 3
0.25 2.73 0.0156 4 1.03 0.225 1 2.75 0.0239 4
0.3 1.61 0.0248 2 1.09 0.225 1 0.92 0.0225 1
0.35 0.93 0.0515 1 1.23 0.197 1 1.05 0.0470 1
0.4 2.08 0.0373 2 1.28 0.161 1 1.16 0.0777 1
0.45 2.00 0.0410 2 1.37 0.176 1 1.21 0.0950 1
0.5 1.93 0.0568 2 6.15 0.167 8 2.40 0.0821 3
0.55 1.80 0.131 2 6.24 0.177 8 1.24 0.113 1
0.6 2.01 0.253 2 6.31 0.150 8 2.01 0.173 2
0.65 2.20 0.688 2 6.36 0.164 8 2.20 0.258 2
0.7 2.29 1.057 3 0.01 0.216 1 2.60 0.861 4
0.75 2.31 1.024 3 0.01 0.241 1 2.63 0.997 4
0.79 2.32 0.988 3 0.01 0.245 1 2.67 1.068 4

170
Table 22: Oscillating frequencies, amplitude response and identified modes for the
Pipe31 model in 90◦ flow at position of accelerometer three with CDrag = 1.1

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.13 0.0001 1 0.13 0.0001 1 0.13 0.0001 1
0.1 0.27 0.0026 1 0.27 0.0026 1 0.27 0.0026 1
0.15 0.41 0.0050 1 0.89 0.0232 1 0.41 0.0052 1
0.2 2.41 0.0104 3 0.95 0.0271 1 2.41 0.0117 3
0.25 2.75 0.0158 4 3.25 0.0857 5 2.75 0.0153 4
0.3 0.92 0.0158 1 3.35 0.133 5 0.87 0.0180 1
0.35 1.05 0.0246 1 3.40 0.164 5 1.05 0.0303 1
0.4 3.52 0.0438 5 3.44 0.185 5 3.52 0.0504 5
0.45 3.64 0.0592 6 3.49 0.197 5 3.64 0.0727 6
0.5 2.40 0.0517 3 3.55 0.210 5 2.40 0.0505 3
0.55 1.80 0.0589 2 3.60 0.195 6 1.80 0.0705 2
0.6 2.01 0.0751 2 3.67 0.195 6 2.01 0.106 2
0.65 2.20 0.196 2 3.79 0.189 7 2.20 0.176 2
0.7 2.60 0.661 4 4.29 0.193 7 2.60 0.626 4
0.75 2.63 0.722 4 2.63 0.243 4 2.63 0.716 4
0.79 2.67 0.761 4 2.67 0.200 4 2.67 0.769 4

171
Table 23: Oscillating frequencies, amplitude response and identified modes for the
Pipe31 model in 90◦ flow at position of accelerometer seven with CDrag = 1.2

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.13 0.0002 1 0.13 0.0002 1 0.13 0.0002 1
0.1 0.27 0.0012 1 0.27 0.0012 1 0.27 0.0012 1
0.15 2.17 0.0063 2 0.88 0.222 1 0.42 0.0100 1
0.2 0.55 0.0012 1 0.95 0.234 1 2.42 0.0220 3
0.25 2.75 0.0123 4 1.05 0.201 1 2.75 0.0232 4
0.3 1.62 0.0137 2 2.27 0.0158 3 0.93 0.0159 1
0.35 0.95 0.0425 1 1.23 0.0328 1 1.08 0.0410 1
0.4 2.08 0.0276 2 1.30 0.0419 1 1.15 0.0438 1
0.45 2.07 0.0427 2 1.40 0.0588 1 1.20 0.0514 1
0.5 1.88 0.0410 2 6.15 0.102 8 1.15 0.0750 1
0.55 1.80 0.133 2 6.23 0.117 8 1.27 0.0930 1
0.6 2.02 0.240 2 6.32 0.141 8 2.02 0.155 2
0.65 2.20 0.637 2 6.37 0.148 8 2.20 0.227 2
0.7 2.27 0.852 3 6.37 0.184 8 2.45 0.529 3
0.75 2.30 0.945 3 2.62 0.153 4 2.62 0.910 4
0.79 2.32 0.942 3 0.02 0.209 1 2.65 0.994 4

172
Table 24: Oscillating frequencies, amplitude response and identified modes for the
Pipe31 model in 90◦ flow at position of accelerometer three with CDrag = 1.2

x-direction y-direction z-direction

Uc fosc A/D mode fosc A/D mode fosc A/D mode
[m/s] [Hz] [-] [-] [Hz] [-] [-] [Hz] [-] [-]
0.05 0.13 0.0001 1 0.13 0.0001 1 0.13 0.0001 1
0.1 0.28 0.0018 1 0.28 0.0018 1 0.28 0.0018 1
0.15 0.41 0.0052 1 0.89 0.0215 1 0.41 0.0052 1
0.2 2.43 0.0102 3 0.95 0.0258 1 2.43 0.0112 3
0.25 2.76 0.0149 4 3.25 0.0834 5 2.76 0.0167 4
0.3 0.93 0.0125 1 3.67 0.0972 6 0.93 0.0134 1
0.35 1.03 0.0245 1 3.40 0.156 5 1.03 0.0295 1
0.4 3.52 0.0427 5 3.44 0.170 5 3.52 0.0503 5
0.45 3.65 0.0583 6 3.49 0.179 5 3.65 0.0712 6
0.5 2.40 0.0510 3 3.56 0.191 6 2.40 0.0474 3
0.55 1.81 0.0659 2 3.61 0.178 6 1.81 0.0702 2
0.6 2.03 0.0768 2 3.69 0.175 6 2.03 0.103 2
0.65 2.20 0.192 2 3.81 0.181 7 2.20 0.168 2
0.7 2.44 0.421 3 4.25 0.192 7 2.44 0.378 3
0.75 2.61 0.660 4 2.61 0.239 4 2.61 0.657 4
0.79 2.64 0.705 4 2.64 0.205 4 2.64 0.715 4

173
Appendix C: Response model calculations
C.1 10◦ flow direction

Figure 24: IL response model for the 10◦ flow direction

Figure 25: CF response model for the 10◦ flow direction

174
 Table 25: IL unit response amplitudes for 10◦ flow with Pipe34 model

mode 2 mode 4 mode 6 mode 9

Uc Vred (f0 ) A/D Vred (f0 ) A/D Vred (f0 ) A/D Vred (f0 ) A/D
[m/s] [-] [-] [-] [-] [-] [-] [-] [-]
0.05 0.43 0 0.34 0 0.24 0 0.12 0
0.1 0.86 0 0.67 0 0.48 0 0.25 0
0.15 1.30 0.0296 1.01 0.00111 0.71 0 0.37 0
0.2 1.73 0.0728 1.35 0.0348 0.95 0 0.50 0
0.25 2.16 0.1160 1.69 0.0685 1.19 0.0190 0.62 0
0.3 2.59 0.1376 2.02 0.1022 1.43 0.0427 0.75 0
0.35 3.02 0.1238 2.36 0.1359 1.66 0.0665 0.87 0
0.4 3.46 0.1101 2.70 0.1343 1.90 0.0903 1.00 0
0.45 3.89 0.0964 3.03 0.1235 2.14 0.1141 1.12 0.0125
0.5 4.32 0 3.37 0.1128 2.38 0.1379 1.25 0.0250
0.55 4.75 0 3.71 0.1021 2.61 0.1368 1.37 0.0374
0.6 5.18 0 4.04 0.0859 2.85 0.1292 1.50 0.0499
0.65 5.62 0 4.38 0 3.09 0.1217 1.62 0.0624
0.7 6.05 0 4.72 0 3.33 0.1141 1.75 0.0749
0.75 6.48 0 5.06 0 3.57 0.1065 1.87 0.0874
0.8 6.91 0 5.39 0 3.80 0.0990 2.00 0.0999
0.85 7.34 0 5.73 0 4.04 0.0859 2.12 0.1124
0.9 7.78 0 6.07 0 4.28 0 2.25 0.1249
0.95 8.21 0 6.40 0 4.52 0 2.37 0.1374
0.98 8.47 0 6.61 0 4.66 0 2.45 0.1421

