0 Голоса «за»0 Голоса «против»

Просмотров: 0341 стр.adequate linearsation and tow analysis

Jul 25, 2018

Adequate Linearization Scheme for a Jackup

© © All Rights Reserved

PDF, TXT или читайте онлайн в Scribd

adequate linearsation and tow analysis

© All Rights Reserved

Просмотров: 0

Adequate Linearization Scheme for a Jackup

adequate linearsation and tow analysis

© All Rights Reserved

- The Law of Explosive Growth: Lesson 20 from The 21 Irrefutable Laws of Leadership
- Hidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space Race
- Hidden Figures Young Readers' Edition
- The E-Myth Revisited: Why Most Small Businesses Don't Work and
- Micro: A Novel
- The Wright Brothers
- The Other Einstein: A Novel
- State of Fear
- State of Fear
- The Power of Discipline: 7 Ways it Can Change Your Life
- The Kiss Quotient: A Novel
- Being Wrong: Adventures in the Margin of Error
- Algorithms to Live By: The Computer Science of Human Decisions
- The 6th Extinction
- The Black Swan
- The Art of Thinking Clearly
- The Last Battle
- Prince Caspian
- A Mind for Numbers: How to Excel at Math and Science Even If You Flunked Algebra
- The Theory of Death: A Decker/Lazarus Novel

Вы находитесь на странице: 1из 341

fatigue assessments using a linear

stochastic fatigue analyses

Marine Technology

Submission date: June 2017

Supervisor: Sverre Kristian Haver, IMT

Co-supervisor: Jørgen Amdahl, IMT

Department of Marine Technology

Preface

This Master Thesis is the final project after 5 years at the Norwegian University

of Technology and Science. This has been a 5 year journey with the best of what

Norway has to offer when it comes to technology and science. Five years that has

provided me with huge amounts of knowledge in highly specialized fields, but also

in more wide senses. It has been a joy to absorb the greatness of academia in the

professional, but also in the non professional way.

The journey has been long. A lot of people have contributed to make it exciting

and interesting. First of all i would like to thank my supervisors Jørgen Amdahl

and Sverre Haver. I have been extremely privileged to have two of you. The

benefit of discussing with two such knowledgeable people is huge. It is vital for

academic institution that such people with eager for their field and willingness

to teach others are present. Their input has been crucial for the quality of the

thesis. The meetings has been entertaining on all fronts, and I have always looked

forward to these. I have always come out of these meetings with new motivation

and new insight. I wish my supervisors all the best and hope to contact them with

professional questions also in the future.

Antonio Goncalves and DNV GL has been contributing by pointing on relevant

literature and providing me with models and drawings. The help is appreciated.

I would also like to thank my family, who have always been there for me. They

have provided me with financial support such that I have been able to focus full

time on my studies. They have also encouraged me in times when motivation has

been on lower levels. They are inspirational people who i admire.

My colleagues and friends on office C1.084 also contribute to make every day at

the office a joy. They all contributes in their own individual way to make the day

interesting and entertaining. They also possess a lot of knowledge to supplement

with my own in interesting discussions about marine structures as well as world

economics.

To all the lads.

Abstract

this work inspected in terms of fatigue damage. To establish reference response

characteristics, time domain simulations are carried out. Time domain simulations

must be considered as state of the art methods for estimating fatigue damage., but

do however demand huge computational capabilities.

Linear potential theory is used to calculate wave kinematics. To compensate for the

deviation to higher order wave kinematics, adjusted drag coefficients are used. The

JONSWAP spectrum is used to generate stochastic surface elevation and forces,

which is realized using both randomness in phase and amplitude. This insures that

a signal is never repeated within a short term sea state.

Stress concentration factors are used to generate stress processes from beam reac-

tions. To calculate both cycle ranges, and amount of cycles, the rainflow algorithm

is utilized, which result in stress range records that are used as input to SN curves

and miner summation.

Ground conditions are selected to give a highest natural period of 7.67s, which

is within energetic areas of the scatter diagram. Large dynamical amplifications

contribute to move most important fatigue damage sea state down to a spectral

peak period of 8.5s.

The linearization consists of evaluating the response of the structure to different

harmonic input components with different frequencies in order to create transfer

functions. In this regard, the question is how the wave heights used as input to these

harmonic components is selected. Three schemes of selecting these are inspected.

The two most promising are achieved by keeping the steepness or the ratio between

height and period constant. They overestimate the total accumulated damage

during 56 years by 20 % and 100% respectively. The steepness or the constant

height-period ratio is calibrated by matching a spectrally calculated wave action

with a deterministic calculated wave action. This calibration process is working

well, and gives reasonable calibrated values. Both methods tend to be efficient and

gives reasonable results. Whether the constant steepness approach is conservative

might be questioned especially at higher frequencies. The constant height-period

ratio however insures conservatism also at larger frequencies.

It might also be possible to switch the drag term to a linear term and replace the

drag coefficient by a linear drag coefficient. This might open up for the use of

simpler explicit methods to select appropriate calibrated values.

2

Sammendrag

arbeidet inspisert når det gjelder utmattingsskader. For å etablere referansere-

sponser utføres tidsplansimuleringer. Tids plan simuleringer regnes som den mest

eksakte metoden for å estimere utmattingsskader, men krever imidlertid store

beregningsmessige ressurser.

Lineœr potensialteori brukes til å beregne bølgekinematikk. For å kompensere

for avviket til høyere ordens bølgekinematikk, brukes justerte drakoeffisienter.

JONSWAP-spekteret brukes til å generere stokastisk bølgeoverflate og krefter, som

realiseres ved hjelp av både tilfeldighet i fase og amplitude. Dette sikrer at et signal

aldri gjentas i en kort sjøtilstand.

Spenningskonsentrasjonsfaktorer brukes til å generere spenningsprosesser fra bjelk-

ereaksjoner. For å beregne både syklusvidde og mengder sykluser, benyttes rain-

flow algoritmen, noe som resulterer i sykel-vidde data som brukes som input til

SN-kurver og Miner-summering.

Bunnforholdene er valgt for å gi en høyeste egenperiode på 7,67s, som ligger innen-

for det energiske områder av scatterdiagrammet. Store dynamiske forsterkninger

bidrar til å flytte viktigste tretthetskader sjøtilstand ned til en spektral peakperiode

på 8,5 år.

Lineariseringen består av å evaluere strukturens respons til forskjellige harmoniske

inputkomponenter med forskjellige frekvenser for å skape transferfunksjoner. I

denne forbindelse er spørsmålet hvordan bølgehøyder som brukes som input til

disse harmoniske komponentene, velges. Tre måter å bestemme disse på blir in-

spisert. De to mest lovende oppnås ved å holde steilheten eller forholdet mellom

høyde og periode konstant. Det gir et overestimat i skade på henholdsvis 20%

og 100% gjennom 56 år akkumulert skade. Steilhet eller konstant høyde-periode

forhold kalibreres ved å matche en spektralberegnet bølgekraft med en determinis-

tisk beregnet bølgekraft. Denne kalibreringsprosessen fungerer godt, og gir rimelige

kalibrerte verdier.

Begge metodene ser ut til å vœre effektive og gir rimelige resultater. Hvorvidt

den konstante steilhet-tilnœrmingen er konservativ, kan bli stilt spørsmålstegn ved

spesielt ved høyere frekvenser. Den konstante høyde-periode tilnœrmingen sikrer

imidlertid konservatisme også ved større frekvenser.

Det kan også vœre mulig å bytte drag leddet til en lineœr term og erstatte dragko-

effisienten med en lineœr dragkoeffisient. Dette kan åpne for bruk av enklere,

eksplisitte uttrykk for å velge passende kalibrerte verdier.

Contents

1 Introduction 17

1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

1.2 Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

1.3 Deviation from problem description . . . . . . . . . . . . . . . . . . . 18

1.4 Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.1 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

2.2 Design challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3 Metocean data 22

3.1 Location . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

3.2 Short term wave statistics . . . . . . . . . . . . . . . . . . . . . . . . 23

3.3 Long term wave statistics: scatter diagram . . . . . . . . . . . . . . 24

3.4 Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

3.5 Wind . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

4.1 Equation of motion for dynamic and quasi static analysis . . . . . . 27

4.1.1 Mass matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

4.1.2 Stiffness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

4.1.3 Damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

4.1.4 Forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

4.1.5 Linear wave kinematics . . . . . . . . . . . . . . . . . . . . . 32

4.1.6 Drag coefficient for linear wave theory and higher order wave

theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

4.2 Spring connections to ground . . . . . . . . . . . . . . . . . . . . . . 39

4.3 Eigenvalue analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

4.4 Solving the dynamic and quasi static equation of motion . . . . . . . 42

4.5 Overshoot and errors in initial responses . . . . . . . . . . . . . . . . 45

4.6 Irregular sea surface . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

4.6.1 Alternative integration technique . . . . . . . . . . . . . . . . 49

5.1 Fatigue damage in this work . . . . . . . . . . . . . . . . . . . . . . . 53

5.2 Joint selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

5.3 Miner summation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

5.4 SN curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

5.5 Stresses in joints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

5.5.1 Stress concentration factors and super positioning of stresses 58

6

CONTENTS

5.7 Stochastic fatigue analysis in thr frequency domain . . . . . . . . . . 64

5.7.1 Stress response spectrum . . . . . . . . . . . . . . . . . . . . 64

5.7.2 Discrete Fourier transform of signal . . . . . . . . . . . . . . 65

6.1 Hydrodynamic damping, and dynamic influence . . . . . . . . . . . . 68

6.2 Fitting the short term stress range with Weibull probability distri-

bution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

6.3 Damage of same sea state with different realizations . . . . . . . . . 72

6.4 Weibull parameters for several short term sea states . . . . . . . . . 76

6.5 Fatigue damage from short term sea states . . . . . . . . . . . . . . . 76

6.6 Damage in the scatter format . . . . . . . . . . . . . . . . . . . . . . 78

6.7 Conclusive remarks for time domain results . . . . . . . . . . . . . . 79

7 Linearization 80

7.1 Motivation for proper drag force treatment . . . . . . . . . . . . . . 80

7.2 Harmonic inputs used to create transfer functions . . . . . . . . . . . 83

7.2.1 Linearization schemes included in analysis . . . . . . . . . . . 83

7.2.2 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

7.2.3 Calibration at the center of fatigue . . . . . . . . . . . . . . . 89

7.2.4 Static transfer functions comparison . . . . . . . . . . . . . . 94

7.2.5 Dynamic transfer functions . . . . . . . . . . . . . . . . . . . 98

7.2.6 Calibration for every sea state . . . . . . . . . . . . . . . . . 102

7.3 Linearization by changing the drag coefficient . . . . . . . . . . . . . 104

8 Conclusion 108

Appendices 113

A.1 Calibration ratios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

A.2 Short term sea state . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

A.3 Long term sea state (56 years) . . . . . . . . . . . . . . . . . . . . . 117

7

List of Figures

2.2 Chords and braces on a truss . . . . . . . . . . . . . . . . . . . . . . 20

3.2 Scatter diagram of the Ekofisk field during 56 years [Aarsnes, 2015] . 24

3.3 Oscillating drag term for different currents . . . . . . . . . . . . . . . 25

4.2 Decomposition of the drag term . . . . . . . . . . . . . . . . . . . . . 32

4.3 Chord cross sectional area. The different color at the tips indicates

that these are the areas which the gears are . . . . . . . . . . . . . . 35

4.4 Wave particle velocity under crest and through for 5th order stoke

waves, and stretched airy waves . . . . . . . . . . . . . . . . . . . . . 36

4.5 Difference in forces for linear theory and higher order wave theory,

for one wave component. Discontinuities are due to change in com-

ponents at +2m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

4.6 Different types of ground fixation . . . . . . . . . . . . . . . . . . . . 39

4.7 Bottom model of one of the legs. Left: rotation springs, Right:

Translation springs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

4.8 Error in amplitude and period from [Hughes, 2000] . . . . . . . . . . 43

4.9 Algorithmic damping. From: [Hughes, 2000] . . . . . . . . . . . . . . 44

4.10 Relative periodical error. From: [Hughes, 2000] . . . . . . . . . . . . 45

4.11 Comparison of energy overshoot for n time steps, from [Hilber and

Hughes, 1978] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

4.12 Initial stress with one harmonic input component . . . . . . . . . . . 47

4.13 PM spectrum for different sea states and summation technique . . . 48

4.14 Integration techniques. Without randomness in amplitude (left)

with some randomness in amplitude (right) . . . . . . . . . . . . . . 49

4.15 Autocorrelation of surface elevation With and without randomness

in amplitude . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

4.16 Autocorrelation of surface elevation . . . . . . . . . . . . . . . . . . . 51

4.17 Linearized Rayleigh distributions . . . . . . . . . . . . . . . . . . . . 52

5.2 Example on allowable combinations of mean stress and stress am-

plitude . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

5.3 Superposition of joint contributions and the eight locations of stress

evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

5.4 Chord cross sectional geometry . . . . . . . . . . . . . . . . . . . . . 60

5.5 Stress process over a short, and shorter time interval . . . . . . . . . 61

8

LIST OF FIGURES

5.7 Rainflow counting scheme . . . . . . . . . . . . . . . . . . . . . . . . 62

5.8 WAFO cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

5.9 Zero up crossing counting. Zero up crossings marked as black dots . 63

5.10 Cumulative distribution in Weibull paper of zero up crossing count-

ing vs rainflow counting. D is the total accumulated damage during

3h . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

5.11 Example of spectrum with and without averaging. JONSWAP with

HS = 4.5, TP = 8.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

6.1 Short term stress repsonse with and without hydrodynamic damping

and static analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

6.2 Linearized 3 parameter weibull plot of stress range during a 3h re-

alization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

6.3 Linearized 3 parameter weibull plot of stress range during a 3h re-

alization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

6.4 Distribution of total damage for different realizations of the same

sea state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

6.5 Weibull plot of the two realizations that gives most and least damage

during a 3h simulation, fitted with 3 parameter Weibull distribution.

Keep in mind that ln(∆σ) is on the axis, not the ln(∆σ − λ) . . . . 75

6.6 Distribution of the three parameters in the Weibull fit for different

realizations. A total of 48 realizations of the same sea state are

included . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

6.7 Weibull parameter values for different sea states. Data points are

indicated with circular dots . . . . . . . . . . . . . . . . . . . . . . . 76

6.8 Number of cycles for different short term 3h sea states, Data points

are indicated with circular dots . . . . . . . . . . . . . . . . . . . . . 77

6.9 Short term damage calculated by both the probability distribution

and by counting, Data points are indicated with circular dots . . . . 78

7.1 Ratio between amplitudes of drag force and inertia force as a func-

tion of wave height . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

7.2 Short term stress response fro comparison of drag vs no drag. HS =4.5,

TP =8.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

7.3 Short term stress distribution (3h) for drag vs no drag . . . . . . . . 83

7.4 Constant wave steepness curves with cutoff at one year return period 85

7.5 Wave heighs used as input given constant ratio between wave height

and period . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

7.6 Transfer functions from wave elevation to wave action with different

constant steepness . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

7.7 Most probable maximum of total applied wave action (quasi-static

base shear). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

7.8 Proportional accumulated damage for different sea states. The cen-

tre of fatigue contains a red dot. Ekofisk scatter diagram is used . . 90

7.9 Dynamic stress transfer functions, for the calibrated steepness . . . . 91

9

LIST OF FIGURES

7.10 Damage during short term sea states (3h) using transfer functions at

the center of fatigue and constant wave steepness. Be aware of the

different values on the color axis. Keep in mind that the values of m

is only used to estimate the center of fatigue, and are not involved

in the damage calculation. . . . . . . . . . . . . . . . . . . . . . . . . 92

7.11 Total damage during 56 years, linearized method. Keep in mind

that the values of m is only used to estimate the center of fatigue,

and are not involved in the damage calculation. . . . . . . . . . . . . 93

7.12 Comparison of steepness-linearized static transfer function of stress

with stochastic time domain . . . . . . . . . . . . . . . . . . . . . . . 95

7.13 Comparison of height-period-linearized static transfer function of

stress with stochastic time domain . . . . . . . . . . . . . . . . . . . 96

7.14 Static stress transfer functions with JONSWAP spectrum used to

generate input wave heights, compared with time domain simula-

tions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

7.15 SComparison of all schemes, static transfer functions . . . . . . . . . 98

7.16 Dynamic stress transfer functions with constant wave steepness,

compared with time domain simulations. . . . . . . . . . . . . . . . 99

7.17 Dynamic stress transfer functions with constant height period ratio,

compared with time domain simulations. . . . . . . . . . . . . . . . 100

7.18 Dynamic stress transfer functions with JONSWAP spectrum used

to generate input wave heights, compared with time domain simu-

lations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

7.19 Comparison of dynamic transfer functions from all schemes . . . . . 102

7.20 Comparison of short term damage for different linearizations (c =

H/T ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

7.21 Comparison of short term damage for different linearizations (c =

H/T ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

7.22 Linearization factors for Morison’s equation for estimating expected

fatigue damage when the SN-slope, m=1, 3, 4 and 5 represented

by full, dotted, dashed and dash-dotted lines, respectively. From:

[Wolfram, 1998] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

10

List of Tables

4.2 Coefficients in Morison’s equation for perfectly cylindrical members . 34

4.3 Coefficients in Morison’s equation for braces . . . . . . . . . . . . . . 34

4.4 Drag coefficients for chords, z is the distance from the mean surface

level, positive upwards. . . . . . . . . . . . . . . . . . . . . . . . . . . 35

4.5 Drag coefficients for chords, with the use of NORSOK scaling to

compensate for Gaussian sea surface. . . . . . . . . . . . . . . . . . . 36

4.6 Spring coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

4.7 Eigenperiods and eigenmodes . . . . . . . . . . . . . . . . . . . . . . 41

4.8 Coefficients for numerical integration . . . . . . . . . . . . . . . . . . 43

6.1 Short term sea states (3h) with and without hydrodynamic damping

in addition to a pure static simulation . . . . . . . . . . . . . . . . . 69

6.2 Standrad deviations for stress

q process and surface

q elevation process

1

PN 2 1

PN 2

HS = 4.5, TP = 8.5. sσ = N i=1 σi , sζ = N i=1 ζi . . . . . 74

6.3 Short term damage of structure (3h) . . . . . . . . . . . . . . . . . . 78

6.4 Long term damage of structure during 56 years . . . . . . . . . . . . 79

TP = 8.5, HS = 4.5. . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

7.2 Significant wave height with one year return period for the Ekofisk

field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

7.3 Deterministic most probable wave action, Hcutof f = 9.26m . . . . . 89

7.4 Centre of fatigue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

7.5 Calibrated steepness . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

7.6 The ratio between structural response velocity amplitude and water

particle velocity amplitude for 2 cases, with harmonic input, mea-

sured in surface area (drag coefficient of 1.15 is used for the whole

structure) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

H

A.1 Steepness, defined as T2 . . . . . . . . . . . . . . . . . . . . . . . . . 115

11

List of symbols

TP Spectral peak period

m0 Variance of a spectrum

Sζζ Surface elevation spectrum function

f Frequency

γ Peakedness parameter

g Gravity

uw wave particle velocity

uc Current velocity

V wind velocity

τ Constant

z0 Constant

F Force vector

M Mass matrix

m Element mass matrix

C Damping matrix

K Stiffness matrix

r, ṙ, r̈ Response vector, velocity and acceleration

ρw Density sea water

D Diameter

U Internal strain energy

E Modulus of elasticity

u,x , v,x , w,x Derivative of displacements

Ix , Iy Second moment of area

H Potential of external loads

δ Notation for virtual of the following measure (e.g. work, displacement)

N Normal force

A Cross sectional area

12

LIST OF TABLES

α1 , α2 Damping coefficients

ξ Damping ratio

ω Circular frequency

CD , CM Drag and inertia coefficients

u, w Fluid velocity in x and z direction

ax , az Fluid acceleration in x and z direction

T Period

φ Velocity potential

ζ Surface elevation

x,z Horizontal and vertical coordinates

z0 Stretched z coordinate

d Depth

CD0 Inline drag coefficient for cylinder with rack mount

CD1 Drag coefficient with rack perpendicular to flow

FD Drag force

FM Inertia force

kx , ky , kz , kθx , kθy , kθz Spring coefficients for ground

ωn Natural frequency

rn Natural mode

α, γ, β Parameters of time integration algorithm

¯

xi Algorithmic damping

T̄ Algorithmic period

S(ω) Input surface elevation spectrum

∆ω Frequency increment

φi Random phase

Trep Repetition period of signal

R(τ ) Auto correlation

FH (h) Cumulative probability distribution of h

D Damage

σ Stress

σa Stress amplitude

N Allowable cycles

log(ā) Constant in SN-curve

k Thickness exponent

σ1 − σ8 Locations for stress measurement

SCF Stress concentration factor

σx , σmy , σmz Axial and bending stress

Tn Natural period

∆σ Stress range

f∆σ Probability distribution of stress range

13

LIST OF TABLES

s Standard deviation

µ Mean

η Degree of freedom in χ dist.

χ2 Chi distribution

F∆σ Cumulative distribution of stress range

β, λ, α Distribution parameters in weibull 3 parameter dist.

FD,a , DM,a Drag and inertia amplitude

sσ Standard deviation in stress

G Parameter for scaling a JONSWAP like input wave-height

∆FM P M Range in most probable maximum in wave actions (base shear)

HM P M Most probable maximum wave height

C(C,L) Equivalent linear drag coefficient

Fp Force peak

C (m) Ratio to calculate equivalent drag coefficient

γ(, ), Γ(, ) lower and upper incomplete gamma functions

Γ() Gamma function

σu2 Variance of velocity

K Parameter to select equivalent drag coefficient

14

Abbreviations

JONSWAP Joint North Sea Wave Project

SCF Stress concetration factor

15

Chapter 1

Introduction

1.1 Background

Fatigue is dependent on the stress range of the cyclic loading rather than the stress

itself. This means that materials subject to cyclic stresses well bellow yield, may

experience critical cracks after a certain amount of cycles. This becomes a vital

phenomenon for offshore structures due to the cyclic nature of the wave loadings.

In stochastic fatigue analysis the relation between forces in each members and

wave height is calculated with the assumption that there exist a linear relationship

between the wave height and the resulting force. However, typically jacket and jack-

up with lattice leg design are drag dominated. The drag forces are proportional to

the square of the wave particle velocity. For such structures, the wave height to

wave force relation is therefore not linear.

In offshore industry, linearization with respect to wave height is generally used. It

is based on the selection of a characteristic wave height for each wave frequency

of interest. Constant wave steepness is frequently used as such characteristics

to select the linearized wave height. However, there are no proper guidelines or

recommendations for selection the correct wave steepness for a specific sea state.

Furthermore constant wave steepness usually results in over predicted drag at small

wave frequencies and under predicted drag at large wave frequencies.

The fatigue damage at jack-ups are strongly dependent on the assumptions made

in the linearization. Hence this might result in adding to much conservatism in

order to rely on the analysis. An appropriate linearization scheme is needed in

order to get more confidence in the results.

17

1.2 Scope

1.2 Scope

The scope of this thesis will be to investigate the linearization scheme with con-

stant steepness, but also look into others. The goal is to come up with schemes that

gives the correct amount of damage. To estimate the correct amount of damage,

time domain simulations with stochastic input loads should be used. The com-

puter program USFOS is used in order to perform these calculations. Such rime

domain simulations require huge amounts of computational efforts and data han-

dling. Hence they are not widely used in the offshore industry to estimate fatigue

damage. In order to deal with these large computational demands, the need of

multicore processing power is essential. Parallel processing will be carried out by

running USFOS in parallel through MATLAB.

Detailed screening of critical heading and location of critical stress will not be

conducted. The focus will be on the linearization for the particular selected joint.

Furthermore there will be no effort given to use second order surface elevation.

The final goal for the work is to come up with linearization schemes designed

for each block in the scatter diagram. These schemes should give a conservative

amount of damage.

Together with the supervisor it has been decided to exclude the following tasks

from the problem description due to work load. See appendix C for the full problem

description.

• Detailed analysis of current, and how current will invoke with linearizations,

and the importance of current is not considered, however a brief discussion

of the importance of current is included.

• Dividing the weather in different directional sectors will not be carried out.

1.4 Contributions

USFOS analysis has been implemented in MATLAB with the use of the parallel

toolbox. This will enable future students to run USFOS through a familiar pro-

graming environment, and at the same time having access to the power of multi-core

CPUs. The real benefit of this is when USFOS executions are part of perplexed

algorithms.

More confidence on linearization techniques used in offshore industry, can from

this work be taken. In addition a more conservative linearization scheme has been

developed by using height-period ratio as tuning parameter instead of steepness.

18

Chapter 2

Jack-up rig CJ-62

In general a jack-up is a platform that has the ability elevate the topside relative

to the legs. In addition to this it has the ability to float by itself when the legs are

not in contact with the bottom. Different jack-up concepts exist where differences

exist in both size and amount of legs, as well as in other parameters.

The jack-ups are usually seen as mobile drilling units, however they can also be used

for installation of wind turbines and demolition of aging offshore installations. In

operational phase the platforms are attached to the sea floor, making them stable

compared to floating units of similar size, and suitable for delicate work. This

combination of stability and mobility are what makes jack-ups attractive.

19

2.1 Design

2.1 Design

The deck structure of the platform is formed as a equilateral triangle, and has three

elevatable legs. This particular model is designed by the engineering company

GustoMSC, and is classified by DNV GL.

The outer dimensions of the legs seen from above is equivalent to a equilateral

triangle with 16 meter sides. Between the center of two of these legs, there are 62

meters. Furthermore the platform is designed to operate on depth as large as 130

meters, making this a large piece of construction.

A typical jack-up platform, including this, has legs made of trusses. These trusses

has slender members compared to wavelength. They are also in fact slender com-

pared to typical jacket structures. Leg design is based on triangular trusses with

x-braces. In addition to braces, the trusses consist of chords. The difference be-

tween a brace and a chord is illustrated figure 2.2.

Chord

Brace

The ability to elevate the deck structure comes with certain aspects. This includes

a system to elevate the whole deck structure on three legs. Keeping in mind that

the platform deck is a heavy structure this is a demanding task. This task is

accomplished by attaching racks to the chords. The platform deck can then be

elevated using gears. This means that in total nine racks are used to elevate the

whole platform. These also introduces certain hydrodynamical properties. Non-

circular geometry on the chords results in larger drag forces.

Depth ranges between 70m to 75m in the Ekofisk area where the platform location is

assumed. However this particular jack-up model support water depth considerably

deeper. At larger water depths the wave actions will inherit more overturning

moment. Hence this analysis is investigating a larger water depth than what is

present at the Ekofisk field.

In a design process wave heading direction must be accounted for. The directions

might be splitted in sectors on 15 degrees. Due to symetry for this particular

triangular platform, only inspection of a heading of 0 to 60◦ are needed. In this

work only the wave heading of 0◦ is inspected.

20

2.2 Design challenges

110m 43m 0◦

The jack-up design comes with certain challenges. The relatively thin leg design

with no stiffening in between legs makes the platform flexible compared to other

fixed offshore installation. With low stiffness the natural period of the platform

becomes significantly higher than jacket structures, and moves the period closer to

high energy regions of the scatter diagram.

This induces large dynamic responses in translation of the deck as well as twisting.

This demands a proper dynamical treatment of the motion of the platform in the

analysis perspective.

Another aspect of the jack-up design is that it has slender members. Structural

members becomes more and more drag dominated as they become thinner. Drag

forces posses a nonlinear nature, and can not be treated in the frequency domain in

the same manner as linear forces. This introduces assumptions and simplifications.

Another difficulty with the jack-up’s is the fact that they are mobile. In the design

phase it is not always known where the rig will operate during its life span. This

also contributes to a complicated fatigue analysis, and in order to do proper fatigue

analysis the load condition history is necessary since the loads will differ depending

on location. In addition the transportation of the platform will add a share of

fatigue damage.

21

Chapter 3

Metocean data

sensor measurements and statistical models. Environmental loads includes loads

from current, wind and waves. There is without doubt necessary to estimate these

loads in a accurate way.

3.1 Location

Wind, waves and current will vary depending on location. For instance the weather

in the North Sea is much harsher than in the Gulf of Mexico.

The wave record is used to describe the different wave statistics. The location for

these measurements is the Ekofisk field, which is located at 56.549197N,3.209986E,

south of the the Norwegian coast line, and can be seen in figure 3.1.

22

3.2 Short term wave statistics

In offshore industry a short term sea state is typically defined as a surface elevation

process with a 3h duration. The two main parameters used to describe this surface

is the significant wave height and spectral peak period. Significant wave height

can be estimated by HS = 4m0 , where m0 is the variance of the surface elevation

process [Myrhaug, 2007]. The spectral peak period can be considered as the average

wave period during these 3h, or more precisely the period at which the peak of the

wave specter occur.

A wave spectra is a mathematical description of the short term sea state behavior.

These relations describes the energy present at every frequency present during

these 3h. Wave spectra from measurements are not always on hand in a design

phase. Standardized spectra is then frequently used to describe the surface. They

depend upon what kind of sea conditions that is present. A typical breakdown is

the following three wave conditions [Haver, 2017].

Wind sea

Swell sea

This is a sea condition where the waves are not generated by local wind. It can

typically be described as a rather narrow banded wave spectrum.

Combined sea

In practice almost every sea state will have one component generated from local

wind and one from swell.

In the combined case, the waves generated form local wind will in general follow

the direction of the wind. In addition the components from swell sea can typically

propagate in another direction. For most applications conservatism is achieved

by assuming that these will follow the same direction. For cases of combined

sea, the Torsethaugen spectrum should be applied, which is a two peak spectrum.

Depending on the how much the two peaks differ, this might induces problems with

narrow banded assumptions.

In this project it is assumed that local wind is the only contributing condition, and

that the sea is growing. This implies that the JONSWAP spectrum is suitable.

23

3.3 Long term wave statistics: scatter diagram

North Sea sin the area of the Ekofisk field. This spectrum is given by equation 3.1

f −fp 2

exp(−0.5( σfp ) )

Sζζ (f ) = 0.3125HS2 TP−4 f −5 exp(−1.25t−4

p f

−4

)(1 − 0.287 ln(γ))γ

(3.1)

Where fp = 1/TP and σ is 0.07 for f ≤ fp and 0.09 otherwise. The peakedness

parameter can be defined as in equation 3.2 [Haver, 2017]

2πHS 6/7

γ = 42.2( ) (3.2)

gTP2

For simplicity the peakedness parameter will be taken as γ = 3.3 in this work.

Hindcast data is created for every 3h. Each of these short sea states comes with

a significant wave height and a spectral peak period. These data can be grouped

together for values of significant wave height and spectral peak period of similar

magnitude in order to form a scatter diagram. A scatter diagram for the Ekofisk

for 56 years is seen in figure 3.2.

2<TP<3

2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 13.5 14.5 15.5 16.5 17.5 18.5 19.5 20.5

0<HS<1 0.5 84 2440 9795 8508 4072 2487 1958 1335 846 473 302 159 75 66 43 18 13 5 2 32681

1.5 109 3925 14361 18546 11615 5862 4112 3278 1846 865 299 183 96 59 19 9 1 65185 Cutoff

2.5 7 641 7358 13621 7142 2828 1662 1439 993 412 155 83 18 7 4 0 36370

3.5 1 256 3121 8071 3685 1239 546 363 246 125 62 27 11 4 0 17757

4.5 1 55 2120 3936 1505 512 185 63 46 48 8 5 0 1 8485

5.5 1 172 1135 1455 582 189 29 24 11 8 1 3607

6.5 142 449 417 221 42 13 7 0 1 1292

7.5 7 70 143 112 44 19 6 0 1 402

8.5 8 29 61 31 12 8 3 152

9.5 1 6 17 25 14 13 1 77

10.5 1 7 9 7 7 1 32

11.5 3 5 4 0 12

12.5 1 1

84 2549 13727 23511 30233 30900 25325 17180 10513 5994 3315 1362 678 411 169 63 30 7 2 166053

Figure 3.2: Scatter diagram of the Ekofisk field during 56 years [Aarsnes, 2015]

This data is essential for fatigue analysis. The values of TP and HS is distributed

within each sea states, however the mean values is the only values considered in

this work, i.e. for a sea state with a period in the range TP ∈ [2, 3] seconds will be

treated as 2.5, even though the values will be spread throughout the whole 2 to 3

interval. Significant wave heights bellow 2m is considered to be uninteresting and

are not included in the analysis.

3.4 Current

Current creates forces on the members introduced in the drag term as an additional

velocity. Currents are slowly varying with time, and can be considered steady

24

3.4 Current

equation 3.3. Wherein the drag term of Morison’s equation is proportional to the

square of velocity. In the most critical sea conditions regarding fatigue for this

particular platform the wave induced water particle speed in the upper part of the

water column is in the range of 1.3m/s to 2m/s.

(uw + uc )2 = u2w + 2uw uc + u2c (3.3)

uw is wave induced water particle velocity and uc is velocity from current. As the

current is constant in nature it is possible to think that it will have low effect on

the fatigue damage. However as illustrated in equation 3.3 there exist a cross term.

This term will be oscillating due to the wave induced velocity, and be amplified

by the current. For a case of a typical current of 0.3 m/s in combination with an

important sea state of HS = 4.5 and TP = 8.5 gives a characteristic wave particle

velocity amplitude in the upper part of the water column of uw =1.7 m/s. This

will give the following amplitudes on the oscillating part of the drag force.

u2w + 2uw uc = 1.72 + 2(1.7)(0.3) = 2.9 + 0.9 (3.4)

For this rare current condition, the current will contribute to increase the total

oscillating wave particle amplitude by 30%. However the drag term also has a sign

change which for low currents will reduce the importance of the cross term.

Assuming that the current is lower than the wave particle amplitude the range

between an amplitude and a valley in equation 3.5 will hence be

(uw + uc )2 + (uw − uc )2 = u2w + u2c (3.6)

Which is still relying on the current. However if the current is small compared

with the wave particle velocity it will no be as important. Figure 3.3 is included

to illustrate this. Due to the sign change, the the amplitude of the valleys will

increase and the range will hence not grow as fast as if the sign change did not

take place.

4

|uw cos(t)+uc|(uw cos(t)+uc)

u c/uw=0, max-min=2

3 u c/uw=0.2, max-min=2.1

u c/uw=0.4, max-min=2.3

2 u c/uw=0.6, max-min=2.7

u c/uw=0.8, max-min=3.3

1

-1

0 5 10 15

t

25

3.5 Wind

Considering that the most important sea state has low to moderate wave particle

velocity there is a possibility that current will contribute a lot to fatigue damage.

In general, fatigue damage is proportional to the stress amplitude to the power of

m. Where m is typically between 3 and 5. In practical terms this means that a

stress increase of 15% will result in 1.155 = 2 times the damage. In terms of load

amplitude, 15% increase is present when the current is 40% of the wave particle

amplitude.

Even though current is an important parameter for drag, it also introduces certain

complexity to the process of linearizing the problem. Hence current is neglected in

this work.

3.5 Wind

Wind has a much lower density than water, and gives in general lower magnitude

of forces than wave actions. Wind do however add certain damage to a structure.

What is interesting in this regard is the huge air gap of 43 meters. This might be

even larger at certain locations. This is interesting because wind speed increases

with the height above the sea. A logarithmic increase in wind speed are commonly

used. Such a logarithmic relation can be seen in equation 3.7 [DNV GL, 2014a].

z

V (z) = τ ln (3.7)

z0

Where U is wind speed, z distance from mean sea level, τ and z0 are constants.

Most of the wind forces will be induced on the deck hull due to its structural shape,

and this is the highest part of the structure. This gives a higher wind force than

a regular air gap of 20 meters. However the logarithmic function will not grow in

huge scales by increasing z. The more important is the additional height added

to the overturning moment arm. Large wind speeds tends to coincide with large

sea states, and hence they should be included in an analysis. However wind is not

considered in this work.

26

Chapter 4

Loads and responses

analysis

Newton’s second law of motion form the basis of the dynamic equilibrium equation.

F = ma (4.1)

F = Sum of forces

m = mass

a = acceleration

By moving damping forces and restoring forces to the same side as the mass and

acceleration, the discrete dynamic equation for the finite element method becomes

equation 4.2.

r, ṙ, r̈ = response, and time derivatives of response

K = Stiffness matrix

M = Mass matrix

C = Damping matrix

f (t) = Load vector

t = time

The quasi static equation is governed under the assumption of negligible accelera-

tions and velocities. This results in equation 4.3. The matrices and force vector in

these equations are discussed in the context of USFOS in the following sections.

27

4.1 Equation of motion for dynamic and quasi static analysis

The mass matrix have two contributions, the mass of the structure itself, and

the hydrodynamic added mass. The structural mass matrix can be set to either

consistent or or lumped in USFOS [Søreide et al., 1993]. The consistent, which

is used in this project, is based on interpolation functions for linear 3D beam.

This implies that it is not a truly “consistent formulation” with the more complex

shape functions used for the stiffness matrix. This is still accurate enough for most

practical purposes [Marintek, 2001]. The consistent mass matrix for a 6 dof beam

element is given by [Søreide et al., 1993]

140 0 0 70 0 0

0 156 -22l 0 54 13l

2

m̄l 0

-22l 4l 0 -13l -3l2

m= (4.4)

420 70 0 0 140 0 0

0 54 -13l 0 156 22l

0 13l -3l2 0 22l 4l2

m̄ = mass per unit length for beam

l = length of beam

The added mass forces have its origin in the radiation problem. The added mass

forces are proportional to the accelerations of the members. Thus it behaves as a

“added” mass in the dynamic equilibrium equation. The added mass is calculated

by the following equation in USFOS [Marintek, 2010].

ρw πD2

dFA = (CM − 1)dz r̈ (4.5)

| 4 {z }

Added mass

ρw = density of water

CM = Innertia coefficient

D = Hydrodynamic diameter

dz = infinitesimal lenght

r̈ = acceleration of element

4.1.2 Stiffness

For an elastic beam element the internal strain energy is written in equation 4.6

[Marintek, 2001].

1 l 1 l

Z Z

1 2 1 2 2 2 2

U= EA(u,x + v,x + w,x ) dx + (EIz v,xx + EIy w,xx )dx (4.6)

2 0 2 2 2 0

| {z } | {z }

Axial stiffness Bending stiffness

28

4.1 Equation of motion for dynamic and quasi static analysis

E = modulus of elesticity

A = cross sectional area of beam

l = length of beam element

v, w = lateral displacement

u = longitudinal displacement

Iz , Iy = Second moment of area

The potential of the external loads is written [Marintek, 2001].

Z l Z l Z l

H = −(Fi ui + qx udx + qy vdx + qz wdx) (4.7)

0 0 0

ui = displacement direction i

qx , qy , qz = Distributed loads

The variation of strain energy given in equation 4.8 is used to formulate the stiffness

matrix.

Z l Z l

N

δU = EAu,x δu,x dx + EIz (v,xx δv,xx −

v,x δv,x )dx

0 0 EIz

Z l Z l

N

+ EIy (w,xx δw,xx w,x δw,x )dx − (N + EAu,x )δu,x dx (4.8)

0 EIy 0

are also expressed in the same way.

u(x) = φT qu (4.9)

T

v(x) = φ qv (4.10)

T

w(x) = φ qw (4.11)

For transverse displacement, the shape functions from the exact solution to the 4th

order differential equation for beams is used. For compression this reads,

And for tension;

φT = [cos(kx), sin(kx), x/L, 1] (4.13)

|N |

k2 = EIz

Similar expressions are used for the displacement fields w(x) and u(x). A great

benefit for this type of element is that it allows simple modeling e.g. one element

per member [Marintek, 2001].

29

4.1 Equation of motion for dynamic and quasi static analysis

4.1.3 Damping

namic damping is included in the force vector due to the relative velocity in Morri-

son’s equation. Rayleigh damping is used for structural damping. Raleigh damping

holds important orthogonality properties.

C = α1 M + α2 K (4.14)

α2 = coefficient

α1 = coefficient

M = Mass matrix

C = Damping matrix

K = Stiffness matrix

The Rayleigh damping has one term that is proportional to the mass matrix and

one that is proportional to the stiffness matrix as seen in equation 4.14. Mass and

stiffness are orthogonal, thus the damping matrix formed by Raleigh damping must

be orthogonal. This enables the writing of equation 4.15.

1 α1

ξ= + ωα2 (4.15)

2 ω

ξ = Damping ratio

ω = Frequency

The two coefficients can be used to specify the damping ratio at two frequencies.

The structural damping ratio should be 2-3% and soil damping 0-2% according

to [DNV GL, 2015]. Soil damping is neglected in this work. By specifying the

damping ratio at two frequencies, two equations can be established, and used to

find α1 and α2 .

ξ f

0.3 0.05

0.3 1

The damping ratio curve generated by Table 4.1 is depicted in figure 4.1, where

the red points are the ones given in Table 4.1. It is important to have structural

damping also for the larger sea states, thus the frequency of 0.05 is selected as

the lower limit, corresponding to a period of 20 s. Waves of significance are not

asociated with frequencies above 1 Hz (period of 1 s), thus this is selected as the

upper limit. The damping is at least 1.5% in the most important interval between

0.05 and 1 Hz, and not above 3 which is the upper limit in [DNV GL, 2015].

30

4.1 Equation of motion for dynamic and quasi static analysis

0.05

0.04

Damping ratio, 9

0.03

0.02

0.01

0 0.5 1 1.5

f (Hz)

4.1.4 Forces

Strip theory is assumed sufficient to describe the forces acting on the cylindrical

members. Strip theory is used together with Morisons’s equation to describe the

forces acting on the structure. Morison’s equation is given in equation 7.1 [Marin-

tek, 2010].

1 1

dF = ρπD2 CM a + CD ρDurel |urel | (4.16)

|4 {z } |2 {z }

Inertia Drag

ρ = Density of fluid

D = Diameter of pipe

CM = Inertia coefficient

CD = Drag coefficient

a = Fluid acceleration perpendicular to cylinder

urel = ux − ṙ Relative fluid speed perpendicular to cylinder

In order for the Morison’s equation to be valid, long waves are required [Faltinsen,

λ

1993]. In practical application the validity limit D > 5 is often used. The largest

structural members of the jack-up have a diameter of 0.75m. For linear potential

theory with deep water assumptions this implies that the lowest valid wave period

is according to equation 4.17. Wave periods bellow this limit is associated with

small amounts of energy and are not critical.

s

10πD

T > = 1.5s (4.17)

g

T =Wave period

D = Diameter of pipe

g = gravity acceleration

31

4.1 Equation of motion for dynamic and quasi static analysis

What should be noted is that the drag term has a nonlinear dependency on fluid

speed. It also results in force components oscillating with different frequencies. A

decomposition of the drag term results in the followingg different frequencies.

8 8 8

sin(x) −

sin(x)|sin(x)| = sin(3x) + sin(5x)... (4.18)

3π 15π 105π

Equation 4.18 and figure 4.18 illustrates how the drag term contains several fre-

quency components. This induces responses in several frequencies for a structure

subject to a regular wave with one frequency component. Hence in general, care

should be taken at waves that oscillates at three times the eigenperiod of the

structure. However for this structure the period of the waves at three times the

egienperiod of the structure will be 23s, which is both a period that is rare, and

that are seen in combinations with small wave heights.

1.5 1

sin(x)|sin(x)| 8/3pi sin(x)

1 8/3pi sin(x)-8/15pi sin(3x)+8/105pi sin(5x) -8/15pi sin(3x)

0.5 8/105pi sin(5x)

0.5

0 0

y

-0.5

-0.5

-1

-1.5 -1

0 5 10 15 0 5 10 15

x x

Linear theory is used in order to describe wave kinematics. Linear theory is not

considered to give a sufficient representation of the kinematics in the surface area,

and at least second order wave kinematics should be used in a design process. In

this work the drag coefficients are tweaked in order to compensate for the difference

between linear and higher order theory. Linear theory is based on certain basic

assumptions.

• Incompressible fluid

• Inviscid fluid

• Irrotational fluid motion

• A velocity potential can be used to describe the fluid field

In addition to this kinematic boundary conditions and dynamic free-surface condi-

tion are applied. For a regular wave component this results in the velocity potential,

32

4.1 Equation of motion for dynamic and quasi static analysis

wave elevation, fluid particle speed, and acceleration found in equation 4.19 to 4.24

[Faltinsen, 1993]. This are kinematics acting in the xz plane.

gζa kz

φ= e cos(ωt − kx) (4.19)

ω

ζ(t) = ζa sin(ωt − kx) (4.20)

Due to linear theory several frequency components can be achieved by simply

superposition of waves with different frequencies.