175
 Table 26: CF unit response amplitudes for 10◦ flow with Pipe34 model

mode 1 mode 3 mode 5

Uc Vred (f0 ) A/D Vred (f0 ) A/D Vred (f0 ) A/D mode with max A/D
[m/s] [-] [-] [-] [-] [-] [-] [-]
0.05 1.02 0 0.39 0 0.25 0 no CF mode
0.1 2.05 0.00701 0.79 0 0.50 0 mode 1
0.15 3.07 0.169 1.18 0 0.74 0 mode 1
0.2 4.09 0.463 1.57 0 0.99 0 mode 1
0.25 5.12 0.757 1.96 0 1.24 0 mode 1
0.3 6.14 0.919 2.36 0.0534 1.49 0 mode 1
0.35 7.16 0.919 2.75 0.112 1.73 0 mode 1
0.4 8.19 0.919 3.14 0.190 1.98 0 mode 1
0.45 9.21 0.919 3.53 0.302 2.23 0.034 mode 1
0.5 10.23 0.919 3.93 0.415 2.48 0.072 mode 1
0.55 11.26 0.880 4.32 0.528 2.73 0.109 mode 1
0.6 12.28 0.690 4.71 0.640 2.97 0.146 mode 1
0.65 13.30 0.500 5.10 0.753 3.22 0.212 mode 3
0.7 14.33 0.310 5.50 0.866 3.47 0.283 mode 3
0.75 15.35 0.120 5.89 0.919 3.72 0.355 mode 3
0.8 16.37 0 6.28 0.919 3.96 0.426 mode 3
0.85 17.40 0 6.68 0.919 4.21 0.497 mode 3
0.9 18.42 0 7.07 0.919 4.46 0.568 mode 3
0.95 19.44 0 7.46 0.919 4.71 0.639 mode 3
0.98 20.06 0 7.70 0.919 4.86 0.682 mode 3

176
Table 27: Ca -Correction and CF-induced IL modes for 10◦ flow with Pipe34 model

mode 1 mode 3
Uc Ca,CF-RES fCF-RES 2×fCF-RES Ca,CF-RES fCF-RES 2×fCF-RES CF-IL candidate
[m/s] [-] [Hz] [Hz] [-] [Hz] [Hz] [-]
0.05 1 0.808 1.62 1 2.105 4.21 -
0.1 1.42 0.771 1.54 1 2.105 4.21 2
0.15 4.93 0.587 1.17 1 2.105 4.21 2
0.2 3.91 0.626 1.25 1 2.105 4.21 2
0.25 2.88 0.676 1.35 1 2.105 4.21 2
0.3 1.86 0.738 1.48 4.204 1.600 3.20 2
0.35 0.84 0.823 1.65 5.251 1.500 3.00 2
0.4 -0.19 0.945 1.89 4.859 1.535 3.07 2
0.45 -0.50 1.00 1.99 4.466 1.573 3.15 2
0.5 1 0.81 1.62 4.073 1.614 3.23 2
0.55 1 0.808 1.62 3.681 1.658 3.32 2
0.6 1 0.808 1.62 3.288 1.706 3.41 2
0.65 1 0.808 1.62 2.895 1.759 3.52 6
0.7 1 0.808 1.62 2.503 1.817 3.63 6
0.75 1 0.808 1.62 2.110 1.880 3.76 6
0.8 1 0.808 1.62 1.717 1.951 3.90 6
0.85 1 0.808 1.62 1.325 2.031 4.06 6
0.9 1 0.808 1.62 0.932 2.121 4.24 6
0.95 1 0.808 1.62 0.539 2.225 4.45 6
0.98 1 0.808 1.62 0.304 2.295 4.59 6

177
C.2 90◦ flow direction

Figure 26: IL response model for the 90◦ flow direction

Figure 27: CF response model for the 90◦ flow direction

178
 Table 28: IL unit response amplitudes for 90◦ flow with Pipe34 model

mode 1 mode 3 mode 5 mode 7

Uc Vred (f0 ) A/D Vred (f0 ) A/D Vred (f0 ) A/D Vred (f0 ) A/D
[m/s] [-] [-] [-] [-] [-] [-] [-] [-]
0.05 1.02 0 0.39 0 0.25 0 0.22 0
0.1 2.05 0.095 0.79 0 0.50 0 0.44 0
0.15 3.07 0.101 1.18 0.008 0.74 0 0.67 0
0.2 4.09 0.002 1.57 0.047 0.99 0 0.89 0
0.25 5.12 0 1.96 0.086 1.24 0.014 1.11 0.001
0.3 6.14 0 2.36 0.116 1.49 0.039 1.33 0.023
0.35 7.16 0 2.75 0.108 1.73 0.063 1.56 0.046
0.4 8.19 0 3.14 0.099 1.98 0.088 1.78 0.068
0.45 9.21 0 3.53 0.091 2.23 0.113 2.00 0.090
0.5 10.23 0 3.93 0.083 2.48 0.114 2.22 0.112
0.55 11.26 0 4.32 0 2.73 0.108 2.45 0.114
0.6 12.28 0 4.71 0 2.97 0.103 2.67 0.110
0.65 13.30 0 5.10 0 3.22 0.098 2.89 0.105
0.7 14.33 0 5.50 0 3.47 0.092 3.11 0.100
0.75 15.35 0 5.89 0 3.72 0.087 3.33 0.095
0.79 16.37 0 6.28 0 3.96 0.067 3.56 0.091

179
 Table 29: CF unit response amplitudes for 90◦ flow with Pipe34 model

mode 2 mode 4 mode 6

Uc Vred (f0 ) A/D Vred (f0 ) A/D Vred (f0 ) A/D mode with max A/D
[m/s] [-] [-] [-] [-] [-] [-] [-]
0.05 0.43 0 0.34 0 0.24 0 no CF mode
0.1 0.86 0 0.67 0 0.48 0 no CF mode
0.15 1.30 0 1.01 0 0.71 0 no CF mode
0.2 1.73 0 1.35 0 0.95 0 no CF mode
0.25 2.16 0.024 1.69 0 1.19 0 mode 2
0.3 2.59 0.089 2.02 0.0033 1.43 0 mode 2
0.35 3.02 0.156 2.36 0.054 1.67 0 mode 2
0.4 3.46 0.280 2.70 0.104 1.90 0 mode 2
0.45 3.89 0.404 3.03 0.159 2.14 0.0212 mode 2
0.5 4.32 0.528 3.37 0.255 2.38 0.057 mode 2
0.55 4.75 0.652 3.71 0.352 2.62 0.093 mode 2
0.6 5.18 0.776 4.04 0.449 2.85 0.128 mode 2
0.65 5.62 0.900 4.38 0.545 3.09 0.176 mode 2
0.7 6.05 0.900 4.72 0.642 3.33 0.244 mode 2
0.75 6.48 0.900 5.06 0.739 3.57 0.312 mode 2
0.79 6.91 0.900 5.39 0.836 3.81 0.380 mode 2

180
Table 30: Ca -Correction and CF-induced IL modes for 90◦ flow with Pipe34 model

mode 2
Uc Ca,CF-RES fCF-RES 2×fCF-RES CF-IL candidate
[m/s] [-] [Hz] [Hz] [-]
0.05 1 1.913 3.83 -
0.1 1 1.913 3.83 -
0.15 1 1.913 3.83 -
0.2 1 1.913 3.83 -
0.25 2.44 1.603 3.21 5
0.3 5.41 1.261 2.52 3
0.35 4.98 1.298 2.60 3
0.4 4.54 1.338 2.68 3
0.45 4.11 1.38 2.76 5
0.5 3.68 1.43 2.86 5
0.55 3.25 1.483 2.97 5
0.6 2.82 1.544 3.09 5
0.65 2.38 1.612 3.22 5
0.7 1.95 1.690 3.38 5
0.75 1.52 1.781 3.56 7
0.79 1.09 1.889 3.78 7

181
C.3 45◦ flow direction

Figure 28: IL response model for the 45◦ flow direction

Figure 29: CF response model for the 45◦ flow direction

182
 Table 31: IL unit response amplitudes for 45◦ flow with Pipe34 model

mode 1 mode 2 mode 3 mode 5

Uc Vred (f0 ) A/D Vred (f0 ) A/D Vred (f0 ) A/D Vred (f0 ) A/D
[m/s] [-] [-] [-] [-] [-] [-] [-] [-]
0.05 1.02 0 0.43 0 0.39 0 0.25 0
0.1 2.05 0.0947 0.86 0 0.79 0 0.50 0
0.15 3.07 0.0886 1.30 0.0196 1.18 0.0078 0.74 0
0.2 4.09 0.0022 1.73 0.0628 1.57 0.0471 0.99 0
0.25 5.12 0 2.16 0.0947 1.96 0.0863 1.24 0.0139
0.3 6.14 0 2.59 0.0918 2.36 0.0934 1.49 0.0386
0.35 7.16 0 3.02 0.0890 2.75 0.0908 1.73 0.0634
0.4 8.19 0 3.46 0.0861 3.14 0.0882 1.98 0.0882
0.45 9.21 0 3.89 0.0832 3.53 0.0856 2.23 0.0942
0.5 10.23 0 4.32 0 3.93 0.0830 2.48 0.0926
0.55 11.26 0 4.75 0 4.32 0 2.73 0.0909
0.6 12.28 0 5.18 0 4.71 0 2.97 0.0893
0.65 13.30 0 5.62 0 5.10 0 3.22 0.0876
0.7 14.33 0 6.05 0 5.50 0 3.47 0.0860
0.75 15.35 0 6.48 0 5.89 0 3.72 0.0844
0.8 16.37 0 6.91 0 6.28 0 3.96 0.0670
0.85 17.40 0 7.34 0 6.68 0 4.21 0
0.9 18.42 0 7.78 0 7.07 0 4.46 0
0.95 19.44 0 8.21 0 7.46 0 4.71 0
0.98 20.06 0 8.47 0 7.70 0 4.86 0

183
 Table 32: CF unit response amplitudes for 45◦ flow with Pipe34 model

mode 1 mode 2 mode 3

Uc Vred (f0 ) A/D Vred (f0 ) A/D Vred (f0 ) A/D mode with max A/D
[m/s] [-] [-] [-] [-] [-] [-] [-]
0.05 1.02 0 0.43 0 0.39 0 no CF mode
0.1 2.05 0.0070 0.86 0 0.79 0 mode 1
0.15 3.07 0.169 1.30 0 1.18 0 mode 1
0.2 4.09 0.463 1.73 0 1.57 0 mode 1
0.25 5.12 0.757 2.16 0.0240 1.96 0 mode 1
0.3 6.14 0.900 2.59 0.089 2.36 0.0534 mode 1
0.35 7.16 0.900 3.02 0.156 2.75 0.112 mode 1
0.4 8.19 0.900 3.46 0.280 3.14 0.190 mode 1
0.45 9.21 0.900 3.89 0.404 3.53 0.302 mode 1
0.5 10.23 0.900 4.32 0.528 3.93 0.415 mode 1
0.55 11.26 0.880 4.75 0.652 4.32 0.528 mode 1
0.6 12.28 0.690 5.18 0.776 4.71 0.640 mode 2
0.65 13.30 0.500 5.62 0.900 5.10 0.753 mode 2
0.7 14.33 0.310 6.05 0.900 5.50 0.866 mode 2
0.75 15.35 0.120 6.48 0.900 5.89 0.900 mode 2 and 3
0.8 16.37 0 6.91 0.900 6.28 0.900 mode 2 and 3
0.85 17.40 0 7.34 0.900 6.68 0.900 mode 2 and 3
0.9 18.42 0 7.78 0.900 7.07 0.900 mode 2 and 3
0.95 19.44 0 8.21 0.900 7.46 0.900 mode 2 and 3
0.98 20.06 0 8.47 0.900 7.70 0.900 mode 2 and 3

184
Table 33: Ca -Correction and CF-induced IL modes for 45◦ flow with Pipe34 model

mode 1 mode 2
Uc Ca,CF-RES fCF-RES 2×fCF-RES Ca,CF-RES fCF-RES 2×fCF-RES
[m/s] [-] [Hz] [Hz] [-] [Hz] [Hz]
0.05 1 0.81 1.62 1 1.91 3.83
0.1 1 0.76 1.52 1 1.91 3.83
0.15 4.93 0.55 1.10 1 1.91 3.83
0.2 3.91 0.59 1.19 1 1.91 3.83
0.25 2.88 0.65 1.30 2 1.60 3.21
0.3 1.86 0.72 1.44 5.41 1.26 2.52
0.35 0.84 0.83 1.66 4.98 1.30 2.60
0.4 -0.19 1.00 2.00 4.54 1.34 2.68
0.45 -0.50 1.08 2.16 4.11 1.38 2.76
0.5 1 0.81 1.62 3.68 1.43 2.86
0.55 1 0.81 1.62 3.25 1.48 2.97
0.6 1 0.81 1.62 2.82 1.54 3.09
0.65 1 0.81 1.62 2.38 1.61 3.22
0.7 1 0.81 1.62 1.95 1.69 3.38
0.75 1 0.81 1.62 1.52 1.78 3.56
0.8 1 0.81 1.62 1.09 1.89 3.78
0.85 1 0.81 1.62 0.66 2.02 4.04
0.9 1 0.81 1.62 0.22 2.18 4.36
0.95 1 0.81 1.62 -0.21 2.38 4.77
0.98 1 0.81 1.62 -0.47 2.54 5.08

185
Table 34: Ca -Correction and CF-induced IL modes for 45◦ flow with Pipe34 model

mode 3
Uc Ca,CF-RES fCF-RES 2×fCF-RES CF-IL candidate
[m/s] [-] [Hz] [Hz] [-]
0.05 1 2.10 4.21 -
0.1 1 2.10 4.21 2
0.15 1 2.10 4.21 1
0.2 1 2.10 4.21 1
0.25 1 2.10 4.21 1
0.3 4.20 1.51 3.02 2
0.35 5.25 1.40 2.80 2
0.4 4.86 1.44 2.88 2
0.45 4.47 1.48 2.96 3
0.5 4.07 1.52 3.05 2
0.55 3.68 1.57 3.15 2
0.6 3.29 1.63 3.25 2
0.65 2.90 1.69 3.37 5
0.7 2.50 1.75 3.50 5
0.75 2.11 1.83 3.65 6
0.8 1.72 1.91 3.82 7
0.85 1.32 2.01 4.02 7
0.9 0.93 2.13 4.25 7
0.95 0.54 2.26 4.53 7
0.98 0.30 2.36 4.72 8

186
 Appendix D: Matlab routine for fatigue damage
Main body
1 clc ; clear ; c l f ;
2
3 D_out = 0 . 0 6 0 5 ; %m
4 D_inn = 0 . 0 5 5 ; %m
5 R_out=D_out / 2 ; %m
6 t =(D_out−D_inn ) / 2 ; %m
7 I _ r i n g=pi ∗R_out^3∗ t ; %m^4
8 J_x=2∗ I _ r i n g ; %m^4
9 W_ring=I _ r i n g /R_out ; %m^3
10 Time=linspace ( 0 , 2 0 0 0 , 200000) ; %s e c
11 Timesec=linspace ( 0 , 2 0 0 0 , 150020) ; %s e c
12 Time1=linspace ( 1 , 7 5 0 1 , 7 5 0 1 ) ; %s e c ( f o r one l o a d c a s e )
13 n c u r r =20;
14
15 t h e t a _ i=pi / 8 ∗ [ 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 1 6 ] ; %rad
16 Uc = [ 0 . 0 5 0 . 1 0 . 1 5 0 . 2 0 0 . 2 5 0 . 3 0 0 . 3 5 0 . 4 0 . 4 5 0 . 5 0 . 5 5 0 . 6
0 . 6 5 0 . 7 0 . 7 5 0 . 8 0 . 8 5 0 . 9 0 . 9 5 0 . 9 8 ] ; %c u r r e n t s p e e d i n
m/ s
17 %f 0 =[0.837709304194545 , 2.19914089174982 , 2.28591606607719 ,
2. 62849624468059 , 3.38003525448659 , 3.72608396290345 ,
3. 76461284710912 , 6.34128947642239 , 7 . 1 6 1 3 9 2 7 3 3 9 0 4 4 5 ] ; %
Pipe31 model
18 f 0 = [ 0 . 8 0 7 5 7 9 5 8 9 6 7 2 3 8 1 , 1 . 9 1 3 0 3 0 4 8 8 7 0 5 9 3 , 2 . 1 0 4 7 2 4 6 1 0 0 4 9 7 5 ,
2.45217339029321 , 3.33582603189090 , 3.47389120695643 ,
3.71780380586922 , 6.27390743330342 , 6.61419043114963]; %
Pipe34 model
19
20 %FFT i n p u t
21 dt = 0 . 0 1 ;
22 fMaxM=1/dt ;
23 fnyq=fMaxM / 2 ;
24 Ndt =7501;
25 F=0:fMaxM/ ( Ndt−1) : fMaxM ;
26 %p l o t (F , a b s ( b ) )
27 %%