4.1.6 Drag coefficient for linear wave theory and higher or-

der wave theory

The general consensus is that higher order wave theories is necessary to insure a

correct representation of the wave kinematics. However this is a procedure that

increase the computational demand significantly. If linear potential theory is used,

a simplification that can be used to compensate for nonlinear effects is to increase

the drag coefficients in a way that gives similar forces [NORSOK, 2007]. In linear

potential theory the wave kinematics are just defined up to the mean surface eleva-

tion meaning that values above this needs to be extapolated somehow. The method

used in this work, and that NORSOK has defined in combination with increased

drag coefficient is by wheeler stretching. This both stretches the kinematics up

to wave crest level, but also down to wave through. The stretching is carried out

according to equation 4.25.

d

z 0 = (z − ζ) (4.25)

d+ζ

z0 = new depth variable

z = old depth variable

ζ = surface elevation

d = depth

33

4.1 Equation of motion for dynamic and quasi static analysis

The drag coefficients recommended in NORSOK for both the stretched first order,

and higher order theory is given in table [NORSOK, 2007].

CD z values

1st order stretched 1.15 z > 2(m)

1st order stretched 1.15 z ≤ 2(m)

Higher order wave theory 0.65 z > 2(m)

Higher order wave theory 1.05 z ≤ 2(m)

4.1.6.1 Braces

The braces have regular circular cross section, which makes it possible to select

drag coefficients according to Table 4.3. The coefficients used for the braces in this

work, is given in Table 4.3, this is based upon what is recommended in [NORSOK,

2007], and [Goncalves, 2017].

CD CM z values

1.15 2.0 z > 2(m)

1.15 1.8 z ≤ 2(m)

4.1.6.2 Chords

The chords are not perfectly tubular, due to the rack that are mounted on the

chords. These racks are used to elevated the legs. However these racks induces

irregularities in the legs geometry and hydrodynamic properties. The legs will

have a different hydrodynamic profile depending on direction. This cross sectional

area can be seen in figure 4.3.

34

4.1 Equation of motion for dynamic and quasi static analysis

Figure 4.3: Chord cross sectional area. The different color at the tips indicates

that these are the areas which the gears are

For such cross sectional areas as the on above the relation in equation 4.26 [DNV

GL, 2014a]

W 9

CD = CD0 + (CD1 − CD0 ) sin2 ( (θ − 20◦ )) (4.26)

D 7

For cases where 20◦ ≤ θ ≤ 90◦ , W and D is defined as in figure 4.3, and for

this structure the ratio between them is W D = 1.05. The CD0 is the the regular

coefficient used for cirvular cylinders, and the CD1 is the drag coefficient for flow

normal to the rack (θ = 90◦ ). For a wave heading in the x-direction, the chords in

this structure will either have θ equal 90◦ or 60◦ . For the structure this results in

the drag coefficients given in Table 4.4.

2.04 0.65 z > 2(m) 2.0 2.14 1.56

2.04 1.05 z ≤ 2(m) 1.8 2.14 1.71

Table 4.4: Drag coefficients for chords, z is the distance from the mean surface

level, positive upwards.

35

4.1 Equation of motion for dynamic and quasi static analysis

compensate for this, an increased drag coefficient can be used [NORSOK, 2007].

This is achieved by letting the drag coefficient be 1.15 over the whole depth for

regular cylinders. However in this case the drag depend on the angle and geometry.

Thus the same scaling is used for the drag coefficient for the chords as for regular

cylinders. This involves a scaling of 1.15

1.05 for the original coefficients valid for depths

above +2m, and let this drag coefficient be valid for the whole depth.

z > 2(m) 2.0 2.34 1.87

z ≤ 2(m) 1.8 2.34 1.87

Table 4.5: Drag coefficients for chords, with the use of NORSOK scaling to com-

pensate for Gaussian sea surface.

kinematics

In order to tell weather or not the adjusted drag coefficients are conservative or

not, the wave kinematics and drag forces for one case is investigated. 5th order

stoke waves are used as reference, and a regular wave component is stepped through

the structure, and investigated under the crest and through. This results in the

velocity to depth relations given in figure 4.4.

20 20

0 0

-20 -20

depth (m)

depth (m)

-40 -40

-60 -60

-80 -80

stetch stech

-120 -120

0 0.5 1 1.5 2 2.5 3 -2.5 -2 -1.5 -1 -0.5 0

Wave particle velocity (m/s) Wave particle velocity (m/s)

Figure 4.4: Wave particle velocity under crest and through for 5th order stoke

waves, and stretched airy waves

The nonlinear wave has a larger maximum wave particle speed achieved under the

crest. Under the through however the stretching is the conservative. The resulting

forces are also of interest to compare. By dividing the drag forces by the terms not

of interest, a comparison can easily be carried out. This is the parameters except

drag coefficient and velocity, and is given in equation 4.27.

36

4.1 Equation of motion for dynamic and quasi static analysis

FD

= CD |u|u (4.27)

0.5Dρ

The results for drag forces acting on the chords that has a 60 degree angle on the

flow, the chords with 90 degrees angle on the flow and the braces are all included

under the though and crest in figure 4.5. The maximum forces are experienced

under the wave crest, and under the wave crest, the adjusted coefficient for linear

theory closes the gap in force between linear and stokes 5th order wave kinematics.

It is still not conservative when the largest force is considered. However under

the through, the adjusted coefficient with stretching gives much larger forces than

the higher order wave kinematics with regular drag coefficient. Considering the

high degree of dynamics that this structure inherit, it is not unlikely that this is

actually conservative due to the larger change in force between through and crest.

In addition to this, fatigue is sensitive to stress range, and not peak. If the range

in force is larger, the range in stress is also probably larger. Hence it is assumed

that the adjusted drag coefficient is a conservative approach for fatigue analysis.

Considering the scope of this work, it will neither be constructive usage of time to

work with higher order irregular wave kinematics.

37

4.1 Equation of motion for dynamic and quasi static analysis

H=6.5, T=8.5, Chords 90, crest H=6.5, T=8.5, Chords 90, through

10 10

0 0

-10 -10

depth (m)

depth (m)

-20 -20

-30 -30

stoke stoke

-40 stetch regular coef -40 stetch regular coef

stretch norsok adjusted stretch norsok adjusted

-50 -50

0 5 10 15 20 -15 -10 -5 0

F /(0.5 D ) F /(0.5 D )

D D

H=6.5, T=8.5, Chords 60, crest H=6.5, T=8.5, Chords 60, through

10 10

0 0

-10 -10

depth (m)

depth (m)

-20 -20

-30 -30

stoke stoke

-40 stetch regular coef -40 stetch regular coef

stretch norsok adjusted stretch norsok adjusted

-50 -50

0 2 4 6 8 10 12 -10 -8 -6 -4 -2 0

FD/(0.5 D ) FD/(0.5 D )

10 10

0 0

-10 -10

depth (m)

depth (m)

-20 -20

-30 -30

stoke stoke

-40 stetch regular coef -40 stetch regular coef

stretch norsok adjusted stretch norsok adjusted

-50 -50

0 2 4 6 8 -8 -6 -4 -2 0

FD/(0.5 D ) FD/(0.5 D )

Figure 4.5: Difference in forces for linear theory and higher order wave theory,

for one wave component. Discontinuities are due to change in components at +2m

38

4.2 Spring connections to ground

Different types of ground fixation exist for jackups. Two such types are exemplified

in figure 4.6. These systems utilizes the platform weight to be pushed down into

the ground, which fixates the legs in terms of bending and translation.

responce at resconance etc. are all influenced by the bottom conditions [DNV GL,

2015]. Bottom conditions vary from location to location, and should be modeled

accordingly [DNV GL, 2015]. With no specific location for this report, the values in

Table 4.6 are used. These values gives quite a dynamic structure with a resonance

period of above 7 seconds.

𝑘𝜃𝑦

𝑘𝜃𝑥 𝑘𝑦 𝑘𝑥

𝑘𝑧

𝑘𝜃𝑧

Figure 4.7: Bottom model of one of the legs. Left: rotation springs, Right:

Translation springs

The bottom conditions are modeled as in figure 4.7. The bottom end of the leg are

modeled with larger pipe sections. In addition rotational and translational springs

are attached to the bottom node.

39

4.3 Eigenvalue analysis

kx 4.000E+10 (N/m)

ky 4.000E+10 (N/m)

kz 5.000E+10 (N/m)

kθx 5.157E+10 (Nm)

kθy 5.157E+10 (Nm)

kθz 1.500E+10 (Nm)

Eigenfrequency or natural frequency are the oscillating periods the structure expe-

riences in the absence of damping and external forces. These periods are important

structural properties, and force excitation oscillating at these periods will result in

large responses. Equation 4.28 defines the eigenfrequencies [Langen and Sigbjorns-

son, 1986], and these can be solved in USFOS.

The amount of wetted surface and hence the water depth has a large influence on

the eigenfrequency due to the change in the added mass, which is also included

in the mass term, M. The eigenfrequency is also heavily reliant on the ground

conditions, due to the boundary conditions huge influence on the stiffness matrix

K. The 3 largest eigenperiods are presented in Table 4.7 alongside with the mode

shape.

A lot of smaller eigenperiods also exist for the structure. These are presented in

the bottom of Table 4.7 and have different modes. As described in Section 4.1.4,

excitations with smaller periods than the wave period are present in the drag term

in Morison’s equation. This might result in smaller amplitude oscillations that

might hit the smaller eigenperiods. However these oscillations are so small in size,

that it is probably not of any concern.

40

4.3 Eigenvalue analysis

Eigenmode Eigenperiod

Translation in yz plane 7.67s

0.609s

0.607s

0.590s

0.588s

0.587s

Smaller eigenperiods

0.565s

0.536s

0.534s

0.534s

0.533s

41

4.4 Solving the dynamic and quasi static equation of motion

of motion

The solving of quasi static equilibrium is rather easy, and is in principle the same

as solving a static problem at every time step. For linear analysis this consist

of inverting the stiffness matrix and multiply the inverted matrix with the force

vector. This is done at every time step, and the stiffness matrix does not change

with time as long as buckling and yielding does not occur.

The solution of the dynamic equation 4.2 is much more complex. In order to solve

the dynamic equation, time integration schemes are utilized. A lot of different

time integration schemes exists, with different properties regarding accuracy and

speed. The newmark beta method is often used in this context, however it has

several drawbacks, and can’t have both sufficient numerical damping, second order

accuracy and be unconditionally stable [Hughes, 2000].

Second order accuracy is beneficial, and considered a necessity if an algorithm

should even be considered. Second order accuracy will in practice translate to

reducing the error to a 14 by halving the time increment. This allows larger time

step to be used compared with first order accurate algorithms, which in turn lead

to faster analysis.

Numerical damping is also a vital concept, and necessary to exclude high-frequency

oscillations and vibrations not of interest or introduced by finite element discretiza-

tion.

USFOS uses the Hilber-Hughes-Taylor method, or often refereed to as α-Method.

This scheme inherit the advantages from the trapezoidal method of a second order

accuracy, while overcoming the problems related to the lack of numerical damping

[Hughes, 2000]. The numerical scheme is given by equation 4.30 to 4.32 [Hughes,

2000].

∆t2

rn+1 = rn + ∆tṙn + (1 − 2β)r̈n + ∆t2 βr̈n+1 (4.31)

2

ṙn+1 = ṙn + ∆t(1 − γ)r̈n + ∆tγr̈n+1 (4.32)

This integration scheme will be unconditionally stable if α is between 0 and − 31

and the other Newmark parameters are determined by equation 4.33 and 4.34.

1

γ= (1 − 2α) (4.33)

2

42

4.4 Solving the dynamic and quasi static equation of motion

1

β= (1 − α)2 (4.34)

4

In this work α is set to -0.1, resulting in the parameters in Table 4.8.

α β γ

-0.1 0.3 0.6

The selection of time step is important in order to insure good accuracy, but it

also has impact of the analysis time. The ideal time step is the largest time step

that gives an acceptable accuracy. It is necessary to have a decent accuracy both

in amplitude decay and relative period error. These are indicated in figure 4.8 as,

amplitude decay: AD, and relative periodical error: T̄ −T

T . Under the assumption

that the system is undamped and freely oscilating, the amplitude decay during one

cycle is proportional to the algorithmic damping by a constant 2π.

domain simulations is in this work going to be used as a reference. Hence a quite

good accuracy is needed. Engineering insight reveals that the response will be

primarily in the first 3 modes, where the first three modes is defined as 0 < ω1 <

ω2 < ω3 , and can be found in Section 4.3 as ωn = T2πn . Enforcing a maximum of

1% relative period error and amplitude decay per cycle is considered to give a good

and sufficient accuracy. This results in the criteria given in equation 4.35 and 4.36.

43

4.4 Solving the dynamic and quasi static equation of motion

T̄ − T

< 0.01 (4.36)

T

The algorithmic damping and periodical error per cycle is given in Figure 4.10

and 4.9, for the alpha method of α = −0.05 and α = −0.3. By using the graph

of α = −0.3 the limit of ∆t will be conservative. The requirement on amplitude

decay from equation 4.35 result in equation 4.37 by using Figure 4.9

∆t

ξ¯ < 0.0016 ⇒ < 0.08 (4.37)

T

For the relative periodical error, similar result is present in equation 4.38. Here

Figure 4.10 is utilized.

T̄ − T ∆t

< 0.01 ⇒ < 0.03 (4.38)

T T

Meaning that relative periodical error will be limiting the time increment. The

lowest of the relevant eigenperiod is T3 = 6.75, and hence the time increment is set

to

∆t = 0.03T3 = 0.2s (4.39)

44

4.5 Overshoot and errors in initial responses

In the initial phase of the analysis the structure has no response displacement or

velocities. This in term contributes to a different ratio between loads and responses

compared to later in an analysis where the structure is moving along with the

waves. There exists different methods to apply the loads in USFOS, they can be

gradually applied by an S-curve with a given duration. However this will still not

give a true signal since the loads are just partly applied at the initial first seconds.

Another aspect that is present is the HHT-α method tendency to overshoot the

velocities in the initial first analysis steps. It does not overshoot displacement by

nature [Hilber and Hughes, 1978]. However the velocities is used in the forcing

term and the damping of the structure. This will in turn will overshoot damping

in the initial phase. The overshoot by the HHT-α will eventually die after certain

analysis steps as illustrated in figure 4.11. Keep in mind that the scale on the

y-axis is logarithmic.

45

4.5 Overshoot and errors in initial responses

Figure 4.11: Comparison of energy overshoot for n time steps, from [Hilber and

Hughes, 1978]

These two factors will influence the stress in the initial phase of the simulation. An

example case is illustrated in figure 4.12. This is solved by letting the structure

stabilize a certain period of time before the time signal is counted. For a 3h

stochastic simulation the first 2min is not counted.

When harmonic inputs are used to find dynamic transfer functions, the amplitudes

are of interest. In this case the sensitivity is even greater. Hence 4 minutes of

stabilization is used in order to capture the response amplitudes.

46

4.6 Irregular sea surface

4

t =0.2

2

(Pa)

-2

-4

0 50 100 150 200 250 300 350 400

t (s)

In offshore environments the sea surface does not behave as an regular sine wave.

Waves with such sine surface elevation is referred to as regular waves. Regular

waves have simple kinematic relations. A common approach to obtain a surface

close to reality, is by summation of several regular wave components.

Sensors have been monitoring the surface elevation in the north sea for decades.

This enables the establishment of statistical descriptions of the sea surface. This

statistic is utilized in the summation of regular wave components, such that they

they as a sum, reflects reality. Equation 4.40 states how such a summation is

carried out.

N

X

ζ(t) = ζa,i cos(ωi t + φi ) (4.40)

i

p

ζa,i = 2S(ωi )∆ωi (4.41)

47

4.6 Irregular sea surface

ζ = Wave elevation

ζa,i = Amplitude of component i

ωi = Frequency of component i

∆ωi = Frequency increment i

φi = Random phase angle between 0 and 2π

N = Number of components

S(ω) = Wave spectrum function

Amplitude and frequency of each wave component are determined by the use of

short term statistics. 3h is the most used duration of a short term sea state. The

description of these short term sea states is in offshore industry introduced as spec-

trum functions. The spectrum used in this work is the JONSWAP spectrum. This

spectrum applies when the growth of the waves is limited bythe size of generation

sea Almar-Næss [1985]. The JONSWAP spectrum is given in equation 4.42, and

for different values of period and wave height can be seen in figure 4.13.

( −4 )

−5 ω 2

/2σ 2 )

S(ω) = αg ω 2

exp −1.25 γ exp(−(ω/ωp −1) (4.42)

ωp

H2

α = 5.061(1 − 0.287logγ) T 4S

P

γ = 3.3

(a commonly used value, however not always the case)

σ=0.07, ω ≤ ωP

σ =

σ=0.09, ω > ωP

ω = angular frequency of wave

ωP = T2πP

g = Gravitational acceleration

The representation of the sea surface is carried out by dividing the spectrum in

several parts and create one wave component for each part. The amplitude of these

components is given in equation 4.41. This splitting is illustrated in figure 4.13.

50

TP=10 (s), HS=3 (m),

40 TP=10 (s), HS=6 (m),

TP=6 (s), HS=3 (m),

TP=10 (s), HS=10 (m),

30

S(!)

S(!)

20

10

0

0 1 2 3 4

! !

Figure 4.13: PM spectrum for different sea states and summation technique

48

4.6 Irregular sea surface

The phase angle gives randomness to the sea surface. This makes sure that each

surface realization is different from each other, even though they represents the

same sea state.

One issue that arises in the context of spectrum integration is repetition within

one signal. The amount of frequencies that is used to represent the sea surface is

directly linked to the period when the signal repeat itself.

1 2π

Trep = = (4.43)

∆f ∆ω

When the left integration technique illustrated in figure 4.14 is use, i.e. the simplest,

equation 4.43 describes the period at when the signal is repeated. This gives a

certain requirement on how many wave components necessary to create a true

3h sea surface, which is not repeated. For 500 wave components, with frequency

range at about 1Hz, the repetition period is about 500s, which is far bellow 3h.

Increasing the amount of wave components is one solution, but this will increase

the computation time by a large amount, and are not feasible. In order to deal with

this issue in another way, the integrating technique can be modified as indicated in

the left of figure 4.14. By introducing some randomness within each interval, the

surface should not repeat itself in such a manner.

Δ𝜔

𝑟𝑎𝑛𝑑

2

Δ𝜔 Δ𝜔 Δ𝜔 Δ𝜔 Δ𝜔 Δ𝜔

some randomness in amplitude (right)

In the standard procedure, ωi is picked as the middle value within one interval.

Here, ωi can be written as in 4.44, where ω0 is the smallest frequency in the domain.

49

4.6 Irregular sea surface

ωi = ω0 + (i + 0.5)∆ω (4.44)

The method illustrated in the right of figure 4.14, introduces a random location of

ωi within one interval. This can in turn be written as equation 4.45.

ωi = ω0 + (i + rnd)∆ω (4.45)

The only difference between these two is the rnd, which is a random number be-

tween 0 and 1. The one with randomness is supposed not to give the same kind of

repeating elevation after Trep . The usual way to give a measurement of repetitive-

ness is by use of the auto correlation function. For a stationary process it can be

written as in equation 4.46, where τ is the lag, and Xt is the signal at time t.

R(τ ) = (4.46)

σ2

For values of τ which gives R(τ ) = 1, a perfect reproduction of the signal is present,

with a time lag of τ . In a discrete sense, this can be calculated by equation 4.47

and 4.48 [Box et al., 1994].

ck

Rk = (4.47)

c0

T −k

1 X

ck = (Xt − µ)(Xt+k − µ) (4.48)

T − 1 t=1

c0 is the variance of the time signal and T is the discrete size of the whole signal.

The discrete autocorrelation is used to verify that the method with randomness

does not reproduce signal in the same sense as the standard procedure.

1 1

wave elevation autocorrelation

0.5 0.5

0

0

-0.5

-0.5

-1

0 2000 4000 6000 8000 10000 12000 0 2000 4000 6000 8000 10000 12000

Lag (s) Lag (s)

in amplitude

50

4.6 Irregular sea surface

In figure 4.15 the autocorrelation for both cases are plotted. In this example, 500

wave components are used and the frequency bonds are set to f ∈ [0.04, 1]Hz. This

gives a period of repetition of Trep = 520s. In the approach without the amplitude

randomness there are large peaks in the autocorrelations at approximately every

520(s) of lag. This indicates that a huge degree of repetition are present at periods

equal this lag, and agrees well with previously stated.

The attempt to improve the drawback by adding randomness to the amplitude is

clearly working. A true reproduction of the surface is never present within the 3h.

The autocorrelation is never above 0.5, which is far from perfectly correlation at

1. Hence this approach is beneficial. Zoomed in version of figure 4.15 is found in

figure 4.16. Each of the peaks visible in figure 4.15 has an similar build up as in

figure 4.16.

1 1

wave elevation autocorrelation

0.5 0.5

0

0

-0.5

-0.5

-1

480 500 520 540 560 480 500 520 540 560

Lag (s) Lag (s)

The distribution of wave heights should follow a Rayleigh distribution, given that

the process is narrow banded [Myrhaug, 2016]. The JONSWAP spectrum can be

classified, or at least close to considered narrow banded. It is important that the al-

ternative integration technique tends to follow the Rayleigh distribution. The best

graphical way to illustrate how the elevation follow a distribution is by linearizing

the distribution, and plot the cumulative distribution in a linearized probability

paper.

h2

FH (h) = 1 − exp − (4.49)

8m0

The Rayleigh cumulative distribution given in equation 4.49 can be linearized by

letting the y axis in the plots bee as illustrated in equation 4.50. Then straight lines

will be Rayleigh distributed. In these equations, m0 is a distribution parameter,

and h individual wave heights.

51

4.6 Irregular sea surface

p h

y= −ln(1 − FH (h)) = √ (4.50)

8m0

In figure 4.17 this is done for one sea state for both methods of integration. It

turns out that the alternative method with amplitude randomness outperforms the

standard method when it comes to follow the Rayleigh distribution i.e. be a straight

line. By increasing the number of components, the distributions tends to be more

straight for both integration techniques. However, at only 500 components, the

alternative method is already very close to straight, which the standard method

is far from. The reason for the vertical behavior of the case without amplitude

randomness is linked to the repeated surface elevation. The largest value in every

repeated signal will be the same, and hence the resulting vertical lines.

3 3

2.5 2.5

sqrt(-ln(1-F(H)))

sqrt(-ln(1-F(H)))

2 2

1.5 1.5

1 1

Number of components: 100 Number of components: 100

0.5 Number of components: 500 0.5 Number of components: 500

Number of components: 1000 Number of components: 1000

0 0

0 5 10 15 0 5 10 15

H H

Based on the above argumentation, 500 components in combination with the al-

ternative integration technique are considered to give sufficient results for the rest

of the work. The arguments can be summarized as follows:

• The sea surface is not repeated

• It follows the Rayleigh distribution impressively well

• It is enough components to capture the spectrum function well

52

Chapter 5

Fatigue analysis of tubular joints

than material strength. Fatigue damage is present when the material is subject to

cyclic loading. Fatigue damages are a result of a three stage process consisting of

the following three stages Almar-Næss [1985].

1. Initiation or crack nucleation

2. Crack growth

3. Final failure

Fatigue is dependent on the stress range of the cyclic loading rather than the stress

itself. This means that materials subject to cyclic stresses well bellow yield, may

experience critical cracks after a certain amount of cycles. This becomes a vital

phenomenon for offshore structures due to the cyclic nature of the wave loadings.

This work concerns jack-ups which are mobile units. Their mobility involves several

phases of loadings. A few are listed bellow.

• Transit i.e. floating, traveling to new site

• Lifting and lowering legs

• Operation at site, attached to the seafloor

These phases will have different influence on the fatigue life of the structure. The

most critical fatigue spots in the process of elevating the legs might be the gears

that are used for lifting, while the transit phase might have a part of the hull as

53

5.2 Joint selection

critical spot. The scope of this work however, will be the fatigue damage in leg

joints during operational phase.

The fatigue assessments in this work is carried out by the use of Miner summation

and SN-curves. The logic of this process can be split in four steps, which are further

discussed in details.

1. Beam reactions are achieved by finite element analysis

2. Stress concentration factors (SCF) are used to generate stress processes

3. The amount of cycles at different stress ranges are calculated by the use of

rainflow counting

4. Cumulative damage is estimated by the use of Miner summation together

with SN-curves

In order to select critical joints with respect to fatigue damage, a quite detailed

screening is necessary in a design process. Considering the scope of the work in this

thesis, only a simple screening is carried out. This screening consist of comparing

the stresses at the bottom of the legs, and the top of the legs (right under the deck)

for a few waves. The joint that turns out to be critical is at the top of the legs.

This might be changed if the springs that models the attachment to sea floor is

stiffer than the ones used in this work. Then it is believable that the bottom joints

also be significantly subject to fatigue damage.

For simplicity only one of the joints in the top area of the legs is investigated

further. The joint selected is indicated with red color in Figure 5.1. The braces

used to estimate stresses is indicated with red color.

54

5.3 Miner summation

Wave heading

The load history acting on marine structures has a stochastic nature. They vary

in both amplitude and frequency over time. This establishes the need of rules on

how to count the cumulative fatigue damage due to the load history. Numerous

such rules or theories may be found in the literature Almar-Næss [1985], but Miner

summation stand out as a simple and good approach.

The basic assumption for the Miner summation is that the damage contribution

for a load cycle is constant at a given stress amplitude. The damage from one such

load cycle is given in equation 5.1.

1

D= (5.1)

N

Where D is Damage, and N Endurance of the material at the given amplitude. The

cumulative damage from a load history of several different stress ranges can then

be written as a sum as given in equation 5.2. Failure will occur when this exceeds

1. The endurance of the material at a given stress amplitude is often described by

what is called SN-curves. The SN-curves are described in the next section.

X ni

D= (5.2)

i

Ni

Where D is Damage, Ni Endurance of the material at stress amplitude i and ni

number of cycles at stress amplitude i

55

5.4 SN curves

5.4 SN curves

In order to obtain material characteristics of fatigue SN, testing is performed. SN

testing deals with machined specimens which are exposed to cyclic stress usually at

constant amplitudes. However, to a certain level, the mean stress in the material

also to some extend contribute to the fatigue life. Several models have been pro-

posed to deal with this Almar-Næss [1985]. The simple idea is displayed in Figure

5.2 where the the allowable ratio between amplitude and mean stress are displayed

for different models.

250

Goodman

200 Gerber

Sorberg

150

<a

100

50 Allowable stresses

0

0 500 1000

<m

Figure 5.2: Example on allowable combinations of mean stress and stress ampli-

tude

These three models are given in equation 5.3 Almar-Næss [1985], where σa,N are

the limiting stress amplitude, σu ultimate stress and σy yield stress. None of these

models are valid in general, and are obtained under different test conditions Almar-

Næss [1985]. In a jack-up rig there will be mean stresses different from zero in the

legs. This is due to constant reactions, including gravity buoyancy etc. However

due to the scope of this work, these aspects are not covered, and the problem will

be treated as it has zero mean stresses.

σm

Modified Goodman: σa = σa,N (1 − )

σu

σm 2

Gerber relation: σa = σa,N (1 − ) (5.3)

σu

σm

Soderberg relation: σa = σa,N (1 − )

σy

Design code SN-curves describes the allowable amount of cycles at a given stress

range. They are material and detail specific, and obtained by tests. The standard

version of a SN-curve is given in equation 5.4 Almar-Næss [1985]. Modifications to

this is made in DNV GL design rules with respect to plate thickness effects DNV

GL [2015] as seen in eqution 5.5. Furthermore the different parameters depend on

56

5.4 SN curves

the classification society. The values recommended by DNV GL, will be used in

this work.

" k #

t

log(N ) = log(ā) − mlog ∆σ (5.5)

tref

N = Allowable cycles

log(ā) = Interception of the log(N ) axis

m = The inverse negative slope of the SN-curve, can be two slopes, m1 and m2

∆σ = Stress range

tref = For tubular joints the reference thickness is 32 mm

t = Thickness through which a crack will most likely grow.

t = tref is used for thickness less than tref

k = Thickness exponent.

The SN-curves that DNV GL recommends are highly dependent on the local ge-

ometry of the particular structural detail. The scope of this work is not related

to the actual design of the structure, but rather the investigation of linearizations

and how these influence the fatigue life. The choice of SN-curve will not be as

important in this context as long as the same SN-curve are used for all the differ-

ent comparisons. The S-N curve for the detail category T in DNV GL [2014b] is

selected for analysis by request from DNV GL.

(

3.0 , N ≤ 106

m= (5.6)

5.0 , N > 106

(

11.764 , N ≤ 106

log(ā) = (5.7)

15.606 , N > 106

k = 0.25 (5.8)

The equivalent to equation 5.6 with stress as limit instead of N is given in equation

5.9

(

3.0 , ∆σ > 86.26 MPa

m= (5.9)

5.0 , ∆σ ≤ 86.26 MPa

57

5.5 Stresses in joints

Joints and welds are critical locations for development of fatigue cracks. At these

locations the concentration of stress are typically a lot larger than locations with a

certain distance to the weld. This leads to the need of a careful consideration of the

particular joint. As this is not a design paper, large focus will not be given to to

describe these stresses in details. Local finite element models with a very fine mesh

is typically used to capture detailed stresses in joints. However this introduces a lot

of modeling work, as well as increased analysis time. Known relations for tubular

joints is here used as an alternative as described in the next section.

stresses

Experimental data, analyzes and tests forms the basis of the DNV GL fatigue

recommendations. Within this work, known relations for different tubular joints

are utilized. Empirical models and relations are made for standard, widely used

joints. By the use of these models, no local detailed finite element model is needed.

The joint problem is broken down by super positioning the different bending- and

axial contributions. As seen in Figure 5.3 the total load is a sum of axial, in plane

and out of plane bending moments.

1

8 2

7 3

= + +

𝑀𝐼𝑃 𝑀𝑂𝑃

6 4 𝑁

5

Figure 5.3: Superposition of joint contributions and the eight locations of stress

evaluation

The stress level is most intensive along the weld of the joint. In order to obtain a

good description of these stresses, they are not only evaluated at the saddle and

the crown, but also an intermediate value between the saddle and the crown. The

58

5.5 Stresses in joints

value at these points are derived by a linear interpolation for axial stress, and

harmonic interpolation for the bending stress between the saddle and crown. This

leads to 8 locations as indicated in Figure 5.3. Here 1 and 5 are the crown, and 3

and 7 the saddle. The rest of the locations are intermediate locations. The stresses

at these locations are given by equations containing nominal stresses and stress

concentration factors. In equation 5.10 these relations are given.

1 1√ 1√

σ2 = (SCFAC + SCFAS )σx + 2SCFMIP σmy − 2SCFMOP σmz

2 2 2

σ3 = SCFAS σx − SCFMOP σmz

1 1√ 1

σ4 = (SCFAC + SCFAS )σx − 2SCFMIP σmy − SCFMOP σmz

2 2 2 (5.10)

σ5 = SCFAC σx − SCFMIP σmy

1 1√

σ6 = (SCFAC + SCFAS σx − 2SCFMOP σmz

2 2

σ7 = SCFAS σx + SCFMOP σmz

1 1√ 1√

σ8 = (SCFAC + SCFAS )σx + 2SCFMIP σmy + 2SCFMOP σmz

2 2 2

In the above relations, SCF are stress concentration factors, and σ nominal stresses.

AC, AS, MIP and MOP stands for axial crown, axial saddle, Moment in plane,

moment out of plane respectively.

These SCF’s can be picked from tables for known joint sections. However the

geometry for the cross sectional area in the chords is not perfectly circular which

illustrated in Figure 5.4. This cross section has a flat bar welded in between the

two half parts of the pipe. This flat bar is due to the gearing system used to

lift and lower the legs, which introduces more stiffness to the chord. Due to this

complex geometry, the stress concentration in the braces are selected for further

analysis, and the chord is treated as a circular pipe. The stress concentration in

the braces also rely on the stiffness in the chord, but it is assumed that this is not

to important. If this was a design paper, this method is not recommended, but for

research on linearization it should be sufficient.

59

5.5 Stresses in joints

Braces

Chord

Jacking gears

The SCF’s are picked from tables in [DNV GL, 2014b]. These are long equations,

and reference is here made to appendix B, tubular K-joints, in DNVGL RP-C203

DNV GL [2014b]. For the braces this result in the SCF given in Table 5.1.

SCF Value

SCFAC 3.2155

SCFAS 3.2283

SCFMIP 1.9032

SCFMOP 1.8839

As seen in Table 5.1, the SCF for axial stresses are close to equal in size. This

makes the first axial terms in equation 5.10 close to similar. The difference in the

8 spots is thus more influenced by the bending stresses. In Figure 5.5 σ1 to σ8 is

plotted for a time snip. Bending stress is clearly present, due to the different mean

values for all 8 different stresses. What is also possible to see, is that the variation

behaves quite similarly.

60

5.5 Stresses in joints

2 2

<1 <1

1 <2 1 <2

<3 <3

0 0

Stress (Pa)

Stress (Pa)

<4 <4

-1 <5 -1 <5

<6 <6

-2 -2

<7 <7

-3 <8 -3 <8

-4 -4

-5 -5

300 350 400 450 500 305 310 315 320

time (s) time (s)

Figure 5.5: Stress process over a short, and shorter time interval

This is investigated further by subtracting the signal mean. Evident in Figure 5.6

where the same stresses are plotted, but with the mean subtracted, is that the

signals are very close to equal. This indicates that bending stresses are mainly

due to constant loads, such as gravity. This is very beneficial, because then their

variation will behave similarly, and variation is the important in context of fatigue.

Thus only one of them needs to be concerned. For further analysis only σ1 will be

used.

1 1

<1 -mean(<1 ) <1 -mean(<1 )

<2 -mean(<2 ) <2 -mean(<2 )

0.5 <3 -mean(<3 ) 0.5 <3 -mean(<3 )

Stress (Pa)

Stress (Pa)

<5 -mean(<5 ) <5 -mean(<5 )

0 <6 -mean(<6 )

0 <6 -mean(<6 )

<7 -mean(<7 ) <7 -mean(<7 )

<8 -mean(<8 ) <8 -mean(<8 )

-0.5 -0.5

-1 -1

300 350 400 450 500 305 310 315 320

time (s) time (s)

Figure 5.6: Stress process over a short, and shorter time interval

It should be noted that the estimated damage is extremely sensitive to the stress

amplitudes. The stress concentration factors needs to be carefully considered, and

61

5.6 Cycle counting using rainflow and the WAFO toolbox

in design processes local modeling of this non standard cross section should be

considered.

toolbox

For fatigue, the stress range within a cycle, and amounts of cycles are essential.

There exist different ways to count cycles for a stochastic process. What is challeng-

ing with a stochastic signal compared with regular sine waves are their local peaks

and slowly varying cycles. General consensus is that the rainflow counting method

is superior to other methods for welded joints in high cycle fatigue Almar-Næss

[1985]. The rules of rainflow counting are the following Almar-Næss [1985]:

1. Rain flow down the roof, starting in either the peak or the valley. It drips

down when it reaches an edge

2. The rain stops, and the cycle is completed, when the flow reaches another

flow falling from above.

3. It also stops when a peak/valley of larger amplitude than where the it started

from is encountered.

𝜎 𝜎

1

6

2

7

3

8

4

9

5

𝑡 𝑡

Figure 5.7 illustrates how this algorithm works. In blue, the water is flowing from

valleys, and green from peaks. Each of the colored flows represent a half cycle.

Cycle 1, 3, 6, 9 are examples on when rule 3 is present. In 4, 5, 8 rule 2 stops the

cycle. The WAFO toolbox for MATLAB is used for cycle counting [WAFO-group,

2011]. This toolbox uses the rainflow counting algorithm. One of the difficulties

62

5.6 Cycle counting using rainflow and the WAFO toolbox

with the rainflow algorithm might be unpaired half cycles Almar-Næss [1985]. The

WAFO toolbox gives an output consisting of cycles. In some cases this means that

different half cycles must be paired. The stress ranges have to be sorted in groups,

i.e. discretized. For this work 200 levels of stress range will be used. This means

that between the largest and the smallest stress range within the time signal, 200

levels are present. A simple signal, consisting of two regular sine waves are used in

order to give a certain idea on how this algorithm detects cycles.

2

sin(3x)+sin(0.5*x)

-1

-2

0 2 4 6 8 10

x

In Figure 5.8 the different cycles, detected by the WAFO toolbox, are included

with colors. The associated signal range are also indicated. If water flows from

the peaks, this is exactly the peaks generated by the use of the 3 rules of rainflow

algorithm.

It is interesting to compare the rainflow counting with a simple zero up crossing

counting. Zero up crossing counting takes the stress range as the maximum minus

the minimum value of the signal between two zero up crossings. This is illustrated

in Figure 5.9.

Δ𝜎𝑖

Figure 5.9: Zero up crossing counting. Zero up crossings marked as black dots

The two methods gives very close to the same distributions evident in Figure 5.10.

The stress range distributions are plotted in a linearized Weibull probability paper,

a plot technique described further in Section 6.2. As expected the largest stress

range from the rainflow method is larger than the largest from the zero up crossing

63

5.7 Stochastic fatigue analysis in thr frequency domain

method, however the difference is small. The largest stress range may significantly

contribute to the fatigue damage and might be important to capture. The total

damage is calculated using miner summation and the SN-curve. The rainflow

counting gives a larger damage than the zero up crossing method. However this

mighe be considered negligible in this case. Rainflow counting gives 5% more

damage.

HS=4.5 TP=8.5

4

0

ln(-ln(1-F))

-2

-4

Zero upcrossing, D=5.3e-07

-6

Rainflow, D=5.6e-07

-8

13 14 15 16 17 18

ln( ) Pa

ing vs rainflow counting. D is the total accumulated damage during 3h

domain

A frequency domain analysis is often preferable. For linear problems this will be

the most efficient way to carry out a fatigue analysis. In order for a frequency

analysis to be conducted the response spectrum has to be known.

how much energy the system is subject to under different frequency influences. In

linear systems the relation between any input spectrum and thr response spectrum

can be described by a single transfer function. If the response spectrum is known,

two ways can be used to calculate fatigue damage described in Section 5.7.1.2 and

5.7.1.1.

64

5.7 Stochastic fatigue analysis in thr frequency domain

For a narrow band process, which is the result of a resonant system subject to a

certain excitation. In such systems there will be 10800

Tn cycles during a 3h short

term sea state on average. The encountered fatigue damage can then be calculated

using equation 5.11 [Newland, 2005].

Z ∞

10800 1

f∆σ (∆σ)d∆σ (5.11)

Tn 0 N (∆σ)

Where N (∆σ) is the SN-curve given in equation 5.5. For a kinked SN-curve this

involves splitting the integral in two. The probability distribution will for a Gaus-

sian narrow banded process be Rayleigh distributed, and the parameter can be

calculated from the response spectrum.

Even dough Section 5.7.1.1 might be the easiest and most convenient way to es-

timate the damage, it involves assumptions of Gaussian narrow banded process.

This will not always be the case when the system is subject to a sharply peaked

input spectrum, which might result in a double peaked response spectrum.

Instead, the stress process can be realized in similar manner as the surface proses is

realized described in Section 4.6. Without inverting any stiffness matrices or doing

any kind of time integration, a realization of the stress process can be created given

that the response spectrum is known. This is obviously a much faster approach

than doing time domain analysis. The difficulties however lies within how these

response spectra are generated, which will be discussed throughout this thesis. For

the cases where they are known, and a realization is generated, the same technique

of calculating fatigue as for time domain simulations can be used. This is by using

rainflow counting and miner summation.

The spectra covered in Section 4.6, creates the foundation of generating time sig-

nals. However it is sometimes convenient to go the other way around, to generate

spectra from time signals. This includes creating response spectra from response

signals.

All signals processed in this work is discrete, hence the Discrete Fourier transforms

are used to generate spectra from time signals. When Discrete Fourier transforms

are used to create spectra, the estimated spectra becomes very noisy, and are

not usable to any practical purposes. This noise however hold certain statistical

properties. The distribution of the signals in the estimated spectrum from Fourier

65

5.7 Stochastic fatigue analysis in thr frequency domain

of freedom Newland [2005].

One of the easiest ways to accommodate the noise is by averaging adjacent esti-

mated spectrum values. This is in fact the only possible thing to do in order to

improve statistical accuracy [Newland, 2005]. As an example to underline that this

will provide a better estimate, the following averaging can be used as an example.

1 1 1

S̃(ωi−1 ) + S̃(ωi ) + S̃(ωi+1 )

S(ωi ) = (5.12)

3 3 3

The distribution of the error of this linear combination will be a chi-square dis-

tribution χ2η with η = 6 degrees of freedom. For the general case, when 2n + 1

adjacent values are averaged their distribution will have a chi-square distribution

with η = 4n + 2 degrees of freedom making it a much more statistical reliable value

for increased values n. This is clearly evident in the ratio between standard devi-

ation, s and mean, µ in equation 5.13, which is true for a chi square distribution.

By increasing η there is more confidence that the sample measurement lies close to

the true mean.

r

s 2

= (5.13)

µ η

However this can only be achieved at the expense of frequency resolution [Newland,

2005]. In order to get insight in how many adjacent points used in averaging that

are beneficial, the JONSWAP spectrum is realized with 5000 wave components.

The raw one sided spectrum from Fourier transform can be seen in Figure 5.11 as

the spectrum with no smoothing.

6 25

n=20

5 Analytical

n=50 20

n=800

4

15

S( )

S( )

3

10

2

5

1

0 0

0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5

Figure 5.11: Example of spectrum with and without averaging. JONSWAP with

HS = 4.5, TP = 8.5

66

5.7 Stochastic fatigue analysis in thr frequency domain

When 5000 components are included, n=50 seem like a good value to use in av-

eraging . The higher values of n, gives bad results due to insufficient frequency

resolution.

67

Chapter 6

Time domain simulation

In this chapter, the results from time domain simulations are presented. Time do-

main simulations are considered the state of the art method to estimate the fatigue

damage, and this will be used as reference when linearizations are considered.

The importance of different phenomenons are also investigated such as hydrody-

namic damping and dynamic amplification. In addition different realizations of the

same sea state are compared.

ence

The hydrodynamic damping is included as relative velocity in the drag forces. This

damping may significantly contribute to reduce the fatigue damage by damping out

the response. Hydrodynamic damping is without doubt present, but whether or

not this hydrodynamic model is conservative or not may be questioned. In order to

establish a certain understanding on how much damage that is added by removing

this damping, a few cases are compared. The stress response from two sea states are

included in figure 6.1, and it might be hard to see the difference between damped

and undamped response, but the undamped response is slightly larger than the

damped. However this small difference might be significant to the fatigue damage.

68

6.2 Fitting the short term stress range with Weibull probability distribution

4 5

undamp dyn undamp dyn

2

(Pa)

(Pa)

damp dyn damp dyn

static 0 static

0

-2 -5

0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3

t (h) t (h)

Figure 6.1: Short term stress repsonse with and without hydrodynamic damping

and static analysis

In table 7.6 some of the most critical sea states when fatigue is considered are

included. The damping clearly removes a lot of damage, and about 40-50% of the

damage is avoided when the damping is included.

The fact that these particular sea states are the most important with their as-

sociated low periods gives rise to another aspect. The largest eigenperiod of the

structure is within this region. The largest eigenperiods are 7.67s and 6.75s. Due

to this, it is expected that the dynamic effects in the motion will be significant. In

figure 6.1 the difference between dynamic and static is huge. It is also interesting

to see how these dynamical effects influence the experienced damage. The same

sea states are also included by the use of static analysis in table 7.6. It turns out

that the dynamical effects are very important in terms of fatigue damage.

HS (m) TP (s) Damage with hyd. damping w/o hyd. damping Static

4.5 9.5 1.4e-07 2.4e-07 2.4e-09

3.5 7.5 3.6e-07 5.6e-07 1.4e-09

3.5 8.5 1e-07 1.6e-07 1.7e-09

5.5 9.5 6.2e-07 1.2e-06 3.9e-09

6.5 11.5 1.4e-06 3e-06 8.3e-09

Table 6.1: Short term sea states (3h) with and without hydrodynamic damping in

addition to a pure static simulation

rest of the work is based on simulations with hydrodynamic damping included.

probability distribution

The two parameter weibull distribution has shown to represent the stress range

in jackets in a fairly sufficient manner [Tahery, 2015], however certain mismatch

between statistical and true distribution showed to be present. Jackets do have

69

6.2 Fitting the short term stress range with Weibull probability distribution

more stiffness than jack-up’s, but share a lot of structural properties with jack-ups.

They both have thin structural members. A 3-parameter weibull distribution will

improve the lower tail behavior. Much of the mismatch in the two-parameter dis-

tribution also might be corrected by the third parameter. The use of the Weibull

model reduces the complexity of the stress range to three parameters. The cumu-

lative distribution function is given in equation 6.1.

( β )

∆σ − λ

F∆σ (∆σ) = 1 − exp (6.1)

α

The might most practical way to estimate these parameter is by the use of method

of moments. The idea is to let the mean, variance and the coefficient of skewness

be the same in the data and the fitted distribution. The skewness is for a data set

is given by equation 6.2, where g1 is the skewness of the data , ∆σk is a member

of the sample of stress ranges, and µ∆σ and s∆σ the sample mean and standard

deviation.

1

PN

N k=1 (∆σk − m∆σ )3

g1 = (6.2)

(s2∆σ )3/2

For a 3 parameter Weibull distribution the mean, variance and skewness is given by

equation 6.3 to 6.5. By equating g1 and γ1 , β can be found by numerical equation

solvers, and then α and λ can be found explicit.

2

= α2 Γ(1 + 2/β) − Γ2 (1 + 1/β)

σ∆σ (6.4)

γ1 = (6.5)

[Γ(1 + 2/β) − Γ2 (1 + 1/β)]3/2

These three parameters are found for 93 different sea states. Dynamic simulations

with 3h duration are used to generate the stress history, and rainflow is used for

cycle counting. Stresses bellow 1 Mpa is left out because they give minimal damage,

and are associated with noise. To get an idea on how well these samples are fitted

with a 3-parameter Weibull distribution, a linearized Weibull plot is used.

By doing some algebraic manipulation, equation 6.1 can be changed such that the

right hand side behaves linearly. By letting the y axis be the same as the left hand

side of equation 6.6 and the x axis be ln(∆σ −λ), a cumulative Weibull distribution

will behave linearly.

70

6.2 Fitting the short term stress range with Weibull probability distribution

The cumulative distribution of a sample is given by equation 6.7, where ∆σk is the

k’th member of the sample, and n is the size of the sample.

0 when ∆σ < ∆σ1

F∆σ (∆σ) = nk when ∆σk ≥ ∆σ < ∆σk+1 (6.7)

1 when ∆σ > ∆σn

of the sample will behave as a straight line. It is also of interest to see how the

fitted distributions compares to the histograms of the actual cycle history. If the

probability distribution is multiplied with the width of a bars in the histogram in

addition with the total number of cycles as in equation 6.8, the expected number

of cycles can be plotted.

Two such distributions with linearized Weibull distribution in addition to the his-

togram of the data is given in figure 6.2 and 6.3. The other of the 93 sea states

can be seen in appendix B. The lower tail has a low degree of importance for the

fatigue damage. In the lower tail a certain degree of missmatch is present, but this

area is not of importance. In the upper tail, the Weibull distribution seem to give

good results. It is concluded that 3 parameter Weibull distribution is adequate to

represent the stress range process.

3 180

Data

2 160 weibull 3param

Expected number of cycles

1 140

0 120

ln(-ln(1-F("<)))

-1 100

-2 80

-3 60

-4 40

Data

-6 0

14 14.5 15 15.5 16 16.5 17 17.5 0 0.5 1 1.5 2 2.5 3 3.5 4

ln("<-6) (Pa) "< (Pa) #10 7

Figure 6.2: Linearized 3 parameter weibull plot of stress range during a 3h real-

ization

71

6.3 Damage of same sea state with different realizations

3 200

Data

2 weibull 3param

1 150

ln(-ln(1-F("<)))

-1 100

-2

-3 50

Data

-5 0

13 14 15 16 17 18 0 1 2 3 4 5

ln("<-6) (Pa) "< (Pa) #10 7

Figure 6.3: Linearized 3 parameter weibull plot of stress range during a 3h real-

ization

izations

Different realizations of the same sea state give different elevation at a given point

in time. Different surface realizations can be generated by re-generate the random

phase angles and the random amplitudes in the sea surface generation. It is known

that different realizations will give different maximum surface elevation even dough

the sea state is the same.

However it is unclear weather or not this also has an influence on the fatigue dam-

age. It is possible to think that the influence of these random selected phases will

not be of importance for fatigue damage since fatigue is related to the stress range

over time, and that these will be equal when the duration of the simulation is as

long as 3h. It is however important to keep in mind that large waves will give much

larger damage than small ones due to the exponent in the SN-curve. In order to

draw any conclusions, the effect of different realizations on fatigue damage is inves-

tigated further. Figure 6.4 clearly illustrates how much the particular realization

might influence the total damage.

72

6.3 Damage of same sea state with different realizations

Figure 6.4: Distribution of total damage for different realizations of the same sea

state

The variation of damage is large, and the most damage encountered during a real-

ization is more than twice the smallest value. However most of the realizations gives

damage that is within a reasonable deviation of the mean. Due to the time con-

sumption of running several realizations of each sea state this will not be preferred

for other sea states, but is included for this sea state to underline the importance

of different realizations. If the time consumption of time domain simulation had

been smaller, at least three surface realizations of the same sea state should be

averaged in order to obtain more statistical accuracy.

The standard deviation of the stress processes along with the standard deviation

in the surface elevation process is included in Table 6.2. The deviation between

maximum and minimum standard deviation in stress is not large. However keep

in mind that the slope of the SN-curve is 5 for the most stress ranges. This means

that the damage is proportional to the the stress to the power of 5. The largest

standard deviation is 21% larger then the smallest. With 1.215 = 2.7 the the huge

variation in damage in Figure 6.4 seem reasonable

73

6.3 Damage of same sea state with different realizations

1 5.19e+06 1.13 2 5.14e+06 1.12

3 4.83e+06 1.12 4 5.11e+06 1.13

5 5.05e+06 1.13 6 5.14e+06 1.12

7 4.96e+06 1.12 8 5.79e+06 1.14

9 4.97e+06 1.13 10 5.05e+06 1.13

11 4.87e+06 1.13 12 5.27e+06 1.12

13 5.29e+06 1.11 14 4.75e+06 1.12

15 4.94e+06 1.13 16 5.19e+06 1.12

17 5.10e+06 1.12 18 5.11e+06 1.12

19 4.96e+06 1.12 20 5.09e+06 1.13

21 5.32e+06 1.13 22 5.08e+06 1.12

23 4.97e+06 1.12 24 5.16e+06 1.13

25 4.98e+06 1.13 26 5.02e+06 1.12

27 5.05e+06 1.13 28 5.02e+06 1.13

29 5.06e+06 1.13 30 5.11e+06 1.13

31 4.79e+06 1.12 32 4.82e+06 1.12

33 5.42e+06 1.12 34 5.01e+06 1.13

35 5.07e+06 1.13 36 4.86e+06 1.12

37 5.15e+06 1.13 38 5.14e+06 1.12

39 5.29e+06 1.13 40 4.96e+06 1.13

41 5.13e+06 1.13 42 4.80e+06 1.12

43 5.03e+06 1.12 44 5.10e+06 1.12

45 5.10e+06 1.13 46 5.09e+06 1.13

47 5.03e+06 1.12 48 5.02e+06 1.12

q P for stress process

q P and surface elevation process

1 N 1 N

HS = 4.5, TP = 8.5. sσ = N i=1 σi , sζ = N i=1 ζi2

2

The distribution of stress range in these 48 realizations of the same sea state will

differ slightly. It is interesting to compare these in order to get a picture on where

the different damage contributions are located. The stress range of the realization

that gives the most, and the one that gives the least damage is depicted in figure

6.5. In this figure, linearized probability plot of the cumulative Weibull is used.

74

6.3 Damage of same sea state with different realizations

4

0

ln(-ln(1-F))

-2

-4

-6 Realization 32

Realization 8

-8

14 15 16 17 18

ln( ) Pa

Figure 6.5: Weibull plot of the two realizations that gives most and least damage

during a 3h simulation, fitted with 3 parameter Weibull distribution. Keep in mind

that ln(∆σ) is on the axis, not the ln(∆σ − λ)

Even dough the two realizations might seem to give quite similar distributions, the

accumulated damage from these realizations differs by a factor of more than two.

A 3 parameter Weibull distribution is used to fit the different realizations, and

explains the curved fitting tails. The fit results in three parameters, α, β and λ.

The cumulative Weibull distribution is given in equation 6.9

( β )

∆σ − λ

F∆σ (∆σ) = 1 − exp (6.9)

α

Each realization can be described by these three parameters. This means that also

the parameters will differ for the different realizations. In figure 6.6 the parameters

for all the 48 different realization are included in histograms.

Figure 6.6: Distribution of the three parameters in the Weibull fit for different

realizations. A total of 48 realizations of the same sea state are included

75

6.4 Weibull parameters for several short term sea states

states

The stress range history has been generated for numerous short term sea states.

The corresponding three parameters for the Weibull distribution has been calcu-

lated according to Section 6.2. It is possible to believe that the nature will behave

smooth to some extend, and that these parameters can be described by a quite

smooth surface. Piecewise linear interpolation polynomials are used to represent

this surface. The three parameters are visualized in Figure 6.7. Both the β and

the α parameters behaves very well, and gives a smooth surface. The λ surface

tends to be less smooth. However since the resolution of data points is so good,

the model is assumed to give a good representation of the stress range process.

Figure 6.7: Weibull parameter values for different sea states. Data points are

indicated with circular dots

The total fatigue damage is next to be calculated for every sea state described

in previous section. This can be carried out in two ways, where one consist of a

76

6.5 Fatigue damage from short term sea states

explicit expression of the the Weibull parameters and SN-curve and the other of

miner summation of the cycles. The expression for the explicit case can be seen in

equation 6.10 [Larsen, 2014].

Z ∞

f∆σ (∆σ)d∆σ

E3h [D] = N3h (6.10)

0 N (∆σ)

Where N3h is the total amount of cycles within a 3h sea state, N (∆σ) the SN

curve as a function of stress range and f∆σ (∆σ) the probability distribution of

the stress range, which in this case is the 3 parameter Weibull probability density

function. This integral has a simple explicit expression as long as simple SN-curves

are used. However a kinked SN-curve is used in this work, and hence the simpler

implementation is numerical integration of the integral. The total number of cycles

are gathered from the rainflow counting, and will vary with sea state. However the

dominant resonance period will have the largest influence on this. This number is

10800

Tn = 1408. By evaluating the rainflow counting record, it turns out to be correct

that these variations are quite small for most of the sea states. This can be seen

in figure 6.8. The majority of the sea states has a N3h of about 1500.

Figure 6.8: Number of cycles for different short term 3h sea states, Data points

are indicated with circular dots

The other method consist of using miner summation to count the cycles as outlined

in chapter 5. Both these methods should give close to the same damage. The results

are illustrated in figure 6.9.

77

6.6 Damage in the scatter format

Figure 6.9: Short term damage calculated by both the probability distribution and

by counting, Data points are indicated with circular dots

More specific the error between these methods is of order 10−6 . As expected sea

states with high HS /TP ratio will generate the most damage.

The short term damage resulting from time domain simulations is presented in the

scatter format in Table 6.3

2<TP<3

2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 13.5 14.5 15.5 16.5 17.5 18.5 19.5 20.5

0<HS<1 0.5

1.5 Cutoff

2.5 2.22E-09 3.96E-09 1.18E-08 4.54E-08 1.56E-08 5.56E-09 4.30E-09 3.47E-09 2.85E-09 2.56E-09 2.28E-09 2.00E-09 1.81E-09 1.68E-09 1.62E-09 1.07E-07

3.5 1.36E-08 7.68E-08 3.59E-07 1.02E-07 2.70E-08 1.74E-08 1.10E-08 7.28E-09 5.51E-09 4.28E-09 3.30E-09 2.77E-09 2.42E-09 2.26E-09 6.35E-07

4.5 4.83E-07 2.05E-06 5.59E-07 1.41E-07 8.55E-08 4.58E-08 2.56E-08 1.68E-08 1.10E-08 7.14E-09 5.04E-09 4.03E-09 0.00E+00 3.38E-09 3.44E-06

5.5 9.19E-06 2.42E-06 6.18E-07 3.72E-07 1.84E-07 9.50E-08 5.72E-08 3.32E-08 1.88E-08 1.17E-08 8.54E-09 1.30E-05

6.5 2.28E-06 1.37E-06 6.72E-07 3.31E-07 1.94E-07 1.06E-07 5.70E-08 0.00E+00 2.21E-08 5.03E-06

7.5 7.35E-06 4.44E-06 2.12E-06 1.04E-06 5.96E-07 3.18E-07 1.65E-07 0.00E+00 6.14E-08 1.61E-05

8.5 1.27E-05 6.00E-06 2.99E-06 1.72E-06 9.03E-07 4.54E-07 2.53E-07 2.50E-05

9.5 3.18E-05 1.54E-05 7.68E-06 4.57E-06 2.36E-06 1.20E-06 6.63E-07 6.37E-05

10.5 3.51E-05 1.80E-05 1.09E-05 5.91E-06 3.00E-06 1.62E-06 7.45E-05

11.5 2.37E-05 1.33E-05 6.96E-06 0.00E+00 4.40E-05

12.5 8.47E-06 8.47E-06

0 0 2.22E-09 1.75E-08 5.72E-07 1.16E-05 3.1E-06 1.04E-05 5.08E-05 5.95E-05 3.02E-05 4.18E-05 2.29E-05 1.19E-05 1.1E-05 1E-07 3.88E-09 3.38E-09 0 0.000254

The long term damage takes into account how many of each sea states that occurs

over a period of time. The scatter diagram describes how many of each sea state

that is present for a period of time. The history is usually generated by physically

measure the surface elevation at the site. The scatter used in this work is the

scatter for the Ekofisk field, generated by measurements over 56 years.

It is possible to estimate the damage encountered during these 56 years in each

block by multiplying the number of short term sea states encountered during 56

years with the damage from each short term sea state. The damage from each

short term sea state is given in figure 6.3. As expected wave with a high height to

amplitude ratio gives the largest damage.

78

6.7 Conclusive remarks for time domain results

This results in the the total accumulated damage given in figure 6.4. The total

damage encountered from all block is quite low, and indicate low degree of damage,

maybe lower than one could expect. However simplifications are made throughout

the work.

2<TP<3

2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 13.5 14.5 15.5 16.5 17.5 18.5 19.5 20.5

0<HS<1 0.5

1.5 Cutoff

2.5 1.32E-08 1.23E-06 6.34E-05 0.000583 9.19E-05 1.03E-05 4.05E-06 2.42E-06 1.31E-06 5.35E-07 2.36E-07 1.72E-07 4.73E-08 2.28E-08 1.55E-08 0 7.59E-04

3.5 1.00E-08 1.80E-05 0.001118 0.000801 9.21E-05 1.85E-05 4.43E-06 1.67E-06 7.38E-07 2.64E-07 9.76E-08 3.78E-08 1.51E-08 6.55E-09 0 2.06E-03

4.5 4.56E-07 0.000113 0.001187 0.000555 0.000125 2.13E-05 3.98E-06 8.00E-07 3.34E-07 2.02E-07 2.19E-08 1.04E-08 0 1.76E-09 2.01E-03

5.5 9.16E-06 0.00042 0.000716 0.000544 0.000104 1.70E-05 1.49E-06 6.58E-07 1.60E-07 6.81E-08 5.83E-09 1.81E-03

6.5 0.000332 0.000621 0.000278 7.21E-05 7.79E-06 1.26E-06 3.50E-07 0 1.86E-08 1.31E-03

7.5 5.28E-05 0.000312 0.000302 0.000116 2.57E-05 5.74E-06 9.21E-07 0 5.67E-08 8.15E-04

8.5 0.000101 0.000172 0.000181 5.20E-05 1.04E-05 3.47E-06 7.30E-07 5.22E-04

9.5 3.17E-05 9.15E-05 0.000129 0.000111 3.16E-05 1.49E-05 6.42E-07 4.10E-04

10.5 3.52E-05 0.000126 9.48E-05 3.92E-05 1.99E-05 1.57E-06 3.16E-04

11.5 7.01E-05 6.31E-05 2.62E-05 0 1.59E-04

12.5 8.06E-06 8.06E-06

0 0 1.32E-08 1.24E-06 8.19E-05 0.001823 0.0025 0.001758 0.001757 0.001011 0.000648 0.000365 0.000153 6.63E-05 1.12E-05 1.3E-07 2.21E-08 1.76E-09 0 0.010176

The damage encountered during 56 years at Ekofisk is lower than one would expect.

However no detailed screening is present. The sea state that turns out to be most

critical, might have a lower significant wave height than one could expect. The

natural period of the structure is accountable for this. It induces an increased

damage in a region of the scatter that has a high occurrence frequency. It must

also be stated that more statistical reliable results can be obtained by at least

average three realizations of the same sea state.

79

Chapter 7

Linearization

Linearization is a commonly used feature to avoid the time consuming time domain

simulation. In short it involves the establishment of a linear relationship between

response and wave elevation. It is commonly used in fatigue life estimation. In

fatigue damage, moderate sea states are the most critical, thus making it easier

to linearize waves about these most critical values. Linearization might even give

very accurate response estimate for certain structures. However for jack-ups, which

are heavily drag dominated due to the small dimensions of the members, it is not

possible to describe the response perfectly by a linear relationship. This is due to

the fact that drag forces are nonlinear. These nonlinear forces can be seen in the

drag term of Morison’s equation 7.1.

1 1

dF = ρπD2 CM a + CD ρDurel |urel | (7.1)

4

| {z } |2 {z }

Inertia term Drag term

ρ = Density of fluid

D = Diameter of pipe

CM = Inertia coefficient

CD = Drag coefficient

a = Fluid acceleration perpendicular to cylinder

urel = Relative fluid speed perpendicular to cylinder

In order to make proper assumptions in the linearization process it is important

to understand how important the drag term is. The ratio between drag force

amplitude and inertia force amplitude is an parameter that is easy to calculate,

80

7.1 Motivation for proper drag force treatment

and gives a good first impression of the importance of drag. In general the ratio

between drag force amplitude and inertia force amplitude can be written as in

equation 7.2 if the structure self velocity is neglected in the relative velocity.

1 2 2

FD,a 2 CD Dζa ω CD

= 1 2 2

= H (7.2)

FM,a 4 CM πD ζa ω

CM πD

Drag coefficient vary with the members. The chords that is 90◦ aligned with the

flow has a larger drag coefficient than the ones that are 60◦ . The braces on the other

hand has a lower drag coefficient, but also a smaller diameter. In Figure 7.1 the

ratios for the different types of members are plotted as a function of wave height.

The dashed lines indicates where the drag forces becomes larger than the inertia

forces. For this particular jackup this is experienced at quite low wave heights. In

comparison a typical jacket member with a diameter of 2m and a drag coefficient

of 1.15, will give a limit wave height of about 10m for where drag becomes more

important than inertia.

6

amplitude ratio

Chords 90°

2 Chords 60°

FD /FM

Braces

Ratio=1

0

0 2 4 6 8 10

H (m) wave height

Figure 7.1: Ratio between amplitudes of drag force and inertia force as a function

of wave height

A significant wave height of 4.5m is the height of one of the most critical sea states

for fatigue damage. At these wave heights the drag force amplitude is about twice

the amplitude of the inertia force for all the members. The braces a bit more.

This underlines the importance of the drag forces, and that they are included in a

sufficient manner.

The importance of drag is investigated further by a case. The jack-up response from

the same stochastic surface elevation with and without drag forces are compared.

81

7.1 Motivation for proper drag force treatment

107

4

CD = Included

2 CD = 0

(Pa)

-2

-4

0 0.5 1 1.5 2 2.5 3

t (h)

Figure 7.2: Short term stress response fro comparison of drag vs no drag.

HS =4.5, TP =8.5

In Figure 7.2 the stress response for a short term sea state is included. The case

with drag forces clearly gives larger stress than for the case without drag. However

the difference might not be as large as one could believe from Figure 7.1. It should

also be kept in mind that turning on drag also turns on hydrodynamic damping in

this regard.

v

u

u1 X N

sσ = t σ2 (7.3)

N i=1 i

signal and is given in equation 7.3. Both the standard deviation of his process and

the damage calculated using rainflow counting can be seen in Table 7.1.

CD =Included CD = 0

sσ 4.9e06 3.7e06

Damage 0.55e-06 0.13e-06

Table 7.1: Damage and standard deviation during 3h for different values of CD .

TP = 8.5, HS = 4.5.

Even dough the ratio between the standard deviation for the case with and without

drag is only 1.3, it increases the damage by a significant amount. Keep in mind

that the kink of the SN-curve is located at 86 MPa, which means that all the

stress cycles for this short term sea state will be in the area where the slope of

82

7.2 Harmonic inputs used to create transfer functions

the SN-curve is m=5. This suggests that the ratio between damage from the case

with and without drag should be 1.325 = 4.1. This corresponds well with what is

conducted with the rainflow counting. Increasing the damage by a factor of 4 must

be considered significant.

The stress range distribution for the two cases seen in Figure 7.3. Keeping in mind

the logarithmic scale, the difference is actually quite large. It also underlines how

important the slope in the weibull distribution is in the context of fatigue.

HS=4.5 TP=8.5

5

ln(-ln(1-F))

-5

CD = 0

CD = Included

-10

14 15 16 17 18

ln( ) Pa

Figure 7.3: Short term stress distribution (3h) for drag vs no drag

tions

In offshore industry it is common practice to linearize for wave height. This is done

by calculating the response amplitude for regular harmonic inputs with different

frequencies and use this to estimate a transfer function. A major question in this

context is how to select the wave heights for the harmonic inputs in order to capture

the nonlinearities due to drag forces in a good way. Different methods can be used

to select proper values of wave height for the harmonic input wave. Constant wave

steepness is frequently used to select these wave heights. However there are no

proper guidelines on how to select the wave steepness.

Three different schemes are investigated. These schemes describes what wave

heights to be used as inputs when transfer functions are established.

83

7.2 Harmonic inputs used to create transfer functions

Usually the ratio between wave height and wave length is defined as steepness.

However for deep water this value is proportional to the wave height divided by

the period squared, which is more convenient in this work, and are hence used as

a measurement of wave steepness as given in equation 7.4.

H H

∝κ= 2 (7.4)

λ T

Constant wave steepness is frequently used to select wave heights. Much of the

reasoning behind this might be explained by the ratio between drag force amplitude

and wave height.

FD,a

∝ Hω 2 (7.5)

H

By inserting wave heights determined by a constant wave steepness given in equa-

tion 7.6 in equation 7.5,

κ

H = κT 2 = (7.6)

4π 2 ω 2

the drag force divided by wave height will behave as a constant independent of ω.

FD,a

∝κ (7.7)

H

When transfer functions are established the response amplitude is divided by the

wave height. Hence by selecting input wave heights from constant steepness a

constant transfer function of drag is achieved. This might be an adequate property

for use in linearization.

Keeping the steepness constant will give unrealistically large responses at low fre-

quencies. Hence a cutoff might be necessary. The low frequencies should be capped

at wave height equal the one year return period [ISO, 2007]. For the Ekofisk field,

this can be calculated from the Weibull distribution as in equation 7.8 with ρ = 1.99

and β = 1.35 [Aarsnes, 2015].

β1 !

1

HS,1year = ρ ln (7.8)

2920

β ρ HS,1year

1.35 1.99 9.26m

Table 7.2: Significant wave height with one year return period for the Ekofisk field

However as it will turn out, this cutoff limit is not suitable for all sea states. It

is for instance to low for a sea state with a significant wave height of 7m. When

84

7.2 Harmonic inputs used to create transfer functions

the cutoff is applied, the wave heights used in the harmonic inputs will look like

the lines in Figure 7.4. It is worth noting that the waves will be breaking with a

steepness value of κ = 0.23 (this corresponds to H 1

λ = 7 ). The goal is to calibrate

κ in such a way that is will produce similar responses as reality.

10

Steepnes: 0.005

9 Steepnes: 0.162

Steepnes: 0.35

8

7

Wave height

0

0 1 2 3 4 5 6 7

!

Figure 7.4: Constant wave steepness curves with cutoff at one year return period

7.2.1.2 Using constant wave height-period ratio for selecting wave heights

Instead of tuning the model for wave steepness, the ratio between wave amplitude

and period might be an option, considering that drag is the most important param-

eter. By letting the wave height be determined by having a constant ratio between

wave height and period certain properties are achieved. This ratio is described as

c in equation 7.9.

H

⇒ H = cTc= (7.9)

T

For each value of c the drag force amplitude will behave constant at all values of

ω. The amplitude of the drag force will then be constant for the wave heights used

as input.

FD,a ∝ (ζω)2 ∝ c2 (7.10)

In Figure 7.5 some of these constant height-period curves are included. Also in this

case a cutoff height might be necessary. The goal is to calibrate c in such a way

that is will produce similar responses as reality.

85

7.2 Harmonic inputs used to create transfer functions

20

c: 0.05

18 c: 0.936

c: 2

16

14

Wave height

12

10

0

0 1 2 3 4 5 6 7

!

Figure 7.5: Wave heighs used as input given constant ratio between wave height

and period

Instead of letting the wave height used in input be equal to constant wave steepness

other possibilities may be considered. An approach that seem reasonable is to

letting the wave heights be somewhat equal to the different sectors in the wave

spectrum. This will result in the wave heights used in the input having similar

characteristics as the stochastic signal. This means that the high energetic sectors

of the spectrum will have a higher input wave height. A good first starting point

in order to do this is by considering the relation 7.11.

p

H(ωk ) = 2 2S(ωk )∆ω (7.11)

used to estimate the transfer functions, the wave height will decrease (∆ω decrease).

The wave heights used as inputs should not decrease by counting more frequencies.

By tweaking the relation with a constant G to form equation 7.12 the representation

of the wave heights will still be similarly distributed as the stochastic process. The

constant G can then be determined by using the calibration technique given in

Section 7.2.2 p

H(ωk ) = G S(ωk ) (7.12)

7.2.2 Calibration

The value of the wave parameters (κ, c and G) can be calibrated by the two main

steps in Section 7.2.2.1 and 7.2.2.2. In Section 7.2.2.1 spectral methods are used

to calculate the most probable maximum wave action (base shear) for several vales

of the parameter. Then a deterministic most probable maximum wave action is

calculated in Section 7.2.2.2. The deterministic value is then matched with the

spectral calculated value in order to find a calibrated parameter. The parameter

will in turn be used to generate stress transfer functions. Here steepness is used

86

7.2 Harmonic inputs used to create transfer functions

to explain the calibration process, but is works in the same manner for the other

schemes.

The total applied wave actions are used for this calibration i.e. quasi-static base

shear. The goal is to calculate the most probable maximum (MPM) value of the

wave actions for different steepness values. This is done by establishment of the

transfer function of the wave actions with a constant wave steepness. These transfer

functions are established by looking at one frequency at the time. The wave height

for these harmonic inputs are determined by the constant steepness. The output

base shear amplitude is then used to calculate the transfer function value for a

given frequency. The response amplitude divided by the input wave amplitude

in order to get the value of the transfer function at the particular frequency and

steepness.

2Famplitude

hζF (ωk ) = (7.13)

Hinput

Where F is the base shear, and Hinput is the input wave height which will depend

on what steepness value that is selected. This results in several transfer functions

for different values of steepness. Three of them are included in Figure 7.6. Note

that these are quasi static transfer functions, thus no resonant areas are present.

However cancellation and amplification is still present, which is the reason for the

peaks that is present.

5

#10

4

Steepnes: 0.005

3.5 Steepnes: 0.162

Steepnes: 0.35

3

h1 F (!) transfer function

2.5

1.5

0.5

0

0 1 2 3 4 5 6 7

!

Figure 7.6: Transfer functions from wave elevation to wave action with different

constant steepness

87

7.2 Harmonic inputs used to create transfer functions

The next step is to use these transfer functions to calculate the value of the most

probable maximum of wave actions depending on steepness. This is carried out by

standard spectral methods. First by calculating the variance of the process, which

is the integral of the response spectrum, in this case the spectrum of applied total

wave action.

Z ∞

m0 = SF F (ω)dω (7.14)

0

The response spectrum can be found using the transfer function according to equa-

tion 7.15. The square of the absolute value multiplied with the input spectrum will

give the spectrum of the output.

The input spectrum will be the particular sea state of interest. The most probable

maximum range in wave action can then be calculated from equation 7.16 for large

values of N [Myrhaug, 2007], where N is the number of waves, which can be taken

as N = 10800

TP . It is important to point out that this is the range in wave action.

p

∆FM P M ≈ 2 2m0 ln N (7.16)

Two resulting curves for most probable maximum wave action range as a function

of steepness is given in Figure 7.7. The reason for these to be different is that

different input spectra are used to calculate the wave action.

10 1.8

9 1.6

MPM wave action

8

1.4

7

1.2

6

1

5

4 0.8

3 0.6

0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4

steepnes steepnes

Figure 7.7: Most probable maximum of total applied wave action (quasi-static

base shear).

These curves are the relevant results of this first step of calibration. It might seem

strange that these curves do flat out for larger values of steepness, since the forces

will be higher with larger values of steepness. However the reason for the flat out

is the cut off height from Table 7.2 and Figure 7.4. At a certain steepness almost

all harmonic components will be limited by the cutoff frequency, and hence an

88

7.2 Harmonic inputs used to create transfer functions

increasing steepness will not result in an increasing height used in the harmonic

input components.

tion

monic input, and stepping one wave through the structure. The period of the input

should be the same as the zero mean up crossing period as the sea state of inter-

est, and the wave height should be the wave height that corresponds to the most

probable maximum wave height for that particular sea state. The most probable

maximum wave height is calculated using equation 7.17 where N is the number of

zero up crossings [Myrhaug, 2007].

r

ln N

HM P M = HS (7.17)

2

The deterministic range of most probable maximum wave action is the maximum

minus the minimum of the resulting wave action. For certain sea states this results

is given in Table 7.3.

HS (m) TP (s) T (s) HM P M (m) ∆FM P M (N)

3.5 8.5 8.5 6.6 4.9e+05

4.5 9.5 9.5 8.4 1.1e+06

5.5 9.5 9.5 10.3 1.6e+06

steepness. It is an very fast way to estimate the fatigue damage. However it turns

out to be extremely sensitive to initial assumptions.

This method is described in [ISO, 2007]. The overall idea is to create one transfer

function that has a good accuracy for the most important sea states when fatigue

is considered. The most important sea state is not known in an initial phase.

However by assuming a straight SN-curve, and a linear relationship between stress

range and wave height, there is possible describe a ratio that is proportional to the

damage [ISO, 2007]. There will then be proportionality between damage and wave

height by the power of m (The slope of the SN curve). There will also be an inverse

proportionality to the mean zero crossing period (TP ) due to the number of cycles

during 3h. This relation does not support kinked SN-curves with several values of

89

7.2 Harmonic inputs used to create transfer functions

m. The SN curve used in this work has a slope of 3 or 5 depending on the stress

range.

N (HS )m

D∝ (7.18)

TP

The total accumulated damage for a certain sea state is then given by the pro-

portionality in equation 7.18 [ISO, 2007], where N is the number of occurrences in

the scatter diagram. The sea state considered the most important, and where the

transfer function is calculated, is for the HS value where there are equal amount

of damage for larger HS values as for smaller HS values. The same thing goes for

TP . This most important region is called the center of fatigue.

0 4 0

7

2 3.5 2

6

3

4 4

proportional damage

proportional damage

5

2.5

6 6

4

HS

HS

2

8 8

3

1.5

10 10

1 2

12 0.5 12 1

14 14

0 5 10 15 20 25 0 5 10 15 20 25

TP TP

Figure 7.8: Proportional accumulated damage for different sea states. The centre

of fatigue contains a red dot. Ekofisk scatter diagram is used

The proportional damage for different sea states is plotted in Figure 7.8. The center

of fatigue turns out to be in the same area as the sea state that contributes with

the most proportional damage for both values of m. The center of fatigue for both

cases is given in Table 7.4 along with the intermediate value m=4, and the result

from time domain simulation. It is hard to tell up front which of the slopes that

will be contributing the most to the damage of a structure. However in this case,

time domain simulations are in hand, and in most sea states all the stress cycles

are bellow the limit of the kink in the SN-curve. This is evident in appendix B,

where stresses bellow 86.2 MPa will have m=5 as slope. This indicates that m=5

might give the best approximation.

m HS TP

3 3.5 8.5

4 4.5 9.5

5 5.5 9.5

time domain 4.5 8.5

90

7.2 Harmonic inputs used to create transfer functions

However as seen in Table 7.4 the cases for m=3 and m=4 are closest to the true

center of fatigue (the one from time domain simulations). m=5 however gives to

large values of both period and wave height and is expected to overestimate the

stresses.

For these cases the cutoff height of 9.26m is used. This results in the steepness

values given in Table 7.5. These vales are used when the stress transfer function is

calculated. The stress transfer function is calculated with the same method as for

the total wave force, but using the calibrated value of steepness. Dynamic effects

are also included in the calculation of the stress transfer function.

m HS TP Calibrated steepness, κ

3 3.5 8.5 0.05

4 4.5 9.5 0.10

5 5.5 9.5 0.29

The steepness value of the case with m=5 is beyond the breaking limit. This indi-

cates that the cutoff height might be to low for this particular sea state. With the

steepness given in Table 7.5, the stress transfer functions are calculated. Evident

in Figure 7.9 is that the larger values of steepness results in larger values in the

stress transfer function.

#10 7

3.5

h1< (! ) Stress transfer function

steepnes: 0.05

3 steepnes: 0.10

steepnes: 0.29

2.5

1.5

0.5

0

0 2 4 6 8

!

Figure 7.9: Dynamic stress transfer functions, for the calibrated steepness

91

7.2 Harmonic inputs used to create transfer functions

When the stress transfer function is known, the stress response spectrum can be

calculated in a similar way as in equation 7.15. This spectrum can then be used

to directly create a realization of the stress process as described in Section 5.7.1.2.

On these time series the same procedure as for time domain time series is used

to calculate damage, i.e. rainflow counting and mine summation with kinked SN-

curve.

The resulting short term damage depicted in Figure 7.10. The damage resulting

from time domain simulations is also included as reference. These are only short

term sea states, and the goal for the linearization technique is not that the esti-

mates shall bee accurate for all sea states. More important is that the regions of

importance, i.e. the regions around the centre of fatigue is about right. The case

when center of fatigue is estimated with m=4 gives the best results in this area.

This is also the case which has the same significant wave height at the center of

fatigue in time domain simulation (Table 7.4).

Figure 7.10: Damage during short term sea states (3h) using transfer functions

at the center of fatigue and constant wave steepness. Be aware of the different

values on the color axis. Keep in mind that the values of m is only used to estimate

the center of fatigue, and are not involved in the damage calculation.

In order to get the long term damage, it is necessary to multiply the damage from

each sea state by the number of occurrences over a period of time. In this case

the scatter diagram for Ekofisk is used for the purpose. It is given for 56 years.

The total damage experienced from one sea state over these 56 years are calculated

92

7.2 Harmonic inputs used to create transfer functions

by the multiplying number occurrences of short term sea states with the damage

achieved by one short term sea state. The total damage is depicted in Figure 7.11.

Figure 7.11: Total damage during 56 years, linearized method. Keep in mind that

the values of m is only used to estimate the center of fatigue, and are not involved

in the damage calculation.

On the top of the plots in Figure 7.11 the total damage from all sea states is

indicated. This is calculated simply by adding the contribution from all sea states.

As mentioned, the case of m=4 and m=3 gives a center of fatigue that is closer

to the true center of fatigue. Thus it is reasonable to believe that these will give

better results than for the case of m=5, which in fact is the case. However m=5 is

the only case which actually over predicts the fatigue damage.

It turns out that the total damage collected from all sea states is about right

in the case when m=4 decides the center of fatigue. It coincides with the total

damage resulting from the time domain simulation. Even though the total damage

is correct in this case, the technique gives a wrong impression on how much each

sea state contributes to the total damage. That being said, the technique gives a

good overall indication on the fatigue damage.

What turns out to be a very important parameter for this technique is the location

of the center of fatigue. The simplicity of the technique is its huge advantage.

However it might be a bit optimistic to let only one transfer function estimate the

93

7.2 Harmonic inputs used to create transfer functions

response for all sea states. It might be possible to extend this to several transfer

functions, and split up the domain in this manner. This is investigated in Section

7.2.6

The calibration is going to be carried out for every sea state in Section 7.2.6. In

this regard it is of interest to investigate the static transfer functions of stress.

However to generate transfer functions from time domain simulations, for compar-

ison, requires huge computational effort. The reason for this is that the frequency

resolution need to be considerably higher in order to generate a transfer function.

Thus this is only carried out for a few cases in this section.

If the response spectrum is known together with the input spectrum, and the

response is stationary, then there exist a relationship between them. Equation 7.19

describes the relation between stress response- and input spectrum [Conte, 2016].

It is possible to establish a response spectrum according to Section 5.7.2. This

demands a time signal of stress response, which can be achieved by time domain

simulation.

The transfer function can then be calculated by equation 7.20. However in order

to achieve a decent response spectrum, the frequency resolution needs to be well

above the 500 components used in previous sections. In this section 5000 wave

components are used which seem to produce a response spectrum of good quality.

Averaging by the 50 adjacent points is utilized, as it worked well in Section 5.7.2

for a signal of 5000 components.

s

Sσσ (ω)

|hζω (ω)| = (7.20)

Sζζ (ω)

Constant steepness

Here the same technique with constant steepness is here inspected, however with

larger cutoff heights. Both of these cutoffs allows quite high sea states to be cal-

ibrated. The results obtained from linearization are compared with the one from

time domain simulations.

In Figure 7.12 a comparison between the two cutoff frequencies and time domain

simulation is present. The most important region for this platform is the region

close to the resonance, which is at ω = 0.82. The region between 0.6 and 1.2 is

thus the most important. As expected the method over predicts at low frequencies,

and under predicts at high frequencies. The under prediction at high frequencies

is not associated with any large responses. For the smaller sea states considerably

94

7.2 Harmonic inputs used to create transfer functions

accurate results is achieved in the important region. For larger sea states the

deviation in low frequency response increases. In all cases is seem to be a certain

degree of conservatism in the important regions. The largest cutoff height gives

better results for frequencies between ω = 0.5 and 1.

The cutoff height of 30m seem to give reasonable results for small sea states. This

indicates that the selection of cutoff height might not as important as long as it is

taken high enough.

10 15

h ( ) stress transfer function

Cutoff=15 Cutoff=15

Cutoff=30 Cutoff=30

8 Time domain Time domain

10

6

4

5

0 0

0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3

2 2.5

h ( ) stress transfer function

Cutoff=15 Cutoff=15

Cutoff=30 Cutoff=30

Time domain 2 Time domain

1.5

1.5

1

1

0.5

0.5

0 0

0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3

with stochastic time domain

The scheme of selecting wave heights outlined in Section 7.2.1.2 of having constant

wave height to period ratio is here investigated in the terms static transfer functions.

This can bee seen in figure 7.13.

Also in this case the overshoot in low frequencies increases for for higher sea states.

The higher frequency response seem to be better approximated than for the con-

95

7.2 Harmonic inputs used to create transfer functions

stant steepness. Thus height period ratio might be worth considering for platforms

with lower natural periods.

10 10

h ( ) stress transfer function

Cutoff=20 Cutoff=20

Time domain Time domain

8 8

6 6

4 4

2 2

0 0

0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3

2 2

h ( ) stress transfer function

Cutoff=20 Cutoff=20

Time domain Time domain

1.5 1.5

1 1

0.5 0.5

0 0

0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3

stress with stochastic time domain

The scheme of selecting wave heights outlined in Section 7.2.1.3 by using the square

root of the spectrum and scale this to achieve a proper wave height is here inspected

in terms of static stress transfer functions.

This scheme give a large peak at the peak period of the spectra. It also seem to

under predict the high frequency response for all cases.

96

7.2 Harmonic inputs used to create transfer functions

12 14

h ( ) stress transfer function

Heights using spectrum Heights using spectrum

10 Time domain 12 Time domain

10

8

8

6

6

4

4

2 2

0 0

0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3

2.5 2.5

h ( ) stress transfer function

Time domain Time domain

2 2

1.5 1.5

1 1

0.5 0.5

0 0

0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3

Figure 7.14: Static stress transfer functions with JONSWAP spectrum used to

generate input wave heights, compared with time domain simulations.

All the schemes above are plotted together in Figure 7.15 for easier comparison.

97

7.2 Harmonic inputs used to create transfer functions

12 14

h ( ) stress transfer function

Heights using spectrum Heights using spectrum

10 const c cutoff=20 12 const c cutoff=20

const steep cutoff=30 const steep cutoff=30

Time domain 10 Time domain

8

8

6

6

4

4

2 2

0 0

0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3

2.5 2.5

h ( ) stress transfer function

const c cutoff=20 const c cutoff=20

2 const steep cutoff=30 2 const steep cutoff=30

Time domain Time domain

1.5 1.5

1 1

0.5 0.5

0 0

0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3

Due to the large degree of dynamic amplification the most important transfer

function is the dynamic transfer function. It is of interest to see how the deviation

in static transfer function influence the dynamic. The accuracy of the response will

be largely influenced by how well the transfer function in the static and dynamic

case preform at the natural frequency.

Constant steepness

The constant steepness approach show promising results for the dynamic transfer

functions. Evident in Figure 7.16 is that it preforms well on low frequencies. At

natural period it overshoots by a significant amount, and this is expected as it

also is present in the static transfer function. The overshoot at resonance will

contribute to add conservatism. At the larger sea the under prediction at high

frequencies increases. This might be important to be aware of if stiffer installations

98

7.2 Harmonic inputs used to create transfer functions

are considered.

12 12

h ( ) stress transfer function

Cutoff=30 Cutoff=30

10 Time domain 10 Time domain

8 8

6 6

4 4

2 2

0 0

0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3

12 12

h ( ) stress transfer function

Cutoff=30 Cutoff=30

10 Time domain 10 Time domain

8 8

6 6

4 4

2 2

0 0

0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3

Figure 7.16: Dynamic stress transfer functions with constant wave steepness,

compared with time domain simulations.

For constant wave height period ratio, the low frequency part is overestimated by a

significant amount compared to constant steepness. This is evident in figure 7.17.

However the high frequency response seem to be better approximated. In the case

of stiff installations with natural periods bellow 4s this scheme might be considered.

99

7.2 Harmonic inputs used to create transfer functions

12 14

h ( ) stress transfer function

Cutoff=20 Cutoff=20

10 Time domain 12 Time domain

10

8

8

6

6

4

4

2 2

0 0

0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3

2 2

h ( ) stress transfer function

Time domain Time domain

1.5 1.5

1 1

0.5 0.5

0 0

0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3

Figure 7.17: Dynamic stress transfer functions with constant height period ratio,

compared with time domain simulations.

The results for this scheme can bee seen in Figure 7.18. When the JONSWAP

spectrum is used to determine the input wave heights the dynamic response seem

to give good results at lower sea states. However at low frequencies at higher sea

states a second peak is present. This is a result of the peak in the spectrum used to

determine the input wave heights. This peak will continue to increase for larger sea

states and will give a large over estimate of response at the spectral peak frequency

of the spectrum. This method will not be investigated any further.

100

7.2 Harmonic inputs used to create transfer functions

14 14

h ( ) stress transfer function

Heights using spectrum Heights using spectrum

12 Time domain 12 Time domain

10 10

8 8

6 6

4 4

2 2

0 0

0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3

15 15

h ( ) stress transfer function

Time domain Time domain

10 10

5 5

0 0

0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3

Figure 7.18: Dynamic stress transfer functions with JONSWAP spectrum used

to generate input wave heights, compared with time domain simulations.

All the schemes above are plotted together in Figure 7.19 for easier comparison.

101

7.2 Harmonic inputs used to create transfer functions

14 14

h ( ) stress transfer function

Heights using spectrum Heights using spectrum

12 const c cutoff=20 12 const c cutoff=20

const steep cutoff=30 const steep cutoff=30

10 Time domain 10 Time domain

8 8

6 6

4 4

2 2

0 0

0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3

2 2

h ( ) stress transfer function

const c cutoff=20 const c cutoff=20

1.5 const steep cutoff=30 1.5 const steep cutoff=30

Time domain Time domain

1 1

0.5 0.5

0 0

0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3

Instead of using one transfer function to represent the response for every sea state,

it is here created one transfer function for each sea state. This means that the

steepness and height-period parameter is calibrated for every sea state. This should

result in significant improvements compared to only using one transfer function.