187
28 %c a l l i n g d a t a from DYNPLOT. For f i l e e x t e n s i o n A: C_Drag
=1.1 , f o r B: C_Drag=1.2
29 [ Mx_05 , Mx_10 , Mx_15 , Mx_20 , Mx_25 , Mx_30 , Mx_35 , Mx_40 , Mx_45 ,
Mx_50 , . . .
30 Mx_55 , Mx_60 , Mx_65 , Mx_70 , Mx_75 , Mx_80 , Mx_85 , Mx_90 , Mx_95 ,
Mx_98 , . . .
31 My_05 , My_10 , My_15 , My_20 , My_25 , My_30 , My_35 , My_40 , My_45
, My_50 , . . .
32 My_55 , My_60 , My_65 , My_70 , My_75 , My_80 , My_85 , My_90 , My_95 ,
My_98 , . . .
33 Mz_05 , Mz_10 , Mz_15 , Mz_20 , Mz_25 , Mz_30 , Mz_35 , Mz_40 , Mz_45 ,
Mz_50 , . . .
34 Mz_55 , Mz_60 , Mz_65 , Mz_70 , Mz_75 , Mz_80 , Mz_85 , Mz_90 , Mz_95 ,
Mz_98]= calling_stress_data_EL1_A ( ) ;
35
36 Mx=[Mx_05 Mx_10 Mx_15 Mx_20 Mx_25 Mx_30 Mx_35 Mx_40 Mx_45
Mx_50 Mx_55 Mx_60 , . . .
37 Mx_65 Mx_70 Mx_75 Mx_80 Mx_85 Mx_90 Mx_95 Mx_98 ] ;
38 My=[My_05 My_10 My_15 , My_20 , My_25 , My_30 , My_35 , My_40 , My_45 ,
My_50 , My_55 , My_60 , . . .
39 My_65 , My_70 , My_75 , My_80 , My_85 , My_90 , My_95 , My_98 ] ;
40 Mz=[Mz_05 Mz_10 Mz_15 , Mz_20 , Mz_25 , Mz_30 , Mz_35 , Mz_40 , Mz_45 ,
Mz_50 , Mz_55 , Mz_60 , . . .
41 Mz_65 , Mz_70 , Mz_75 , Mz_80 , Mz_85 , Mz_90 , Mz_95 , Mz_98 ] ;
42
43 Pos0 =[1 10001 20001 30001 40001 50001 60001 70001 80001
90001 100001 110001 120001 130001 140001 150001 160001
170001 180001 1 9 0 0 0 1 ] ;
44 Pos1 =[10000 20000 30000 40000 50000 60000 70000 80000 90000
100000 110000 120000 130000 140000 150000 160000 170000
180000 190000 2 0 0 0 0 0 ] ;
45
46 P o s 0 s e c =[1 7502 15003 22504 30005 37506 45007 52508 60009
67510 75011 82512 90013 97514 105015 112516 120017
127518 135019 1 4 2 5 2 0 ] ;
47 P o s 1 s e c =[7501 15002 22503 30004 37505 45006 52507 60008
67509 75010 82511 90012 97513 105014 112515 120016
127517 135018 142519 1 5 0 0 2 0 ] ;
48

188
49 %%
50 %p r e a l l o c a t e s i z e s o f a r r a y s
51 MXX=zeros ( 1 0 0 0 0 , 2 0 ) ;
52 MXXtot=zeros ( 2 0 0 0 0 0 , 1 ) ;
53 sectionMX=zeros ( 7 5 0 1 , 2 0 ) ;
54 sectionMX0=zeros ( 7 5 0 1 , 2 0 ) ;
55 sectionMXtot=zeros ( 1 5 0 0 2 0 , 1 ) ;
56 MaxMX=zeros ( 1 , 2 0 ) ;
57 FFTiX=zeros ( 7 5 0 1 , 2 0 ) ;
58 FFTabsX=zeros ( 7 5 0 1 , 2 0 ) ;
59 FFTMaxMX=zeros ( 1 , 2 0 ) ;
60 f_oscX=zeros ( 1 , 2 0 ) ;
61
62 MYY=zeros ( 1 0 0 0 0 , 2 0 ) ;
63 MYYtot=zeros ( 2 0 0 0 0 0 , 1 ) ;
64 sectionMY=zeros ( 7 5 0 1 , 2 0 ) ;
65 sectionMY0=zeros ( 7 5 0 1 , 2 0 ) ;
66 sectionMYtot=zeros ( 1 5 0 0 2 0 , 1 ) ;
67 MaxMY=zeros ( 1 , 2 0 ) ;
68 FFTiY=zeros ( 7 5 0 1 , 2 0 ) ;
69 FFTabsY=zeros ( 7 5 0 1 , 2 0 ) ;
70 FFTMaxMY=zeros ( 1 , 2 0 ) ;
71 f_oscY=zeros ( 1 , 2 0 ) ;
72
73 MZZ=zeros ( 1 0 0 0 0 , 2 0 ) ;
74 MZZtot=zeros ( 2 0 0 0 0 0 , 1 ) ;
75 sectionMZ=zeros ( 7 5 0 1 , 2 0 ) ;
76 sectionMZ0=zeros ( 7 5 0 1 , 2 0 ) ;
77 s e c t i o n M Z t o t=zeros ( 1 5 0 0 2 0 , 1 ) ;
78 MaxMZ=zeros ( 1 , 2 0 ) ;
79 FFTiZ=zeros ( 7 5 0 1 , 2 0 ) ;
80 FFTabsZ=zeros ( 7 5 0 1 , 2 0 ) ;
81 FFTMaxMZ=zeros ( 1 , 2 0 ) ;
82 f_oscZ=zeros ( 1 , 2 0 ) ;
83
84 sigma_xx_MY=zeros ( 7 5 0 1 , 2 0 , 1 6 ) ;
85 sigma_xx_MZ=zeros ( 7 5 0 1 , 2 0 , 1 6 ) ;
86 sigma_xxi=zeros ( 7 5 0 1 , 2 0 , 1 6 ) ;
87 %%

189
88 for i =1: n c u r r
89 %s c a l e moments around x−a x i s ( c o n d i t i o n when i n i t i a l
curvature is applied )
90 MXX( : , i ) =(Mx( : , i )−Mx( 1 6 0 , i ) ) ;
91 MXXtot( Pos0 ( i ) : Pos1 ( i ) )=MXX( : , i ) ;
92 MYY( : , i ) =(My( : , i )−My( 1 6 0 , i ) ) ;
93 MYYtot( Pos0 ( i ) : Pos1 ( i ) )=MYY( : , i ) ;
94 MZZ( : , i ) =(Mz ( : , i )−Mz( 1 6 0 , i ) ) ;
95 MZZtot ( Pos0 ( i ) : Pos1 ( i ) )=MZZ( : , i ) ;
96
97 %moments o v e r 75 s e c o n d s ( s t e a d y c u r r e n t c o n d i t i o n )
98 sectionMX ( : , i )=MXX( 1 0 0 0 : 8 5 0 0 , i ) ; %s e l e c t d a t a o v e r 75
seconds
99 sectionMX0 ( : , i )=sectionMX ( : , i )−mean( sectionMX ( : , i ) ) ; %s c a l e
i t around x−a x i s ( z e r o mean )
100 sectionMXtot ( P o s 0 s e c ( i ) : P o s 1 s e c ( i ) )=sectionMX0 ( : , i ) ;
101 MaxMX( i )=max( sectionMX0 ( : , i ) ) ; %maximum moment due t o VIV
response
102 %FFT
103 FFTiX ( : , i )=f f t ( sectionMX0 ( : , i ) ) ;
104 FFTabsX ( : , i )=abs (FFTiX ( : , i ) ) ;
105 [ FFTMaxX, indexX ] = max(FFTabsX ( : , i ) ) ;
106 f_oscX ( i )= F( indexX ) ;
107
108 sectionMY ( : , i )=MYY( 1 0 0 0 : 8 5 0 0 , i ) ; %s e l e c t d a t a o v e r 75
seconds
109 sectionMY0 ( : , i )=sectionMY ( : , i )−mean( sectionMY ( : , i ) ) ; %s c a l e
i t around z e r o mean
110 sectionMYtot ( P o s 0 s e c ( i ) : P o s 1 s e c ( i ) )=sectionMY0 ( : , i ) ;
111 section_MY_nzm ( P o s 0 s e c ( i ) : P o s 1 s e c ( i ) )=sectionMY ( : , i ) ;
112 MaxMY( i )=max( sectionMY0 ( : , i ) ) ; %maximum moment due t o VIV
response
113 %FFT
114 FFTiY ( : , i )=f f t ( sectionMY0 ( : , i ) ) ;
115 FFTabsY ( : , i )=abs (FFTiY ( : , i ) ) ;
116 [ FFTMaxY, indexY ] = max(FFTabsY ( : , i ) ) ;
117 f_oscY ( i ) = F( indexY ) ;
118