This will in turn increase the computational demand, but still be considerably

lower than the time domain simulation. In the following two sections the results

are presented with colors. For numerical results in tables see appendix A.

The constant steepness scheme gives very promising results. It do however over

predict the damage at resonance compared to the time domain simulation. For

calibration of c the values are in general to large. However it does give as large

amplifications in the resonant area as the the steepness.

102

7.2 Harmonic inputs used to create transfer functions

H/T )

In this section the long term damage is presented. This is the total damage expe-

rienced during 56 years in the Ekofisk area from each block in the scatter diagram.

As indicated in the previous section the amplitudes of scheme with constant height-

period ratio (c) gives larger damages, this is also the case for the long term damage.

However the total damage is not to far from what is calculated in time domain sim-

ulations. The scheme with constant steepness gives total damage slightly higher

than the time domain simulation. However in this case the damage is much more

concentrated. The degree of conservatism might bee too small in order to rely on

the steepness in the general case.

103

7.3 Linearization by changing the drag coefficient

H/T )

In a general case there can hardly be possible to come up with empirical simple

mathematical relations describing the calibration of parameters such as steepness.

The main aspect leading to this is the differences from design to design. Dimen-

sional properties vary from design to design, and will influence the performance of

a certain steepness value.

Another take on the problem might be in changing the drag coefficient. A method

of linearizing the problem is illustrated in equation 7.21, by simply changing the

drag term to a linear. This involves changing the drag coefficient. The huge

benefit of this method is its local impact. It opens up for more generalized ways

of selecting appropriate values for the linearized problem since the drag coefficient

can be customized for each member.

Here one such method is outlined. This method utilizes the standard deviation of

the wave particle velocity in addition to diameter in order to select drag coefficient.

The standard deviation in particle velocity will rely on the sea state as well as depth.

104

7.3 Linearization by changing the drag coefficient

1 lin 1

FD =CD ρD|u|u → FD = C(D,L) ρDu (7.21)

2 2

In equation 7.21, the C(D,L) is a linear drag coefficient. This coefficient should be

selected such that the total force gives similar characteristics as the nonlinear. One

way to select this coefficient is such that the integral of the deviation between the

linear and nonlinear drag force during one cycle is minimized. However for fatigue

damage the stress range is of importance, and not the deviation during a cycle.

Instead of minimize the deviation for one cycle, it seem more reasonable to select

the linear drag coefficient to give an equal amount of fatigue damage. By assuming

that the fatigue damage can be estimated from the individual peak values of the

stress process it is with certain assumptions possible to establish a drag coefficient

that gives equal damage as the true drag.

By assuming that there exist a linear relationship between force peak and stress

peak, equation 7.22 can be used to describe the damage from one peak.

Di = ν(Fp,i )m (7.22)

Here m is the slope of the SN-curve, ν a constant, and Fp,i force peak number i.

By assuming that the velocities follows a Rayleigh distribution, and requiring that

the expected value of damage from forces with linear drag term es equal to the true

forces as in equation 7.23,

m

E[Fp,L ] = E[Fpm ] (7.23)

C(D,L)

the ratio C = C D σu can be given on a closed form as in equation 7.24 [Wolfram,

1998].

( "m

(m) 2 1 X m! 1 m−p 0.5p

C (K) = Γ(1 + p, ) /2

Γ(1 + 0.5m) p=0 (m − p)!p! 8K 2

(7.24)

2/m

1 m 1

+ γ(1 + 0.5m, )/K − 2

8K 2 K

Where Γ() is the gamma function, and γ(, ) and Γ(, ) is the lower and upper incom-

plete gamma function given by equation 7.25 and 7.26 [Abramowitz and Stegun,

1972].

Z x

γ(a, x) = exp(−t)ta−1 dt (7.25)

0

Z ∞

Γ(a, x) = exp(−t)ta−1 dt (7.26)

x

K is given in equation 7.27

105

7.3 Linearization by changing the drag coefficient

CD σu TP

K= (7.27)

π 2 CM D

The standard deviation in wave particle velocity will obviously vary with sea state.

However it will also vary with depth. Members in the deep part of the structure

will have smaller standard deviation in velocity compared to the ones in the surface

area. The standard deviation of velocity can be calculated according to equation

7.28. Other properties such as angle on current will also contribute to the velocity

standard deviation.

Z ∞ Z ∞

ω2 z

σu2 = e 2kz 2

ω Sζζ (ω)dω = e2 g ω 2 Sζζ (ω)dω (7.28)

0 0

The solution procedure will then be to first calculate K from equation 7.27 and

use this to calculate a drag coefficient according to equation 7.29. Both K and

C (m) (K) has to be calculated for all members due to the standard deviation in

velocity and its dependency on depth.

C (m) (K) is plotted in Figure 7.22. For the members in the surface area K will be

typically above 1 for the most important sea states (region HS =4.5,TP = 8.5).

Figure 7.22: Linearization factors for Morison’s equation for estimating expected

fatigue damage when the SN-slope, m=1, 3, 4 and 5 represented by full, dotted,

dashed and dash-dotted lines, respectively. From: [Wolfram, 1998]

106

7.3 Linearization by changing the drag coefficient

In the above method relative velocity are not accounted for. However it might

give reasonable values of the linear drag coefficient considering that the structural

velocities are small compared to wave particle velocities as seen in Table 7.6.

H T ṙa /ua

7m 13s ≈ 10−3

3m 5s ≈ 10−4

Table 7.6: The ratio between structural response velocity amplitude and water

particle velocity amplitude for 2 cases, with harmonic input, measured in surface

area (drag coefficient of 1.15 is used for the whole structure)

source code has to be made. The program need to be able to change to a linear drag

term, which is not a feature at the present point in time. Hence an implementation

of this method is not acomplished.

107

Chapter 8

Conclusion

The three parameter Weibull distribution fitted by the method of moments gives

adequate statistical descriptions of the short term stress processes. The parameters

can be presented as smooth surfaces over the HS − TP plane.

The damage encountered is influenced by the particular realization of each sea

state. In order to ensure that the estimated damage has statistical reliability several

realizations of the same sea state should be averaged.

Drag forces are found to be a lot more important than inertia forces. This is due

to the small dimensions on the structural members, in addition to the large drag

coefficient for chords induced by the rack.

If the response is represented by one transfer function at the center of fatigue the

results are highly dependent on the initial assumptions that estimates the location

of the center of fatigue. This method might give good results if the center of fatigue

is estimated correctly, however this is not an straight forward process. Different

aspects complicates this process. Kinked SN-curves and nonlinear relationships

between wave height and stress is among them. If the center of fatigue is not

located properly the method will give poor results. To add conservatism it is

possible to move the estimated center of fatigue to a higher significant wave height.

It is however strongly recommended to calibrate the wave heights for every sea

state. This demands more computational efforts, but is still very manageable. It

gives a much better impression of how the structure is behaving at each sea state.

For this particular jack-up this method gives conservative results, and in general

the calibration procedure must be considered to give good results. If the cutoff

height selected is to low, the calibration procedure might not be possible for sea

states with larger significant wave heights.

Both steepness and constant height-period ratio has shown promising results. The

108

constant wave height-period ratio is giving more accurate results for periods of less

than 4s, and should be considered for stiffer drag dominated structures.

Another aspect is the rapid improvement in computational capacity. This makes

time domain simulations more and more affordable. The method used for the

realization of the waves allows for less wave components and hence saving a lot

of simulation time. More computational time must however be expected by using

second order irregular wave kinematics instead of first order.

109

Chapter 9

Recommendations for further work

Other critical spots for stress concentrations than the joint selected might be criti-

cal. There is a high possibility that a critical spot might be on rack or in the gearing

system used to elevate and lock the racks in place. This will require detailed finite

element modeling of the rack and a gear system. The dimensional properties of

gears are such that 3 dimensional solid elements should be used for modeling. This

is not always supported software wise in the offshore industry, which usually sticks

to plates and beams.

The linearization techniques might also be viewed in the context of heading. By

dividing the heading in different sectors, and perhaps calculate the calibrated steep-

ness for each heading.

The spectrum used here is a sharply peaked JONSWAP spectrum. It might be

interesting to look into the case of a double peaked spectrum. This perhaps intro-

duces new difficulties regarding the linearization process.

A equivalent drag coefficient might be another way to approach the problem. This

will however require software to support such linear drag terms. It will open up

for a far more detailed linearization process, and may also by simple mathematical

relations.

An extended analysis might also be carried out to insure sufficient statistical accu-

racy in terms of realizations. This can be carried out by averaging several realiza-

tions of the same sea state.

What is the might most important aspect to get more insight into, is however,

the effect of current. As indicated in the beginning of the thesis, the current may

amplify the amplitudes of the forces significantly.

110

Bibliography

Application of Stochastic Methods, Department of Marine Technology NTNU.

Abramowitz, M. and Stegun, I. A. [1972], Handbook of Mathematical Functions with

Formulas, Graphs, and Mathematical Tables, U.S. Department of Commerce,

NIST.

Almar-Næss, A. [1985], Fatigue Handbook: Offshore Steel Structures, 1 edn, Tapir.

Box, G. E. P., Jenkins, G. M. and Reinsel, G. C. [1994], Time Series Analysis:

Forecasting and Control, Prentice Hall.

ing.

DNV GL [2014a], DNV-RP-C205: Environmental Conditions and Environmental

Loads, july edn, DNV GL.

july edn, DNV GL.

DNV GL [2015], DNVGL-RP-C104: Self-elevating units, july edn, DNV GL.

Faltinsen, O. M. [1993], Sea Loads on Ships and Offshore Structures, 1 edn, Cam-

bridge University Press.

Haver, S. [2016], Stochastic Description of Ocean Waves and Response Analysis

and Prediction of Extremes, NTNU.

Haver, S. [2017], Metocean Modelling and Prediction of Extremes, Draft version 1,

Univeristy in Stavangerm NTNU.

111

BIBLIOGRAPHY

for Time Integration Schemes in Structural Dynamics, Earthquake Engineering

and Structural Dynamics 6 p. 99-118.

Hughes, T. J. R. [2000], The Finite Element Method Linear Static and Dynamic

Finite Element Analysis, 1 edn, Dover.

ISO [2007], ISO 19902 Petroleum and natural gas industries – Fixed steel offshore

structures, International Organization for Standardization.

Langen, I. and Sigbjornsson, R. [1986], Dynamisk Analyse av marine konstruk-

sjoner, Tapir.

Marine Technology.

Marintek [2001], USFOS Getting Started, SINTEF group.

Marintek [2010], USFOS Hydrodynmaics, Theory, Description of use, Verification,

SINTEF group.

Myrhaug, D. [2007], TMR 4180 Marin Dynamikk uregelmessig sjø, NTNU.

Myrhaug, D. [2016], TMR4235: Stochastic Theory of Sea Loads. Statistics of Nar-

row Band Processes and Equivalent Linearization, Department of Marine Tech-

nology NTNU.

Dover Publications.

NORSOK [2007], N-003.

Søreide, T. H., Amdahl, J., Eberg, E., Holmås, T. and Øyvind Hellan [1993],

USFOS - A computer Program for Progressive Collapse Analysis of Steel Offshore

Structures. Theory Manual, SINTEF Division of Structural Engineering.

Tahery, H. M. [2015], Investigation of which sea states yield the dominating con-

tribution to fatigue accumulation in offshore structure, Department of marine

technology, NTNU.

and Loads - A Tutorial, Math. Stat., Center for Math. Sci., Lund Univ.

Wolfram, J. [1998], On alternative approaches to linearization and Morison’s equa-

tion for wave forces, Department of Civil and Offshore Engineering, Heriot- Watt

University, Edinburgh EH9 1.

112

Appendices

113

Appendix A

Linearization and time domain results and

tables

In this appendix the numerical results are presented in tables. These tables are in the same format

as the scatter diagram.

114

A.1

2<TP<3

2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 13.5 14.5 15.5 16.5 17.5 18.5 19.5 20.5

0<HS<1 0.5

1.5 Cutoff

2.5 1.57E-01 1.84E-01 1.55E-01 6.75E-02 5.37E-02 3.99E-02 4.22E-02 3.63E-02 3.60E-02 3.09E-02 2.72E-02 2.41E-02 2.25E-02 2.15E-02 2.13E-02 0

3.5 2.55E-01 2.09E-01 1.07E-01 4.86E-02 6.10E-02 5.92E-02 5.11E-02 4.75E-02 4.01E-02 3.45E-02 3.03E-02 2.82E-02 2.69E-02 2.67E-02 0

4.5 2.75E-01 1.43E-01 6.72E-02 7.98E-02 7.53E-02 6.39E-02 5.83E-02 5.03E-02 4.35E-02 3.82E-02 3.54E-02 3.37E-02 0 3.60E-02

fined as TH2

5.5 1.76E-01 9.39E-02 1.02E-01 9.40E-02 7.92E-02 7.05E-02 6.02E-02 5.29E-02 4.69E-02 4.33E-02 4.12E-02

6.5 1.19E-01 1.09E-01 9.15E-02 8.17E-02 6.98E-02 6.13E-02 5.49E-02 0 4.85E-02

7.5 1.37E-01 1.25E-01 1.04E-01 9.28E-02 7.93E-02 6.97E-02 6.25E-02 0 5.59E-02

8.5 1.46E-01 1.21E-01 1.06E-01 9.10E-02 7.93E-02 7.12E-02 6.64E-02

9.5 1.63E-01 1.37E-01 1.20E-01 1.02E-01 8.94E-02 7.88E-02 7.38E-02

10.5 1.53E-01 1.34E-01 1.13E-01 9.97E-02 8.89E-02 8.24E-02

11.5 1.30E-01 1.12E-01 9.99E-02 0

12.5 1.03E-01

Calibration ratios

2<TP<3

2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 13.5 14.5 15.5 16.5 17.5 18.5 19.5 20.5

0<HS<1 0.5

1.5 Cutoff

2.5 7.78E-01 1.01E+00 9.36E-01 4.71E-01 4.39E-01 4.07E-01 4.75E-01 4.53E-01 4.89E-01 4.54E-01 4.28E-01 3.95E-01 3.79E-01 3.71E-01 3.80E-01

3.5 1.35E+00 1.24E+00 7.22E-01 4.08E-01 6.14E-01 6.61E-01 6.24E-01 6.32E-01 5.85E-01 5.40E-01 4.98E-01 4.81E-01 4.74E-01 4.85E-01

4.5 1.62E+00 9.85E-01 5.82E-01 7.83E-01 8.34E-01 7.86E-01 7.77E-01 7.16E-01 6.69E-01 6.22E-01 6.02E-01 5.94E-01 0 6.64E-01

5.5 1.20E+00 7.76E-01 1.00E+00 1.05E+00 9.69E-01 9.42E-01 8.62E-01 8.04E-01 7.51E-01 7.27E-01 7.20E-01

6.5 1.17E+00 1.21E+00 1.12E+00 1.09E+00 1.00E+00 9.35E-01 8.76E-01 0 8.38E-01

7.5 1.35E+00 1.38E+00 1.28E+00 1.24E+00 1.14E+00 1.07E+00 1.00E+00 0 9.67E-01

8.5 1.76E+00 1.59E+00 1.50E+00 1.36E+00 1.27E+00 1.18E+00 1.14E+00

9.5 1.98E+00 1.78E+00 1.68E+00 1.53E+00 1.40E+00 1.30E+00 1.27E+00

T

ratio, defined as H

10.5 2.00E+00 1.89E+00 1.71E+00 1.57E+00 1.47E+00 1.42E+00

11.5 2.41E+00 2.03E+00 1.78E+00 0

12.5 2.50E+00

115

A.1 Calibration ratios

2<TP<3

2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 13.5 14.5 15.5 16.5 17.5 18.5 19.5 20.5

0<HS<1 0.5

1.5 Cutoff

A.2

2.5 1.67E-10 7.26E-09 7.10E-08 6.02E-08 7.65E-08 2.31E-09 1.34E-09 6.60E-10 3.91E-10 1.97E-10 1.10E-10 6.21E-11 3.74E-11 2.32E-11 1.62E-11 2.20E-07

3.5 1.40E-07 1.07E-06 8.39E-07 5.73E-08 2.15E-08 1.18E-08 5.53E-09 3.20E-09 1.62E-09 8.57E-10 4.88E-10 3.02E-10 1.92E-10 1.41E-10 2.15E-06

4.5 1.03E-05 6.52E-06 3.33E-07 1.34E-07 7.17E-08 3.14E-08 1.74E-08 9.13E-09 5.00E-09 2.96E-09 1.82E-09 1.19E-09 0 8.14E-10 1.74E-05

5.5 3.38E-05 1.76E-06 6.47E-07 3.40E-07 1.50E-07 8.22E-08 4.12E-08 2.39E-08 1.40E-08 8.93E-09 6.00E-09 3.68E-05

6.5 2.42E-06 1.22E-06 5.21E-07 3.03E-07 1.54E-07 9.33E-08 5.89E-08 0 2.53E-08 4.80E-06

ing steepness

7.5 7.65E-06 3.92E-06 1.61E-06 9.41E-07 5.10E-07 3.11E-07 2.08E-07 0 9.41E-08 1.52E-05

8.5 1.22E-05 5.19E-06 2.89E-06 1.60E-06 1.04E-06 6.78E-07 4.67E-07 2.41E-05

9.5 3.19E-05 1.40E-05 8.27E-06 4.48E-06 3.02E-06 1.97E-06 1.36E-06 6.49E-05

10.5 3.46E-05 2.08E-05 1.18E-05 8.15E-06 5.69E-06 3.75E-06 8.48E-05

11.5 3.22E-05 2.19E-05 1.56E-05 0 6.97E-05

12.5 2.66E-05 2.66E-05

0 0 1.67E-10 1.47E-07 1.14E-05 4.12E-05 2.23E-06 1.09E-05 4.96E-05 5.61E-05 3.33E-05 5.08E-05 3.46E-05 2.42E-05 3.22E-05 1.27E-07 1.57E-10 8.14E-10 0 0.000347

2<TP<3

2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 13.5 14.5 15.5 16.5 17.5 18.5 19.5 20.5

0<HS<1 0.5

1.5 Cutoff

2.5 2.22E-09 3.96E-09 1.18E-08 4.54E-08 1.56E-08 5.56E-09 4.30E-09 3.47E-09 2.85E-09 2.56E-09 2.28E-09 2.00E-09 1.81E-09 1.68E-09 1.62E-09 1.07E-07

Short term sea state

3.5 1.36E-08 7.68E-08 3.59E-07 1.02E-07 2.70E-08 1.74E-08 1.10E-08 7.28E-09 5.51E-09 4.28E-09 3.30E-09 2.77E-09 2.42E-09 2.26E-09 6.35E-07

4.5 4.83E-07 2.05E-06 5.59E-07 1.41E-07 8.55E-08 4.58E-08 2.56E-08 1.68E-08 1.10E-08 7.14E-09 5.04E-09 4.03E-09 0.00E+00 3.38E-09 3.44E-06

simulations

5.5 9.19E-06 2.42E-06 6.18E-07 3.72E-07 1.84E-07 9.50E-08 5.72E-08 3.32E-08 1.88E-08 1.17E-08 8.54E-09 1.30E-05

6.5 2.28E-06 1.37E-06 6.72E-07 3.31E-07 1.94E-07 1.06E-07 5.70E-08 0.00E+00 2.21E-08 5.03E-06

7.5 7.35E-06 4.44E-06 2.12E-06 1.04E-06 5.96E-07 3.18E-07 1.65E-07 0.00E+00 6.14E-08 1.61E-05

8.5 1.27E-05 6.00E-06 2.99E-06 1.72E-06 9.03E-07 4.54E-07 2.53E-07 2.50E-05

9.5 3.18E-05 1.54E-05 7.68E-06 4.57E-06 2.36E-06 1.20E-06 6.63E-07 6.37E-05

10.5 3.51E-05 1.80E-05 1.09E-05 5.91E-06 3.00E-06 1.62E-06 7.45E-05

11.5 2.37E-05 1.33E-05 6.96E-06 0.00E+00 4.40E-05

12.5 8.47E-06 8.47E-06

0 0 2.22E-09 1.75E-08 5.72E-07 1.16E-05 3.1E-06 1.04E-05 5.08E-05 5.95E-05 3.02E-05 4.18E-05 2.29E-05 1.19E-05 1.1E-05 1E-07 3.88E-09 3.38E-09 0 0.000254

2<TP<3

2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 13.5 14.5 15.5 16.5 17.5 18.5 19.5 20.5

0<HS<1 0.5

1.5 Cutoff

2.5 1.74E-10 4.73E-09 4.67E-08 5.39E-08 5.86E-09 3.04E-09 2.14E-09 1.07E-09 7.17E-10 3.59E-10 1.93E-10 1.03E-10 5.97E-11 3.64E-11 2.56E-11 1.19E-07

3.5 8.39E-08 6.52E-07 6.82E-07 6.41E-08 3.70E-08 2.46E-08 1.20E-08 7.30E-09 3.60E-09 1.85E-09 9.68E-10 5.78E-10 3.64E-10 2.66E-10 1.57E-06

4.5 6.22E-06 5.67E-06 4.30E-07 2.25E-07 1.55E-07 7.33E-08 4.35E-08 2.21E-08 1.23E-08 6.98E-09 4.23E-09 2.69E-09 0 1.86E-09 1.29E-05

5.5 2.71E-05 2.16E-06 1.35E-06 8.60E-07 4.05E-07 2.30E-07 1.11E-07 6.10E-08 3.42E-08 2.14E-08 1.42E-08 3.24E-05

116

A.2 Short term sea state

6.5 4.91E-06 3.20E-06 1.47E-06 8.61E-07 4.50E-07 2.55E-07 1.49E-07 0 6.13E-08 1.14E-05

7.5 1.66E-05 1.04E-05 4.89E-06 2.86E-06 1.47E-06 8.71E-07 5.50E-07 0 2.41E-07 3.79E-05

8.5 4.54E-05 2.00E-05 1.10E-05 5.78E-06 3.50E-06 2.17E-06 1.38E-06 8.93E-05

9.5 1.13E-04 5.20E-05 2.92E-05 1.58E-05 9.67E-06 6.07E-06 4.03E-06 2.30E-04

height-period ratio

10.5 1.24E-04 7.08E-05 3.99E-05 2.53E-05 1.69E-05 1.13E-05 2.88E-04

11.5 1.46E-04 8.03E-05 4.99E-05 0 2.76E-04

12.5 9.24E-05 9.24E-05

0 0 1.74E-10 8.86E-08 6.92E-06 3.35E-05 2.66E-06 2.31E-05 0.000173 0.000203 0.000115 0.000209 0.00012 7.58E-05 0.000109 3.2E-07 2.91E-10 1.86E-09 0 0.001072

2<TP<3

2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 13.5 14.5 15.5 16.5 17.5 18.5 19.5 20.5

0<HS<1 0.5

A.3

1.5 Cutoff

2.5 1.17E-09 4.65E-06 5.22E-04 8.20E-04 5.47E-04 6.53E-06 2.23E-06 9.49E-07 3.88E-07 8.13E-08 1.70E-08 5.15E-09 6.73E-10 1.62E-10 6.48E-11 0 1.90E-03

3.5 1.40E-07 2.74E-04 2.62E-03 4.62E-04 7.92E-05 1.46E-05 3.02E-06 1.16E-06 3.99E-07 1.07E-07 3.02E-08 8.16E-09 2.11E-09 5.64E-10 0 3.45E-03

4.5 1.03E-05 3.59E-04 7.06E-04 5.29E-04 1.08E-04 1.61E-05 3.23E-06 5.75E-07 2.30E-07 1.42E-07 1.45E-08 5.95E-09 0 8.14E-10 1.73E-03

5.5 3.38E-05 3.03E-04 7.34E-04 4.94E-04 8.74E-05 1.55E-05 1.19E-06 5.75E-07 1.54E-07 7.15E-08 6.00E-09 1.67E-03

6.5 3.44E-04 5.48E-04 2.17E-04 6.69E-05 6.49E-06 1.21E-06 4.12E-07 0 2.53E-08 1.18E-03

ing steepness

7.5 5.36E-05 2.74E-04 2.30E-04 1.05E-04 2.24E-05 5.92E-06 1.25E-06 0 9.41E-08 6.93E-04

8.5 9.76E-05 1.51E-04 1.76E-04 4.97E-05 1.25E-05 5.42E-06 1.40E-06 4.94E-04

9.5 3.19E-05 8.38E-05 1.41E-04 1.12E-04 4.23E-05 2.55E-05 1.36E-06 4.37E-04

10.5 3.46E-05 1.46E-04 1.07E-04 5.71E-05 3.98E-05 3.75E-06 3.87E-04

11.5 9.66E-05 1.10E-04 6.22E-05 0 2.68E-04

12.5 2.66E-05 2.66E-05

0 0 1.17E-09 4.79E-06 0.000807 0.003831 0.002018 0.001746 0.001571 0.000824 0.000655 0.000396 0.00023 0.000135 3.32E-05 1.34E-07 6.28E-10 8.14E-10 0 0.012251

2<TP<3

2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 13.5 14.5 15.5 16.5 17.5 18.5 19.5 20.5

0<HS<1 0.5

1.5 Cutoff

2.5 1.32E-08 1.23E-06 6.34E-05 0.000583 9.19E-05 1.03E-05 4.05E-06 2.42E-06 1.31E-06 5.35E-07 2.36E-07 1.72E-07 4.73E-08 2.28E-08 1.55E-08 0 7.59E-04

3.5 1.00E-08 1.80E-05 0.001118 0.000801 9.21E-05 1.85E-05 4.43E-06 1.67E-06 7.38E-07 2.64E-07 9.76E-08 3.78E-08 1.51E-08 6.55E-09 0 2.06E-03

4.5 4.56E-07 0.000113 0.001187 0.000555 0.000125 2.13E-05 3.98E-06 8.00E-07 3.34E-07 2.02E-07 2.19E-08 1.04E-08 0 1.76E-09 2.01E-03

simulations

5.5 9.16E-06 0.00042 0.000716 0.000544 0.000104 1.70E-05 1.49E-06 6.58E-07 1.60E-07 6.81E-08 5.83E-09 1.81E-03

6.5 0.000332 0.000621 0.000278 7.21E-05 7.79E-06 1.26E-06 3.50E-07 0 1.86E-08 1.31E-03

7.5 5.28E-05 0.000312 0.000302 0.000116 2.57E-05 5.74E-06 9.21E-07 0 5.67E-08 8.15E-04

8.5 0.000101 0.000172 0.000181 5.20E-05 1.04E-05 3.47E-06 7.30E-07 5.22E-04

9.5 3.17E-05 9.15E-05 0.000129 0.000111 3.16E-05 1.49E-05 6.42E-07 4.10E-04

10.5 3.52E-05 0.000126 9.48E-05 3.92E-05 1.99E-05 1.57E-06 3.16E-04

11.5 7.01E-05 6.31E-05 2.62E-05 0 1.59E-04

12.5 8.06E-06 8.06E-06

0 0 1.32E-08 1.24E-06 8.19E-05 0.001823 0.0025 0.001758 0.001757 0.001011 0.000648 0.000365 0.000153 6.63E-05 1.12E-05 1.3E-07 2.21E-08 1.76E-09 0 0.010176

Long term sea state (56 years)

2<TP<3

2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 13.5 14.5 15.5 16.5 17.5 18.5 19.5 20.5

0<HS<1 0.5

1.5 Cutoff

2.5 1.22E-09 3.03E-06 3.44E-04 7.34E-04 4.18E-05 8.59E-06 3.56E-06 1.54E-06 7.12E-07 1.48E-07 2.99E-08 8.57E-09

1.07E-09 2.55E-10 1.02E-10 1.14E-03

3.5 8.39E-08 1.67E-04 2.13E-03 5.17E-04 1.36E-04 3.04E-05 6.57E-06 2.65E-06 8.86E-07 2.31E-07 6.00E-08

1.56E-08 4.00E-09 1.06E-09 2.99E-03

4.5 6.22E-06 3.12E-04 9.12E-04 8.87E-04 2.33E-04 3.75E-05 8.04E-06 1.39E-06 5.67E-07 3.35E-07

3.39E-08 1.34E-08 0 1.86E-09 2.40E-03

5.5 2.71E-05 3.71E-04 1.53E-03 1.25E-03 2.36E-04 4.34E-05 3.23E-06 1.46E-06 3.77E-07

1.71E-07 1.42E-08 3.46E-03

117

A.3 Long term sea state (56 years)

6.5 6.97E-04 1.44E-03 6.15E-04 1.90E-04 1.89E-05 3.31E-06 1.05E-06 0 6.13E-08 2.96E-03

Table A.8:

7.5 1.16E-04 7.30E-04 6.99E-04 3.21E-04 6.48E-05 1.66E-05 3.30E-06 0 2.41E-07 1.95E-03

8.5 3.63E-04 5.80E-04 6.73E-04 1.79E-04 4.21E-05 1.74E-05

4.14E-06 1.86E-03

9.5 1.13E-04 3.12E-04 4.97E-04 3.95E-04 1.35E-04 7.90E-05

4.03E-06 1.54E-03

height-period ratio

10.5 1.24E-04 4.95E-04 3.59E-04 1.77E-04 1.18E-04

1.13E-05 1.29E-03

11.5 4.37E-04 4.02E-04 2.00E-04 0 1.04E-03

12.5 9.24E-05 9.24E-05

0 0 1.22E-09 3.11E-06 0.000517 0.003203 0.001842 0.003374 0.004161 0.00261 0.002231 0.00146 0.000779 0.000419 0.000112 3.34E-07 1.17E-09 1.86E-09 0 0.020711

Constant

Appendix B

Stress range distributions from 93 sea

states

HS=2.5, TP=4.5

4

0

ln(-ln(1-F("<)))

-2

-4

-6

3param Weibull fit

Data

-8

6 8 10 12 14 16

ln("<-6) (Pa)

Figure B.1: Probability distributions of stress range during 3h realizations HS:2.5 TP:4.5

118

HS=2.5, TP=5.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

3param Weibull fit

Data

-4

11 12 13 14 15 16 17

ln("<-6) (Pa)

Figure B.2: Probability distributions of stress range during 3h realizations HS:2.5 TP:5.5

HS=2.5, TP=6.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

12 13 14 15 16 17

ln("<-6) (Pa)

Figure B.3: Probability distributions of stress range during 3h realizations HS:2.5 TP:6.5

119

HS=2.5, TP=7.5

2

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

13.5 14 14.5 15 15.5 16 16.5 17

ln("<-6) (Pa)

Figure B.4: Probability distributions of stress range during 3h realizations HS:2.5 TP:7.5

HS=2.5, TP=8.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

13.5 14 14.5 15 15.5 16 16.5 17

ln("<-6) (Pa)

Figure B.5: Probability distributions of stress range during 3h realizations HS:2.5 TP:8.5

120

HS=2.5, TP=9.5

2

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

3param Weibull fit

Data

-5

13 13.5 14 14.5 15 15.5 16 16.5

ln("<-6) (Pa)

Figure B.6: Probability distributions of stress range during 3h realizations HS:2.5 TP:9.5

HS=2.5, TP=10.5

2

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

3param Weibull fit

Data

-5

13 13.5 14 14.5 15 15.5 16 16.5

ln("<-6) (Pa)

Figure B.7: Probability distributions of stress range during 3h realizations HS:2.5 TP:10.5

121

HS=2.5, TP=11.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

12.5 13 13.5 14 14.5 15 15.5 16 16.5

ln("<-6) (Pa)

Figure B.8: Probability distributions of stress range during 3h realizations HS:2.5 TP:11.5

HS=2.5, TP=12.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

3param Weibull fit

Data

-4

12.5 13 13.5 14 14.5 15 15.5 16

ln("<-6) (Pa)

Figure B.9: Probability distributions of stress range during 3h realizations HS:2.5 TP:12.5

122

HS=2.5, TP=13.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

3param Weibull fit

Data

-4

12 12.5 13 13.5 14 14.5 15 15.5 16

ln("<-6) (Pa)

Figure B.10: Probability distributions of stress range during 3h realizations HS:2.5 TP:13.5

HS=2.5, TP=14.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

3param Weibull fit

Data

-4

11 12 13 14 15 16

ln("<-6) (Pa)

Figure B.11: Probability distributions of stress range during 3h realizations HS:2.5 TP:14.5

123

HS=2.5, TP=15.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

3param Weibull fit

Data

-4

11 12 13 14 15 16

ln("<-6) (Pa)

Figure B.12: Probability distributions of stress range during 3h realizations HS:2.5 TP:15.5

HS=2.5, TP=16.5

2

0

ln(-ln(1-F("<)))

-1

-2

-3

3param Weibull fit

Data

-4

11 12 13 14 15 16

ln("<-6) (Pa)

Figure B.13: Probability distributions of stress range during 3h realizations HS:2.5 TP:16.5

124

HS=2.5, TP=17.5

2

0

ln(-ln(1-F("<)))

-1

-2

-3

3param Weibull fit

Data

-4

10 11 12 13 14 15

ln("<-6) (Pa)

Figure B.14: Probability distributions of stress range during 3h realizations HS:2.5 TP:17.5

HS=2.5, TP=18.5

2

0

ln(-ln(1-F("<)))

-1

-2

-3

3param Weibull fit

Data

-4

10 11 12 13 14 15

ln("<-6) (Pa)

Figure B.15: Probability distributions of stress range during 3h realizations HS:2.5 TP:18.5

125

HS=3.5, TP=5.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

11 12 13 14 15 16 17

ln("<-6) (Pa)

Figure B.16: Probability distributions of stress range during 3h realizations HS:3.5 TP:5.5

HS=3.5, TP=6.5

4

0

ln(-ln(1-F("<)))

-2

-4

-6

3param Weibull fit

Data

-8

10 11 12 13 14 15 16 17 18

ln("<-6) (Pa)

Figure B.17: Probability distributions of stress range during 3h realizations HS:3.5 TP:6.5

126

HS=3.5, TP=7.5

2

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

13.5 14 14.5 15 15.5 16 16.5 17 17.5

ln("<-6) (Pa)

Figure B.18: Probability distributions of stress range during 3h realizations HS:3.5 TP:7.5

HS=3.5, TP=8.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

14 14.5 15 15.5 16 16.5 17 17.5

ln("<-6) (Pa)

Figure B.19: Probability distributions of stress range during 3h realizations HS:3.5 TP:8.5

127

HS=3.5, TP=9.5

2

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

3param Weibull fit

Data

-5

13.5 14 14.5 15 15.5 16 16.5 17

ln("<-6) (Pa)

Figure B.20: Probability distributions of stress range during 3h realizations HS:3.5 TP:9.5

HS=3.5, TP=10.5

2

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

3param Weibull fit

Data

-5

13.5 14 14.5 15 15.5 16 16.5 17

ln("<-6) (Pa)

Figure B.21: Probability distributions of stress range during 3h realizations HS:3.5 TP:10.5

128

HS=3.5, TP=11.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

13 13.5 14 14.5 15 15.5 16 16.5 17

ln("<-6) (Pa)

Figure B.22: Probability distributions of stress range during 3h realizations HS:3.5 TP:11.5

HS=3.5, TP=12.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

13 13.5 14 14.5 15 15.5 16 16.5

ln("<-6) (Pa)

Figure B.23: Probability distributions of stress range during 3h realizations HS:3.5 TP:12.5

129

HS=3.5, TP=13.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

3param Weibull fit

Data

-4

13 13.5 14 14.5 15 15.5 16 16.5

ln("<-6) (Pa)

Figure B.24: Probability distributions of stress range during 3h realizations HS:3.5 TP:13.5

HS=3.5, TP=14.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

12 13 14 15 16 17

ln("<-6) (Pa)

Figure B.25: Probability distributions of stress range during 3h realizations HS:3.5 TP:14.5

130

HS=3.5, TP=15.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

11 12 13 14 15 16

ln("<-6) (Pa)

Figure B.26: Probability distributions of stress range during 3h realizations HS:3.5 TP:15.5

HS=3.5, TP=16.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

3param Weibull fit

Data

-4

11 12 13 14 15 16

ln("<-6) (Pa)

Figure B.27: Probability distributions of stress range during 3h realizations HS:3.5 TP:16.5

131

HS=3.5, TP=17.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

3param Weibull fit

Data

-4

11 12 13 14 15 16

ln("<-6) (Pa)

Figure B.28: Probability distributions of stress range during 3h realizations HS:3.5 TP:17.5

HS=3.5, TP=18.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

11 12 13 14 15 16

ln("<-6) (Pa)

Figure B.29: Probability distributions of stress range during 3h realizations HS:3.5 TP:18.5

132

HS=4.5, TP=6.5

4

0

ln(-ln(1-F("<)))

-2

-4

-6

3param Weibull fit

Data

-8

11 12 13 14 15 16 17 18

ln("<-6) (Pa)

Figure B.30: Probability distributions of stress range during 3h realizations HS:4.5 TP:6.5

HS=4.5, TP=7.5

2

-1

ln(-ln(1-F("<)))

-2

-3

-4

-5

Data

-7

13 14 15 16 17 18

ln("<-6) (Pa)

Figure B.31: Probability distributions of stress range during 3h realizations HS:4.5 TP:7.5

133

HS=4.5, TP=8.5

3

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

14 14.5 15 15.5 16 16.5 17 17.5

ln("<-6) (Pa)

Figure B.32: Probability distributions of stress range during 3h realizations HS:4.5 TP:8.5

HS=4.5, TP=9.5

2

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

13.5 14 14.5 15 15.5 16 16.5 17 17.5

ln("<-6) (Pa)

Figure B.33: Probability distributions of stress range during 3h realizations HS:4.5 TP:9.5

134

HS=4.5, TP=10.5

2

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

3param Weibull fit

Data

-5

13.5 14 14.5 15 15.5 16 16.5 17 17.5

ln("<-6) (Pa)

Figure B.34: Probability distributions of stress range during 3h realizations HS:4.5 TP:10.5

HS=4.5, TP=11.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

13.5 14 14.5 15 15.5 16 16.5 17

ln("<-6) (Pa)

Figure B.35: Probability distributions of stress range during 3h realizations HS:4.5 TP:11.5

135

HS=4.5, TP=12.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

13.5 14 14.5 15 15.5 16 16.5 17

ln("<-6) (Pa)

Figure B.36: Probability distributions of stress range during 3h realizations HS:4.5 TP:12.5

HS=4.5, TP=13.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

13 13.5 14 14.5 15 15.5 16 16.5 17

ln("<-6) (Pa)

Figure B.37: Probability distributions of stress range during 3h realizations HS:4.5 TP:13.5

136

HS=4.5, TP=14.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

12 13 14 15 16 17

ln("<-6) (Pa)

Figure B.38: Probability distributions of stress range during 3h realizations HS:4.5 TP:14.5

HS=4.5, TP=15.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

12 13 14 15 16 17

ln("<-6) (Pa)

Figure B.39: Probability distributions of stress range during 3h realizations HS:4.5 TP:15.5

137

HS=4.5, TP=16.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

12 13 14 15 16 17

ln("<-6) (Pa)

Figure B.40: Probability distributions of stress range during 3h realizations HS:4.5 TP:16.5

HS=4.5, TP=17.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

3param Weibull fit

Data

-4

12 13 14 15 16 17

ln("<-6) (Pa)

Figure B.41: Probability distributions of stress range during 3h realizations HS:4.5 TP:17.5

138

HS=4.5, TP=19.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

3param Weibull fit

Data

-4

11 12 13 14 15 16 17

ln("<-6) (Pa)

Figure B.42: Probability distributions of stress range during 3h realizations HS:4.5 TP:19.5

HS=5.5, TP=7.5

3

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

13 14 15 16 17 18 19

ln("<-6) (Pa)

Figure B.43: Probability distributions of stress range during 3h realizations HS:5.5 TP:7.5

139

HS=5.5, TP=8.5

3

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

14 14.5 15 15.5 16 16.5 17 17.5 18

ln("<-6) (Pa)

Figure B.44: Probability distributions of stress range during 3h realizations HS:5.5 TP:8.5

HS=5.5, TP=9.5

2

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

14 14.5 15 15.5 16 16.5 17 17.5

ln("<-6) (Pa)

Figure B.45: Probability distributions of stress range during 3h realizations HS:5.5 TP:9.5

140

HS=5.5, TP=10.5

2

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

3param Weibull fit

Data

-5

13.5 14 14.5 15 15.5 16 16.5 17 17.5

ln("<-6) (Pa)

Figure B.46: Probability distributions of stress range during 3h realizations HS:5.5 TP:10.5

HS=5.5, TP=11.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

13.5 14 14.5 15 15.5 16 16.5 17 17.5

ln("<-6) (Pa)

Figure B.47: Probability distributions of stress range during 3h realizations HS:5.5 TP:11.5

141

HS=5.5, TP=12.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

13.5 14 14.5 15 15.5 16 16.5 17 17.5

ln("<-6) (Pa)

Figure B.48: Probability distributions of stress range during 3h realizations HS:5.5 TP:12.5

HS=5.5, TP=13.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

13 14 15 16 17 18

ln("<-6) (Pa)

Figure B.49: Probability distributions of stress range during 3h realizations HS:5.5 TP:13.5

142

HS=5.5, TP=14.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

12 13 14 15 16 17

ln("<-6) (Pa)

Figure B.50: Probability distributions of stress range during 3h realizations HS:5.5 TP:14.5

HS=5.5, TP=15.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

12 13 14 15 16 17

ln("<-6) (Pa)

Figure B.51: Probability distributions of stress range during 3h realizations HS:5.5 TP:15.5

143

HS=5.5, TP=16.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

12 13 14 15 16 17

ln("<-6) (Pa)

Figure B.52: Probability distributions of stress range during 3h realizations HS:5.5 TP:16.5

HS=5.5, TP=17.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

12 13 14 15 16 17

ln("<-6) (Pa)

Figure B.53: Probability distributions of stress range during 3h realizations HS:5.5 TP:17.5

144

HS=6.5, TP=9.5

2

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

14 14.5 15 15.5 16 16.5 17 17.5 18

ln("<-6) (Pa)

Figure B.54: Probability distributions of stress range during 3h realizations HS:6.5 TP:9.5

HS=6.5, TP=10.5

2

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

13 14 15 16 17 18

ln("<-6) (Pa)

Figure B.55: Probability distributions of stress range during 3h realizations HS:6.5 TP:10.5

145

HS=6.5, TP=11.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

13 14 15 16 17 18

ln("<-6) (Pa)

Figure B.56: Probability distributions of stress range during 3h realizations HS:6.5 TP:11.5

HS=6.5, TP=12.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

13.5 14 14.5 15 15.5 16 16.5 17 17.5

ln("<-6) (Pa)

Figure B.57: Probability distributions of stress range during 3h realizations HS:6.5 TP:12.5

146

HS=6.5, TP=13.5

3

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

13 14 15 16 17 18

ln("<-6) (Pa)

Figure B.58: Probability distributions of stress range during 3h realizations HS:6.5 TP:13.5

HS=6.5, TP=14.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

12 13 14 15 16 17 18

ln("<-6) (Pa)

Figure B.59: Probability distributions of stress range during 3h realizations HS:6.5 TP:14.5

147

HS=6.5, TP=15.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

12 13 14 15 16 17 18

ln("<-6) (Pa)

Figure B.60: Probability distributions of stress range during 3h realizations HS:6.5 TP:15.5

HS=6.5, TP=17.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

12 13 14 15 16 17

ln("<-6) (Pa)

Figure B.61: Probability distributions of stress range during 3h realizations HS:6.5 TP:17.5

148

HS=7.5, TP=9.5

2

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

14 14.5 15 15.5 16 16.5 17 17.5 18

ln("<-6) (Pa)

Figure B.62: Probability distributions of stress range during 3h realizations HS:7.5 TP:9.5

HS=7.5, TP=10.5

3

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

13 14 15 16 17 18

ln("<-6) (Pa)

Figure B.63: Probability distributions of stress range during 3h realizations HS:7.5 TP:10.5

149

HS=7.5, TP=11.5

3

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

13 14 15 16 17 18

ln("<-6) (Pa)

Figure B.64: Probability distributions of stress range during 3h realizations HS:7.5 TP:11.5

HS=7.5, TP=12.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

14 14.5 15 15.5 16 16.5 17 17.5 18

ln("<-6) (Pa)

Figure B.65: Probability distributions of stress range during 3h realizations HS:7.5 TP:12.5

150

HS=7.5, TP=13.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

13 14 15 16 17 18

ln("<-6) (Pa)

Figure B.66: Probability distributions of stress range during 3h realizations HS:7.5 TP:13.5