190
119 sectionMZ ( : , i )=MZZ( 1 0 0 0 : 8 5 0 0 , i ) ; %s e l e c t d a t a o v e r 75
seconds
120 sectionMZ0 ( : , i )=sectionMZ ( : , i )−mean( sectionMZ ( : , i ) ) ; %s c a l e
i t around z e r o mean
121 s e c t i o n M Z t o t ( P o s 0 s e c ( i ) : P o s 1 s e c ( i ) )=sectionMZ0 ( : , i ) ;
122 MaxMZ( i )=max( sectionMZ0 ( : , i ) ) ; %maximum moment due t o VIV
response
123 %FFT
124 FFTiZ ( : , i )=f f t ( sectionMZ0 ( : , i ) ) ;
125 FFTabsZ ( : , i )=abs ( FFTiZ ( : , i ) ) ;
126 [ FFTMaxZ, indexZ ] = max( FFTabsZ ( : , i ) ) ;
127 f_oscZ ( i ) = F( indexZ ) ;
128 end
129
130 %%
131 %S t r e s s e s from moments
132 %f o r 16 p o i n t s around c i r c u m f e r e n c e
133 for i =1:16 %l o o p o v e r p o i n t s around c i r c u m f e r e n c e
134 %w i t h moments s c a l e d around z e r o mean
135 %f l e x u r a l s t r e s s e s
136 sigma_xx_MY ( : , : , i )= sectionMY0 ( : , : ) . / I _ r i n g . ∗ R_out . ∗ 1E
−6;
137 sigma_xx_MZ ( : , : , i )= sectionMZ0 ( : , : ) . / I _ r i n g . ∗ R_out . ∗ 1E
−6;
138 sigma_xxi ( : , : , i )= − sigma_xx_MY ( : , : , i ) . ∗ sin ( t h e t a _ i ( i ) )
+ sigma_xx_MZ ( : , : , i ) . ∗ cos ( t h e t a _ i ( i ) ) ; %e l . end 1
139 % %u n s c a l e d :
140 % sigma_xx_MY_nzm ( : , : , i )= sectionMY ( : , : ) . / I _ r i n g . ∗ R_out
. ∗ 1E−6;
141 % sigma_xx_MZ_nzm ( : , : , i )= sectionMZ ( : , : ) . / I _ r i n g . ∗ R_out
. ∗ 1E−6;
142 % sigma_xxi_nzm ( : , : , i )=−sigma_xx_MY_nzm ( : , : , i ) . ∗ s i n (
t h e t a _ i ( i ) )+sigma_xx_MZ_nzm ( : , : , i ) . ∗ c o s ( t h e t a _ i ( i ) ) ;
143 end
144
145 %t o r s i o n s t r e s s and max . f l e x u r a l s t r e s s
146 tau_x=sectionMX0 . / J_x . ∗ R_out . ∗ 1E−6;
147 sigma_princ_MY=max(sigma_xx_MY , [ ] , 3 ) ;
148 sigma_princ_MZ=max( sigma_xx_MZ , [ ] , 3 ) ;

191
149 [ sigma_xx , index_xx ]=max( sigma_xxi , [ ] , 3 ) ;
150
151 %p o i n t around c i r c u m f e r e n c e where sigma_xx i s max :
152 theta_xx=rad2deg ( t h e t a _ i ( mode ( index_xx , 1 ) ) ) ;
153 theta_xx_index=mode ( index_xx , 1 ) ;
154
155 %p r i n c i p a l s t r e s s e s and p r i n c i p a l s t r e s s d i r e c t i o n (2D
section )
156 sigma1=(sigma_xx ) ./2+ sqrt ( ( sigma_xx . / 2 ) .^2+tau_x . ^ 2 ) ;
157 sigma2=(sigma_xx ) ./2 − sqrt ( ( sigma_xx . / 2 ) .^2+tau_x . ^ 2 ) ;
158 t h e t a 1 r a d=atan ( 2 . ∗ tau_x . / ( sigma_xx ) ) . / 2 ;
159 t h e t a 1=rad2deg ( t h e t a 1 r a d )+theta_xx ;
160 theta1min=min( t h e t a 1 ) ;
161 theta1max=max( t h e t a 1 ) ;
162 t h e t a 1 r a n g e=theta1max−theta1min ;
163 for i =1: n c u r r
164 t h e t a _ i _ t o t ( P o s 0 s e c ( i ) : P o s 1 s e c ( i ) )=t h e t a 1 ( : , i ) ;
165 end
166
167 % f i g u r e (4)
168 % p l o t ( Timesec , deg2rad ( t h e t a _ i _ t o t ) )
169 % s e t ( gca , ’ YTick ’ , 0 : p i / 2 : 2 ∗ p i )
170 % s e t ( gca , ’ YTickLabel ’ , { ’ 0 ’ , ’ p i / 2 ’ , ’ pi ’ , ’ 3 ∗ p i / 2 ’ , ’ 2 ∗ pi ’ } )
171 % t i t l e ( ’ P r i n c i p a l s t r e s s d i r e c t i o n d u r i n g VIV ’ )
172 % y l a b e l ( ’ theta_ { Sigma1 } [ rad ] ’ )
173 % x l a b e l ( ’ Time [ s e c ] ’ )
174
175 [ Seq_sigma , ni75sec_sigma ] = rainflow_and_Seq ( sigma_xx ,
Time1 ) ; %f l e x u r a l s t r e s s
176 [ Seq_tau , n i 7 5 s e c _ t a u ] = rainflow_and_Seq ( tau_x , Time1 ) ; %
torsion stress
177 [ Seq_1 , n i 7 5 s e c _ 1 ] = rainflow_and_Seq ( sigma1 , Time1 ) ; %f i r s t
principal stress
178 %%
179 %f a t i g u e damage
180 Tyear =365∗24∗60∗60; %[ s ]
181 cp =1; %P r o b a b i l i t y o f occurance
182 ni_sigma=ni75sec_sigma . / 7 5 . ∗ Tyear . ∗ cp ; %number o f c y c l e s per
year

192
183 ni_tau=n i 7 5 s e c _ t a u . / 7 5 . ∗ Tyear . ∗ cp ;
184 ni_1=n i 7 5 s e c _ 1 . / 7 5 . ∗ Tyear . ∗ cp ;
185 f_sigma_RF=ni75sec_sigma . / 7 5 ; %c y c l e −c o u n t i n g f r e q u e n c y
186 f_tau_RF=n i 7 5 s e c _ t a u . / 7 5 ;
187
188 % %SN c u r v e D w i t h s l o p e −1/5
189 m=5;
190 a2bar =10^15.606;
191
192 %SN c u r v e a p p l i e d on Seq
193 N_SN_sigma=a2bar . ∗ Seq_sigma .^( −m) ;
194 N_SN_tau=a2bar . ∗ Seq_tau .^( −m) ;
195 N_SN_1=a2bar . ∗ Seq_1.^( −m) ;
196 Dminersum_sigma=ni_sigma . / N_SN_sigma ;
197 Dminersum_tau=ni_tau . / N_SN_tau ;
198 Dminersum=ni_1 . /N_SN_1;
Import time signals
1 function [ Mx_05 , Mx_10 , Mx_15 , Mx_20 , Mx_25 , Mx_30 , Mx_35 , Mx_40 ,
Mx_45 , Mx_50 , . . .
2 Mx_55 , Mx_60 , Mx_65 , Mx_70 , Mx_75 , Mx_80 , Mx_85 , Mx_90 , Mx_95 ,
Mx_98 , . . .
3 My_05 , My_10 , My_15 , My_20 , My_25 , My_30 , My_35 , My_40 , My_45
, My_50 , . . .
4 My_55 , My_60 , My_65 , My_70 , My_75 , My_80 , My_85 , My_90 , My_95 ,
My_98 , . . .
5 Mz_05 , Mz_10 , Mz_15 , Mz_20 , Mz_25 , Mz_30 , Mz_35 , Mz_40 , Mz_45 ,
Mz_50 , . . .
6 Mz_55 , Mz_60 , Mz_65 , Mz_70 , Mz_75 , Mz_80 , Mz_85 , Mz_90 , Mz_95 ,
Mz_98]= calling_stress_data_EL1_A ( )
7
8 %c a l l i n g v a r i a b l e s from d y n p l o t
9 %moments around x−a x i s
10 MX_05=importdata ( ’EL1_MX_05A. t x t ’ , ’ ␣ ’ , 8 ) ;
11 Mx_05=MX_05. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
12 MX_10=importdata ( ’EL1_MX_10A. t x t ’ , ’ ␣ ’ , 8 ) ;
13 Mx_10=MX_10. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
14 MX_15=importdata ( ’EL1_MX_15A. t x t ’ , ’ ␣ ’ , 8 ) ;
15 Mx_15=MX_15. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;