HS=7.5, TP=14.5

3

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

12 13 14 15 16 17 18

ln("<-6) (Pa)

Figure B.67: Probability distributions of stress range during 3h realizations HS:7.5 TP:14.5

151

HS=7.5, TP=15.5

3

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

12 13 14 15 16 17 18

ln("<-6) (Pa)

Figure B.68: Probability distributions of stress range during 3h realizations HS:7.5 TP:15.5

HS=7.5, TP=17.5

4

0

ln(-ln(1-F("<)))

-2

-4

-6

3param Weibull fit

Data

-8

11 12 13 14 15 16 17 18

ln("<-6) (Pa)

Figure B.69: Probability distributions of stress range during 3h realizations HS:7.5 TP:17.5

152

HS=8.5, TP=10.5

3

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

13 14 15 16 17 18 19

ln("<-6) (Pa)

Figure B.70: Probability distributions of stress range during 3h realizations HS:8.5 TP:10.5

HS=8.5, TP=11.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

14 15 16 17 18 19

ln("<-6) (Pa)

Figure B.71: Probability distributions of stress range during 3h realizations HS:8.5 TP:11.5

153

HS=8.5, TP=12.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

14 14.5 15 15.5 16 16.5 17 17.5 18

ln("<-6) (Pa)

Figure B.72: Probability distributions of stress range during 3h realizations HS:8.5 TP:12.5

HS=8.5, TP=13.5

3

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

13 14 15 16 17 18

ln("<-6) (Pa)

Figure B.73: Probability distributions of stress range during 3h realizations HS:8.5 TP:13.5

154

HS=8.5, TP=14.5

4

0

ln(-ln(1-F("<)))

-2

-4

-6

3param Weibull fit

Data

-8

12 13 14 15 16 17 18

ln("<-6) (Pa)

Figure B.74: Probability distributions of stress range during 3h realizations HS:8.5 TP:14.5

HS=8.5, TP=15.5

4

0

ln(-ln(1-F("<)))

-2

-4

-6

-8

Data

-12

8 10 12 14 16 18

ln("<-6) (Pa)

Figure B.75: Probability distributions of stress range during 3h realizations HS:8.5 TP:15.5

155

HS=8.5, TP=16.5

3

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

12 13 14 15 16 17 18

ln("<-6) (Pa)

Figure B.76: Probability distributions of stress range during 3h realizations HS:8.5 TP:16.5

HS=9.5, TP=10.5

3

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

13 14 15 16 17 18 19

ln("<-6) (Pa)

Figure B.77: Probability distributions of stress range during 3h realizations HS:9.5 TP:10.5

156

HS=9.5, TP=11.5

3

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

14 15 16 17 18 19

ln("<-6) (Pa)

Figure B.78: Probability distributions of stress range during 3h realizations HS:9.5 TP:11.5

HS=9.5, TP=12.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

14 15 16 17 18 19

ln("<-6) (Pa)

Figure B.79: Probability distributions of stress range during 3h realizations HS:9.5 TP:12.5

157

HS=9.5, TP=13.5

3

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

12 13 14 15 16 17 18 19

ln("<-6) (Pa)

Figure B.80: Probability distributions of stress range during 3h realizations HS:9.5 TP:13.5

HS=9.5, TP=14.5

4

0

ln(-ln(1-F("<)))

-2

-4

-6

-8

3param Weibull fit

Data

-10

10 11 12 13 14 15 16 17 18

ln("<-6) (Pa)

Figure B.81: Probability distributions of stress range during 3h realizations HS:9.5 TP:14.5

158

HS=9.5, TP=15.5

3

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

12 13 14 15 16 17 18

ln("<-6) (Pa)

Figure B.82: Probability distributions of stress range during 3h realizations HS:9.5 TP:15.5

HS=9.5, TP=16.5

4

0

ln(-ln(1-F("<)))

-2

-4

-6

3param Weibull fit

Data

-8

10 11 12 13 14 15 16 17 18

ln("<-6) (Pa)

Figure B.83: Probability distributions of stress range during 3h realizations HS:9.5 TP:16.5

159

HS=10.5, TP=11.5

3

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

14 15 16 17 18 19

ln("<-6) (Pa)

Figure B.84: Probability distributions of stress range during 3h realizations HS:10.5 TP:11.5

HS=10.5, TP=12.5

3

1

ln(-ln(1-F("<)))

-1

-2

-3

Data

-5

14 15 16 17 18 19

ln("<-6) (Pa)

Figure B.85: Probability distributions of stress range during 3h realizations HS:10.5 TP:12.5

160

HS=10.5, TP=13.5

4

0

ln(-ln(1-F("<)))

-2

-4

-6

-8

3param Weibull fit

Data

-10

10 12 14 16 18 20

ln("<-6) (Pa)

Figure B.86: Probability distributions of stress range during 3h realizations HS:10.5 TP:13.5

HS=10.5, TP=14.5

4

0

ln(-ln(1-F("<)))

-2

-4

-6

3param Weibull fit

Data

-8

12 13 14 15 16 17 18 19

ln("<-6) (Pa)

Figure B.87: Probability distributions of stress range during 3h realizations HS:10.5 TP:14.5

161

HS=10.5, TP=15.5

3

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

12 13 14 15 16 17 18

ln("<-6) (Pa)

Figure B.88: Probability distributions of stress range during 3h realizations HS:10.5 TP:15.5

HS=10.5, TP=16.5

3

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

12 13 14 15 16 17 18

ln("<-6) (Pa)

Figure B.89: Probability distributions of stress range during 3h realizations HS:10.5 TP:16.5

162

HS=11.5, TP=13.5

4

0

ln(-ln(1-F("<)))

-2

-4

-6

3param Weibull fit

Data

-8

12 13 14 15 16 17 18 19

ln("<-6) (Pa)

Figure B.90: Probability distributions of stress range during 3h realizations HS:11.5 TP:13.5

HS=11.5, TP=14.5

4

0

ln(-ln(1-F("<)))

-2

-4

-6

3param Weibull fit

Data

-8

11 12 13 14 15 16 17 18 19

ln("<-6) (Pa)

Figure B.91: Probability distributions of stress range during 3h realizations HS:11.5 TP:14.5

163

HS=11.5, TP=15.5

4

0

ln(-ln(1-F("<)))

-2

-4

-6

3param Weibull fit

Data

-8

11 12 13 14 15 16 17 18 19

ln("<-6) (Pa)

Figure B.92: Probability distributions of stress range during 3h realizations HS:11.5 TP:15.5

HS=12.5, TP=16.5

3

0

ln(-ln(1-F("<)))

-1

-2

-3

-4

Data

-6

12 13 14 15 16 17 18 19

ln("<-6) (Pa)

Figure B.93: Probability distributions of stress range during 3h realizations HS:12.5 TP:16.5

164

Appendix C

165

Problem description

Title: Adequate linearization scheme for a jack-up in order to obtain sufficiently accurate fatigue

assessments using a linear stochastic fatigue analyses

Background

The topic to be investigated in this thesis is fatigue assessment of a typical jack-up platform.

Properties of a jack-up structure making a fatigue assessment cost effective are:

• Largest natural periods are typically well inside the energetic frequency band of the wave

spectrum – even for storm seas – which may result in considerable dynamic amplifications of

stress width for a broad range of sea states.

The overall aim of this thesis is to establish some guidelines on how to linearize the hydrodynamic

loads such that a stochastic fatigue analysis will give results of sufficient accuracy.

The topic for the thesis is suggested by MsC Antonio Goncalves, DNV-GL. His background for the

thesis is given below:

“In stochastic fatigue analysis the relation between forces in each members and wave height is

calculated with the assumption that exist a linear relationship between the wave height and the

resulting force. However, typically jacket and jack-up with lattice leg design are drag dominated;

where the drag forces are proportional to the square of the wave particle velocity. For such

structures, the wave height to wave force relation is therefore not linear.

In offshore industry, the linearization with respect to wave height is generally used. It is based on the

selection of a characteristic or linearization wave height for each wave frequency of the scatter

diagram. Constant wave steepness is frequently used for select the linearize wave height. However,

there are no proper guidelines or recommendations for selection the correct wave steepness for a

specific location, and constant wave steepness usually results in over predicted drag at small wave

frequencies and under predicted drag at large wave frequencies.

Due to fatigue results at typically jack-ups are strongly dependent on assumptions made in the wave

linearization; an appropriate linearization wave height method is highly required.

Our proposal is to create an analytical method that estimates a linearization wave height that

globally produces the same fatigue damage as the waves defined in the scatter diagram, for a given

frequency, by using an equivalent jack-up leg model. The equivalent linearization wave height

estimated with the analytical method will be then applied in the stochastic fatigue FE analysis.”

166

Suggested work plan

Below a possible division into sub-tasks is given.

1. Present a brief overview of methods for assessing fatigue damage due to waves for jacket

and jack-ups. Discuss pro and contra for the methods in view of linear versus non-linear

loading and static versus dynamic behaviour.

2. Present in detail the stochastic fatigue analysis in frequency domain. Discuss possible

approaches for linearizing the non-linear hydrodynamic load. A topic that should be

considered is consequence of current for the linearization? Can we neglect current for

fatigue assessment?

3. In order to evaluate the accuracy of fatigue assessment for various linearization methods,

we an approach that we will consider state-of-the-art fatigue damage. For a stationary short

term sea state this would be a proper time domain solution of the structural problem

together with a rain-flow counting of accumulated fatigue damage for the sea state. Discuss

and describe in detail how this can be extended to a full long term fatigue assessment.

4. Do a long term fatigue assessment for the jack-up under consideration using approaches

outlined in 3.. It shall be discussed how many sectors the weather is divided into. A sector

width of 30o or 45o seems reasonable. A proper modelling of wave conditions for the various

sectors maybe found in literature or it may have to be established by the candidate using

hindcast data (NORA10) for a proper location. (One must consider the work load of including

all sectors. If found convenient one can limit the assessment to the worst sector. If time

permits one can include more sectors.)

5. Do a long term stochastic fatigue analysis using linearized response amplitude operators for

the stress process. Compare with estimated damage in 4.. Test eventually several

approaches for the linearization. Recommend a best approach for the platform under

consideration. Discuss how general you thing recommendation is.

The candidate may of course select another scheme as the preferred approach for his work.

The work may show to be more extensive than anticipated. Some topics may therefore be

left out after discussion with the supervisor without any negative influence on the grading.

The candidate should in his report give a personal contribution to the solution of the problem

formulated in this text. All assumptions and conclusions must be supported by mathematical

models and/or references to physical effects in a logical manner. The candidate should apply

all available sources to find relevant literature and information on the actual problem.

The report should be well organised and give a clear presentation of the work and all

conclusions. It is important that the text is well written and that tables and figures are used

to support the verbal presentation. The report should be complete, but still as short as

possible. The final report must contain this text, an acknowledgement, summary, main body,

conclusions, suggestions for further work, symbol list, references and appendices. All figures,

tables and equations must be identified by numbers. References should be given by author

167

and year in the text, and presented alphabetically in the reference list. The report must be

submitted in two copies unless otherwise has been agreed with the supervisor. From the

report it should be possible to identify the work carried out by the candidate and what has

been found in the available literature. It is important to give references to the original source

for theories and experimental results.

The supervisor may require that the candidate should give a written plan that describes the

progress of the work after having received this text. The plan may contain a table of content

for the report and also assumed use of computer resources. As an indication such a plan

should be available by end of March.

The report must be signed by the candidate, include this text, appear as a paperback, and - if

needed - have a separate enclosure (binder, diskette or CD-ROM) with additional material.

Co-supervisors Jørgen Amdahl, NTNU

Universität Stuttgart

Baustatik und Baudynamik

A HEAVY-LIFT VESSEL

submitted by

Prakash Mohanasundaram

in

July 2009

Prof. Dr.-Ing. habil. Manfred Bischoff

STRUCTURAL ANALYSIS OF A

HEAVY-LIFT VESSEL

requirements for the degree of

Master of Science in

Computational Mechanics of Materials and Structures (COMMAS)

Breda, The Netherlands

In co-operation with

University of Stuttgart

I

Supervisors

Prof. Dr.-Ing. habil. Manfred Bischoff

Institute of Structural Mechanics

University of Stuttgart

Pfaffenwaldring 7, 70550 Stuttgart, Germany

Engineering Manager,

Dockwise Shipping B.V,

Lage Mosten, Breda, The Netherlands

II

Acknowledgement

Firstly, I would like to thank Prof. Dr.-Ing. habil. Manfred Bischoff for giving me the

opportunity to do my Master Thesis under his institute.

carry out my master thesis with Dockwise Shipping. I express my deep sense of

gratitude to my guide Mr. Jacco van de Vreugde, Senior Structural Engineer, Dockwise

Shipping for his valuable encouragement and guidance during my master thesis.

I would also like to thank my university supervisor Dipl.-Ing. Johannes Irslinger for his

guidance and support in completing this work.

Finally, I would like to thank my family members and my colleagues who have been a

constant source of support and encouragement in all respects during my entire course

work.

-Prakash.E.M

III

ABSTRACT

Dockwise is market leader in transport of extremely large and heavy cargoes. The world's largest

heavy transport carrier Blue Marlin is able to carry ultra large and heavy floating production and

drilling platforms up to 73,000tons. The Blue Marlin enables oil companies to build fully integrated

units anywhere in the world and transport them to the final offshore destination, in order to limit

hook up and commissioning at the offshore location.

Even though Blue Marlin is capable of transporting ultra heavy loads, the load-out of SPAR buoy

weighing 20,000tons by skidding over the stern of the vessel poses difficulties to structure at the

stern section mainly due to the engine room which is located on the stern and is quite a large

section with out any bulkheads to carry the load.

The purpose of the thesis is to investigate the possibility of transporting a Future SPAR which

weighs 35855tons. The structure of the vessel is investigated using FEA software FEMAP with

Nx Nastran solver to study the feasibility of transporting “Future SPAR”. An approximate model of

the future SPAR is created in FEMAP environment and the model is used to load the vessel

structure.

Several steps are linearly analyzed, based on the parameters such as position of SPAR on

vessel deck, number of skid beams and method used for loading-out. By studying the results of

the analysis the best possible method to load-out the SPAR has been recommended.

IV

Declaration

independently and no sources have been used other than the stated references.

.................................................. ..................................................

Place/Date Signature

V

Baustatik und Baudynamik

und Baudynamik

Prof. Dr.-Ing. habil. M. Bischoff

Master Thesis

Dockwise is market leader in transport of extremely large and heavy cargoes. Dockwise ser-

ves its clients with a fleet of 22 semi-submersible vessels of different types and designs. The

world's largest heavy transport carrier Blue Marlin is able to carry ultra large and heavy float-

ing production and drilling platforms up to 73,000tons. The Blue Marlin enables oil compa-

nies to build fully integrated units anywhere in the world and transport them to the final off-

shore destination, in order to limit hook up and commissioning at the offshore location.

Even though Blue Marlin is capable of transporting ultra heavy loads, the load-out of SPAR

buoy weighing 20,000tons by skidding over the stern of the vessel poses difficulties to struc-

ture at the stern section which causes bending and bucking of the deck. The structure of the

vessel has to be investigated to find out constraints affecting the integration of the structure

during such load-outs and explore ways to improve and optimize it. And also to increase the

load carrying capacity of the vessel’s structure during stern skidded load-out of SPAR buoy’s

from 20,000tons to 40,000tons.

• Review of existing structure of the vessel and study the process of the stern skidded

load-out to find the parameters involved.

• Analysis of the vessel model using Finite Element Method based on the parameters

collected using the commercial software package FEMAP with integrated solver

NX Nastran.

• Investigation of the results obtained from the analysis to identify the constraints affecting

the structure during loading process.

• Explore ways to improve and optimize the structure and depicting recommendations.

Student: Prakash Mohanasundaram

VI

CONTENTS

Abstract IV

Declaration V

Problem Statement VI

1 Introduction

1.1 Motivation 1

1.2 Objective and overview of Thesis 1

1.3 Methodology 2

2 SPAR

2.1 SPAR types and description 3

2.2 SPAR construction 5

3 BLUE MARLIN

3.1 Definition of a Heavy-lift Vessel 6

3.2 Semi-submersible heavy-lift vessel 6

3.3 Heavy-lift carrier MV Blue Marlin 7

4 Load-out Procedure

4.1 Vessel capabilities & limitations 9

4.2 SPAR particulars 11

4.2.1 Weight distribution of SPAR 12

4.3 Load-out 12

4.3.1 Types of load-out methods 13

4.3.2 Types of skidding methods 15

4.4 Stowage position 16

4.4.1 Possible stowage positions 16

4.4.2 Selection of Stowage position for analysis 17

4.4.3 No. of skid beams 18

4.4.4 Location of skid beams 18

4.4.5 Supporting the SPAR for load-out 19

4.5 Selections of steps to be analyzed 19

5.1 Features of FEMAP 22

5.2 Elements description 24

5.2.1 Line Elements 24

5.2.2 Plane Elements 26

5.2.3 Volume Elements 27

5.2.4 Other Elements 28

6.1 Blue Marlin Model. 30

6.1.1 Coordinate system and units 30

6.1.2 FE Model details & properties 31

6.1.2.1 Keel 31

VII

6.1.2.2 Bottom structure 32

6.1.2.3 Bulkheads 34

6.1.2.4 Stiffeners 35

6.1.2.5 Deck 35

6.1.2.6 Bow & superstructure 36

6.1.2.7 Stern 37

6.1.2.8 Floatation 38

6.1.2.9 Draft and Water Pressure 39

6.1.2.10 Water Ballast. 40

6.1.2.11 Skid beams 42

6.1.3 Discrepancies in the model 43

6.2.1 Coordinate system and units 45

6.2.2 Material Specification 45

6.2.3 Model details & properties 46

6.2.3.1 Hard tank 46

6.2.3.2 Soft tank 47

6.2.3.3 Heave plates 47

6.2.3.4 Truss 48

6.2.3.5 Supports 50

6.2.3.6 Outfittings 53

7.1 Allowable stress levels 55

7.2 Uniform loading by Hydraulics 56

7.2.1 Method of load application 56

7.2.2 Constraints in uniform loading 57

7.2.3 Draft 58

7.2.4 Ballast condition 59

7.2.5 Load calculation for the uniform loading cases. 61

7.2.6 Analysis of uniform loading cases 65

7.2.7 Location of stress components 68

7.2.8 Stress components of analysis 69

7.3 SPAR Model Loading 71

7.3.1 Hard Tank on the Vessel by standard skidding 71

7.3.1.1 Connecting SPAR with vessel 71

7.3.1.2 Load case specific constraints 72

7.3.1.3 Draft and Ballast Condition 73

7.3.1.4 Analysis of the step hard tank on vessel 73

7.3.1.5 Load on skid beams 78

7.3.1.6 Location of stress components 83

7.3.1.7 Stress components of analysis 83

7.3.2 Entire SPAR on Vessel 85

7.3.2.1 Position of the SPAR 86

7.3.2.2 Connecting SPAR with vessel 87

7.3.2.3 Load case specific constraints 87

7.3.2.4 Draft and ballast Condition 87

7.3.2.5 Analysis of the step entire SPAR on vessel 90

7.3.2.6 Location of stress components 95

7.3.2.7 Stress components of analysis 95

7.3.3 Final stowage position of SPAR on vessel 97

VIII

8 Conclusion 101

9 APPENDIX 103

10 Bibliography 159

IX

CHAPTER 1

INTRODUCTION OF THE THESIS

1.1 Motivation

In the summer of 2001 Dockwise carried

out the first single piece of Truss SPAR

transport across the Atlantic from

Technip fabrication yard in Pori, Finland

to Gulf of Mexico using its heavy lift

carrier vessel Might Servant 1. The

weight of the SPAR transported was

21390 Tons.

existence when Dockwise shipping B.V

head office in Breda, The Netherlands

was contacted by Technip Offshore Finland to transport a SPAR from Pori in Finland to

US Gulf with an expected departure date of mid April 2012. This time Technip came up

with a proposal requesting Dockwise to evaluate the feasibility of transporting a SPAR

weighing 35000Tons. The spar, if realized, will be the largest ever. The SPAR is termed

by Dockwise as “Future SPAR”.

Engineering Team at Dockwise revised the technical details of the SPAR received from

Technip and came to a decision that the SPAR can not be transported by the vessel

Mighty Servant 1 as its deadweight carrying capacity is 40,000 Tons. Transporting a

SPAR of 35,000 Tons with all the other accessories, sea fastening supports and grillages

would push the structure of Mighty Servant 1 to its limits. So the SPAR can only be

transported by its Heavy lift carrier vessel MV Blue Marlin which has a capacity of

carrying ultra heavy load of up to 73,000 tons.

Even though Blue Marlin is capable of transporting ultra heavy loads, engineers at

Dockwise are expecting some heavy loading on the structure at the stern section of the

ship. Load-out of heavy cargoes over the stern of the vessel as it would be done with

Future SPAR might induce heavy stresses on the structure, mainly because of two

reasons:

1

1. The engine room which is located on the stern of the vessel and is quite a large

section with out any bulkheads to carry the load.

2. The stern section is an overhang above the propeller which allows Blue Marlin to

dock with keyside along the stern to load-out the cargoes. If translated, which also

means that the water pressure which acts as a support under the vessel could be

less in the stern section compared to the mid ship or even no water contact

depending on the draft of the vessel. This exposes the structure of the vessel in stern

section for heavy loading condition.

The vessel structure is investigated in the frame work of Finite Element Analysis using

package FEMAP with Nx Nastran.

1.3 Methodology

The load-out conditions are re-created as a Finite Element model in FEMAP with Nx

Nastran software package and the structure of the vessel is investigated to study the

feasibility of transporting “Future SPAR” which weighs almost 40,000 tons and to find out

the weak spots affecting the vessel’s structure during load-out of the “Future SPAR” and

explore ways to improve/optimize it.

1. Based on the criteria and parameters of the SPAR and the vessel, the possible

stowage positions of the SPAR on the vessel deck are investigated.

2. The most critical stowage position, in terms of structural loads is identified and

possible load-out methods of the SPAR are studied to find out the most feasible load-

out conditions.

3. The conditions are then further analyzed using Finite Element Method.

2

CHAPTER 2

SPAR

drilling and production. It is a cylindrical hollow structure with

inner structural members to form tanks which creates buoyancy

to keep the weight of the topside and self weight of the SPAR

floating in the high seas. It is designed in such a way to adapt

to the ocean currents and provide damping for the whole

structure with the topside to be stable in the sea during all

weather conditions. As of now there are more than 15 SPAR’s

installed on the world’s oceans, most of the SPAR’s are built by

Technip, a world leader in the offshore structures. SPAR will be

moored to the sea bed for preventing the motion due to the

ocean currents.

Fig 2.1 SPAR with production platform [3]

There are three types of production spars built so far, they are classic, truss and cell spars. They

differ in design due to the technological advances in the field. The classic SPAR is rather a

cylindrical structure from the top to bottom. A truss SPAR has a cylindrical tank at the top and a

truss in the middle section and a square tank at the bottom. The cell SPAR is a new generation of

the SPAR’s it is much cheaper and easier to build.

1. Hard tank

2. Middle section

3. Soft tank

The name hard tank means that the stiffness of the section is high as it contains internal

structural members to make the tank into compartments to be able to withstand the hydrodynamic

pressure imposed by the waves. The hard tank provides the sufficient buoyancy to support the

weight of the topside and also the self weight of the SPAR. There are usually five to six levels of

tanks inside separated by watertight decks. Each level is further divided in to four compartments.

Usually only the bottom level of the compartments is used for ballasting of the SPAR, other

compartments are empty spaces to provide the floatation to the SPAR structure.

3

4

Figure 2.2 Progressions of Spar (Technip Offshore) [4]

The midsection section below the hard tank is designed to give the deep draft necessary for the

SPAR to prevent the swaying of the SPAR in ocean currents. The midsection of the a classic

SPAR is just an extension of the outer cylindrical shell of the hard tank. There are no internal

structures. The truss SPAR’s as the name indicated contains a truss in the midsection, the truss

SPAR reduced the weight of the SPAR and is also less expensive than a classic SPAR and has

less drag and also reduces loads on the mooring lines.

The soft tank is designed for providing floatation for the SPAR during installation process where

it is floatation horizontally. It also has compartments for the fixed ballast to keep the SPAR

upended.

SPAR’s are always built horizontally on the land and transported in the same state to its final

destination where it is upended while floating. The important factor for fabricating the SPAR

horizontally is the usage of the standard fabrication yard equipment and shipyard techniques.

Over 70% of the spars built till date has been constructed at the Mantiluoto, Finland yard owned

by Technip Offshore.

The SPAR structure are divided in to subassemblies and fabricated in a shop equipped with

automatic plate cutting and welding equipment. The subassemblies are joined to form half round

ring sections of the spar. After construction, the spar is loaded out onto a heavy lift vessel for

transportation using hydraulics and skid beams to move the hull.

5

CHAPTER 3

BLUE MARLIN

3.1 Definition of a Heavy-lift Vessel

A Heavy Lift Vessel is designed to carry extremely large and/or heavy items that cannot be

transported in or on other vessels. Semi-submersible types are designed to transport very large

floating objects, including damaged ships. Ro-Ro heavy lift vessels also exist.

Heavy lift vessels have extremely large cargo decks, unusual deck and pilothouse configurations

to accommodate unusual types of cargo and very high-capacity on-deck cranes

Size or Length varies by vessel and type of cargo. There are different types of heavy-lift vessels

built specific for a purpose. A large heavy lift vessel can measure 584.6ft (178.20m) in length and

137.8ft (42m) in width.

Heavy-lift carriers can transport high-value, complex cargoes such as: transformers, locomotives,

rockets, crushers, semi-submersible rigs, jack-ups, oil platforms, bridge spans, TLPs (tension leg

platforms) and SPAR buoys, etc.

Cargo is loaded by onboard cranes or shore-based cranes. Conventional loading is the over-the-

top Lo-Lo (Load-on/Load-off). Some heavy lift vessels also have Ro-Ro capabilities

[10] A semi-submersible is a watercraft that can put much of its bulk underwater. With a relatively

small area above the water's surface, the semisubmersible is less affected by the waves than a

normal ship, but must be trimmed continuously. Unlike a submarine, such a ship is never entirely

underwater.

A semisubmersible heavy-lift ship, also known as a "flo/flo" (for float-on/float-off), has a long and

low well deck between a forward pilot house and an after machinery space. In superficial

appearance, it is somewhat similar to a dry bulk carrier or some forms of oil tanker. Its ballast

tanks can be flooded to lower the well deck below the water's surface, allowing oil platforms,

other vessels, or other floating cargo to be moved into position for loading. The tanks are then

pumped out, and the well deck rises to shoulder the load. To balance the cargo, the various tanks

can be pumped unevenly.

6

The flo/flo industry's largest customer base is the oil industry. Dockwise have transported many

oil drilling rigs (the flo/flo ships can carry the rigs from their construction site to a drilling site at

roughly three to four times the speed of a self-deploying rig). Rapid deployment of the rig to the

drilling site can translate into major savings to the oil industry. They also transport other outsized

cargo and yachts. [10]

The heavy transport vessel Blue Marlin is an open-deck vessel and able to carry fully integrated

ultra heavy and large floating production and drilling platforms up to 73,000 tons. The Blue Marlin

enables oil companies to build fully integrated units anywhere in the world and transport them to

the final offshore destination, in order to limit hook up and commissioning at the offshore location.

The vessel can be semi-submerged to load floating cargos like rigs even other ships and

submarines.

The Blue Marlin has an open cargo deck of 63 meters at portside and 164.20 meters at

starboard. By moving the portside buoyancy casing the cargo stowage possibilities can be

increased

7

Principal characteristics of Blue Marlin

length b.p. 206.50 m

breadth 63.00 m

depth 13.30 m

draft sailing 10.28 m

max draft submerged 28.40 m

gross tonnage 51,821

deadweight 76,410 tons

deck space 63 x 178.2 m

deck load 27.5 t/sq.m

Service Speed 12.0 Knots

Maximum Speed 13.3 Knots

Range 65 days

Table 3.1 Principal characteristics of Blue Marlin

Propulsion: 12,640 kW MAN-B&W 8S50MC-C, driving one 4 bladed c.p. propeller, One 2,000

KW bow thruster. Three diesel generators of 4,581kVA / 6,600V each, driving two retractable

Azimuth thrusters of 4,500 kW each.

Auxiliary engines: One diesel generator 450 V/60 Hz, 990 kW aft. One emergency generator of

450 V/60 Hz, 200 kW.

Anchors and chains: Anchors fwd (chain 107 mm) 2 pcs of 14.4 tons, Anchor aft (chain 95 mm)

1 pc of 11 tons

8

CHAPTER 4

The SPAR is a very heavy offshore structure used in deep sea oil production. It is fabricated on

land and is dry transported on a heavy-lift vessel such as Blue Marlin to its destination, usually

other part of the globe for its utilization.

The process of transferring the SPAR from land on to a heavy-lift vessel for its final voyage is

called Load-out. This section investigates and explains the procedure involved in transporting the

SPAR from its position in the fabrication yard to the high seas where it is used for oil production.

The Vessel Blue Marlin was built as a submersible heavy lift carrier; it was acquired by Dockwise

and modified. The original vessel is extended on its sides to increase its dead weight capacity.

Figures 4.1 shows web frame of the original vessel and Fig 4.2 shows web frame of the modified

vessel.

9

What makes Blue Marlin capable of carrying ultra heavy loads? The answer lies in its structural

design. The vessel is divided in to a number of individual tanks allowing it to take on ballast water

for adjusting its draft and stability. To separate the tanks from one another and to withstand the

high pressure exerted by the fluid it carries, it has been strengthened with bulkheads both

longitudinally and transversely. These are the structures which help carry heavy loads on Blue

Marlin today.

On the deck of the Blue Marlin, there are two casing at the stern end of the deck. Casings are like

a tower extending from the deck. The casing on the Star Board side is called Star Board casing

(SB casing) and is for engines breathing during a submerging operation and it also has some

tanks for taking in ballast water. The Star Board casing is fixed to the deck and cannot be

removed. The casing on the port side is called port side casing and are just with tanks for ballast

water for compensating the weight of the star board casing and provide stability for the vessel.

The port side casing is not fixed and can be moved to another location on deck when needed.

10

4.2 SPAR particulars

As discussed in chapter 2, the Future SPAR is a truss SPAR. It is fabricated in a yard near by a

port so that it can be transported after its fabrication. Having discussed enough about the SPAR

design and working principle in chapter 2, we will discuss the particulars of the Future SPAR

which has been analyzed in this thesis for transportation on the vessel Blue Marlin.

The SPAR includes a hard tank with a square well connected through a truss with its soft tank

and floatation tank. The spar general arrangement has been shown in fig 4.4 and a detailed

drawing with dimensions is given in Appendix 1.

Weight 30174 Ton

Longitudinal center of gravity (from bottom soft tank) 119.14 Meer

Transverse center of gravity (from centerline) 0.00 Meter

Vertical center of gravity (from side hard tank) 22.25 Meter

Length 81.38 Meter

Diameter 44.5 Meter

Inner well breadth/ height 18.29 Meter

strake height 15 %

Outer diameter 57.85 Meter

Length 7.32 Meter

Breadth/ Height 44.50 Meter

Inner well breadth/ height 18.29 Meter

Centre well breadth dimensions 21.34*24.38Meter

5.49*44.50*13.11

Flotation tank dimensions (L*B*H)

Meter

Total height (including cradles) 51.78 Meter

Cradle height (distance from deck to spar hull) 0.60 Meter

Free floating draft aft 10.6 Meter

Free floating draft mid 10.4 Meter

Free floating draft forward 10.1 Meter

11

SPAR weight including supports

Weight 34855 Ton

Longitudinal center of gravity (from bottom soft tank) 119.14 Meter

Transverse center of gravity (from centerline) 0.00 Meter

Vertical center of gravity (from side hard tank) 20.35 Meter

Free floating draft aft 12.8 Meter

Free floating draft mid 11.9 Meter

Free floating draft forward 11.0 Meter

The weight and CG of the spar has been derived by the designers, and a detailed breakdown has

been given in the table below. It is assumed that the weight of the supports includes the supports

of the hard tank and heave plates.

Hard tank 14304 149.34

Heave plate at 140 ft 529 42.69

Heave plate at 256 ft 529 78.04

Truss 3292 63.39

Soft tank 2315 5.26

Outfitting 9205 127.53

Spar weight 30174 119.14

Supports 4681 119.14

Spar weight including supports 34855 119.14

Table 4.1 Weight of SPAR components

General procedure of SPAR fabrication is discussed in chapter 2, the same fabrication technique

will be used in Future SPAR fabrication. The SPAR is built in a number of sections and is

assembled under a crane on rails. The same rails are later used in the load out process of the

SPAR on to the vessel.

The weight distribution of the whole SPAR with its supports and outfittings based on its support

location are provided by the designers, and is listed below;

Percentage

Support Location Weight (mt)

distribution %

Hard Tank 68.22% 23776.8

Upper Heave Plate 11.00% 3834.1

Lower Heave Plate 20.78% 7244.2

Table 4.2 Weight distribution of SPAR

4.3 Load-Out

Load-out as the name implies, means loading an object from its origin to a vehicle for

transportation to its destination. The term load-out is often used in heavy-lift industry during its

handling - heavy structure moved from the place where it is fabricated to a marine vessel (a ship

or barge) for transportation to its final destination.

12

There are quite a lot of ways used to load-out cargoes on a normal scale. For example, forklifts,

cranes, trolleys, conveyors etc. But when it comes to heavy-lift there are very few methods

employed to do the task.

1. Lifting

2. Ro-Ro using Modular trailers

3. Floating

4. Skidding method

4.3.1.1 Lifting

Method of loading-out heavy objects using gantry or cranes is called lifting. Gantry are

equipments usually used to upend the object in the same position, it can move the lifted objects

horizontally. So use of gantry for loading-out SPAR is not possible

[11] Cranes are distinguished from jacks and simple hoists in that they not only have the

capability of lifting a load but also can move a load horizontally and set it down again. The

operation is usually performed with a hook and line from above the object being moved. Cranes

can be classified into several broad categories such as mobile cranes, tower cranes, derricks,

bridge/gantry cranes, container cranes, barge cranes, etc. Tower cranes and derricks are referred

to as fixed cranes.

Tower crane and mobile crane design is continually improving, which greatly increases their lifting

capacity and have made them an invaluable tool in heavy rigging work operations. [11]

Lifting of the SPAR weighing 35000tons is not possible as there are no cranes available

to lift more than 5000Tons.

Roll-on/roll-off (RORO or ro-ro) ships are vessels designed to carry wheeled cargo such as

automobiles, trucks, semi-trailer trucks, trailers or railroad cars that are driven on and off the ship

on their own wheels.

13

In the heavy-lift scenario, the wheeled vehicles are replaced by Modular trailers which are

capable of carrying ultra heavy loads. The modern modular trailers are called SPMT. The self-

propelled modular transporter, or SPMT, is a flat bed high capacity transporter with 4 or 6 axle

lines that can each accommodate a load of 30 tons. Each axle is mounted such that sharp turns,

side way movements or even ‘carousels’ on the spot can be executed. Integrated computer

control calculates each intended move to the right setting of any axle line. Hydraulic suspensions

allow for very slow movements and the positioning of a load within millimeters of tolerance. The

modular design allows combining individual SPMT’s into large configurations that may consist of

several clusters. Fig 4.5 shows Roll-off operation of a heavy structure on SPMT

Even though modular trailers can be configured to carry any amount of loads, they have a

limitation. The cargo has to be jacked up to a height more than the height of the trailer to load the

cargo on its top. In case of the SPAR, it is such a huge structure, it is too costly affair either to

fabricate the SPAR at such heights or to jack up the SPAR to the required height after fabrication,

so other methods of Load-outs have to be adopted.

4.3.1.3 Floating.

Since SPAR is a floating structure, it can be fabricated in a dry dock as like fabricating a seagoing

vessel. Once fabricated the dry dock is flooded to float the SPAR to the ocean. The SPAR is then

towed on to the semi-submerged vessel like Blue Marlin to load the SPAR on its deck and

transport to its destination.

But fabricating a SPAR in a workshop is much easier and cost effective than building a dry dock.

So the Floating of the SPAR for Load-out is ruled out.

An alternative method for horizontal transport is skidding. The load is fitted with skid shoes that

move in skid tracks with a fairly low resistance. Power is provided by hydraulics but may also

involve the use of strand jack systems to pull the load along its way. Simple and non-invasive,

skidding has proven to be a very economical method of transporting heavy loads over short

distances.

14

To load-out the SPAR from its fabrication yard to the vessel deck, skidding method is feasible and

cost effective way to move such heavy loads. The SPAR is normally fabricated in a location near

to a jetty or port for ease of loading–out to the heavy-lift vessel. The separately fabricated

sections of SPAR are assembled in position on a skid rails placed under it. Rollers are placed in

between the SPAR and skid beams to reduce the friction and to facilitate easy movement of load

on the beams. The SPAR is pushed by means of hydraulics power on the Skid beams until it

reaches its final stowage position on the deck of the vessel.

Skidding is the most feasible method to load-out the SPAR from its fabrication yard to the vessel

Blue Marlin.

The standard skidding method utilizes only 2 skidding beams to support the heavy load, as

usually the center of gravity of the load will be in the center and the load can be equally

distributed between the skid beams. The beams also serve as a spreader to distribute the load on

the vessel structure.

The skidding can further be divided into 2 categories based on the loading condition

1. Standard skidding

The standard skidding is the usual way of doing skidding with out any additional equipment, the

load is placed on skid beam, rollers are placed in between the heavy object (load) and the beams

for reducing friction and hydraulic power is applied to push the load. The loads on the beams are

in relation to the properties of the cargo moved. The loads are not same all along the skid tracks.

The distribution of the SPAR mass on the skidding beam can also be controlled and distributed

evenly all along the skid beams by means of hydraulic smart skidding system. Hydraulic

cylinders are placed all along the line in between the cargo supports and skid rollers as shown in

fig 4.11. The hydraulic cylinders connected to a computer controls and maintains the load exerted

on the skid beam below it, to be even all along the beam. As the methods implies, it involves

more equipments and time consuming to do the same skidding operation and is comparatively

costly.

15

4.4 Stowage Position

The position of the cargo on the vessel is referred to as stowage position. Depending on the

SPAR parameters the possible stowage position of the Future SPAR on the deck of the vessel is

investigated.

The Hard tank of the SPAR is fabricated with strakes around it, which provide stability to SPAR

from ocean currents and provides damping from wave loads. Since the most of the strakes are

fabricated with the hard tank in the yard and transported together, a part of the strakes under the

hard tank where transportation supports are located will be installed after the transportation.

Removing other sections of the strakes for transportation purposes is also possible and

investigated in this section.

The preferable stowage position for all kind of heavy lift transportation of such huge scale will be

that the centre line (where the COG acts) of the cargo and vessel should be on the same plane.

But in our case SPAR’s COG cannot be positioned with vessel’s COG as Blue Marlin has a fixed

SB casing. The casing cannot be removed to accommodate the SPAR as air intake and exhaust

of the engine runs through this casing.

As the SPAR has to be positioned away from the center, now the possible stowage positions of

the SPAR on vessel deck mainly depends on the whether the strakes will be removed or not.

1. The SPAR 1.7 m off vessel centre towards the port side

2. The SPAR 8.3 m off vessel centre towards the port side

In both cases the hard tank and heave plates are ~0.3 m clear from the starboard casing. In order

to load the spar, based on the given spar width, the vessel PS casing needs to be removed and

will be placed in front of the SB casing. For both spar stowage positions the ballast capacity of the

Blue Marlin is sufficient for stern load out.

16

The SPAR centre is offset from the centre line of the vessel by 1.7m on the port side as shown in

the figure 4.8. The strakes on the star board side of the vessel have to be removed to prevent the

intervention of the Star Board casing.

The SPAR centre is offset from the centre line of the vessel by 8.3m on the port side as shown in

the figure 4.9. This stowage position does not require any removal of strakes as the SPAR is

offset enough on the vessel to accommodate it with the strakes attached. But most of the SPAR

load is on the port side causing a lot of shear and bending stress on the vessel structure.

The ballast capacity of Blue Marlin is investigated and found that it is sufficient for the stern load

out of the SPAR.

As per the weight distribution of the SPAR from chapter 4.2.1, it is clear that the weight of the

hard tank with its supports and outfittings is the largest load acting of the vessel structure. And

the step, hard tank just passed vessel stern is the most critical step due to the weakness present

in the stern section of the vessel, i.e, engine room. In this step the moment of the spar weight that

has been skidded onto the vessel is the largest.

The stowage position 8.3m offset to the vessel center is the most critical position as the moment

of the SPAR weight on the vessel when the SPAR is stowed at 8.3m offset is higher than the

1.7m offset.

So we can say in any case, the loads exerted by the stowage position where the SPAR is 8.3m

offset is the critical loading case. It is also desirable to transport the SPAR with as much strakes

as possible to reduce the amount of work needed to fix the strakes when the SPAR is floating on

the ocean.

The load-out of SPAR with 1.7m offset will be taken into consideration for analysis, if the results

of the 8.3m offset position of the SPAR are not favorable.

As the SPAR stowage position of 8.3m offset from the vessel center is considered as the critical

and desired load-out position, the same position is chosen for our analysis.

17

4.4.3 Number of skid beams

As the “Future SPAR” is extremely large in terms of size and mass compared to its predecessors,

skidding with just 2 skid beams could lead to over loading the vessels structure. So the number of

skid beams might have to be increased to distribute the loads to other parts of the vessel.

Increasing the number of skid beams could also be a solution for making the load-out possible.

Number of skid beams used for load-out is also an influencing factor. But, can 3 beams do the job

or 4? Has to be found out in the analysis

The skid beams bears and transfers all the weight of the SPAR to the vessel structure; so it has

to be positioned on top bulkheads which are vertical partitions in a ship arranged transversely and

longitudinally. The bulkheads are of greater structural importance, it holds all the sections of the

vessel together.

Since the load-out is carried out over the stern of the vessel, the skid beams have to be placed

longitudinally on the vessel deck from running from stern end towards the bow. Skid beams must

be located on top of the longitudinal bulkheads as much as possible to avoid loading of the vessel

on other section.

The bulkheads for supporting the load on vessel are chosen according to the stowage position of

the SPAR. The location of the skid beam on bulkheads based on the stowage position of 8.3m

offset of the SPAR is shown in the fig. 4.10

SB

casing

1 2 3 4

The location of the skid beams number 2 and 3 are very obvious, above the longitudinal

bulkheads under the SPAR. Skid beam number 2 is located on the second longitudinal bulkhead

from the center on the port side of the vessel, which is also a side wall structure of the vessel

before its modified. Skid beam number 3 is located on the first longitudinal bulkheads from the

center on the starboard side of the vessel.

The location of skid beam number 1 is on the side wall structure of the vessel, which is also a

strong member of the structure similar to a bulkhead except that it has no stiffeners on the

outside wall of the vessel.

18

The specified location of 20.5750m from the center of SPAR is chosen based on the length of

skid beams, at this location the Skid beam is well within the supports of the stiffeners on the side

wall and also the length of the skid beam at this location is the maximum, which is 163.2m.

As the next bulkhead on the star board side is on the line of the star board casing, it is not

possible to support the SPAR on that location, so a different location is chosen based on

strongest members on the vessel. The location shows in the fig 4.10 for skid beam number 4 is

chosen as the material used in that section is thicker, so it can carry more loads.

The same system of numbering of skid beams have been adopted throughout this thesis for

identifying the skid beams.

As the locations of the skid beams have been investigated in the previous section, it is very clear

that the supports under the SPAR should be conforming to the dimensions in relation to the

SPAR centre and skid beams.

The length of the SPAR is more than the length of the vessel deck, we can only support the hard

tank and the truss section on the vessel deck, soft tank will be hanging during the transportation.

Through past experiences in similar transportations projects, it is decided to leave the SPAR

hanging as the stiffness of the truss will be able to hold the soft tank mass within the allowable

deflection. So the idea of supporting the soft tank on vessel deck is not possible and is omitted.

But during previously executed projects, the soft tank will be temporarily supported during load-

out until the SPAR reaches its final stowage position. The soft tank will be supported on the skid

beams on the same plane of the hard tank and truss supports, but the supports under the soft

tank are not fixed or temporarily fixed. When the SPAR reaches its final stowage position on the

vessel deck, the supports under the soft tank will be released and the draft of the vessel will be

decreased so that the soft tank lifts-off the supports and will be hanging over the vessel stern.