193
16 MX_20=importdata ( ’EL1_MX_20A. t x t ’ , ’␣ ’ ,8) ;
17 Mx_20=MX_20. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
18 MX_25=importdata ( ’EL1_MX_25A. t x t ’ , ’␣ ’ ,8) ;
19 Mx_25=MX_25. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
20 MX_30=importdata ( ’EL1_MX_30A. t x t ’ , ’␣ ’ ,8) ;
21 Mx_30=MX_30. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
22 MX_35=importdata ( ’EL1_MX_35A. t x t ’ , ’␣ ’ ,8) ;
23 Mx_35=MX_35. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
24 MX_40=importdata ( ’EL1_MX_40A. t x t ’ , ’␣ ’ ,8) ;
25 Mx_40=MX_40. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
26 MX_45=importdata ( ’EL1_MX_45A. t x t ’ , ’␣ ’ ,8) ;
27 Mx_45=MX_45. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
28 MX_50=importdata ( ’EL1_MX_50A. t x t ’ , ’␣ ’ ,8) ;
29 Mx_50=MX_50. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
30 MX_55=importdata ( ’EL1_MX_55A. t x t ’ , ’␣ ’ ,8) ;
31 Mx_55=MX_55. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
32 MX_60=importdata ( ’EL1_MX_60A. t x t ’ , ’␣ ’ ,8) ;
33 Mx_60=MX_60. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
34 MX_65=importdata ( ’EL1_MX_65A. t x t ’ , ’␣ ’ ,8) ;
35 Mx_65=MX_65. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
36 MX_70=importdata ( ’EL1_MX_70A. t x t ’ , ’␣ ’ ,8) ;
37 Mx_70=MX_70. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
38 MX_75=importdata ( ’EL1_MX_75A. t x t ’ , ’␣ ’ ,8) ;
39 Mx_75=MX_75. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
40 MX_80=importdata ( ’EL1_MX_80A. t x t ’ , ’␣ ’ ,8) ;
41 Mx_80=MX_80. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
42 MX_85=importdata ( ’EL1_MX_85A. t x t ’ , ’␣ ’ ,8) ;
43 Mx_85=MX_85. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
44 MX_90=importdata ( ’EL1_MX_90A. t x t ’ , ’␣ ’ ,8) ;
45 Mx_90=MX_90. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
46 MX_95=importdata ( ’EL1_MX_95A. t x t ’ , ’␣ ’ ,8) ;
47 Mx_95=MX_95. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
48 MX_98=importdata ( ’EL1_MX_98A. t x t ’ , ’␣ ’ ,8) ;
49 Mx_98=MX_98. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
50
51 %moments around y−a x i s
52 MY_05=importdata ( ’EL1_MY_05A. t x t ’ , ’ ␣ ’ , 8 ) ;
53 My_05=MY_05. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
54 MY_10=importdata ( ’EL1_MY_10A. t x t ’ , ’ ␣ ’ , 8 ) ;

194
55 My_10=MY_10. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
56 MY_15=importdata ( ’EL1_MY_15A. t x t ’ , ’␣ ’ ,8) ;
57 My_15=MY_15. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
58 MY_20=importdata ( ’EL1_MY_20A. t x t ’ , ’␣ ’ ,8) ;
59 My_20=MY_20. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
60 MY_25=importdata ( ’EL1_MY_25A. t x t ’ , ’␣ ’ ,8) ;
61 My_25=MY_25. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
62 MY_30=importdata ( ’EL1_MY_30A. t x t ’ , ’␣ ’ ,8) ;
63 My_30=MY_30. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
64 MY_35=importdata ( ’EL1_MY_35A. t x t ’ , ’␣ ’ ,8) ;
65 My_35=MY_35. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
66 MY_40=importdata ( ’EL1_MY_40A. t x t ’ , ’␣ ’ ,8) ;
67 My_40=MY_40. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
68 MY_45=importdata ( ’EL1_MY_45A. t x t ’ , ’␣ ’ ,8) ;
69 My_45=MY_45. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
70 MY_50=importdata ( ’EL1_MY_50A. t x t ’ , ’␣ ’ ,8) ;
71 My_50=MY_50. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
72 MY_55=importdata ( ’EL1_MY_55A. t x t ’ , ’␣ ’ ,8) ;
73 My_55=MY_55. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
74 MY_60=importdata ( ’EL1_MY_60A. t x t ’ , ’␣ ’ ,8) ;
75 My_60=MY_60. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
76 MY_65=importdata ( ’EL1_MY_65A. t x t ’ , ’␣ ’ ,8) ;
77 My_65=MY_65. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
78 MY_70=importdata ( ’EL1_MY_70A. t x t ’ , ’␣ ’ ,8) ;
79 My_70=MY_70. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
80 MY_75=importdata ( ’EL1_MY_75A. t x t ’ , ’␣ ’ ,8) ;
81 My_75=MY_75. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
82 MY_80=importdata ( ’EL1_MY_80A. t x t ’ , ’␣ ’ ,8) ;
83 My_80=MY_80. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
84 MY_85=importdata ( ’EL1_MY_85A. t x t ’ , ’␣ ’ ,8) ;
85 My_85=MY_85. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
86 MY_90=importdata ( ’EL1_MY_90A. t x t ’ , ’␣ ’ ,8) ;
87 My_90=MY_90. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
88 MY_95=importdata ( ’EL1_MY_95A. t x t ’ , ’␣ ’ ,8) ;
89 My_95=MY_95. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
90 MY_98=importdata ( ’EL1_MY_98A. t x t ’ , ’␣ ’ ,8) ;
91 My_98=MY_98. data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
92
93 %moments around z−a x i s

195
94 MZ_05=importdata ( ’EL1_MZ_05A . t x t ’ , ’␣ ’ ,8) ;
95 Mz_05=MZ_05 . data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
96 MZ_10=importdata ( ’EL1_MZ_10A . t x t ’ , ’␣ ’ ,8) ;
97 Mz_10=MZ_10 . data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
98 MZ_15=importdata ( ’EL1_MZ_15A . t x t ’ , ’␣ ’ ,8) ;
99 Mz_15=MZ_15 . data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
100 MZ_20=importdata ( ’EL1_MZ_20A . t x t ’ , ’␣ ’ ,8) ;
101 Mz_20=MZ_20 . data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
102 MZ_25=importdata ( ’EL1_MZ_25A . t x t ’ , ’␣ ’ ,8) ;
103 Mz_25=MZ_25 . data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
104 MZ_30=importdata ( ’EL1_MZ_30A . t x t ’ , ’␣ ’ ,8) ;
105 Mz_30=MZ_30 . data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
106 MZ_35=importdata ( ’EL1_MZ_35A . t x t ’ , ’␣ ’ ,8) ;
107 Mz_35=MZ_35 . data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
108 MZ_40=importdata ( ’EL1_MZ_40A . t x t ’ , ’␣ ’ ,8) ;
109 Mz_40=MZ_40 . data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
110 MZ_45=importdata ( ’EL1_MZ_45A . t x t ’ , ’␣ ’ ,8) ;
111 Mz_45=MZ_45 . data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
112 MZ_50=importdata ( ’EL1_MZ_50A . t x t ’ , ’␣ ’ ,8) ;
113 Mz_50=MZ_50 . data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
114 MZ_55=importdata ( ’EL1_MZ_55A . t x t ’ , ’␣ ’ ,8) ;
115 Mz_55=MZ_55 . data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
116 MZ_60=importdata ( ’EL1_MZ_60A . t x t ’ , ’␣ ’ ,8) ;
117 Mz_60=MZ_60 . data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
118 MZ_65=importdata ( ’EL1_MZ_65A . t x t ’ , ’␣ ’ ,8) ;
119 Mz_65=MZ_65 . data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
120 MZ_70=importdata ( ’EL1_MZ_70A . t x t ’ , ’␣ ’ ,8) ;
121 Mz_70=MZ_70 . data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
122 MZ_75=importdata ( ’EL1_MZ_75A . t x t ’ , ’␣ ’ ,8) ;
123 Mz_75=MZ_75 . data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
124 MZ_80=importdata ( ’EL1_MZ_80A . t x t ’ , ’␣ ’ ,8) ;
125 Mz_80=MZ_80 . data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
126 MZ_85=importdata ( ’EL1_MZ_85A . t x t ’ , ’␣ ’ ,8) ;
127 Mz_85=MZ_85 . data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
128 MZ_90=importdata ( ’EL1_MZ_90A . t x t ’ , ’␣ ’ ,8) ;
129 Mz_90=MZ_90 . data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
130 MZ_95=importdata ( ’EL1_MZ_95A . t x t ’ , ’␣ ’ ,8) ;
131 Mz_95=MZ_95 . data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
132 MZ_98=importdata ( ’EL1_MZ_98A . t x t ’ , ’␣ ’ ,8) ;