The temporary supports for the soft tank is to prevent the deformation of the SPAR due to the

deflection of the over hang during load-out and also to avoid the irregularities in the loading

condition on the supports under the hard tank and truss due to the deflection, as the SPAR is a

very stiff structure.

The same method will be followed in all the analysis done in this thesis, SPAR’s soft tank will be

supported during the load-out process.

As discussed in the above sections, the factors affecting the load-out of the Future SPAR can be

summarized as:

2. Number of skid beams

3. Stowage position

By investigating the stowage position in the chapter 4.4, it is decided that 8.3m offset of the SPAR

on the vessel deck is considered to be the critical position.

Now the factors influencing can be narrowed to number of skid beams and types of skidding

methods.

19

With the space on the vessel deck, we can rule out the use of 5 skid beams for load-out. So the

2, 3 and 4 skid beams are the possible option concerning the number of skid beams for load-outs.

As discussed in chapter 4.4.2, the step that the complete hard tank just passed vessel stern has

been found to be the most critical step regarding vessel longitudinal strength which can been

seen in attachment 3.

The critical load step together with the influencing factors; number of skid beams and types of

skidding method, the following combination of 9 steps have been chosen for analysis.

1. Hard tank on the vessel with 4 skid beams by uniform loading using hydraulics

2. Hard tank on the vessel with 3 skid beams by uniform loading using hydraulics

3. Hard tank on the vessel with 2 skid beams by uniform loading using hydraulics

4. Hard tank on the vessel with 4 skid beams by SPAR model loading

5. Hard tank on the vessel with 3 skid beams by SPAR model loading

6. Hard tank on the vessel with 2 skid beams by SPAR model loading

7. Entire SPAR on vessel with 4 skid beams by SPAR model loading

8. Entire SPAR on vessel with 3 skid beams by SPAR model loading

9. Entire SPAR on vessel with 2 skid beams by SPAR model loading

The 2 skid beam load-out is also analyzed to make sure whether the load out is possible in such

a case and to compare the results.

The above 9 steps are recreated in the FEMAP environment and analyzed to find the optimal

method to load-out the “Future SPAR”

20

CHAPTER 5

[13] Femap is an advanced engineering finite element analysis environment. It is both CAD and

solver neutral, and has become the world’s most popular engineering analysis environment for

Nastran users. It is widely used by the world’s leading engineering organizations and consultants

to model complex products, systems and processes, including satellites, aircraft, defense,

automotive vehicles, electronics, heavy construction equipment, lift cranes, marine vessels and

process equipment.

Femap is CAD independent and both leverages the Siemens Parasolid® software modeling

kernel that allows direct access to Parasolid data for surface and solid modeling in addition to

providing advanced geometric tools necessary to access non-Parasolid geometry.

From advanced beam modeling, mid-surface extraction and hex meshing, to robust CAD import

and idealization, Femap gives unparalleled model control and flexibility with a broad range of

loads, materials, analysis types and visualization options.

In many cases this may be all that you need. It includes linear statics, normal modes, buckling,

model checkout, spot weld, steady-state and transient heat transfer, basic nonlinear, design

sensitivity and unlimited problem size capabilities.

The high level of analysis functionality exposed by Femap and supported by the user interface

provides great value to full-time analysts and occasional-use multi-disciplinary engineers alike

and cannot be matched by other solution offerings in the industry. This combined with the level of

advanced analysis solutions offered by NX Nastran makes the Femap with NX Nastran

combination able to solve more complex engineering problems in a straightforward manner.

Nastran is traditionally a very capable solver for dynamic response analysis, and additional

dynamics solutions that are supported by Femap with NX Nastran, but not by competitive

offerings include random response, response spectrum, material and geometric nonlinearity with

time-dependent loading and rigid and deformable body contact.

Other types of advanced analysis options offered by NX Nastran, include implicit integration

nonlinearity, advanced thermal analysis solutions and 3D computational fluid dynamics.

22

2. Post-processing versatility

Post solution, Femap offers more results processing tools to enable quick and efficient

understanding of the behavior of the system under analysis. Additional tools including time

animations, streamlines, cutting planes, free-body diagrams, grid point force balance, bar and

beam visualization and shear and bending moment diagrams, and user defined reports are

provided by Femap. Furthermore, Femap offers versatile results processing functionality that

allow the result combinations to be made after the analysis has completed, and data mapping to

create input loading for results quantities for subsequent analysis.

Femap includes an impressive array of versatile modeling and visualization tools to aid FE model

creation and verification prior to the analysis. Additional capabilities include the following:

• Loads and boundary conditions can be applied either directly to the geometry model, or

to the FE entities created after meshing.

• Time and temperature dependent loads can be applied for transient dynamic and heat

transfer analyses.

• Application of loads based on equations, functions.

• Weldment modeling.

• Visualization using model transparency.

From simple linear static analysis right through to robust advanced solutions based computational

fluid dynamics, engineers and analysts use Femap to virtually simulate a complete range of

product behavior. The advanced analysis capabilities available with Femap with NX Nastran

employ the Nastran and TMG solvers and provide proven industrial strength solutions of known

and trusted quality. These advanced solutions enjoy an excellent reputation, to solve the most

demanding real world engineering problems by large engineering companies for many years.

23

5. Scalability of solutions

Femap offers scalable solutions and can be used to simulate everything from simple solid

components to entire spacecraft assemblies, within the same user interface. Femap with NX

Nastran can be tailored to specific customer needs, and can expand to meet future demands

without the need for additional product procurement or retraining.

6. Easy to deploy

With a nativeWindows environment, Femap offers comprehensive analysis functionality that is

easy to use and quick to learn. This brings about a reduced training overhead and allows

engineers to maintain maximum productivity even when Femap is used on an occasional basis.

The depth and true scalability of functionality offered by Femap provides exceptional value.

Femap with NX Nastran offers a complete range of powerful analysis options at competitive

prices.

Femap can be used with a large variety of solvers, allowing access to many types of advanced

analysis solutions. There is no limitation with access to alternative solvers for advanced or

alternative analysis solutions.

9. CAD independent

Femap offers neutral CAD support that allows the analyst or engineer access to any kind of CAD

data.

Femap leverages the Parasolid® modeling kernel that allows direct access to Parasolid data for

surface and solid modeling, and provides robust advanced geometric tools necessary to access

non-Parasolid geometry. In addition, Femap is fully associative with Solid Edge® software.

Femap offers a comprehensive set of customization tools including a fully featured application

programming interface and a facility to record, edit and play user-defined macros. The new

Program File capability is hosted in its own Femap window. User-defined macros can be

recorded, edited, debugged and played back all within the Femap interface. Macros to automate

repetitive modeling tasks for example, can be created in an easy and straightforward manner,

and once created, can be added to any Femap toolbar, providing powerful automation tools that

are easy to use and deploy. In addition to the macro-driven Program Files, a full-featured BASIC

development environment in a separate window is available. Directly from Femap’s user

interface, you can now access the OLE/COM object-oriented FEMAP Application Programming

Interface (API) that provides direct access to all Femap objects and functionality. The BASIC

engine is fully OLE/COM compliant and can interface with Femap as well as any OLE/COM

compliant program such as Word or Excel. You can create custom programs that transfer model

information to Word or Excel to create customized reports. [13]

[14] All line elements structurally connect two nodes. The different types represent

different structural conditions.

1. Rod Element

Description: Uniaxial element with tension, compression and torsional stiffness. It does not have

any bending or shear capability.

24

Fig 5.4 Rod Element

Element Coordinate System: The element X axis goes from the first node to the second.

Properties: Area (of cross-section), Torsional Constant, Coefficient for Torsional Stress,

Nonstructural Mass/Length.

2. Beam Element

Description: Uniaxial element with tension, compression, torsion, and bending capabilities. This

element can be tapered. You can specify different properties at each end of the beam.

Shape: Line, connecting 2 or 3 nodes, A orientaion node can be specified to orient the element Y

axis.

Element Coordinate System: The element X axis goes from the first node to the second. The

element Y axis is perpendicular to the element X axis. It points from the first node toward the

orientation (or third) node. If you use an orientation vector, the Y axis points from the first node in

the direction of the orientation vector. The element Z axis is determined from the cross product of

the element X and Y axes.

25

Properties: Area, Moments of Inertia (I1, I2, I12), Torsional Constant, Shear Areas (Y, Z),

Nonstructural Mass/Length, Warping Constant, Stress Recovery Locations, Neutral Axis Offsets

(Nay, Naz, Nby and Nbz). All required input properties for this element can be automatically

calculated for standard or arbitrary shapes by using the FEMAP beam cross section generator

(accessed under Model, Property, Shape).The Shear Areas calculated by the beam property

section generator and the input to FEMAP are the effective areas for shearing, not a shear factor.

If you are inputting values directly, and has a shear factor, simply multiple it by the actual area to

obtain the shear area. If the beam is tapered, you can specify different properties at each end of

the element.

3. Spring Element

Description: A combined stiffness (spring) and damper element. It can be either axial or torsional.

The DOF spring is an alternative formulation.

Application: Used to represent any purely axial, or purely torsional, structural member.

Element Coordinate System: The element X axis goes from the first node toward the second.

If formulation is 1.. CBUSH then Stiffness and Damping values can be defined for individual

degrees of freedom, Structural Damping, Spring/Damper Location, Orientation Csys,

Stress/Strain recovery coefficients. For Frequency or nonlinear analysis function dependence can

be define for stiffness and damping values.

The plane elements are used to represent membrane, shell, and plate structures. They all follow

the same shape and numbering conventions. The simplest formulations of these elements are

just a three-noded triangle and a four-noded quadrilateral. In addition, six-noded "parabolic"

triangles and eight-noded "parabolic" quadrilaterals are also available.

In most cases, loads on plane elements will be applied to face 1. In this case positive pressure

acts in the same direction as the face normal (as determined by the right-hand rule). Conversely,

if loads are applied to face 2, their positive direction will be opposite to the face normal. Therefore

a positive pressure on face 2 is equivalent to a negative pressure on face 1. If you need to apply

edge loads, they can be applied to faces 3 through 6 as shown. Their positive direction is inward,

toward the element center.

Whenever possible, you should try to use elements which closely resemble equilateral triangles

or squares. These shapes will usually result in the best analysis accuracy. Consult your analysis

program documentation for specific shape limitations of that program.

Plate/Shell Element

Description: A combined planar shell element. This element typically resists membrane (in-plane),

shear, and bending forces. Some analysis programs also include transverse (through the

thickness of the element) capabilities.

quadrilateral. Some shapes are not available for all analysis programs.

26

Element Coordinate System: Refer to the figure 4.6. The material angle can be used to rotate the

element X axis.

Stiffness parameter (Nastran only), Transverse shear/Membrane thickness (Nastran only),

Bending, Shear and Membrane-Bending Coupling Materials (Nastran only), Fiber distances for

stress recovery.

Additional Notes: Many analysis programs do not support tapered plate elements. For those that

do, specify a different thickness for each corner of the plate. You can always specify a single

thickness for all corners simply by entering the average thickness. Plate Offsets (Nastran Only)

can be defined to offset the plate a particular distance from its nodes. Only one offset may be

specified, and it will be in the plate's positive or negative normal direction.

These elements are all used to model three-dimensional solid structures. They can provide very

detailed results, but usually require additional modeling and analysis time and effort.

Solid Element

tetrahedron, fifteen-noded wedge, and twenty-noded brick. Some shapes are not available for all

analysis programs.

27

Fig 5.7 Solid Element

Element Coordinate System: Can be aligned based on the node locations or aligned to a

coordinate system. Check your analysis program documentation for supported options.

Additional Notes: If you want to apply pressure loads to solid elements, you must specify a face

number. The previous and following figures show the face numbers (F1 through F6, in the circles)

for each element shape. Positive pressure is always directed inward, toward the center of the

element.

1. Mass Element

Description: A generalized three-dimensional mass and/or inertia element located at a node. The

center of mass can be offset from the node. An even more general form is the mass matrix

element.

Application: Representing parts of a structure which contain mass, but which do not add any

stiffness.

Element Coordinate System: Aligned with a coordinate system that you specify. Some analysis

programs require that you define masses relative to global rectangular coordinates or the nodal

degrees of freedom.

Properties: Mass (or MassX, MassY, and MassZ for some programs), Inertias (Ixx, Iyy, Izz, Ixy,

Iyz, Izx), Offsets.

2. Rigid Element

Description: Represents a rigid connection between a master node and one or more other nodes.

FEMAP has no limit on the number of additional nodes, or the degrees of freedom which may be

connected on these additional nodes. Weighting factors for these connections may also be

defined. Some analysis programs require that the rigid element connects all six degrees of

freedom. Other programs let you limit the connection to selected degrees of freedom. In addition,

support for the rigid element weighting factors in analysis programs is limited.

28

Application: Modeling connections which are very stiff relative to the remainder of the structure.

Shape: One master node, connected to one, to nineteen, additional nodes. If element formulation

for Nastran is set to 1..RSPLINE then the element will have at least two independent nodes and

at lest 1 dependent node.

29

CHAPTER 6

In this chapter, the FE model of the vessel Blue Marlin and SPAR are discussed. It elaborates on

steps involved in creation of the model and the necessity of doing so.

A standard FE model of Blue Marlin is available with Dockwise engineering department. So the

need for creating the FE model of the vessel is exempted. This section explains the construction

details of the Blue Marlin model.

FEMAP is a full three-dimensional modeling program. All coordinates are always specified with

three coordinates, relative to one of the global or user-defined coordinate systems. The FEMAP

work plane is only used for graphical selections and to orient geometry created by certain

geometry creation commands.

Rectangular coordinate system is used in the FE model of the Blue Marlin and throughout this

thesis. In Rectangular system the coordinate is located by the relative distance from the origin in

X, Y and Z axis as show in the fig. 6.1

The origin (0,0,0) of the Blue Marlin model is located on intersection of the aft perpendicular and

the bottom line of the vessel model. The length of the vessel runs through the X axis, stern on

the origin and the aft facing the positive direction of the X-axis. The port side of the vessel is

located in the positive of the Y-axis and star board side on the negative of the Y-axis.

30

The deck of the vessel is located in the positive z-axis of the coordinate system and the keel on

the negative z-axis. Fig 6.2 shows the FE model of the Blue Marlin and its coordinate system

axes.

The units used in the FE model of the Blue Marlin and the entire analysis are Metric (metres).

The vessel structure is a complex combination of steel of different cross sections and properties.

It has a total of 306 properties with a combination of 4 element types and 13 materials. A

summary of the whole model can be seen in the table below

Total No. of Elements 119486

Total No. of Properties 306

Total No. of Materials 13

Plate, Beam, Spring,

Elements types used

Mass

Table 6.1 Summary of FE model of the vessel

As the vessel model was already available; details about the most important components of the

vessels structure are presented. As the greatest longitudinal bending stresses will occur over

the midship region, the plate thickness of most of the members of the structure in this section

will be thicker compared to other sections. So, Let us have a detailed look in to the design

details of important aspects of the vessel structure at midship section and its corresponding FE

model.

6.1.2.1 Keel

[9] At the centre line of the bottom structure is located the keel, which is often said to form

the backbone of the ship. This contributes substantially to the longitudinal strength and

effectively distributes local loading caused when docking the ship. The commonest form of

keel is that known as the ‘flat plate’ keel, and this is fitted in the majority of ocean-going and

other vessels. A form of keel found on smaller vessels is the bar keel. [9]

31

Blue Marlin has a flat plate keel which is fabricated using a plate thickness of 0.21m & 0.20m.

The flat plate keel of Blue Marlin in the FE model is shown in the Fig.6.3. The FE mesh of keel in

the model is created using plate/shell elements.

Shell / Plate Thickness 7850.kg/m

Keel, flat plate Plate 3

(QUAD4) 0.021m/0.020m

Table 6.2 Properties of FE model of keel

Blue Marlin has a double hull bottom structure. An inner bottom (or tank top) provided at a

minimum height above the bottom shell, and maintained watertight to the bilges. This provides a

considerable margin of safety, since in the event of bottom shell damage only the double bottom

space may be flooded. The space is not wasted but utilized to carry oil fuel and fresh water

required for the ship, as well as providing ballast capacity. Water ballast bottom tanks are

commonly provided right forward and aft for trimming purposes.

[9] Double bottoms may be framed longitudinally or transversely, but where the ship’s length

exceeds 120m it is considered desirable to adopt longitudinal framing. The explanation of this is

that on longer ships tests and experience have shown that there is a tendency for the inner

bottom and bottom shell to buckle if welded transverse framing is adopted. [9]

As Blue Marlin is longer than 200m, it has a longitudinally framed double hull bottom structure.

The extended section of the vessel is a single hull bottom structure. The bottom structure is a

combination of number of different cross sections of girders and frames as shows in the fig 6.4.

The FE model of the vessel bottom structure at midship section consists of the following cross

sections and properties

32

Fig 6.4 FE model of Double hull bottom structure of Blue Marlin

Longitudinal centre Shell / Plate 3

Plate Thickness 0.022m 12575.5kg/m

girder (QUAD4)

Shell / Plate 3

Longitudinal side girder Plate Thickness 0.018m 12575.5kg/m

(QUAD4)

Transverse centre Shell / Plate 3

Plate Thickness 0.017m 12575.5kg/m

girder (QUAD4)

Shell / Plate 3

Transverse side girder Plate Thickness 0.015m 12575.5kg/m

(QUAD4)

33

Transverse frames of

Shell / Plate 3

port & starboard side Plate Thickness 0.022m 12575.5kg/m

(QUAD4)

extension

Inner bottom Offset bulb 3

Beam H*T (0.32*0.012)m 7850kg/m

longitudinals plate

H*T (0.35*0.012)m 3

bottom longitudinals Beam Tee Bar 7850kg/m

W*T(0.15*0.014)m

Table 6.3 Properties of FE model of bottom structure

6.1.2.3 Bulkheads

‘bulkheads’. Those bulkheads which are of greatest importance are the main hull transverse

and longitudinal bulkheads dividing the ship into a number of watertight compartments.

These are the muscles of the vessel which help resist the transverse and longitudinal stress

induced in the vessel structure. Much of this structure, particularly the pillars and to some

extent the transverse strength bulkheads, is responsible for carrying the vertical loading

experienced by the ship. When a huge cargo is loaded on the vessel deck, it is important that

the support of the cargo must be placed above the bulkhead sections.

The bulkheads are fabricated using different properties at different sections of the vessel; the

following table shows properties of transverse and longitudinal bulkheads at midship section.

Shell / Plate 3

Transverse bulkhead Plate Thickness 0.015m 12575.5kg/m

(QUAD4)

Shell / Plate 3

Longitudinal bulkhead Plate Thickness 0.022m 12575.5kg/m

(QUAD4)

Table 6.4 Properties of FE model of bulkhead

34

6.1.2.4 Stiffeners

All the longitudinal and transverse surfaces of the vessel structure are fitted with stiffeners

longitudinally at least on one of its surfaces to provide strength to the structure. It prevents

bucking of plates in loading. The side shell of the vessel also longitudinally framed with

stiffeners. These stiffeners are the nerves of the vessel which helps in distributing the load.

Often profile bulb sections are used for stiffeners. But angle bar and flat bar are used for

stiffening side shell of the vessel and deck. All the stiffeners in FE model of the vessel are

created with beam elements of appropriate cross section.

The following table shows properties of some of the stiffeners used in Blue Marlin FE model

Description Element type Cross section Dimension Material Density

Stiffeners on bottom 3

Beam profile bulb H*T (0.26*0.011)m 7850kg/m

deck

Stiffeners on vertical 3

Beam profile bulb H*T (0.28*0.011)m 7850kg/m

bulkheads

Stiffeners along the H*T (0.35*0.012)m 3

Beam Angle Bar 7850kg/m

side shell W*T(0.10*0.017)m

Stiffeners under top 3

Beam Flat Bar H*T (0.43*0.026)m 7850kg/m

deck

Table 6.4 Properties of FE model of stiffeners

Fig 6.6 FE model of typical Blue Marlin web frame with stiffeners

6.1.2.5 Deck

Similar to the double hull bottom structure of Blue Marlin, the deck is also a double deck to give

the vessel the agility it needed to carry ultra heavy cargos. The deck is strongest structure in the

vessel. It is a combination of longitudinal and transverse plate panels as shown in the fig.6.7.

Unlike other cargo ships Blue Marlin has a fixed horizontal deck, which helps in carrying huge

structures such as offshore oil & gas platforms on its deck. Longitudinal stiffeners run under top

deck plate and above bottom deck plate to reinforce the deck structure further.

The following table shows properties of some of the elements used in modeling deck of the Blue

Marlin FE model.

35

Description Element type Cross section Dimension Material Density

Longitudinal centre Shell / Plate 3

Plate Thickness 0.026m 12575.5kg/m

girder (QUAD4)

Shell / Plate 3

Longitudinal side girder Plate Thickness 0.018m 12575.5kg/m

(QUAD4)

Shell / Plate 3

Transverse girder Plate Thickness 0.018m 12575.5kg/m

(QUAD4)

Inner deck longitudinal Offset bulb 3

Beam H*T (0.32*0.012)m 7850kg/m

stiffeners plate

Top deck longitudinal 3

Beam Flat Bar H*T (0.26*0.011)m 7850kg/m

stiffeners

Table 6.5 Properties of FE model of deck

The bow is the front part of the vessel before the deck, the hull in front of the bow is designed in

such a way to reduce drag during sailing. Bow section of the vessel Blue Marlin is not modeled to

actual design. As the cargo will only be loaded on top of the deck, analysis of the bow section is

not necessary and is of little importance in the model. Since the bottom of the bow section has

ballast tanks, just the outline of the bow section and its divisions are modeled to utilize the ballast

tank space in the model.

The whole bow section is modeled using few properties; the following table shows the properties

of the bow section model.

Description Element type Cross section Dimension Material Density

Shell / Plate 3

Front section Plate Thickness 0.03m 35778.68kg/m

(QUAD4)

Shell / Plate 3

Middle section Plate Thickness 0.03m 3758kg/m

(QUAD4)

Shell / Plate 3

Back section Plate Thickness 0.03m 9955.4kg/m

(QUAD4)

Shell / Plate 3

Hull Side shell Plate Thickness 0.03m 7850kg/m

(QUAD4)

H*T 0.8x0.02 3

stiffeners Beam T-section 7850kg/m

+0.2x0.04 m

Table 6.6 Properties of FE model of bow section

36

Fig 6.8 FE model of the bow section

Superstructure is the tower on top of the bow which houses control tower; crew accommodation,

stores and other necessary spaces. Only the mass of the structure matters when calculating the

ballast necessary for the vessel, so the weight of the superstructure is incorporated in the model

using mass elements

6.1.2.7 Stern

Blue Marlin has a transom stern, which offers a greater deck area aft and can also provide

improved flow around the stern. The stern of the Blue Marlin is the most complicated section of

the vessel; it should incorporate a number of features, engine room, shape of the stern for

propeller and rudder, provision for the drive shaft to connect the engine to the propeller, ballast

tanks, provision for piping of the engine and the vessel etc.

Due to the presence of the engine in the stern section, the engine room section along the centre

line of the vessel, to accommodate the engine is a big space with out any bulkheads or supports.

This makes the structure on the stern section of the vessel vulnerable to heavy loads. Special

Care should be taken when supporting the heavy cargoes on top of the stern section. The

considerable complication of the stern section makes it a hassle to model it. It is modeled using

61 properties with 2 different element types. The most important properties used in the FE model

of the stern of the vessel Blue Marlin is illustrated in the table below.

Description Element type Cross section Dimension Material Density

Shell / Plate 3

Front section Plate Thickness 0.015m 13568.1kg/m

(QUAD4)

Shell / Plate 3

Mid section Plate Thickness 0.012m 20581.2kg/m

(QUAD4)

Shell / Plate 3

Back section Plate Thickness 0.02m 16217.9kg/m

(QUAD4)

Shell / Plate 3

Hull Side shell Plate Thickness 0.035m 7850kg/m

(QUAD4)

3

stiffeners Beam Flat bulb H*T 0.3x0.011m 7850kg/m

3

stiffeners Beam Flat bulb H*T 0.2x0.02m 7850kg/m

Table 6.7 Properties of FE model of stern section

37

Fig 6.9 FE model of stern section showing the profile of the stern

Fig 6.10 FE model of stern section without deck showing the engine room section

6.1.2.8 Floatation

Vessels floatation is provided by spring elements in the FE model. The bottom hull of the vessel’s

FE model is constrained on its nodes using spring elements, which prevents the model from

dropping off to infinity. The spring elements are restricted to axial only deformation, as the vessel

during floatation experience only axial movements not torsional. The stiffness used in the spring

elements represents the stiffness of the buoyancy of the vessel. This stiffness provides upward

force due to buoyancy of the vessel. This upward force in addition to the water pressure

influences the draft of a floating vessel. Fig.6.8 shows the spring elements used in the model.

38

Fig 6.11 Spring elements under the vessel model.

Floatation on the vessel

Spring 21995.86 Axial

model

Table 6.8 Properties of spring element under the vessel model

The stiffness of the springs created under the vessel model is to simulate the buoyancy of the

vessel in the model; it represents the stiffness of the buoyancy. This is the additional upward

force at a given displacement downward.

The height of the vessel immersed in water measured from the midship section is called draft of a

vessel. All vessels have a free floating draft which is the height of vessel under water with out any

additional ballast or load on deck. The ideal sailing draft for Blue Marlin is 10.28 which is a

design parameter of the vessel. The buoyancy as explained in the last section plays an important

role in the draft of the vessel.

The pressure of water acting on the vessel’s hull when the vessel is not sailing is called

hydrostatic pressure. Since the main motive of this thesis is to analyze the load-out of heavy

SPAR when the vessel is docked to the keyside, means the vessel is in static condition, we only

take into account the hydrostatic pressure not the hydrodynamic pressure.

The hydrostatic pressure acting on the vessel hull at a point can be found out. The pressure at a

certain point depends on the draft of the vessel. The pressure will be less on the sides of the hull

near to the water surface and high on the hull bottom. The hydrostatic pressure acting on a unit

area is calculated and applied on the element face as a pressure.

There are quite a lot of software’s available in the market to calculate such marine properties of

the vessel. Dockwise engineers make use of one such software and calculated the hydrostatic

pressure of Blue Marlin for different drafts. To make it much easier they’ve incorporated the

hydrostatic pressure data in an API tool to generate the water pressure on the hull of the vessel

according to its draft.

39

If the draft of the vessel is entered in the tool, the API automatically calculates the pressure on

each element face and creates a load set for hydrostatic pressure and applies it to all the

elements inside the draft line. The effect of waves in sea water can also be set in the model using

the API tool. Fig 6.12 shows the API tool for implanting draft and hydrostatic pressure. When the

draft of the vessel is entered in the tool, it automatically calculates and applies the water pressure

below the draft line.

Fig 6.12 API tool for implementing draft and hydrostatic pressure in the model

A closer look at the applied pressure reveals the amount of pressure applied corresponding to the

depth of water. Bigger the arrow size, higher the pressure. As you can see from the fig 6.14 the

pressure is much higher at the hull bottom than on the sides.

40

6.1.2.10 Water Ballast

Blue Marlin has 84 water ballast tanks, which helps the vessel in maintaining the draft and

stability on the high seas. Other than the water ballast tanks, there are other tanks to store fresh

water, fuel, oil for lubrication of its engines and other equipments onboard.

All the tanks and spaces used to store equipments are also part of the ballast tanks. Even the

crew members are also part of the ballast. When doing a ballast calculation of the vessel all the

above masses are taken into account. All the ballast tanks spaces are shown in the drawing in

the Appendix number 2.

The ballast masses are represented in the model by placing mass elements to the nodes in

appropriate location in the model. Other than water ballast tanks all the other masses are all

permanently included in the FE model of the Blue Marlin. All the masses are represented by

Mass elements. The water mass for ballasting the vessel can be added to the mass elements in

the model according to the ballast calculation created. To do it more easily, an API tool is created.

Each ballast tanks section in the model is added with mass elements on its surfaces to add the

mass of the ballast water. These elements are grouped under the name of the ballast tank in the

API tool. Using the tool, the ballast water masses can be updated on the model quickly and

efficiently. The entered mass for a tank will be equally distributed among all the mass elements

present in that particular tank section. Fig 6.15 shows the API tool used for adding ballast masses

to the model. The mass elements in the FE model can be seen in the fig 6.16

Fig. 6.15 API tool for updating the mass elements in the model

41

Fig. 6.16 Mass elements in the FE model of Blue Marlin

The following table shoes the summary of all the mass elements in the model

Description Element type

spaces masses ballast masses

Table 6.9 Properties of Mass element in vessel model

The number of skid beam and its location is already investigated in chapter 4.3 and 4.4. Based on

the dimensions in the above chapter, skid beams are created on vessel model above the deck.

The location of skid beams are identified and is modeled and meshed based on the nodes

already available on the vessel deck. The properties of the skid beams, like, thickness, height

and width are acquired from the experiences of previous projects. Fig.6.17 shows the modeled

skid beams on the vessel deck.

42

The skid beams are created using the following properties;

Shell / Plate 3

Skid beams Plate Thickness 0.025m 7850kg/m

(QUAD4)

As the most important constituents of the Blue Marlin structural FE model have been detailed

above, the model is not 100% identical to the physical vessel. There are some discrepancies

exist between the physical vessel structural details and the FE model.

The FE model does not include the following details which are present in the physical vessel:

1. Manholes

2. Holes for piping

3. Air/drain holes

4. Holes in the bottom and deck structural frames to interconnect the sections to form the

ballast tanks and

5. Collars in the web frame are not chamfered

6. The bow section of the vessel is not modeled to exact details. As the structure of the vessel

under the deck is the one carrying the load and the bow section has nothing to do with the

load bearing, the details of the bow section are neglected and just an outline of the section

is modeled.

The difference between the physical vessel and FE model can be visualized in the fig 6.18 and

6.19 as it shows the typical web frame of the Physical vessel and FE model.

The mesh of the whole FE model of the vessel is modeled in a coarse manner. The resulting

output from the analysis of the model should be viewed as a representation of the stress of the

whole model rather than the particular region. When the stresses are higher in a particular area,

the model has to be compared with the vessel design drawings and modified to incorporate the

discrepancies in that particular region and further refinement of mesh in that region should be

done and analyzed again.

43

Fig 6.19 Typical web frame details of the FE model

The FE model of the vessel Blue Marlin has the following characteristics values, compared with

the actual designed values as shown in the tale below

Mass of the vessel 34,302.41 Tons 34629.538 Tons

Longitudinal Centre of gravity from aft

102.819m 101.8264m

perpendicular

Table 6.10 Comparison of FE model with the actual vessel.

The mass of vessel includes the weight of light ship, port casing, starboard casing, crew and

effects, provisions, stores, fuel oil, lubricating oil, fresh water. The weight of these components

doesn’t change much for every project performed, there for it is fixed value. These component

masses are permanently included in the FE model of the vessel as mass elements.

From the table above, it is clear that there are some percentage of difference in the weight of the

model and the actual vessel. But the difference in the mass between the actual vessel and FE

model is about 300tons, this difference is mainly due to the discrepancies present in the model as

explained above. But for a model of such a huge scale where the masses are calculated in

thousands, the difference of 300 tons is very less and it can be neglected.

The difference in centre of gravity of the model to the actual vessel is about 1 m, and is due to the

above mentioned discrepancies and the differences in the design to the FE model.

It is very clear from the table above that the FE model of the vessel is almost as accurate to the

actual design and can be safely utilized for the analysis.

Unlike the Blue Marlin Model, which was readily available with Dockwise, the SPAR model was

modeled from scratch. The design data of the SPAR provided by the client is discussed in section

4.2. With that information in hand, the SPAR is modeled in FEMAP.

44

Almost all the section of the SPAR resembles standard geometry like cylinder, tube and box; it is

decided to model the SPAR entirely using beam elements except for the heave plates, to keep

the model simple. Modeling the SPAR entirely using beam elements is due to the fact that it

improves analysis execution speed compared to plate and solid elements and also the absence

of the complete design details of the SPAR.

The data provided by clients about the SPAR is just an approximation to study the feasibility of

the transport, so design details of the SPAR is not a concern, what matters most in our analysis is

stiffness of the whole structure. To get the desired stiffness, it is very important to model the

SPAR with same centre of gravity as per the clients design.

As like the Blue Marlin model, the SPAR was also modeled in rectangular coordinate system. The

origin (0,0,0) of the Spar model is located on the centre line of the SPAR at the Hard tank end.

The centre line of the SPAR starting from the origin runs in the x-axis, to match the centre line of

the vessel. The truss section with heave plates is followed by the hard tank and the SPAR ends

with the soft tank keel.

The units used in the FE model of the SPAR and the entire analysis are Metric (Metres).

As the complete design data about the SPAR, the material specifications and details of the

dimension of the interior sections are not known, we are not following any standards in model of

the SPAR created. The important parameter in the SPAR model is the density of the material

used and the stiffness of the whole SPAR.

So to keep our model simple, we assume that the all the SPAR sections are hollow structure, the

masses due to the internal structural details are added to the outline elements of the section by

adjusting the density of the material specified in analysis. For example the mass of the internal

details of the hard tank and the outfitting mass are added to the beam elements of cylindrical

shape by altering the density of the material.

45

6.2.3 Model details and properties

As the FE model of SPAR is modeled using beam elements and the mass of additional structural

elements are added to it by adjusting the density, so every component of the SPAR has to be fine

tunes to achieve the desired mass by altering the density of the material.

The SPAR model has a total of 13 properties with a combination of 2 element types. A summary

of the whole model can be seen in the table below

Total No. of Elements 4819

Total No. of Properties 13

Total No. of Materials 13

Elements types used Plate, Beam

Table 6.11 Comparison of FE model with the actual vessel

Now, we will have a detailed look in to the modeling details of each section of the SPAR.

The Hard tank is the most important structure in the SPAR; it provides the buoyancy to float on

the high seas and also withstand the loads of topside and also waves on its walls. The hard tank

constitutes for more than 60% of the SPAR mass.

The total mass of the hard tank alone is 14304tons without the supports and outfittings. A line of

81.3m is created and meshed with beam elements with circular tube shape using the following

properties for the achieving the required weight. Fig 6.21 shows the FE model of the hard tank

section of the SPAR.

Radius: 22.2505m 3

Hard tank Beam Circular tube 62889kg/m

Thickness: 0.020m

Table 6.12 Properties of FE model of hard tank

46

6.2.3.2 Soft Tank

The soft tank is used for ballasting the SPAR to maintain its draught and also to provide floatation

during horizontal position before upending.

The soft tank’s mass is 2315 tons. All the time during transportation, the soft tank are suspended

on the stern of the vessel, so it doesn’t has supports underneath it. There will be some additional

masses in the soft tanks for guiding the risers in the center well.

Similar to Hard tank the soft tank is created by utilizing line geometry of length 12.8m and

meshed using the beam element of following properties for achieving the required weight. Fig.

6.22 shows the FE model of the Soft Tank section of the SPAR.

Height * Width:

Rectangular 3

Soft tank Beam 44.501m*44.501m 50820.93kg/m

tube

Thickness: 0.020m

Table 6.13 Properties of FE model of soft tank

Heave plates are placed in between the truss structure of the SPAR to provide stiffness and also

for damping the SPAR vertical motion while floating. The Future SPAR has 2 heave plates in

between the truss structures.

The heave plates are modeled using plate elements with the following properties for achieving the

required mass. Fig. 6.23 shows the FE model of the heave plate. The weight of the heave plate

is 529tons and it has a cutoff section on both sides at the bottom for proving space for the

supports.

Material

Description Element type Cross section Dimension

Density

Height * Width:

Shell / Plate 11877.21

Heave plate Plate 44.501m*44.501m 3

(QUAD4) kg/m

Thickness: 0.025m

Table 6.14 Properties of FE model of Heave plates

47

Fig. 6.23 FE model of Heave plate section of the SPAR

6.2.3.4 Truss

The truss section of the SPAR extends from the hard tank and connects the soft tank. It gives the

SPAR the deep draught it requires for withstanding the heavy ocean currents.

The truss section of the SPAR is similar to a normal truss used for construction. The truss is

fabricated with circular tubes, large tube for 4 legs and smaller tubes for the cross members. The

weight of the truss is 3293 tons and is also modeled by beam elements with circular tube shape

using the following properties for achieving the required mass. Fig 6.24 shows the FE model of

the truss section of the SPAR

48

Description Element type Cross section Dimension Material Density

Radius: 3m 3

Truss legs Beam Circular tube 13245.9kg/m

Thickness: 0.015m

Radius: 1.5m 3

Cross members Beam Circular tube 11521.84kg/m

Thickness: 0.015m

Table 6.15 Properties of FE model of Truss section

The connection between the hard tank and the four truss legs are made by rigid elements, so that

the loads on the hard tank at the end node will be directly transferred to the truss legs and vice

versa. This step is very critical for obtaining the desired stiffness of the SPAR, as there are no

interior components designed in the hard tank model which in reality will make the hard tank

stiffer and hold the truss legs in position. Fig 6.25 shows the line diagram of the hard tank and

truss with rigid element. The rigid element connection between the truss legs and the end node of

the hard tank is visible in the Fig. 6.25

Fig 6.25 Line only FE model assembly of hard tank and truss

49

At this step the weight of the SPAR is 20970 tons, which is the sum of masses of hard tank, soft

tank, truss and 2 heave plates. The weight in the model is checked and verified to make sure that

the weights match with the design. Fig 6.26 shows the 3D view of the hard tank and truss

connected with rigid element.

6.2.3.5 Supports

Now the main sections of the SPAR are modeled and assembled, its time to create the supports.

As there are no constraints regarding the supports from the client’s side, support design depends

on the results of this analysis. Based on the conclusion of this thesis, Dockwise can recommend

the clients number of skid beams configuration needed and also the dimension of the supports in

relation to the center of the SPAR. For now, the supports are modeled appropriately based on the

information from previous projects.

As the hard tank and truss section can be accommodated on the vessel deck area, it is

necessary to support both hard tank and truss section. Since the Hard tank constitutes for more

than 60% of the weight in the SPAR, it is the critical part whose loads have to be distributed on

the vessel properly. So the Hard tank has to be supported throughout its length. As the truss

weigh only 3293 tons, compared to the hard tank’s weight it is very less. So it doesn’t have to be

supported throughout its length. The location where the heave plates are fixed in the truss are the

strongest points, so it has been decided to support the truss at 4 points under its heave plates on

the truss legs, as it was done earlier in previous projects.

The supports for soft tank as discussed in chapter 4.4.5 are not modeled. The soft tank will be

temporarily supported from the start of the load-out operation until the SPAR reaches its final

stowage position, until then it will be supported on the beams on the keyside not on the vessel

deck. So the necessary supporting condition for the soft tank can be applied using the constraints

in the model, the location and design of the supports for soft tank is of no significance.

The weight of the supports recommended by the clients in their design is 4681 tons, and we don’t

have any data regarding how much of the support weight is under the hard tank and how much is

under the truss. The weight of the SPAR specified by clients is just an estimation based on their

past experiences. So without that information it is not possible to model the support individually

for each section, without knowing the weights of each support legs. When the supports are

modeled after the SPAR assembly, the mass of the supports in total can be tuned exactly to

match the specified weight.

As discussed in the topic 4.5, one of the factors affecting the load-out is the number of skid

beams required, the supports design will also change according to the number of skid beams. So

we need to design 3 support configurations for the SPAR as listed below:

1. for 4 skid beams load-out

2. for 3 skid beams load-out

3. for 2 skid beams load-out

As the locations of the skid beams are already known from the chapter 4.2.4, the supports are

modeled as per the dimension in relation to the centre of the SPAR. The interior connections

between the supports and hard tank nodes are meshed with rigid elements to connect the nodes

on the hard tank with the supports and also simulate the stiffness due to the interior structural

details of the SPAR. Fig 6.27 shows the modeled hard tank supports and the rigid element

connection with the hard tank.

The supports under the truss are also modeled using the same skid beam location but with a

different cross section of the beam element. From the past projects information, tapered beams

were used as it tends to distribute the loads more gradually and also withstand SPAR’s self

weight. The length of each support under truss is taken as 15m, by mere approximation on

studying the previous projects. The supports for the truss legs can be seen in the fig 6.28.

50

Fig 6.27 FE model of the hard tank supports

After the supports are modeled, the densities of the different properties used are tuned to achieve

the specified weight. Now the SPAR with its supports weighs 25651 tons. Fig. 6.29 shows the

SPAR with supports for 4 skid beams load-out.

The same procedure is followed to model the support of the SPAR for 3 and 2 skid beams load-

th

out. The 4 skid beam has been removed and the supports on the starboard side of the SPAR

st

have been modified as shown in the Fig. 6.30 for the SPAR with 3 skid beams load-out. The 1

th

and 4 skid beam is removed and the supports are modified as shown in the fig 6.31 for SPAR

with 2 skid beams load-out.

51

Fig 6.29 FE model of SPAR with supports for 4 skid beam load-out

Fig 6.30 FE model of SPAR with supports for 3 skid beam load-out

The following table shows a summary of all the properties used in the SPAR supports modeling

Element Cross

Description Dimension Material Density

type section

Pipe section of the Circular Radius: 0.75m 3

Beam 11328.6kg/m

support tube Thickness: 0.022m

H*W: 3*1.5*1.5m 3

End support beams Beam I beam 11354.3kg/m

T: 0.02*0.02*0.025m

Beam connecting H*W*W: 1.5*1.5*1.5m 3

Beam I beam 8951kg/m

the end supports T: 0.015*0.015*0.025m

Support beam Circular Radius: 1.5m 3

Beam 11521.8kg/m

under heave plate tube Thickness: 0.025m

End A: 5*3*3*0.02*0.02*0.025

Tapered beams for Tapered 3

Beam End B: 61521.8kg/m

truss supports I beam

3*2.5*2.5*0.02*0.02*0.025

Table 6.16 Properties of FE model of SPAR supports

52

Fig 6.31 FE model of SPAR with supports for 2 skid beam load-out

6.2.3.6 Outfittings

Except the hard tank, soft tank, truss, supports, all the other objects which are fixed to the SPAR

structure, for example pipe works, strakes, and additional steel fabrication works on the hard tank

side etc., which doesn’t influence the stiffness of the SPAR are termed as outfittings.

As the outfittings are objects scattered along all the sections of the SPAR and we don’t have any

detail information about the outfittings, it is of no importance to model it in a specific location in

the SPAR model. So we utilize the mass of the outfittings, which is 9205 tons, to tune the COG

specified. By spreading the outfittings mass on hard tank, truss and soft tank and tuning its

densities we can achieve the required COG.

The weight of the outfittings is distributed in hard tank, truss and soft tank by modifying, i.e

increasing the densities of its materials. And finally when the required total mass of the SPAR is

attained, the COG can be moved by replacing the mass in one section to the other, such that the

required location of COG, i.e 119.14, from the soft tank keel (end of soft tank) is achieved.

The final properties of the FE model of the SPAR including all the weights and after tuning the

COG are summarized as follows:

Element

Description Cross section Dimension Material Density

type

Radius: 22.2505m 3

Hard tank Beam Circular tube 91583.57kg/m

Thickness: 0.020m

Height * Width:

Rectangular 3

Soft tank Beam 44.501m*44.501m 65315.8kg/m

tube

Thickness: 0.020m

Shell / Height * Width:

3

Heave plate Plate Plate 44.501m*44.501m 11877.21kg/m

(QUAD4) Thickness: 0.025m

Radius: 3m 3

Truss legs Beam Circular tube 19798kg/m

Thickness: 0.015m

53

Radius: 1.5m 19657.4

Cross members Beam Circular tube 3

Thickness: 0.015m kg/m

Pipe section of the Radius: 0.75m 11328.6

Beam Circular tube 3

support Thickness: 0.022m kg/m

H*W: 3*1.5*1.5m 11354.3

End support beams Beam I beam 3

T: 0.02*0.02*0.025m kg/m

Beam connecting H*W*W: 1.5*1.5*1.5m 8951

Beam I beam 3

the end supports T: 0.015*0.015*0.025m kg/m

Support beam Radius: 1.5m 11521.8

Beam Circular tube 3

under heave plate Thickness: 0.025m kg/m

End A: 5*3*3*0.02*0.02*0.025

Tapered beams for Tapered I 61521.8

Beam End B: 3

truss supports beam kg/m

3*2.5*2.5*0.02*0.02*0.025

Table 6.17 Properties of FE model of SPAR with outfittings for 4 skid beam load-out

The FE model of the SPAR has the following characteristics values, compared with the design

values given in chapter 4.2 as shown in the tale below

Mass of the SPAR 34855 Tons 34855 Tons

Longitudinal Centre of gravity from soft

119.14m 119.14m

tank keel

Table 6.18 Comparison of FE model of SPAR with the design.