196
133 Mz_98=MZ_98 . data ( 5 0 1 : 1 0 5 0 0 , 1 ) ;
134 end

Rainflow counting and equivalent stress ranges

1 function [ Seq , n i 7 5 s e c ]= rainflow_and_Seq ( sigma1 , Time1 )
2
3 % %R a i n f l o w c o u n t i n g
4 %c=c y c l e c o u n t s : count , range , mean , s t a r t , end
5 %h i s t=histogram ,
6 %rmm=h i s t o g r a m b i n edges ,
7 %i d x=l i n e a r i n d i c e s o f r e v e r s a l s
8 % p e r i o d s 1=c1 ( : , 5 )−c1 ( : , 4 ) ; %s e c
9 % f c y c 1 =1./ p e r i o d s 1 ; %1/ s e c
10
11 % %SN c u r v e d a t a
12 m2=5;
13
14 %Uc=0.05
15 [ c1 , h i s t 1 , edges1 , rmm1 , i d x 1 ] = r a i n f l o w ( sigma1 ( : , 1 ) , Time1 ) ;
16 n0_1=sum( h i s t 1 , 2 ) ; %number o f c y c l e s f o r each s t r e s s range
17 S0_1=e d g e s 1 ( 2 : length ( e d g e s 1 ) ) ; %s t r e s s r a n g e s
18 num=0;
19 N1=0;
20 for i =1: length ( S0_1 )
21 num=num+(n0_1 ( i ) . ∗ S0_1 ( i ) . ^m2) ;
22 N1=N1+n0_1 ( i ) ;
23 end
24 Seq_1=(num . / N1) . ^ ( 1 /m2) ;
25 %Uc=0.10
26 [ c2 , h i s t 2 , edges2 , rmm2 , i d x 2 ] = r a i n f l o w ( sigma1 ( : , 2 ) , Time1 ) ;
27 n0_2=sum( h i s t 2 , 2 ) ; %number o f c y c l e s f o r each s t r e s s range
28 S0_2=e d g e s 2 ( 2 : length ( e d g e s 2 ) ) ; %s t r e s s r a n g e s
29 num=0;
30 N2=0;
31 for i =1: length ( S0_2 )
32 num=num+(n0_2 ( i ) . ∗ S0_2 ( i ) . ^m2) ;
33 N2=N2+n0_2 ( i ) ;
34 end
35 Seq_2=(num . / N2) . ^ ( 1 /m2) ;

197
36 %Uc=0.15
37 [ c3 , h i s t 3 , edges3 , rmm3 , i d x 3 ] = r a i n f l o w ( sigma1 ( : , 3 ) , Time1 ) ;
38 n0_3=sum( h i s t 3 , 2 ) ; %number o f c y c l e s f o r each s t r e s s range
39 S0_3=e d g e s 3 ( 2 : length ( e d g e s 3 ) ) ; %s t r e s s r a n g e s
40 num=0;
41 N3=0;
42 for i =1: length ( S0_3 )
43 num=num+(n0_3 ( i ) . ∗ S0_3 ( i ) . ^m2) ;
44 N3=N3+n0_3 ( i ) ;
45 end
46 Seq_3=(num . / N3) . ^ ( 1 /m2) ;
47 %Uc=0.20
48 [ c4 , h i s t 4 , edges4 , rmm4 , i d x 4 ] = r a i n f l o w ( sigma1 ( : , 4 ) , Time1 ) ;
49 n0_4=sum( h i s t 4 , 2 ) ; %number o f c y c l e s f o r each s t r e s s range
50 S0_4=e d g e s 4 ( 2 : length ( e d g e s 4 ) ) ; %s t r e s s r a n g e s
51 num=0;
52 N4=0;
53 for i =1: length ( S0_4 )
54 num=num+(n0_4 ( i ) . ∗ S0_4 ( i ) . ^m2) ;
55 N4=N4+n0_4 ( i ) ;
56 end
57 Seq_4=(num . / N4) . ^ ( 1 /m2) ;
58 %Uc=0.25
59 [ c5 , h i s t 5 , edges5 , rmm5 , i d x 5 ] = r a i n f l o w ( sigma1 ( : , 5 ) , Time1 ) ;
60 n0_5=sum( h i s t 5 , 2 ) ; %number o f c y c l e s f o r each s t r e s s range
61 S0_5=e d g e s 5 ( 2 : length ( e d g e s 5 ) ) ; %s t r e s s r a n g e s
62 num=0;
63 N5=0;
64 for i =1: length ( S0_5 )
65 num=num+(n0_5 ( i ) . ∗ S0_5 ( i ) . ^m2) ;
66 N5=N5+n0_5 ( i ) ;
67 end
68 Seq_5=(num . / N5) . ^ ( 1 /m2) ;
69 %Uc=0.30
70 [ c6 , h i s t 6 , edges6 , rmm6 , i d x 6 ] = r a i n f l o w ( sigma1 ( : , 6 ) , Time1 ) ;
71 n0_6=sum( h i s t 6 , 2 ) ; %number o f c y c l e s f o r each s t r e s s range
72 S0_6=e d g e s 6 ( 2 : length ( e d g e s 6 ) ) ; %s t r e s s r a n g e s
73 num=0;
74 N6=0;

198
75 for i =1: length ( S0_6 )
76 num=num+(n0_6 ( i ) . ∗ S0_6 ( i ) . ^m2) ;
77 N6=N6+n0_6 ( i ) ;
78 end
79 Seq_6=(num . / N6) . ^ ( 1 /m2) ;
80 %Uc=0.35
81 [ c7 , h i s t 7 , edges7 , rmm7 , i d x 7 ] = r a i n f l o w ( sigma1 ( : , 7 ) , Time1 ) ;
82 n0_7=sum( h i s t 7 , 2 ) ; %number o f c y c l e s f o r each s t r e s s range
83 S0_7=e d g e s 7 ( 2 : length ( e d g e s 7 ) ) ; %s t r e s s r a n g e s
84 num=0;
85 N7=0;
86 for i =1: length ( S0_7 )
87 num=num+(n0_7 ( i ) . ∗ S0_7 ( i ) . ^m2) ;
88 N7=N7+n0_7 ( i ) ;
89 end
90 Seq_7=(num . / N7) . ^ ( 1 /m2) ;
91 %Uc=0.40
92 [ c8 , h i s t 8 , edges8 , rmm8 , i d x 8 ] = r a i n f l o w ( sigma1 ( : , 8 ) , Time1 ) ;
93 n0_8=sum( h i s t 8 , 2 ) ; %number o f c y c l e s f o r each s t r e s s range
94 S0_8=e d g e s 8 ( 2 : length ( e d g e s 8 ) ) ; %s t r e s s r a n g e s
95 num=0;
96 N8=0;
97 for i =1: length ( S0_8 )
98 num=num+(n0_8 ( i ) . ∗ S0_8 ( i ) . ^m2) ;
99 N8=N8+n0_8 ( i ) ;
100 end
101 Seq_8=(num . / N8) . ^ ( 1 /m2) ;
102 %Uc=0.45
103 [ c9 , h i s t 9 , edges9 , rmm9 , i d x 9 ] = r a i n f l o w ( sigma1 ( : , 9 ) , Time1 ) ;
104 n0_9=sum( h i s t 9 , 2 ) ; %number o f c y c l e s f o r each s t r e s s range
105 S0_9=e d g e s 9 ( 2 : length ( e d g e s 9 ) ) ; %s t r e s s r a n g e s
106 num=0;
107 N9=0;
108 for i =1: length ( S0_9 )
109 num=num+(n0_9 ( i ) . ∗ S0_9 ( i ) . ^m2) ;
110 N9=N9+n0_9 ( i ) ;
111 end
112 Seq_9=(num . / N9) . ^ ( 1 /m2) ;
113 %Uc=0.50

199
114 [ c10 , h i s t 1 0 , edges10 , rmm10 , i d x 1 0 ] = r a i n f l o w ( sigma1 ( : , 1 0 ) ,
Time1 ) ;
115 n0_10=sum( h i s t 1 0 , 2 ) ; %number o f c y c l e s f o r each s t r e s s range
116 S0_10=e d g e s 1 0 ( 2 : length ( e d g e s 1 0 ) ) ; %s t r e s s r a n g e s
117 num=0;
118 N10=0;
119 for i =1: length ( S0_10 )
120 num=num+(n0_10 ( i ) . ∗ S0_10 ( i ) . ^m2) ;
121 N10=N10+n0_10 ( i ) ;
122 end
123 Seq_10=(num . / N10 ) . ^ ( 1 /m2) ;
124 %Uc=0.55
125 [ c11 , h i s t 1 1 , edges11 , rmm11 , i d x 1 1 ] = r a i n f l o w ( sigma1 ( : , 1 1 ) ,
Time1 ) ;
126 n0_11=sum( h i s t 1 1 , 2 ) ; %number o f c y c l e s f o r each s t r e s s range
127 S0_11=e d g e s 1 1 ( 2 : length ( e d g e s 1 1 ) ) ; %s t r e s s r a n g e s
128 num=0;
129 N11=0;
130 for i =1: length ( S0_11 )
131 num=num+(n0_11 ( i ) . ∗ S0_11 ( i ) . ^m2) ;
132 N11=N11+n0_11 ( i ) ;
133 end
134 Seq_11=(num . / N11 ) . ^ ( 1 /m2) ;
135 %Uc=0.60
136 [ c12 , h i s t 1 2 , edges12 , rmm12 , i d x 1 2 ] = r a i n f l o w ( sigma1 ( : , 1 2 ) ,
Time1 ) ;
137 n0_12=sum( h i s t 1 2 , 2 ) ; %number o f c y c l e s f o r each s t r e s s range
138 S0_12=e d g e s 1 2 ( 2 : length ( e d g e s 1 2 ) ) ; %s t r e s s r a n g e s
139 num=0;
140 N12=0;
141 for i =1: length ( S0_12 )
142 num=num+(n0_12 ( i ) . ∗ S0_12 ( i ) . ^m2) ;
143 N12=N12+n0_12 ( i ) ;
144 end
145 Seq_12=(num . / N12 ) . ^ ( 1 /m2) ;
146 %Uc=0.65
147 [ c13 , h i s t 1 3 , edges13 , rmm13 , i d x 1 3 ] = r a i n f l o w ( sigma1 ( : , 1 3 ) ,
Time1 ) ;
148 n0_13=sum( h i s t 1 3 , 2 ) ; %number o f c y c l e s f o r each s t r e s s range

200
149 S0_13=e d g e s 1 3 ( 2 : length ( e d g e s 1 3 ) ) ; %s t r e s s r a n g e s
150 num=0;
151 N13=0;
152 for i =1: length ( S0_13 )
153 num=num+(n0_13 ( i ) . ∗ S0_13 ( i ) . ^m2) ;
154 N13=N13+n0_13 ( i ) ;
155 end
156 Seq_13=(num . / N13 ) . ^ ( 1 /m2) ;
157 %Uc=0.70
158 [ c14 , h i s t 1 4 , edges14 , rmm14 , i d x 1 4 ] = r a i n f l o w ( sigma1 ( : , 1 4 ) ,
Time1 ) ;
159 n0_14=sum( h i s t 1 4 , 2 ) ; %number o f c y c l e s f o r each s t r e s s range
160 S0_14=e d g e s 1 4 ( 2 : length ( e d g e s 1 4 ) ) ; %s t r e s s r a n g e s
161 num=0;
162 N14=0;
163 for i =1: length ( S0_14 )
164 num=num+(n0_14 ( i ) . ∗ S0_14 ( i ) . ^m2) ;
165 N14=N14+n0_14 ( i ) ;
166 end
167 Seq_14=(num . / N14 ) . ^ ( 1 /m2) ;
168 %Uc=0.75
169 [ c15 , h i s t 1 5 , edges15 , rmm15 , i d x 1 5 ] = r a i n f l o w ( sigma1 ( : , 1 5 ) ,
Time1 ) ;
170 n0_15=sum( h i s t 1 5 , 2 ) ; %number o f c y c l e s f o r each s t r e s s range
171 S0_15=e d g e s 1 5 ( 2 : length ( e d g e s 1 5 ) ) ; %s t r e s s r a n g e s
172 num=0;
173 N15=0;
174 for i =1: length ( S0_15 )
175 num=num+(n0_15 ( i ) . ∗ S0_15 ( i ) . ^m2) ;
176 N15=N15+n0_15 ( i ) ;
177 end
178 Seq_15=(num . / N15 ) . ^ ( 1 /m2) ;
179 %Uc=0.80
180 [ c16 , h i s t 1 6 , edges16 , rmm16 , i d x 1 6 ] = r a i n f l o w ( sigma1 ( : , 1 6 ) ,
Time1 ) ;
181 n0_16=sum( h i s t 1 6 , 2 ) ; %number o f c y c l e s f o r each s t r e s s range
182 S0_16=e d g e s 1 6 ( 2 : length ( e d g e s 1 6 ) ) ; %s t r e s s r a n g e s
183 num=0;
184 N16=0;

201
185 for i =1: length ( S0_16 )
186 num=num+(n0_16 ( i ) . ∗ S0_16 ( i ) . ^m2) ;
187 N16=N16+n0_16 ( i ) ;
188 end
189 Seq_16=(num . / N16 ) . ^ ( 1 /m2) ;
190 %Uc=0.85
191 [ c17 , h i s t 1 7 , edges17 , rmm17 , i d x 1 7 ] = r a i n f l o w ( sigma1 ( : , 1 7 ) ,
Time1 ) ;
192 n0_17=sum( h i s t 1 7 , 2 ) ; %number o f c y c l e s f o r each s t r e s s range
193 S0_17=e d g e s 1 7 ( 2 : length ( e d g e s 1 7 ) ) ; %s t r e s s r a n g e s
194 num=0;
195 N17=0;
196 for i =1: length ( S0_17 )
197 num=num+(n0_17 ( i ) . ∗ S0_17 ( i ) . ^m2) ;
198 N17=N17+n0_17 ( i ) ;
199 end
200 Seq_17=(num . / N17 ) . ^ ( 1 /m2) ;
201 %Uc=0.90
202 [ c18 , h i s t 1 8 , edges18 , rmm18 , i d x 1 8 ] = r a i n f l o w ( sigma1 ( : , 1 8 ) ,
Time1 ) ;
203 n0_18=sum( h i s t 1 8 , 2 ) ; %number o f c y c l e s f o r each s t r e s s range
204 S0_18=e d g e s 1 8 ( 2 : length ( e d g e s 1 8 ) ) ; %s t r e s s r a n g e s
205 num=0;
206 N18=0;
207 for i =1: length ( S0_18 )
208 num=num+(n0_18 ( i ) . ∗ S0_18 ( i ) . ^m2) ;
209 N18=N18+n0_18 ( i ) ;
210 end
211 Seq_18=(num . / N18 ) . ^ ( 1 /m2) ;
212 %Uc=0.95
213 [ c19 , h i s t 1 9 , edges19 , rmm19 , i d x 1 9 ] = r a i n f l o w ( sigma1 ( : , 1 9 ) ,
Time1 ) ;
214 n0_19=sum( h i s t 1 9 , 2 ) ; %number o f c y c l e s f o r each s t r e s s range
215 S0_19=e d g e s 1 9 ( 2 : length ( e d g e s 1 9 ) ) ; %s t r e s s r a n g e s
216 num=0;
217 N19=0;
218 for i =1: length ( S0_19 )
219 num=num+(n0_19 ( i ) . ∗ S0_19 ( i ) . ^m2) ;
220 N19=N19+n0_19 ( i ) ;

202
221 end
222 Seq_19=(num . / N19 ) . ^ ( 1 /m2) ;
223 %Uc=0.98
224 [ c20 , h i s t 2 0 , edges20 , rmm20 , i d x 2 0 ] = r a i n f l o w ( sigma1 ( : , 2 0 ) ,
Time1 ) ;
225 n0_20=sum( h i s t 2 0 , 2 ) ; %number o f c y c l e s f o r each s t r e s s range
226 S0_20=e d g e s 2 0 ( 2 : length ( e d g e s 2 0 ) ) ; %s t r e s s r a n g e s
227 num=0;
228 N20=0;
229 for i =1: length ( S0_20 )
230 num=num+(n0_20 ( i ) . ∗ S0_20 ( i ) . ^m2) ;
231 N20=N20+n0_20 ( i ) ;
232 end
233 Seq_20=(num . / N20 ) . ^ ( 1 /m2) ;
234
235 Seq =[Seq_1 Seq_2 Seq_3 Seq_4 Seq_5 Seq_6 Seq_7 Seq_8 Seq_9
Seq_10 Seq_11 . . .
236 Seq_12 Seq_13 Seq_14 Seq_15 Seq_16 Seq_17 Seq_18 Seq_19
Seq_20 ] ;
237 n i 7 5 s e c =[N1 N2 N3 N4 N5 N6 N7 N8 N9 N10 N11 N12 N13 N14 N15
N16 N17 N18 N19 N20 ] ;
238 end

203
Linda Sieber VIV fatigue of rigid spool for subsea template by a time domain model