The SPAR model is accurate to the design specification provided by the clients. The SPAR model

can be safely utilized for the analysis.

54

CHAPTER 7

The process of load-out is discussed in chapter 4. The reason and factors for choosing skidding

method for load-out of the Future SPAR has been discussed in detail. The skidding method is a

very slow process; the average movement of the SPAR on the deck will be about 75mm per

minute or even less which in turn depends on stroke of the hydraulic cylinders pushing the SPAR.

Since the skidding method is very slow, the load on the vessel structure can be considered as a

static and linear loading as there are no sudden or harmonic changes in the loading pattern. The

term "linear" means that the computed response--displacement or stress, for example, is linearly

related to the applied force. The term "static" means that the forces do not vary with time--or, that

the time variation is insignificant and can therefore be safely ignored.

So in this analysis, it has been assumed that the loading is static and the most critical load steps

on the vessel structure are recreated in FEMAP environment and analyzed by linear static

analysis.

[15] The most suitable strength or stiffness criterion for any structural element or component is

normally some maximum stress or deformation which must not be exceeded. In the case of

stresses the value is generally known as the maximum allowable working stress. Because of

uncertainties of loading conditions, design procedures, production methods, etc., designers

generally introduce a factor of safety into their designs, defined as follows:

However, in view of the fact that plastic deformations are not accepted this definition can be

modified to: Factor of safety = yield stress / allowable working stress

According to the factors taken from allowable stresses in ship construction engineering

guidelines, recommended safety factor is 1.43. [15]

Yield stress for the material used in ship construction = 235Mpa

Allowable stress

stress ( allowable) = 235/1.43 = 164.3Mpa

55

7.2 Uniform loading by hydraulics

Since the hard tank constitutes the most of the mass of the SPAR, the vessel deck will

experience the critical load at any given area. When compared with loads induced by hard tank

mass, the truss and soft tank loads are very small. So the hydraulic cylinders are used only under

the hard tank for leveling out the peak loads exerted by the very stiff hard tank structure and

equally distribute the critical load of the hard tank.

When the SPAR is loaded-out using hydraulics cylinders under the hard tank, the loads on skid

beam surface are equal at any given point. So to analyze the uniform loading scenario the actual

SPAR model is not used, instead the loads on each skid beam are calculated manually and the

resulting loads are equally applied on the skid beam surface as nodal loads. The loads on skid

beam vary according to the number of skid beam and its relative position from the Centre of

gravity of the SPAR. So the loads are calculated separately for each steps analyzed in the

uniform loading case.

Once the load on each skid beam is calculated manually as explained in the next section, to

make sure that the loads are equally distributed along the length of the skid beam, following steps

were utilized.

1. A line is drawn above the skid beams for the length of the hard tank which is 81.382m.

2. The calculated total load on each skid beam is applied to the line as a force per unit length

(which is criteria in uniform loading)

3. The nodes on the centre line of the skid beams are associated with the line.

4. The load on the line, i.e., force per unit length is expanded and transferred to the nodes

associated with the line.

5. Once the loads are transferred to the nodes, the line is detached from the nodes.

By this method the loads are equally distributed on the nodes in regard to its distances. Fig. 7.1

shows the loads applied on the skid beam by the above mentioned steps.

56

7.2.2 Boundary conditions in Uniform loading

Every node in FEMAP has six degrees of freedom; the constraints are applied to the model by

restricting the degrees of freedom of the necessary nodes corresponding to the real life scenario.

The following constraints were utilized in general for all the uniform loading steps:

The vessel’s Y-axis is fixed to arrest the movement of the vessel in Y direction, as the vessel will

not move in transverse direction during load-out; mooring lines are attached to prevent the vessel

from moving. The X axis is not constrained as the vessel is free to deflect longitudinally when the

loads are applied on the deck. The Z-axis is not constrained as the vessel will be floating on the

water and is free to move up and down depending on the load on deck, ballast and tide

conditions. The model is free to rotate in all 3 axes, since the vessel during load-out experience

various rotational movements. Fig. 7.2 shows the constraint in the bow section of the vessel

model. The constraint is applied as a nodal constraint.

Unlike the bow, stern section of the vessel is fixed in Y and X-axis as the vessel is docked along

the stern with the keyside and moored. It can’t move or deflect in X-axis due to the keyside. All

the other DOF are free as the model has to move up and down in the Z-axis and the rotations

should be free to simulate the floating condition of the vessel, where the vessel while floating in

water will experience various rotational movement due to waves and also if not properly

stabilised. Fig. 7.3 shows the Constrains in the stern section of the vessel.

The model is constrained only at 2 nodes as explained because two nodes are located on the

center line of the vessel and the model is free is rotate around the X- axis. Constraining other

nodes may lead to restriction of the rotation movement of whole vessel model which might results

in not providing the actual floatation scenario to the model.

57

Fig 7.3 Constraints in stern section of vessel

The spring elements as explained in chapter 6.1.3.6 are used to make the model stable when not

precisely balanced. The spring elements are created between two node groups, nodes on the

bottom hull of the vessel and one on the space below the vessel. The hanging node below the

vessel is fixed in all 3-axes, as the spring element can take only axial loads, to stabilize the lower

node the additional X and Y axis constraints are needed. Fig. 7.4 shows the constraints of the

spring elements

7.2.3 Draft

During the calculation of the draft and ballast condition using the software as explained in chapter

6.1.2.9, it is found that the vessel draft has to be 11m.

58

Only when the draft is increased to 11.0 m the aft part of the vessel generates sufficient buoyancy

to reduce the shear force down to 99.9% of the allowable limit during the load-out of the SPAR

hard tank. The minimum draft required for load-out of the SPAR is set at 10.95m for the load-out

operation of the hard tank.

The draft is updated in the model using the API tool. When the draft is entered the water line is

drawn on the model to show the level of the draft and the water pressures are applied on the

element faces which are inside the draft level. The water level line serves as a reference in the

model to visualize the trim and rolling movement of the vessel model. The draft and the

hydrostatic pressures applied can be seen in the Fig. 7.5

Fig. 7.5 Draft level and hydrostatic pressure for the step hard tank on the vessel.

As discussed in chapter 6.1.2.10, the mass of ballast water required to maintain the draft and

stability of the vessel during load-out are incorporated as mass elements in the model.

The amount of water necessary for ballast is calculated based on number of parameters like load

on deck and its COG, draft, bending moment in the vessel and stability. It is calculated using

software called GHS, the calculated ballast is updated in the vessel model using the API tool

available. A complete ballast calculation as created from the software GHS for the step hard tank

on the vessel has been attached in appendix 3. The detail of ballast mass needed in each ballast

tank for the load step hard tank on the vessel is given in the table below.

.

Tank Description Name Capacity in kg % filled Mass in kg

1 upper S WBT1U.S 839760 100 839760

1 upper P WBT1U.P 851280 100 851280

1 upper C WBT1U.C 624050 100 624050

2 upper S WBT2U.S 839340 100 839340

2 upper P WBT2U.P 861430 100 861430

2 upper C WBT2U.C 839210 100 839210

3 upper S WBT3U.S 861430 100 861430

3 upper P WBT3U.P 850380 100 850380

3 upper C WBT3U.C 861290 0 0

59

4 upper S WBT4U.S 850840 25 212710

4 upper P WBT4U.P 0 0

4 upper C WBT4U.C 839210 0 0

5 upper S WBT5U.S 862475 0 0

5 upper P WBT5U.P 849650 0 0

5 upper C WBT5U.C 861910 0 0

6 upper S WBT6U.S 585330 0 0

6 upper P WBT6U.P 607420 0 0

6 upper C WBT6U.C 585240 0 0

1 middle S WBT1M.S 2444290 100 2444290

1 middle P WBT1M.P 2444290 100 2444290

1 middle C WBT1M.C 846690 100 846690

2 middle S WBT2M.S 2467700 100 2467700

2 middle P WBT2M.P 2487600 100 2487600

3 middle S WBT3M.S 0 0

3 middle P WBT3M.P 2467700 0 0

3 middle C WBT3M.C 0 0

4 middle S WBT4M.S 2466490 79.6 1963326

4 middle P WBT4M.P 2487610 100 2487610

4 middle C WBT4M.C 2321520 100 2321520

5 middle S WBT5M.S 2487600 0 0

5 middle P WBT5M.P 2467700 0 0

5 middle C WBT5M.C 2332400 0 0

6 middle S WBT6M.S 1711630 0 0

6 middle P WBT6M.P 1732380 0 0

6 middle C WBT6M.C 1588440 0 0

1 lower S WBT1L.S 1214410 100 1214410

1 lower P WBT1L.P 1214410 100 1214410

1 lower C WBT1L.C 1116960 100 1116960

2 lower S WBT2L.S 1302220 100 1302220

2 lower P WBT2L.P 1307690 100 1307690

2 lower C WBT2L.C 1325950 100 1325950

3 lower S WBT3L.S 1309480 100 1309480

3 lower P WBT3L.P 1304220 100 1304220

3 lower C WBT3L.C 1327370 100 1327370

4 lower S WBT4L.S 1303230 90.4 1178119.9

4 lower P WBT4L.P 1310214 88.8 1163470

4 lower C WBT4L.C 1325897 75.8 1005029.9

5 lower S WBT5L.S 1309430 0 0

5 lower P WBT5L.P 1309430 0 0

5 lower C WBT5L.C 1328750 0 0

6 lower S WBT6L.S 782700 0 0

6 lower P WBT6L.P 788170 0 0

6 lower C WBT6L.C 789920 0 0

water bal oflow WBTOVERF.S 0

fwd lower WB P WBTFL.P 0 0

fwd lower WB S WBTFL.S 1592033 54.1 861289.85

fwd upper WB S WBTFU.S 0 0

60

fwd upper WB P WBTFU.P 0 0

aft WB P WBTA.P 0 0

poop aft WB S WBTPA.S 0 0

poop fwd WB S WBTPF.S 0 0

engine RM TOP S WBTERTOP.S 0 0

engine RM TOP P WBTERTOP.P 0 0

aft ballast S WBTALWL.S 0 0

aft peak C WBTAPC.S 0 0

lower peak C WBTLOWPEAK.C 0 0

high peak C WBTHIGHPEAK.C 2995970 0 0

WB overf fwd WBTOVERFWD.C 0 0

aft poop lower WB P WBTAPL.P 0 0

aft poop upper WB P WBTAPU.P 0 0

1 outer wing P WBT10.P 1291440 100 1291440

1 outer wing S WBT10.S 1291470 100 1291470

2 outer wing P WBT20.P 2708360 100 2708360

2 outer wing S WBT20.S 2708360 100 2708360

3 outer wing P WBT30.P 3168290 0 0

3 outer wing S WBT30.S 3168290 100 3168290

4 outer wing P WBT40.P 3262750 0 0

4 outer wing S WBT40.S 3262750 100 3262750

5 outer wing P WBT50.P 3276980 0 0

5 outer wing S WBT50.S 3276980 3.5 114694.3

6 outer wing P WBT60.P 2126910 0 0

6 outer wing S WBT60.S 2126900 0 0

7 outer wing P WBT70.P 2596050 0 0

7 outer wing S WBT70.S 1859960 0 0

Total ballast mass 54418600

Table 7.1 Ballast condition for step hard tank on the vessel.

Some of the masses which are permanent in the vessel like, fuel, lubricating oil, fresh water,

weight of the crew, weight of the materials in stores etc., regardless of the load on deck are

permanently incorporated in the model. But, if needed the permanent masses can also be edited

using the API tool.

The ballast condition above is the used for all the analysis step of hard tank on vessel by uniform

loading, namely, hard tank on vessel with 4 skid beams, hard tank on vessel with 3 skid beams

and hard tank on vessel with 2 skid beams.

In all the cases of the uniform loading, weight distribution of the hard tank taken in to account for

analysis is 25122 tons, which is much higher than the weight distribution of the hard tank

provided by the designers. The weight of the 4 skid beams, which is 327 tons, is already taken in

to account with the SPAR weight. Since the weight of the skid beams are not that much high

compared to the vessel and SPAR weight, the difference of weight arising in the number of skid

beams in every other step can neglected.

When the load-out is done by uniform loading using hydraulics, the loads on the each

side of the supports will be equal on both skid beams. Since the peak loads will be

equally distributed, loads on each skid beam can be calculated using analytical method

and can be used in the model.

61

Hard tank on vessel with 4 skid beams:

The following calculation shows the method of load calculation for the step hard tank on vessel

with 4 skid beams:

The support locations for 4 skid beams on the vessel deck and its dimension in relative to the

SPAR centre is shown in the fig 7.6.

SB

casing

Weight of skid beams from the model = 327 tons

length of hard tank on the skid beams (length of skid beam where load acts) = 81.382 m

weight of hard tank on vessel with outfitting used for ballast calculation = 25122 tons

Assuming that the skid beams are also included in the above weight of hard tank for ballast

calculation,

= 25122 -327 =24795 tons

Since the skid beam is already modeled with the vessel, the load of hard tank is taken as above

for calculation purposes.

Distance between 2 skid beams = 20.575 - 12.7 = 7.875 m

Centre of 2 skid beams where the load acts in case of uniform loading = 7.875/2 = 3.9375 m

Distance between SPAR center and the point of load acting = 12.7 + 3.9375 = 16.6375 m

Distance between 2 skid beams = 23.175 - 15.3 = 7.875 m

62

Centre of 2 skid beams where the load acts in case of uniform loading = 7.875/2 = 3.9375 m

Distance between SPAR center and the point of load acting = 15.3 + 3.9375 = 19.2375 m

Total distance between point of load acting on both sides = 16.6375 +19.2375 = 35.8750 m

Total load on both skid beams = (24795 x 19.2375) / 35.8750 = 13295.9947 tons

Load on each skid beam = 13295.9947 / 2 = 6647.99738 tons

Load per unit length on each skid beam = 6647.99378 / 81.382 = 81.688 ton/m

Total load on both skid beams = (24795 x 16.6375) / 35.8750 = 11499.0052 tons

Load on each skid beam = 11499.0052 / 2 = 5749.50261 tons

Load per unit length on each skid beam = 5749.50261 / 81.382 = 70.648 ton /m

Total load 13295.994 tons 11499.0052 tons

Total load on each skid beam 6647.997 tons 5749.50261 tons

Load per unit length on each skid

81.688 tons/m 70.648 tons/m

beam

Table 7.2 Load calculation for the uniform loading with 4 skid beams load-out

5

The load per unit length on each skid beam in port side is 81.688 ton/m or 8.0136 x10 N/m

5

The load per unit length on each skid beam in starboard side is 70.648 ton/m or 6.930 x 10 N/m

The support locations for 3 skid beams on the vessel deck and its dimensions in relative to the

SPAR centre is shown in the fig 7.7

SB

casing

.

Fig. 7.7 Location of supports for 3 skid beams load-out

Based on fig 7.7, the loads on each skid beams are calculated as in the previous loading case

and the results are displayed in the table below.

63

The weight of 3 skid beams from the model = 247 tons.

Total weight of hard tank on vessel with outfitting in the ballast calculation = 25122 tons

Assuming that the skid beams are also included in the above weight of hard tank for ballast

calculation,

Total load on vessel = weight of hard tank - weight of skid beams = 25122 -247 = 24875 tons.

Total load 11916.634 tons 12958.365 tons

Total load on each skid beam 5958.317 tons 12958.365 tons

Load per unit length on each skid

73.214 tons/m 159.228 tons/m

beam

Table 7.3 Load calculation for the uniform loading with 3 skid beams load-out

5

The load per unit length on each skid beam on port side is 73.214 tons/m or 7.182*10 N/m

6

The load per unit length on each skid beam on starboard side is 159.228 tons/m or 1.562*10 N/m

The support locations for 2 skid beams and its dimension relative to the SPAR centre is shown in

the fig 7.8 It is very obvious that the skid beams are supported on the longitudinal bulkheads.

SB

casing

Based on fig 7.8, the loads on each skid beams are calculated as in the previous loading case

and the results are displayed in the table below.

Total weight of hard tank on vessel with outfitting in the ballast calculation = 25122 tons

Assuming that the skid beams are also included in the above weight of hard tank for ballast

calculation,

Total load on vessel = weight of hard tank - weight of skid beams = 25122 -169 = 24953 tons.

64

Port side (1 skid beam) Starboard side (1 skid beam)

Total load 13635.032 tons 11317.967 tons

Total load on each skid beam 13635.032 tons 11317.967 tons

Load per unit length on each skid

167.543 tons/m 139.072 tons/m

beam

Table 7.4 Load calculation for the uniform loading with 2 skid beams load-out

6

The load per unit length on each skid beam on port side is 167.54 tons/m or 1.64x10 N/m

6

The load per unit length on each skid beam on starboard side is 139.07 tons/m or 1.36x10 N/m

Port side (#1 & #2) Starboard side (#3 & #4)

6647.997 tons / 5749.50261 tons /

4 skid beams load-out

81.688 tons/m 70.648 tons/m

5958.317 tons/ 12958.365 tons/

3 skid beams load-out

73.214 tons/m 159.228 tons/m

13635.032 tons / 11317.967 tons/

2 skid beams load-out

167.543 tons/m 139.072 tons/m

Table 7.5 Summary of load calculation for all 3 steps of uniform loading

The following section describes the analysis and results of each step of uniform loading

cases.

The above calculated load for the load step hard tank on vessel with 4 skid beams using uniform

load together with the ballast condition and draft have been applied in the model of the vessel

with 4 skid beams. Fig 7.9 shows the model of the step hard tank on vessel with 4 skid beams by

uniform loading with the calculated load, draft and ballast condition.

Fig. 7.9 FE model of analysis step hard tank on vessel with 4 skid beams by uniform loading

65

The results of static linear analysis of the step hard tank on vessel with 4 skid beams are

illustrated below.

The total translation in the model is 0.82m. The highest equivalent stress found was about

178.983Mpa.

Fig 7.10 shows the total deformation of the model after the analysis. The scale of the results is

increased by 10% of the actual value to visualize the actual deformation of the model. The red

line shows the draft line of the vessel or the water surface level.

Fig 7.10 Scaled total deformation of the analysis step uniform loading with 4 skid beams

Fig. 7.11 shows the FE model of the Analysis step hard tank on vessel with 3 skid beams. The

application of loads on nodes in the centre line of the skid beams for a length of 81.382 m can be

seen in the figure below.

Fig 7.11 FE model of analysis step hard tank on vessel with 3 skid beams by uniform loading

66

The results of static linear analysis of the step hard tank on vessel with 3 skid beams are

illustrated below.

The total translation in the model is 0.807m. The highest equivalent stress found was about

151.215Mpa.

Fig 7.12 shows the total deformation of the model after the analysis. The scale of the results is

increased by 10% of the actual value to visualize the actual deformation of the model. The red

line shows the draft line of the vessel or the water surface level

Fig 7.12 Scaled total deformation of the analysis step uniform loading with 3 skid beams

Fig. 7.13 shows the FE model of the Analysis step hard tank on vessel with 2 skid beams. The

application of loads on nodes in the centre line of the skid beams for a length of 81.382 m can be

seen in the figure below.

Fig 7.13 FE model of analysis step hard tank on vessel with 2 skid beams by uniform loading

67

The total translation in the model is 0.77m. The highest equivalent stress found was about

149.8Mpa.

Fig 7.14 shows the total deformation of the model after the analysis. The scale of the results is

increased by 10% of the actual value to visualize the actual deformation of the model. The red

line shows the draft line of the vessel or the water surface level

Fig 7.14 Scaled total deformation of the analysis step uniform loading with 2 skid beams

Total translation Highest equivalent stress

uniform loading case

4 skid beams 0.82m 178.983Mpa

Table 7.6 Summary of results of uniform loading cases

During the analysis of all the 3 load steps of the uniform loading, namely 4 skid beams load-out, 3

skid beams load-out and 2 skid beams load-out, the highest equivalent stresses are found in

almost same locations on the vessel model. Fig 7.15 shows the location where the highest

equivalent stresses are found in vessel model.

68

Fig. 7.15 Location of highest equivalent stresses in uniform loading case analysis.

The results of the stress components in the locations mentioned above, on all the 3 steps

analyzed are presented below;

-35.660 -85.429 -45.901 108.831

skid beam #2

Longitudinal bulkhead under

-67.105 -20.385 53.889 110.738

skid beam #3

Stern section under skid

-76.254 -12.952 -94.939 178.983

beam #4

Transverse bulkhead #1 on

-10.033 -90.918 -72.358 152.190

webframe #46

Transverse bulkhead #5 on

-113.543 16.072 1.180 122.391

webframe #83

Hull side shell on starboard

-29.673 -8.366 64.880 115.459

side

Table 7.7 Highest equivalent stresses of the analysis step hard tank on vessel with 4 skid beams.

The highest equivalent stress of 178.983MPa is more than the allowable stress of 164.3MPa,

which suggests that the load-out with 4 skid beams using uniform loading might not be a solution.

But the factors affecting the results can be listed as follows:

1. The peak stress is found on a location of a duct which intersects a longitudinal frame in the

stern to connect two sections to form a ballast tank. And also it is found on a triangular element,

whose results are not so accurate as compared to quadrilateral elements.

2. The critical stress is found exactly under the skid beam no. 4, in location on the deck

where there is no major bulkhead present.

69

The stress plot figures of the stress components location listed in the table 7.7 has been attached

in the appendix no.4.

-31.885 -71.238 -36.686 92.439

skid beam #2

Longitudinal bulkhead under

-116.027 -28.710 33.522 151.215

skid beam #3

Stern section under skid

-16.480 -22.310 72.583 127.305

beam #4

Transverse bulkhead #1 on

-3.159 -84.404 -64.353 138.894

webframe #46

Transverse bulkhead #5 on

-113.197 16.074 1.187 122.048

webframe #83

Hull side shell on starboard

-19.359 -7.222 66.923 117.146

side

Table 7.8 Highest equivalent stresses of the analysis step hard tank on vessel with 3 skid beams

The highest equivalent stress of 151.2Mpa is well below the allowable stress of 164.3MPa, so the

critical step of loading-out hard tank of the SPAR on the vessel with 3 skid beams arrangement is

possible with out causing any problems to the structure.

The highest stress of 151.215Mpa is found in the longitudinal bulkhead under the skid beam

number 3, where there is a discontinuity in the bulkhead in the stern section above the aft

perpendicular to interconnect the sections to form a ballast tank.

The stress plot figures of the locations listed above has been attached in the appendix no.5.

-38.783 -73.229 -44.544 99.895

skid beam #2

Longitudinal bulkhead under

-111.346 -28.435 59.994 144.293

skid beam #3

Stern section under skid

0.161 -10.503 71.040 123.499

beam #4

Transverse bulkhead #1 on

-6.366 -90.168 -69.037 147.969

webframe #46

Transverse bulkhead #5 on

16.127 -113.310 1.205 122.193

webframe #83

Hull side shell on starboard

-30.198 -8.592 68.257 121.258

side

Table 7.9 Highest equivalent stresses of the analysis step hard tank on vessel with 2 skid beams

70

The highest equivalent stress of 147.969Mpa is well below the allowable stress of 164.3MPa, in

contrary to the expected results the load-out of SPAR hard tank with 2 skid beams turn out to be

possible by uniform loading as the complete load is transferred to the longitudinal bulkheads

which is capable of withstanding heavy loads.

Since the load in skid beams are higher, the lowest stress among the stress concentration

location is found on the longitudinal bulkhead under skid beam #2, which is also the case in 3 and

4 skid beams load-outs as well. So the stress concentration on the model reveals that the load-

out using 2 skid beam by uniform loading is very well possible as of the 3 skid beam step.

The stress plot figures of the locations listed above has been attached in the appendix no.6.

The analysis and results of uniform loading with application of hydraulics to level out the peak

loads were discussed in the previous chapter, which doesn’t make use of the SPAR model. In this

chapter the SPAR model is loaded on the vessel deck, to analyze the load-out by standard

skidding method without hydraulics. The necessary details such as constraints, draft, and ballast

condition are discussed in detail.

1. hard tank on vessel to decide whether the load-out has to be carried out using hydraulics

2. Entire SPAR on vessel to see the overall stress in the vessel after the load-out.

Each of the above steps are further subdivided into load-out with 3 and 4 skid beams to decide on

the number of skid beams needed for distribution of the SPAR load during load-out.

Whole length of the SPAR hard tank model is loaded on the vessel model and analyzed to find

the stress components on the vessel structure.

The SPAR model is moved on top of the vessel model such that the entire hard tank is above the

skid beams and aligned so that the SPAR supports center lines are in same plane with the skid

beams center lines.

During the skidding method the contact between the SPAR support and skid beams are by

means of a roller as discussed in the chapter 4.3.1.4. The rollers are used to reduce the friction

between the surfaces for the facilitation of easy movement of the load. The roller contact on the

skid beam and SPAR support surfaces are usually a line contact.

The distance between the skid beams and the SPAR supports are maintained at about 1.5m as

the nodes are located on the center of the beam elements cross section. The height of the

support I-beam section below the node is 1.5m.

Since the friction does not comes in to play that much during the load-out process, the need for

gap elements can be ruled out.

The nodes on the SPAR’s support are edited to match the nodes on the center line of the skid

beams, so that both line of nodes are in the same plane.

71

The nodes on the supports base and the nodes on the skid beams center are connected using

rod elements. The area of cross section of the rod is chosen according the area of contact

surface between the skid beams and SPAR supports.

Rod elements are axial elements which only take tension or compression loads and translates the

loads to the connecting nodes. So the rod elements are very much suitable to simulate the rollers

used during the load-out by skidding method.

Details about the rod element are depicted in the table below.

2

Connecting rod Rod Circular rod 2.25m 0

Table 7.10 Properties of rod elements

Fig 7.16 shows rod elements connecting the nodes on the SPAR supports and skid beam.

Fig 7.16 rod elements connecting nodes on supports and skid beams

All the constraints in the vessel model as explained in the chapter 7.2.2 for uniform loading

condition are also utilized for the analysis using SPAR model. In addition to the constraints on the

vessel model following constraints are also created on the SPAR model for the analysis of the

step hard tank on vessel with SPAR model loading.

The supports under the hard tank are not constrained as the load of the hard tank has to be

transferred to the skid beam for loading the vessel structure.

All nodes on the supports under the heave plates are constrained in all 6 degrees of freedom.

During load-out when the hard tank supports are on the vessel the heave plate supports will still

be on the keyside, so it can’t move or rotate in any direction.

Fig 7.17 shows the constraints of the supports under the heave plates.

72

Fig 7.17 Constraints of the SPAR supports under the heave plates and soft tank

The soft tank doesn’t have any permanent supports under it as it will be hanging out of the vessel

during transportation, but during the load-out it will be supported on the skid beams using

temporary supports until the SPAR is in the final stowage position on the vessel deck. This is

explained in chapters 4.4.5 & 6.2.3.5

All the nodes on the soft tank are fixed in all 6 degrees of freedom as it will be on the skid beams

and all its movements will be locked. The constraints in the soft tank can be visualized in the

figure 7.17.

The draft and ballast condition are same as the uniform loading case. The mass of the hard tank

is same in both the cases only difference is that the load is applied using hydraulics to level out

the peak load and distribute it evenly in uniform loading case. But in this case the actual SPAR

model is used in loading.

The draft of the vessel is 10.95m and the ballast condition is exactly the same as previous

loading case. Please refer to chapter 7.2.3 and 7.2.4 for further details

The following section describes the analysis and results of the steps hard tank on vessel with 4

skid beams, hard tank on vessel with 3 skid beams and hard tank on vessel with 2 skid beams by

SPAR model loading case.

The SPAR model is loaded on the Vessel model so that the hard tank of the SPAR is completely

on the vessel stern and the centre line of the supports is aligned with the centre line of the skid

beams. Nodes on the supports are connected with nodes on skid beams using rod elements.

73

Hard tank on vessel with 4 skid beams

Fig 7.18 shows the analysis step hard tank on vessel with 4 skid beams.

Fig. 7.18 FE model of analysis step hard tank on vessel with 4 skid beams by SPAR model loading

The total translation in the model is 0.437m. The highest equivalent stress found was about

130.575Mpa.

Fig 7.19 shows the total deformation of the model after the analysis. The scale of the results is

increased by 10% of the actual value to visualize the actual deformation of the model. The red

line shows the draft line of the vessel or the water surface level. The total deformation includes

the deformation of the SPAR as well.

Fig 7.19 Scaled total deformation of the analysis step hard tank on vessel with 4 skid beams

74

Fig. 7.20 shows the plot of total translation in Z-axis verses its X-axis coordinates. The resulting

plot reveals the bending of the whole vessel model.

Fig. 7.20 Plot of vessel model translation in Z-axis for the step hard tank on vessel with 4 skid beams

Fig 7.21 shows the analysis step hard tank on vessel with 3 skid beams.

Fig. 7.21 FE model of analysis step hard tank on vessel with 3 skid beams by SPAR model loading

The total translation in the model is 0.41m. The highest equivalent stress found was about

195.718Mpa.

75

Fig 7.22 shows the total deformation of the model after the analysis. The scale of the results is

increased by 10% of the actual value to visualize the actual deformation of the model. The red

line shows the draft line of the vessel or the water surface level. The total deformation includes

the deformation of the SPAR as well.

Fig 7.22 Scaled total deformation of the analysis step hard tank on vessel with 3 skid beams

Fig. 7.23 shows the plot of total translation in Z-axis verses its X-axis coordinates. The resulting

plot reveals the bending of the whole vessel model.

Fig. 7.23 Plot of vessel model translation in Z-axis for the step hard tank on vessel with 3 skid beams

Fig 7.24 shows the analysis step hard tank on vessel with 2 skid beams.

76

Fig. 7.24 FE model of analysis step hard tank on vessel with 2 skid beams by SPAR model loading

The total translation in the model is 0.425m. The highest equivalent stress found was about

180.230Mpa.

Fig 7.25 shows the total deformation of the model after the analysis. The scale of the results is

increased by 10% of the actual value to visualize the actual deformation of the model. The red

line shows the draft line of the vessel or the water surface level. The total deformation includes

the deformation of the SPAR as well.

Fig 7.25 Scaled total deformation of the analysis step hard tank on vessel with 2 skid beams

Fig. 7.26 shows the plot of total translation in Z-axis verses its X-axis coordinates. The resulting

plot reveals the bending of the whole vessel model.

77

Fig. 7.26 Plot of vessel model translation in Z-axis for the step hard tank on vessel with 2 skid beams

Summary of the results of hard tank on vessel by SPAR model loading cases

Total translation Highest equivalent stress

model loading

4 skid beams 0.437m 130.575Mpa

Table 7.11 Summary of results of hard tank on vessel by SPAR model loading cases

Since it cannot be confirmed by viewing the model whether the vessel structure is loaded as

expected, which is, loading by compression forces due the SPAR on top, the stresses in the

connecting rod are evaluated. The tables and figures below show the magnitude and plot of all

the axial forces on the rods, these forces are the loads on the skid beams. The following section

explains the loads on the skid beam in each load step analyzed.

The axial forces on the rods are summed up and the total load acting on each skid beam is given

in the table below.

Force in Newton Load in tons

Skid beam #1 46146891.1 4705.673

Skid beam #2 93538263.1 9538.248

Skid beam #3 80934120.3 8252.983

Skid beam #4 45070108 4595.871

Total load 27092.775 tons

Table 7.12 Load on skid beams in step hard tank on vessel with 4 skid beams

78

Fig. 7.27 Axial forces in rod elements on skid beam #1 – hard tank on vessel with 4 skid beams

Fig. 7.28 Axial forces in rod elements on skid beam #2– hard tank on vessel with 4 skid beams

79

Fig. 7.29 Axial forces in rod elements on skid beam #3– hard tank on vessel with 4 skid beams

Fig. 7.30 Axial forces in rod elements on skid beam #4– hard tank on vessel with 4 skid beams

The axial forces on the rods are summed up and the total load acting on each skid beam is given

in the table below.

Skid beam #1 50832664.4 5183.48

Skid beam #2 82246036.6 8386.76

Skid beam #3 142107838 14490.96

Total load 28061.20 tons

Table 7.13 Load on skid beams in step hard tank on vessel with 3 skid beams

Fig. 7.31 Axial forces in rod elements on skid beam #1 - hard tank on vessel with 3 skid beams

80

Fig. 7.32 Axial forces in rod elements on skid beam #2 - hard tank on vessel with 3 skid beams

Fig. 7.33 Axial forces in rod elements on skid beam #3 - hard tank on vessel with 3 skid beams

The axial forces on the rods are summed up and the total load acting on each skid beam is given

in the table below.

Skid beam #2 146159651 14904.13

Skid beam #3 130400410 13297.14

Total load 28201.27 tons

Table 7.14 Load on skid beams in step hard tank on vessel with 2 skid beams

81

Fig. 7.34 Axial forces in rod elements on skid beam #2 - hard tank on vessel with 2 skid beams

Fig. 7.35 Axial forces in rod elements on skid beam #3 - hard tank on vessel with 2 skid beams

Summary of load skid beams in hard tank on vessel by SPAR model loading case.

beam #1 beam #2 beam #3 beam #4

4 skid beams load-out 4705.673 9538.248 8252.983 4595.871

3 skid beams load-out 5183.48 8386.76 14490.96

2 skid beams load-out 14904.13 13297.14

Table 7.15 Summary of load on skid beams in step hard tank on vessel by SPAR model loading

Negative sign indicates that the loading is compression and positive sign is tension. It is clear

from the figures below that the forces in the rods are compressive and the vessel structure is

loaded as expected.

82

The tensional forces on few nodes at the end of SPAR is due to the bending of the SPAR and

vessel, in reality there will be minimal contact in those areas due to the bending. But in model it is

translated into a tensional force as the nodes are connected. The small forces are due to the

bending of the model can be neglected as it doesn’t create any high stresses in the structure of

the vessel model.

During the analysis of all the 3 load steps of hard tank on vessel by SPAR model loading, namely

4 skid beams load-out, 3 skid beams load-out and 2 skid beams load-out, the highest equivalent

stresses are found in almost same locations on the vessel model. Fig 7.36 shows the location

where the highest equivalent stresses are found in vessel model

Fig. 7.36 Location of highest equivalent stresses in hard tank on vessel by SPAR model loading case analysis.

The results of the stress components in the locations mentioned above, on all the 3 steps

analyzed are presented below;

-44.086 -93.621 -53.116 122.658

skid beam #2

Longitudinal bulkhead under

-121.474 -31.967 1.707 109.103

skid beam #3

Transverse bulkhead #1 on

-14.255 -75.791 63.725 130.575

webframe #46

Transverse bulkhead #5 on

-118.907 16.810 1.394 128.159

webframe #83

-31.512 -8.500 60.363 108.298

side

Table 7.16 Highest equivalent stresses of the analysis step hard tank on vessel with 4 skid beams by SPAR

model loading

83

The highest equivalent stress of 130.575Mpa is well below the allowable stress of 164.3MPa, it is

possible to load-out the SPAR using 4 skid beam without the need for hydraulics for leveling the

peak loads. The highest stress is found on the transverse bulkhead stiffeners, where, on the

backside of the stiffeners the engine room section is present. The stress plot figures of the

locations listed above has been attached in the appendix no.7.

-42.570 -89.147 -50.920 117.231

skid beam #2

Longitudinal bulkhead under

-215.916 -55.936 -14.541 195.718

skid beam #3

Transverse bulkhead #1 on

-8.561 -61.450 -53.632 109.328

webframe #46

Transverse bulkhead #5 on

-119.407 16.820 1.391 128.667

webframe #83

-31.629 -8.561 61.448 110.139

side

Table 7.17 Highest equivalent stresses of the analysis step hard tank on vessel with 3 skid beams by SPAR

model loading

The highest equivalent stress of 195.718Mpa is above the allowable stress of 164.3MPa, the load

imposed on the vessels structure causes high stress concentration, it is not possible to load-out

the SPAR using 3 skid beam without the hydraulics for leveling the peak loads. The high stress

concentration is found on the longitudinal bulkhead under skid beam #3, where there is a opening

in the bulkhead for cables and pipes. The stress plot figures of the locations listed above has

been attached in the appendix no.8.

-111.042 -27.874 63.218 148.330

skid beam #2

Longitudinal bulkhead under

-197.698 -51.111 -17.229 180.230

skid beam #3

Transverse bulkhead #1 on

-11.310 -63.624 -55.042 112.006

webframe #46

Transverse bulkhead #5 on

-119.197 16.764 1.375 128.424

webframe #83

-31.431 -8.555 61.747 110.591

side

Table 7.18 Highest equivalent stresses of the analysis step hard tank on vessel with 2 skid beams by SPAR

model loading

84

The highest equivalent stress of 180.230Mpa is above the allowable stress of 164.3MPa, the load

imposed on the vessels structure causes high stress concentration, it is not possible to load-out

the SPAR using 3 skid beam without the hydraulics for leveling the peak loads. The high stress

concentration is found on the longitudinal bulkhead under skid beam #3, where there is an

opening in the bulkhead for cables and pipes, in the location similar to the 3 skid beam load-out.

The stress plot figures of the locations listed above has been attached in the appendix no.9.

Entire length of the SPAR is loaded on the vessel model and analyzed to find the stress

components on the vessel structure.

The main purpose of analysis of this position of SPAR is to check the capability of the vessel

structure in the final stowage position for voyage. Whether the SPAR is loaded-out using

standard skidding method or uniform loading by using hydraulics, once the SPAR is in the final

stowage position on the vessel deck, the hydraulic cylinders (in case of uniform loading) and the

other equipments used for load-out under the supports will be removed and the SPAR will be

rested on the vessel deck directly. The skid beams may or may not be present under the supports

for final voyage. Then the SPAR will be sea fastened in the final position to prevent the SPAR

from moving during voyage. During this step of the analysis, suitable position for the stowage of

the SPAR for voyage conditions is determined.

As per the data provided by the client regarding the design of the SPAR, the mass of hard tank

with outfittings and supports is 25122 tons which is the critical load acting on the vessel structure.

The loads exerted by the supports under heave plates is, 2958 tons for the upper heave plate and

2957 tons for the lower heave plates when the soft tank is still resting on the key side.

Until the SPAR reaches the final position the load-out is carried by one of the suitable steps

explained in the previous topics (uniform loading or standard skidding). The soft tank will still be

supported on the keyside until the SPAR reaches its final position on deck, so the loads will be

less on the supports under the truss.

The total load on the vessel structure under the heave plates supports is 5915.4 tons when the

SPAR reaches its final position on the vessel deck. During this step of the analysis, the vessel

structure’s capability to withstand the loads due to the heave plates supports is also investigated,

but the load acting on the vessel structure at any unit area due to the combined mass of truss and

heave plates will always be less than the hard tank load when the soft tank is supported on the

key side. So the need to analyze the vessel structure until the SPAR reaches its final position

when the soft tank is fixed is ruled out.

When the SPAR reaches the final stowage position, the SPAR is sea fastened for its voyage and

the vessel’s draft is decreased by discharging the ballast water so that the soft tank is lifted off its

temporary supports from the keyside and left hanging over the vessel stern. Only during this step

the whole weight of the SPAR will be acting on the vessel deck. Especially the loads on the

supports under the truss and heave plates will be increased due to the additional weight of the

soft tank.

In this section the final stowage position for voyage is analyzed to find a suitable position for the

SPAR on vessel deck using 3 different skid beams configuration namely;

2. Load-out using 3 skid beams

3. Load-out using 2 skid beams

85

7.3.2.1 Position of the SPAR

The SPAR model is moved on top of the vessel model such that the entire SPAR is above skid

beams and the hard tank is at the end of the skid beams, the SPAR is aligned so that the SPAR

supports center lines are in same plane with the skid beams center lines. Fig 7.37 shows the

position of hard tank at the end of the skid beams near the bow section of the vessel model.

Fig 7.37 Final position of SPAR hard tank supports on skid beams

It is very clear from the figure 7.37 that the skid beam #1 is at the maximum length of 163.2m on

the vessel deck. It cannot be extended further on the deck to increase the length of the skid

beams, as the vessel structure is converging toward the bow. But the skid beams #2, #3 and #4

can be extended further to increase its length on deck.

Fig. 7.38 shows the position of the truss supports on vessel deck. The SPAR is located almost on

the edge of the stern section.

86

7.3.2.2 Connecting SPAR with skid beams

The connection between the SPAR and skid beams are made using rod elements as explained in

chapter 7.3.1.1.

All the constraints in the vessel model as explained in the chapter 7.2.2 are also utilized for the

analysis using SPAR model. In addition to the constraints on the vessel model following

constraints are also created on the SPAR model for the analysis of the step entire SPAR on

vessel by SPAR model loading.

The supports under the heave are connected to the skid beam using rod elements. The bottom of

the support is fixed in x and y axis to prevent further movement of the SPAR, as the loading will

only be in the z-axis. It is also needed to match the constraints on the vessel stern. Fig 7.70

shows the constraints in the supports under the heave plates.

Fig 7.39 Constraints in supports under heave plates for the step entire SPAR on vessel

The draft of the vessel is same as the other loading cases, which is 10.95m.

The ballast masses will not be the same as the step hard tank on vessel, since the loading is

done with the entire SPAR. This step of the analysis is done for final position of the SPAR on

deck. This analysis step is for the final stowage position of SPAR when the whole weight of the

SPAR (34855 tons) is acting on the vessel structure.

But the ballast condition used for the analysis is when the soft tank is still resting on the keyside

on its temporary supports; this is due to the reason that the ballast calculation was readily

available at the time of the thesis. The total mass of SPAR on the vessel will be 31038 Tons, and

is taken into consideration for calculating the ballast masses required to maintain the vessel in the

required draft of 10.95m.

87

The model will be translating a lot longitudinally from bow to stern due to the difference in the

mass of SPAR used for ballast calculation and the whole SPAR. But this does not affect the

analysis in anyway as the entire SPAR is on the vessel deck and is free to move in all degrees of

freedom except the X and Y axial movements. So the total load of SPAR will be acting on the

vessel model and the vessel structure will be loaded as needed.

Table 7.19 shows the mass of ballast water required in each tank for maintaining the vessel draft

for step entire SPAR on vessel when the soft tank is still supported on the keyside.

1 upper S WBT1U.S 839760 0 0

1 upper P WBT1U.P 851280 0 0

1 upper C WBT1U.C 624050 6.2 38691.1

2 upper S WBT2U.S 839340 100 839340

2 upper P WBT2U.P 861430 100 861430

2 upper C WBT2U.C 839210 0 0

3 upper S WBT3U.S 861430 100 861430

3 upper P WBT3U.P 850380 100 850380

3 upper C WBT3U.C 861290 100 861290

4 upper S WBT4U.S 850840 0 0

4 upper P WBT4U.P 0 0

4 upper C WBT4U.C 839210 0 0

5 upper S WBT5U.S 862475 29.5 254430.1

5 upper P WBT5U.P 849650 28.7 243850

5 upper C WBT5U.C 861910 97.4 839210

6 upper S WBT6U.S 585330 100 585330

6 upper P WBT6U.P 607420 100 607420

6 upper C WBT6U.C 585240 100 585240

1 middle S WBT1M.S 2444290 0 0

1 middle P WBT1M.P 2444290 0 0

1 middle C WBT1M.C 846690 0 0

2 middle S WBT2M.S 2467700 30.6 756070

2 middle P WBT2M.P 2487600 17.4 432870

FO3.C

3 middle S WBT3M.S 0 0

3 middle P WBT3M.P 2467700 0 0

3 middle C WBT3M.C 0 0

4 middle S WBT4M.S 2467710 100 2467710

4 middle P WBT4M.P 2487610 100 2487610

4 middle C WBT4M.C 2321520 100 2321520

5 middle S WBT5M.S 2487600 0 0

5 middle P WBT5M.P 2467700 0 0

5 middle C WBT5M.C 2332400 0 0

6 middle S WBT6M.S 1711630 100 1711630

6 middle P WBT6M.P 1732380 100 1732380

6 middle C WBT6M.C 1588290 100 1588290

1 lower S WBT1L.S 1214410 100 1214410

1 lower P WBT1L.P 1214410 100 1214410

1 lower C WBT1L.C 1116960 100 1116960

2 lower S WBT2L.S 1302220 100 1302220

88

2 lower P WBT2L.P 1307690 100 1307690

2 lower C WBT2L.C 1325950 100 1325950

3 lower S WBT3L.S 1309480 100 1309480

3 lower P WBT3L.P 1304220 100 1304220

3 lower C WBT3L.C 1327370 100 1327370

4 lower S WBT4L.S 1303230 0 0

4 lower P WBT4L.P 1310214 0 0

4 lower C WBT4L.C 1325897 0 0

5 lower S WBT5L.S 1309430 0 0

5 lower P WBT5L.P 1309430 100 1309430

5 lower C WBT5L.C 1328750 0 0

6 lower S WBT6L.S 782700 0 0

6 lower P WBT6L.P 788170 0 0

6 lower C WBT6L.C 789920 0 0

water bal oflow WBTOVERF.S 0

FO1C.S 0

FO2.S 0

FO2.P 0

DOA.P 0

LO.P 0

FWTFW.S 0

FWTFWD.S 0

FWTAFWD.P 0

FWTAFW.S 0

fwd lower WB P WBTFL.P 0 0

fwd lower WB S WBTFL.S 1592033 0 0

fwd upper WB S WBTFU.S 0 0

fwd upper WB P WBTFU.P 0 0

aft WB P WBTA.P 0 0

poop aft WB S WBTPA.S 0 0

poop fwd WB S WBTPF.S 0 0

engine RM TOP S WBTERTOP.S 0 0

engine RM TOP P WBTERTOP.P 0 0

aft ballast S WBTALWL.S 0 0

aft peak C WBTAPC.S 0 0

lower peak C WBTLOWPEAK.C 0 0

high peak C WBTHIGHPEAK.C 2995970 0 0

WB overf fwd WBTOVERFWD.C 0 0

aft poop lower WB P WBTAPL.P 0 0

aft poop upper WB P WBTAPU.P 0 0

1 outer wing P WBT10.P 1291440 0 0

1 outer wing S WBT10.S 1291470 0 0

2 outer wing P WBT20.P 2708360 0 0

2 outer wing S WBT20.S 2708360 0 0

3 outer wing P WBT30.P 3168290 0 0

3 outer wing S WBT30.S 3168290 80.4 2546810

4 outer wing P WBT40.P 3262750 6.5 210730

4 outer wing S WBT40.S 3262750 75 2447063

89

5 outer wing P WBT50.P 3276980 75.2 2464040

5 outer wing S WBT50.S 3276980 100 3276980

6 outer wing P WBT60.P 2126910 0 0

6 outer wing S WBT60.S 2126900 100 2126900

7 outer wing P WBT70.P 2596050 0 0

7 outer wing S WBT70.S 1859960 100 1859960

STORES CREW

STORES ENGINE

Total Mass 48590744kg

Table 7.19 Ballast condition for the step entire SPAR on vessel

The analysis and summary of the results of the step entire SPAR on vessel with 4 skid beam, 3

skid beam and 2 skid beam load-out have been explained below.

Fig 7.40 show the FE model of the step entire SPAR on vessel with 4 skid beam. It has been

modeled with the all the necessary steps and boundary conditions explained earlier in this

chapter.

Fig 7.40 FE model of analysis step entire SPAR on vessel with 4 skid beams.

The above FE model of the step entire SPAR on vessel with 4 skid beams is analyzed by linear

static analysis and the results of the analysis is given below.

The total translation in the model is 12.56m. The highest equivalent stress found was about

225.169Mpa.

The total translation of the model after the analysis can be seen in the fig 7.41. The high

deformation is due to the reason that the ballast condition used is without the mass of the soft

tank (i.e, when the soft tank is still supported on the keyside, which is already explained in the

section draft and ballast condition).

90

So it is very clear from the model that the vessel is tilted towards the stern in Y-axis due to the

additional weight of the soft tank.

Fig 7.41 Total deformation of the analysis step entire spar on vessel with 4 skid beams

Fig 7.42 Axial forces on the rod elements in the analysis step entire spar on vessel with 4 skid beams

Fig. 7.42 shows the axial forces on the rod elements in the analysis step entire SPAR on vessel

with 4 skid beams. It is clear from the figure that the vessel structure is loaded with the entire

weight of the SPAR. There are no tensional forces on any of the rod elements, all the forces are

compressive.

Fig 7.43 show the FE model of the step entire SPAR on vessel with 3 skid beam. It has been

modeled with the all the necessary steps and boundary conditions explained earlier in this

chapter.

91

Fig 7.43 FE model of analysis step entire SPAR on vessel with 3 skid beams.

The above FE model of the step entire SPAR on vessel with 3 skid beams is analyzed by linear

static analysis and the results of the analysis is given below.

The total translation in the model is 11.55m. The highest equivalent stress found was about

283.518MPa.

The total translation of the model after the analysis can be seen in the fig 7.44.

Fig 7.44 Total deformation of the analysis step entire spar on vessel with 3 skid beams

Fig. 7.45 shows the axial forces on the rod elements in the analysis step entire SPAR on vessel

with 3 skid beams. It is clear from the figure that the vessel structure is loaded with the entire

weight of the SPAR. There are no tensional forces on any of the rod elements, all the forces are

compressive.

92

Fig 7.45 Axial forces on the rod elements in the analysis step entire spar on vessel with 3 skid beams

Fig 7.46 show the FE model of the step entire SPAR on vessel with 2 skid beam. It has been

modeled with the all the necessary steps and boundary conditions explained earlier in this

chapter.

Fig 7.46 FE model of analysis step entire SPAR on vessel with 2 skid beams.

The above FE model of the step entire SPAR on vessel with 2 skid beams is analyzed by linear

static analysis and the results of the analysis is given below.

The total translation in the model is 12.46m. The highest equivalent stress found was about

284.992MPa.

The total translation of the model after the analysis can be seen in the fig 7.47.

93

Fig 7.47 Total deformation of the analysis step entire spar on vessel with 2 skid beams

Fig. 7.48 shows the axial forces on the rod elements in the analysis step entire SPAR on vessel

with 2 skid beams. It is clear from the figure that the vessel structure is loaded with the entire

weight of the SPAR. There are no tensional forces on any of the rod elements, all the forces are

compressive.

Fig 7.48 Axial forces on the rod elements in the analysis step entire spar on vessel with 2 skid beams

Summary of the results of entire SPAR on vessel by SPAR model loading cases

Total translation Highest equivalent stress

SPAR model loading

4 skid beams 12.56m 225.169MPa

Table 7.20 Summary of results of entire SPAR on vessel by SPAR model loading cases

94

7.3.2.6 Location of stress components

Fig 7.49 shows the location where the highest equivalent stresses are found in vessel model

during analysis of the entire SPAR on vessel by SPAR model loading case.

Fig. 7.49 Location of highest equivalent stresses in entire SPAR on vessel by SPAR model loading case analysis.

The results of the representative stress components in the locations mentioned above, on all the

3 steps analyzed are presented below;

In this case of the analysis, the skid beams has a lot of triangular elements due to the supports under the

truss. Since the support of the truss are not much easier to modify according to the nodes on the skid beam

as like the hard tank, due to the usage of tapered beam elements. The skid beam nodes have to be modified

to adapt to the nodes on the truss supports.

It is very much clear from the previous analysis cases, hard tank on vessel by uniform loading and hard tank

on vessel by SPAR model loading, that the stresses due to the critical load of the hard tank does not impose

any critical stresses on the skid beam elements if the elements are quadrilateral and follows the webframe

locations of the vessel. If the quadrilateral mesh of the skid beams are modified the stresses are very high

on the triangular elements. The stresses are highly localized to the triangular element, so the stresses on

the triangular elements of the skid beam can be safely omitted from the results of this analysis steps.

10.070 6.153 -51.746 90.058

along the skid beam # 1

Longitudinal bulkhead under

-8.515 -36.169 -74.498 133.127

skid beam #2

Longitudinal bulkhead under

-161.915 -41.453 99.126 225.169

skid beam #3

longitudinal sections under

-22.871 -24.677 78.155 137.449

skid beam # 4

Table 7.21 Highest equivalent stresses of the analysis step entire SPAR on vessel with 4 skid beams

95

The highest equivalent stress of 225.169Mpa is very high above the allowable stress of

164.3MPa; it is not possible to position the SPAR with 4 skid beam in this stowage position.

The highest stress is found on the longitudinal bulkhead, where there is a discontinuity in the

structure for interconnecting two sections to form a ballast tank.

The stress plot figures of the locations listed above has been attached in the appendix no.11.

20.072 2.268 -36.476 65.985

along the skid beam # 1

Longitudinal bulkhead under

-94.293 -14.261 -117.124 221.144

skid beam #2

Longitudinal bulkhead under

-264.736 -7.410 63.781 283.518

skid beam #3

longitudinal sections under

-12.964 -17.759 54.902 96.416

skid beam # 4

Table 7.22 Highest equivalent stresses of the analysis step entire SPAR on vessel with 3 skid beams

The highest equivalent stress of 283.518Mpa is very high above the allowable stress of

164.3MPa; it is not possible to position the SPAR with 3 skid beam in this stowage position.

The stresses above the allowable stress are found in longitudinal bulkhead under skid beam #2

and longitudinal bulkhead under skid beam #3.

The highest stress is found on the longitudinal bulkhead, where there is a discontinuity in the

structure for interconnecting two sections to form a ballast tank. The location of the critical stress

is found in same location as like the step entire SPAR on vessel with 4 skid beams.

The stress plot figures of the locations listed above has been attached in the appendix no.12.

-9.628 -4.402 -30.869 54.115

along the skid beam # 1

Longitudinal bulkhead under

-0.782 -0.940 -16.446 284.992

skid beam #2

Longitudinal bulkhead under

-174.418 -44.231 95.080 227.561

skid beam #3

longitudinal sections under

1.771 -10.414 78.844 137.037

skid beam # 4

Table 7.23 Highest equivalent stresses of the analysis step entire SPAR on vessel with 2 skid beams

The highest equivalent stress of 284.992Mpa is very high above the allowable stress of

164.3MPa; it is not possible to position the SPAR with 2 skid beam in this final stowage position.

96

The stresses above the allowable stress are found in longitudinal bulkhead under skid beam #2

and longitudinal bulkhead under skid beam #3.

The highest stress is found on the longitudinal bulkhead, where there is a discontinuity in the

structure for interconnecting two sections to form a ballast tank. The location of the critical stress

is found in same location as like the step entire SPAR on vessel with 4 and 3 skid beams.

The stress plot figures of the locations listed above has been attached in the appendix no.13.

It turns out that the steps analyzed in the case of entire SPAR on vessel for determining the final

stowage position of the SPAR on the vessel is not possible by any of the skid beam

configurations. The critical or failure stress is found in at least 2 locations of the longitudinal

bulkhead structure at the stern end of the vessel in all the 3 load-out steps. The location of the

critical stress is same in all 3 steps, so it is very clear that the structure in that location is weak for

taking heavy loads.

So the SPAR’s final stowage location has to be changed, The SPAR has to be stowed in a

position by avoiding the use of the particular stern section, it has to be moved more forward

longitudinally. The position chosen for the analysis steps of the case entire SPAR on vessel is

due to the maximum length of the skid beam #1, which is located almost on the edge of the port

side of the vessel. Even though there is 17.5m length of deck area in front of the SPAR, it is not

possible to load-out the SPAR up to that location by using 4 and 3 skid beams configuration due

to the restriction in length of the skid beam #1.

But the 2 skid beams configuration does not make use of skid beam #1, so it is possible to use

the 2 skid beam configuration to change the stowage position of the SPAR and move it forward

longitudinally towards the bow. The step of entire SPAR on vessel with the modified final stowage

position is analyzed in this section.

Fig 7.50 Final stowage position of SPAR hard tank supports on skid beams

97

The additional length of the skid beams are modeled until distance between the end of the deck

and the skid beam is about 5m. SPAR model is moved on top of the vessel model such that the

entire SPAR is above skid beams and the hard tank is at a distance of 5m from the end of the

deck, the SPAR is aligned so that the SPAR supports center lines are in same plane with the skid

beams center lines. Fig 7.50 shows the position of hard tank at the end of the skid beams near

the bow section of the vessel model.

Fig 7.51 Final stowage position of SPAR truss supports on skid beams

Fig. 7.51 shows the position of the truss supports on vessel deck. The SPAR location has been

moved about 12m from its previous position.

The connection between the SPAR and skid beams are made using rod elements as explained in

chapter 7.3.1.1.

The boundary conditions are all same as the analysis case of entire SPAR on vessel explained in

chapter 7.3.2.3.

The draft and ballast conditions are the same as explained in the chapter 7.3.2.4,

7.3.3.5 Analysis

The analysis and summary of the results of the final stowage position of the SPAR using 2 skid

beams have been explained below.

Fig 7.52 show the FE model of the step final stowage position of the SPAR with 2 skid beams. It

has been modeled with the all the necessary steps and boundary conditions explained above.

98

Fig 7.52 FE model of analysis step final stowage position of the SPAR with 2 skid beams.

The above FE model of the step final stowage position of the SPAR with 2 skid beams is

analyzed by linear static analysis and the results of the analysis is given below.

The total translation in the model is 7.003m. The highest equivalent stress found was about

145.239Mpa.

The total translation of the model after the analysis can be seen in the fig 7.53. The high

deformation is due to the reason that the ballast condition used is without the mass of the soft

tank (i.e, when the soft tank is still supported on the keyside, which is already explained in the

section draft and ballast condition). So it is very clear from the model that the vessel is tilted

towards the stern in Y-axis due to the additional weight of the soft tank.

Fig 7.53 Total deformation of the analysis step final stowage position of SPAR with 2 skid beams

99

Fig 7.54 Axial forces on the rod elements in the analysis step final stowage position of SPAR with 2 skid beams

Fig. 7.54 shows the axial forces on the rod elements in the analysis step final stowage position of

SPAR with 2 skid beams. It is clear from the figure that the vessel structure is loaded with the

entire weight of the SPAR. There are no tensional forces on any of the rod elements, all the

forces are compressive.

The location of the stress components are the same as explained in the chapter 7.3.2.6.

The results of the triangular elements on the skid beams are omitted.

-7.634 -3.546 -38.834 67.587

along the skid beam # 1

Longitudinal bulkhead under

-68.499 -0.934 -74.084 145.239

skid beam #2

Longitudinal bulkhead under

-35.314 -10.775 -69.754 124.819

skid beam #3

longitudinal sections under

-4.286 -59.916 -11.355 61.184

skid beam # 4

Table 7.24 Highest equivalent stresses of the analysis step final stowage position of SPAR with 2 skid beams

The highest equivalent stress of 145.239Mpa is well below the allowable stress of 164.3MPa; it is

possible to position the SPAR with 2 skid beam in this final stowage position.

The highest stress is found on the longitudinal bulkhead under skid beam #2, where the supports

under the lower heave plates are positioned. The high stress is right on the longitudinal bulkhead.

The stress plot figures of the locations listed above has been attached in the appendix no.14.

100

CHAPTER 8

CONCLUSION

Comparison and Load-out step selection

In the case of hard tank on vessel by uniform loading, it is very clear from the results that the

load-out is possible with 3 skid beams and 2 skid beams. Since the load on skid beam all along

the length of the beam is equal, there won’t be much difference in the loading condition through

out the load-out process. Load-out with 4 skid beams is also possible, the critical stress is found

on the location where there is a triangular element and it is highly localized, so it can be safely

neglected.

In the analysis case of load-out by standard skidding without hydraulics, only the 4 skid beam

load-out method proved to have well balanced stress distribution within the allowable values. But

the 3 and 2 skid beams load-out causes critical stresses in the longitudinal bulkhead section

under skid beam #3 where there is a duck opening for cables and piping in the vessel. It is clear

from the stress plots that the load distribution for the 3 and 2 skid beam load-out without

hydraulics could be harmful for the vessel structure.

The overall stress components of the load-out step by uniform loading using 2 skid beams turns

out to be the best method when taken into consideration the amount of work and also the

economical point of view. There is not much difference between values of the stress components

of the 2 skid beam uniform loading when comparing with the 3 and 4 skid beams load-out. The

reason for the little difference in the stress values is mainly due to the load distribution on the

supports on skid beam #1 and #4 based on its relation from the centre of gravity of the SPAR. In

the 2 skid beams load-out the load distribution is more even between the 2 supports as the #1

and #4 skid beams and supports are not present.

Load-out by SPAR model loading, i.e, without hydraulics could be the most affordable method.

But considering the enormous weight of the SPAR and the complexity of the weight distribution of

the SPAR when supported using more than 2 skid beams, it should be better avoided.

During the analysis of all the steps discussed in this thesis, the structure of the vessel on the

stern end appears to be weak as the critical stresses are found here; the reason for this is mainly

due to the water pressure under the vessel. The height of the vessel structure at the stern section

gradually decreases from the starting of the vessel stern, the height is lowest at the transom of

the vessel, due to the presence of propeller and rudder below the stern of vessel. This profiling in

the stern section provides the ability for the vessel to dock along the stern to load-out cargoes.

101

Since the height of the vessel structure is decreased and also the water pressure acting under

the stern section will be decreased, as the pressure will be more at the deepest part of the vessel.

This leads to a condition where the support provided by the water pressure under the stern is not

enough to compensate the loading on the deck of the stern. Thus leading to the high stresses in

the stern section especially at the transom of the vessel.

There is an increase in the stress concentration in the vessel structure during the final stowage

position of the SPAR, i.e., when the SPAR is positioned on the deck after removal of the hydraulic

cylinders and skid beam. The stress activity is concentrated on the corners of the supports

towards the soft tank side. This is mainly due to the stiffness of the SPAR, when the soft tank is

allowed to hang over the stern. Even though during the analysis it does not creates critical

stresses in the vessel structure, the loading at the support corners could be prevented by

providing additional supports at the end of hard tank and heave supports.

The SPAR can be loaded out using 2 skid beams by utilizing hydraulics to level out the peak

loads. The following points should be adhered to when the load-out is carried out.

1. Number of skid beams used is 2; the supports should be appropriately designed to match the

location of the skid beams.

2. The skid beams should be positioned on the longitudinal bulkheads.

3. The load-out should be carried by uniform loading by the use of hydraulic cylinders in

between the SPAR and skid beams to distribute the load equally on the skid beam.

4. During the whole load-out operation the soft tank should be temporarily supported on the skid

beams, until the SPAR reached the final stowage position as explained in chapter 7.3.3

5. When the SPAR reaches the final stowage position the hydraulic cylinders must be removed

and the SPAR maybe rested on the skid beam or directly on the vessel deck for sea

fastening and final voyage.

Further analysis

Since the whole analysis is carried out using the estimated design of the SPAR, when the design

details of the actual SPAR is known the vessel should be analyzed again with the exact SPAR

model to make sure the load-out of the real SPAR matches with the results of this thesis. If

possible the 3D model of the SPAR with all the exact details of the inner design of the SPAR

should be utilized to narrow down the assumptions used in this thesis concerning the SPAR

model.

The vessel structure should be investigated for buckling; to make sure the plate sections of the

vessel structure does not buckle during load-out of the SPAR. The vessel structure can be further

investigated and analyzed for the stowage position of the SPAR with 1.7m offset from the centre

of the vessel.

102

APPENDIX 1

SPAR DIMENSIONS

103

APPENDIX 2

Ballast plan of Blue Marlin

105

APPENDIX 3

Ballast calculation for the step hard tank on vessel

107

07/15/08 11:44:44 Dockwise Shipping B.V. Page 16

GHS 10.50 BLUE MARLIN LO03

REFERENCE 0611266, STEP 3

AT A DRAFT OF 10.95M @ APP, THE PROPELLER IMMERSION = 145%

DRAFTS SUMMARY

DRAFT @

PERPENDICULARS-----------DRAFT @ MARKS------------------------------

10.95M @

FPP 206.50f FROM APP 10.95M @ BOW 204.50f FROM APP

10.95M @

MID 103.25f FROM APP 10.95M @ MIDSHIPS 103.25f FROM APP

10.95M @

APP 0.00a FROM APP 10.95M @ RUDDER 3.50a FROM APP

10.95M @ ACCOM.BULKHEAD 174.70f FROM APP

10.95M @ CASING 10.50f FROM APP

------------------------------------------------------------------------------

HYDROSTATIC PROPERTIES

Trim: 0.00/206.50, Heel: Port 0.11 deg., VCG = 13.561

Draft----Weight(MT)----LCB-----VCB-------cm-----LCF---cm trim----GML-----GMT

10.949 113,840.16 99.088f 5.858 121.46 91.356f 1774.28 321.84 24.258

Distances in METERS.-----Specific Gravity = 1.025.-----------Moment in m.-MT.

Trim is per 206.50m.

Draft is from Baseline. Free Surface included. GMT is from RA curve.

Caution: Standard GMT is 24.253

WEIGHT STATUS

Trim: 0.00/206.50, Heel: Port 0.11 deg.

Part------------------------------Weight(MT)----LCG-----TCG-----VCG-------FSM

Total Fixed---------> 54,224.22 69.479f 3.390p 21.404

Total Tanks---------> 59,620.82 126.017f 3.002s 6.428 40485.0

Total Weight--------> 113,845.05 99.088f 0.042p 13.561

Free Surface Adjustment----------> 0.356

Adjusted CG----------------------> 99.088f 0.042p 13.917

Distances in METERS.-----------------------------------------Moments in m.-MT.

DISPLACEMENT STATUS

Baseline draft: 10.948 @ 206.50f, 10.949 @ 0.00

Trim: 0.00/206.50, Heel: Port 0.11 deg.

Part---------------------SpGr------Displ(MT)----LCB-----TCB-----VCB

HULL 1.025 113,840.16 99.088f 0.057p 5.858

PSCASING 1.025 0.00

Total Displacement--> 1.025 113,840.16 99.088f 0.057p 5.858

Distances in METERS.-----------------------------------------------

07/15/08 11:44:44 Dockwise Shipping B.V. Page 17

GHS 10.50 BLUE MARLIN LO03

REFERENCE 0611266, STEP 3

WEIGHT STATUS

Part------------------------------Weight(MT)----LCG-----TCG-----VCG

LIGHT SHIP+ 27,935.69 103.855f 0.255s 9.956

STARBOARD CASING 615.41 0.600f 26.130s 27.040

STARBOARD CASING ADDITION 79.30 3.500f 26.130s 15.800

PORT CASING 0.24 3.150f 26.196p 26.600

PORT CASING ADDITION 0.08 3.150f 26.196p 15.800

CREW AND EFFECTS 5.70 186.600f 0.000 31.910

PROVISIONS 7.69 189.750f 0.000 24.000

STORES FORE 23.45 194.330f 0.000 23.110

STORES AFT 19.19 15.170f 0.000 9.980

SMALL TANK IN E/R (FORE) 52.22 184.720f 0.000 15.410

SMALL TANK IN E/R (AFT) 362.65 22.890f 0.000 6.150

Skid load Hard Tank 25,122.60 33.445f 8.322p 34.250

Total Fixed---------> 54,224.22 69.479f 3.390p 21.404

Distances in METERS.-----------------------------------------------

TANK STATUS

Trim: 0.00/206.50, Heel: Port 0.11 deg.

Part------------Load-----SpGr-----Weight(MT)----LCG-----TCG-----VCG

SEA WATER 0.375 1.025 54,420.03 127.308f 3.236s 6.352

HEAVY FO 0.955 0.991 4,290.77 119.566f 0.537s 6.330

DIESEL OIL 0.980 0.850 308.37 13.696f 14.344p 9.871

LUBE OIL 0.980 0.900 33.69 18.585f 9.910p 6.444

FRESH WATER 1.000 1.000 567.96 118.466f 9.409s 12.569

WATER 1.001 0.000 1.001 0.00

Total Tanks---------> 59,620.82 126.017f 3.002s 6.428

Distances in METERS.-----------------------------------------------

LOAD STATUS

Trim: 0.00/206.50 Heel: Port 0.11 deg.

1 UPPER C WBT1U.C SEA 624.05 100.0 15.710

1 UPPER P WBT1U.P SEA 851.28 100.0 15.710

1 UPPER S WBT1U.S SEA 839.76 100.0 15.710

2 UPPER C WBT2U.C SEA 839.21 100.0 15.710

2 UPPER P WBT2U.P SEA 861.43 100.0 15.710

2 UPPER S WBT2U.S SEA 839.34 100.0 15.710

3 UPPER C WBT3U.C SEA 0.00 0.0 0.000

3 UPPER P WBT3U.P SEA 850.38 100.0 15.710

3 UPPER S WBT3U.S SEA 861.43 100.0 15.710

4 UPPER C WBT4U.C SEA 0.00 0.0 0.000

4 UPPER P WBT4U.P SEA 0.00 0.0 0.000

4 UPPER S WBT4U.S SEA 212.71 25.0 0.574

5 UPPER C WBT5U.C SEA 0.00 0.0 0.000

5 UPPER P WBT5U.P SEA 0.00 0.0 0.000

5 UPPER S WBT5U.S SEA 0.00 0.0 0.000

6 UPPER C WBT6U.C SEA 0.00 0.0 0.000

6 UPPER P WBT6U.P SEA 0.00 0.0 0.000

6 UPPER S WBT6U.S SEA 0.00 0.0 0.000

1 MIDDLE C WBT1M.C SEA 846.69 100.0 22.750

1 MIDDLE P WBT1M.P SEA 2,444.29 100.0 22.750

07/15/08 11:44:44 Dockwise Shipping B.V. Page 18

GHS 10.50 BLUE MARLIN LO03

REFERENCE 0611266, STEP 3

1 MIDDLE S WBT1M.S SEA 2,444.29 100.0 22.750

1 OUTER WING P WBT1O.P SEA 1,291.44 100.0 26.510

1 OUTER WING S WBT1O.S SEA 1,291.47 100.0 26.510

2 MIDDLE P WBT2M.P SEA 2,487.60 100.0 22.750

2 MIDDLE S WBT2M.S SEA 2,467.70 100.0 22.750

2 OUTER WING P WBT2O.P SEA 2,708.36 100.0 26.510

2 OUTER WING S WBT2O.S SEA 2,708.36 100.0 26.510

3 MIDDLE C WBT3M.C SEA 0.00 0.0 0.000

3 MIDDLE P WBT3M.P SEA 0.00 0.0 0.000

3 MIDDLE S WBT3M.S SEA 0.00 0.0 0.000

3 OUTER WING P WBT3O.P SEA 0.00 0.0 0.000

3 OUTER WING S WBT3O.S SEA 3,168.29 100.0 26.510

4 MIDDLE C WBT4M.C SEA 2,321.52 100.0 22.750

4 MIDDLE P WBT4M.P SEA 2,487.61 100.0 22.750

4 MIDDLE S WBT4M.S SEA 1,963.33 79.6 5.524

4 OUTER WING P WBT4O.P SEA 0.00 0.0 0.000

4 OUTER WING S WBT4O.S SEA 3,262.75 100.0 26.510

5 MIDDLE C WBT5M.C SEA 0.00 0.0 0.000

5 MIDDLE P WBT5M.P SEA 0.00 0.0 0.000

5 MIDDLE S WBT5M.S SEA 0.00 0.0 0.000

5 OUTER WING P WBT5O.P SEA 0.00 0.0 0.000

5 OUTER WING S WBT5O.S SEA 114.69 3.5 0.539

6 MIDDLE C WBT6M.C SEA 0.00 0.0 0.000

6 MIDDLE P WBT6M.P SEA 0.00 0.0 0.000

6 MIDDLE S WBT6M.S SEA 0.00 0.0 0.000

6 OUTER WING P WBT6O.P SEA 0.00 0.0 0.000

6 OUTER WING S WBT6O.S SEA 0.00 0.0 0.000

7 OUTER WING P WBT7O.P SEA 0.00 0.0 0.000

7 OUTER WING S WBT7O.S SEA 0.00 0.0 0.000

1 LOWER C WBT1L.C SEA 1,116.96 100.0 26.510

1 LOWER P WBT1L.P SEA 1,214.41 100.0 26.510

1 LOWER S WBT1L.S SEA 1,214.41 100.0 26.510

2 LOWER C WBT2L.C SEA 1,325.95 100.0 26.510

2 LOWER P WBT2L.P SEA 1,307.69 100.0 26.510

2 LOWER S WBT2L.S SEA 1,302.22 100.0 26.510

3 LOWER C WBT3L.C SEA 1,327.37 100.0 26.510

3 LOWER P WBT3L.P SEA 1,304.22 100.0 26.510

3 LOWER S WBT3L.S SEA 1,309.48 100.0 26.510

4 LOWER C WBT4L.C SEA 1,005.03 75.8 2.759

4 LOWER P WBT4L.P SEA 1,163.47 88.8 3.248

4 LOWER S WBT4L.S SEA 1,178.12 90.4 3.321

5 LOWER C WBT5L.C SEA 0.00 0.0 0.000

5 LOWER P WBT5L.P SEA 0.00 0.0 0.000

5 LOWER S WBT5L.S SEA 0.00 0.0 0.000

6 LOWER C WBT6L.C SEA 0.00 0.0 0.000

6 LOWER P WBT6L.P SEA 0.00 0.0 0.000

6 LOWER S WBT6L.S SEA 0.00 0.0 0.000

FWD LOWER WB P WBTFL.P SEA 0.00 0.0 0.000

FWD LOWER WB S WBTFL.S SEA 861.29 54.1 8.439

FWD UPPER WB P WBTFU.P SEA 0.00 0.0 0.000

07/15/08 11:44:44 Dockwise Shipping B.V. Page 19

GHS 10.50 BLUE MARLIN LO03

REFERENCE 0611266, STEP 3

FWD UPPER WB S WBTFU.S SEA 0.00 0.0 0.000

AFT WB P WBTA.P SEA 0.00 0.0 0.000

POOP AFT WB S WBTPA.S SEA 0.00 0.0 0.000

POOP FWD WB S WBTPF.S SEA 0.00 0.0 0.000

AFT L BALLAST S WBTALWL.S SEA 0.00 0.0 0.000

AFT PEAK C WBTAPC.C SEA 0.00 0.0 0.000

AFT POOP LOWER WB P WBTAPL.P SEA 0.00 0.0 0.000

AFT POOP UPPER WB P WBTAPU.P SEA 0.00 0.0 0.000

ENGINE RM TOP P WBTERTOP.P SEA 0.00 0.0 0.000

ENGINE RM TOP S WBTERTOP.S SEA 0.00 0.0 0.000

LOWER PEAK C WBTLOWPEAK.C SEA 0.00 0.0 0.000

HIGH PEAK C WBTHIGHPEAK.C SEA 0.00 0.0 0.000

WB OVRFL FWD WBTOVERFWD.C SEA 0.00 0.0 0.000

WATER BAL OFLOW WBTOVERF.S SEA 0.00 0.0 0.000

Total as loaded----------> 54,418.60

Soundings in m.---------------------------------------------------------------

07/15/08 11:44:44 Dockwise Shipping B.V. Page 20

GHS 10.50 BLUE MARLIN LO03

REFERENCE 0611266, STEP 3

SHELTERED CRITERION LONGITUDINAL STRENGTH SUMMARY at Heel = Port 0.11 deg.

Largest Bending Moment: 318,494 MT-m. at 69.100f (Hogging)

----------------m.------------MT-----%--------MT---------MT-m.----%------MT-m.

Fr.-5 3.500a 181.65 3 6,538.00 621.1 1 119,623.0

Fr.17 11.900f 4,783.89 81 5,888.00 39,799.8 45 89,073.8

Fr.29 20.300f 6,548.95 49 13,275.00 88,068.3 46 191,178.6

Fr.47 34.700f 7,211.03 39 18,703.00 191,501.1 58 329,713.2

Fr.53 49.700f 4,195.02 29 14,682.00 278,242.8 47 597,778.0

Fr.60 67.199f 378.64 3 15,015.00 318,162.2 48 664,318.4

Fr.63 74.700f -2,702.82 19 -14,124.00 311,780.5 45 686,598.2

Fr.73 99.700f -3,068.39 22 -14,124.00 240,033.8 38 635,562.0

Fr.83 124.700f -8,041.25 53 -15,272.00 100,923.0 16 635,669.8

Fr.93 149.700f -1,985.97 13 -14,869.00 -22,806.0 4 -533,195.0

Fr.103 174.700f 2,148.14 39 5,491.00 -12,794.3 5 -237,181.0

Fr.131 194.300f 522.35 12 4,421.00 6,218.8 9 66,269.0

------------------------------------------------------------------------------

SHELTERED CRITERION SUMMARY

Largest Shear: 99.9% at 17.500f

Largest Bending Moment: 65.1% at 39.700f (Hogging)

SHELTERED criterion LONGITUDINAL STRENGTH at 0.11 degrees PORT HEEL

GHS 10.50

07/15/08 11:44:44

300

250

200

2

150

100 3

General

BLUE MARLIN

50

1

Scale

-50

-100

-150

2

-200

-250

-300

Page 21

LO03

0 20 40 60 80 100 120 140 160 180 200

Aft <-- Location (M.) --> Fwd

07/15/08 11:44:44 Dockwise Shipping B.V. Page 22

GHS 10.50 BLUE MARLIN LO03

REFERENCE 0611266, STEP 3

Body @ 5.000f Body @ 30.000f Body @ 62.200f Body @ 90.000f Body @ 120.000f Body @ 175.000f

78

74

77 71

70

79 80 79 80 14 13 15 11 10 12 8 7 9

48 72 83 83

76 39 38 40 34 33 35 29 28 30

48 49 68 69

41 63 36 37 31 32

62 64 42 60 59 61 57 56 58

70 82

79 17 14 11 8 5 2

72 83

76

44 39 34 29 24 20 68 81

66 63 60 57 54 51

82

80 16 13 10 7 4 1

76 19

43 38 33 28 81

84 65 62 59 56 53 50

71

82

80 18 15 12 9 6 3

69

83

76

45 40 35 30 25 21 81

67 64 61 58 55 52

Plan View

77

78 48 46 41 36 31 26 22

72 17

44

66 63

39

14 60

34

11 57

29

8 54

24

5 2

20

51 70

68

79

76 65

43 16 62

38

13 33

10

59 28

56

7 53

4 19 50

1 83 81

82

84

80

67

45

18 64

40

15 61

35

12 58

30

9 55

25

6 52

21

3 69

71

75

49 47 42 37 32 27 23

73

74

1 WBT1U.C.......100% 18 WBT6U.S.........0% 35 WBT4M.S........80% 52 WBT1L.S.......100% 69 WBTFL.S........54%

2 WBT1U.P.......100% 19 WBT1M.C.......100% 36 WBT4O.P.........0% 53 WBT2L.C.......100% 70 WBTFU.P.........0%

3 WBT1U.S.......100% 20 WBT1M.P.......100% 37 WBT4O.S.......100% 54 WBT2L.P.......100% 71 WBTFU.S.........0%

4 WBT2U.C.......100% 21 WBT1M.S.......100% 38 WBT5M.C.........0% 55 WBT2L.S.......100% 72 WBTA.P..........0%

5 WBT2U.P.......100% 22 WBT1O.P.......100% 39 WBT5M.P.........0% 56 WBT3L.C.......100% 73 WBTPA.S.........0%

6 WBT2U.S.......100% 23 WBT1O.S.......100% 40 WBT5M.S.........0% 57 WBT3L.P.......100% 74 WBTPF.S.........0%

7 WBT3U.C.........0% 24 WBT2M.P.......100% 41 WBT5O.P.........0% 58 WBT3L.S.......100% 75 WBTALWL.S.......0%

8 WBT3U.P.......100% 25 WBT2M.S.......100% 42 WBT5O.S.........3% 59 WBT4L.C........76% 76 WBTAPC.C........0%

9 WBT3U.S.......100% 26 WBT2O.P.......100% 43 WBT6M.C.........0% 60 WBT4L.P........89% 77 WBTAPL.P........0%

10 WBT4U.C.........0% 27 WBT2O.S.......100% 44 WBT6M.P.........0% 61 WBT4L.S........90% 78 WBTAPU.P........0%

11 WBT4U.P.........0% 28 WBT3M.C.........0% 45 WBT6M.S.........0% 62 WBT5L.C.........0% 79 WBTERTOP.P......0%

12 WBT4U.S........25% 29 WBT3M.P.........0% 46 WBT6O.P.........0% 63 WBT5L.P.........0% 80 WBTERTOP.S......0%

13 WBT5U.C.........0% 30 WBT3M.S.........0% 47 WBT6O.S.........0% 64 WBT5L.S.........0% 81 WBTLOWPEAK.C....0%

14 WBT5U.P.........0% 31 WBT3O.P.........0% 48 WBT7O.P.........0% 65 WBT6L.C.........0% 82 WBTHIGHPEAK.C...0%

15 WBT5U.S.........0% 32 WBT3O.S.......100% 49 WBT7O.S.........0% 66 WBT6L.P.........0% 83 WBTOVERFWD.C....0%

16 WBT6U.C.........0% 33 WBT4M.C.......100% 50 WBT1L.C.......100% 67 WBT6L.S.........0% 84 WBTOVERF.S......0%

APPENDIX 4

Stress plots of the analysis step hard tank on vessel

with 4 skid beams by uniform loading

115

1. Stress components on longitudinal bulkhead under skid beam #2

116

3. Stress components on stern section under skid beam#4

117

5. Stress components on transverse bulkhead #5 on webframe #83

118

6. Stress components on hull side shell on starboard side

APPENDIX 5

Stress plots of the analysis step hard tank on vessel

with 3 skid beams by uniform loading

119

1. Stress components on longitudinal bulkhead under skid beam #2

120

3. Stress components on stern section under skid beam#4

121

5. Stress components on transverse bulkhead #5 on webframe #83

122

APPENDIX 6

Stress plots of the analysis step hard tank on vessel

with 2 skid beams by uniform loading

123

1. Stress components on longitudinal bulkhead under skid beam #2

124

3. Stress components on stern section under skid beam#4

125

5. Stress components on transverse bulkhead #5 on webframe #83

126

APPENDIX 7

Stress plots of the analysis step hard tank on vessel

with 4 skid beams by SPAR model loading

127

1. Stress components on longitudinal bulkhead under skid beam #2

128

3. Stress components on transverse bulkhead #1 on webframe #46

129

5. Stress components on hull side shell on port side

130

APPENDIX 8

Stress plots of the analysis step hard tank on vessel

with 3 skid beams by SPAR model loading

131

1. Stress components on longitudinal bulkhead under skid beam #2

132

3. Stress components on transverse bulkhead #1 on webframe #46

133

5. Stress components on hull side shell on port side

134

APPENDIX 9

Stress plots of the analysis step hard tank on vessel

with 2 skid beams by SPAR model loading

135

1. Stress components on longitudinal bulkhead under skid beam #2

136

3. Stress components on transverse bulkhead #1 on webframe #46

137

5. Stress components on hull side shell on port side

138

APPENDIX 10

Ballast calculation for the step entire SPAR on vessel

139

07/15/08 11:44:48 Dockwise Shipping B.V. Page 23

GHS 10.50 BLUE MARLIN LO04

REFERENCE 0611266, STEP 4

AT A DRAFT OF 10.96M @ APP, THE PROPELLER IMMERSION = 145%

DRAFTS SUMMARY

DRAFT @

PERPENDICULARS-----------DRAFT @ MARKS------------------------------

10.95M @

FPP 206.50f FROM APP 10.95M @ BOW 204.50f FROM APP

10.96M @

MID 103.25f FROM APP 10.96M @ MIDSHIPS 103.25f FROM APP

10.96M @

APP 0.00a FROM APP 10.96M @ RUDDER 3.50a FROM APP

10.95M @ ACCOM.BULKHEAD 174.70f FROM APP

10.96M @ CASING 10.50f FROM APP

------------------------------------------------------------------------------

HYDROSTATIC PROPERTIES

Trim: 0.00/206.50, Heel: Stbd 0.07 deg., VCG = 14.963

Draft----Weight(MT)----LCB-----VCB-------cm-----LCF---cm trim----GML-----GMT

10.956 113,932.59 99.080f 5.862 121.46 91.359f 1764.22 319.76 22.777

Distances in METERS.-----Specific Gravity = 1.025.-----------Moment in m.-MT.

Trim is per 206.50m.

Draft is from Baseline. Free Surface included. GMT is from RA curve.

Caution: Standard GMT is 22.760

WEIGHT STATUS

Trim: 0.00/206.50, Heel: Stbd 0.07 deg.

Part------------------------------Weight(MT)----LCG-----TCG-----VCG-------FSM

Total Fixed---------> 60,139.63 106.017f 3.875p 22.668

Total Tanks---------> 53,792.91 91.326f 4.396s 6.350 48691.6

Total Weight--------> 113,932.55 99.081f 0.030s 14.963

Free Surface Adjustment----------> 0.427

Adjusted CG----------------------> 99.081f 0.030s 15.391

Distances in METERS.-----------------------------------------Moments in m.-MT.

DISPLACEMENT STATUS

Baseline draft: 10.955 @ 206.50f, 10.957 @ 0.00

Trim: 0.00/206.50, Heel: Stbd 0.07 deg.

Part---------------------SpGr------Displ(MT)----LCB-----TCB-----VCB

HULL 1.025 113,932.59 99.080f 0.041s 5.862

PSCASING 1.025 0.00

Total Displacement--> 1.025 113,932.59 99.080f 0.041s 5.862

Distances in METERS.-----------------------------------------------

07/15/08 11:44:48 Dockwise Shipping B.V. Page 24

GHS 10.50 BLUE MARLIN LO04

REFERENCE 0611266, STEP 4

WEIGHT STATUS

Part------------------------------Weight(MT)----LCG-----TCG-----VCG

LIGHT SHIP+ 27,935.69 103.855f 0.255s 9.956

STARBOARD CASING 615.41 0.600f 26.130s 27.040

STARBOARD CASING ADDITION 79.30 3.500f 26.130s 15.800

PORT CASING 0.24 3.150f 26.196p 26.600

PORT CASING ADDITION 0.08 3.150f 26.196p 15.800

CREW AND EFFECTS 5.70 186.600f 0.000 31.910

PROVISIONS 7.69 189.750f 0.000 24.000

STORES FORE 23.45 194.330f 0.000 23.110

STORES AFT 19.19 15.170f 0.000 9.980

SMALL TANK IN E/R (FORE) 52.22 184.720f 0.000 15.410

SMALL TANK IN E/R (AFT) 362.65 22.890f 0.000 6.150

Skid load Hard Tank 25,122.60 128.255f 8.322p 34.250

Skid load UHP 2,958.39 55.961f 8.322p 34.250

Skid load LHP 2,957.01 20.608f 8.322p 34.250

Total Fixed---------> 60,139.63 106.017f 3.875p 22.668

Distances in METERS.-----------------------------------------------

TANK STATUS

Trim: 0.00/206.50, Heel: Stbd 0.07 deg.

Part------------Load-----SpGr-----Weight(MT)----LCG-----TCG-----VCG

SEA WATER 0.335 1.025 48,592.13 89.059f 4.807s 6.257

HEAVY FO 0.955 0.991 4,290.77 119.566f 0.543s 6.330

DIESEL OIL 0.980 0.850 308.37 13.694f 14.335p 9.871

LUBE OIL 0.980 0.900 33.69 18.585f 9.907p 6.444

FRESH WATER 1.000 1.000 567.96 118.466f 9.409s 12.569

WATER 1.001 0.000 1.001 0.00

Total Tanks---------> 53,792.92 91.326f 4.396s 6.350

Distances in METERS.-----------------------------------------------

LOAD STATUS

Trim: 0.00/206.50 Heel: Stbd 0.07 deg.

1 UPPER C WBT1U.C SEA 38.81 6.2 0.071

1 UPPER P WBT1U.P SEA 0.00 0.0 0.000

1 UPPER S WBT1U.S SEA 0.00 0.0 0.000

2 UPPER C WBT2U.C SEA 0.00 0.0 0.000

2 UPPER P WBT2U.P SEA 861.43 100.0 15.710

2 UPPER S WBT2U.S SEA 839.34 100.0 15.710

3 UPPER C WBT3U.C SEA 861.29 100.0 15.710

3 UPPER P WBT3U.P SEA 850.38 100.0 15.710

3 UPPER S WBT3U.S SEA 861.43 100.0 15.710

4 UPPER C WBT4U.C SEA 0.00 0.0 0.000

4 UPPER P WBT4U.P SEA 0.00 0.0 0.000

4 UPPER S WBT4U.S SEA 0.00 0.0 0.000

5 UPPER C WBT5U.C SEA 839.21 97.4 2.347

5 UPPER P WBT5U.P SEA 243.85 28.7 0.664

5 UPPER S WBT5U.S SEA 254.43 29.5 0.670

6 UPPER C WBT6U.C SEA 585.24 100.0 15.710

6 UPPER P WBT6U.P SEA 607.42 100.0 15.710

6 UPPER S WBT6U.S SEA 585.33 100.0 15.710

07/15/08 11:44:48 Dockwise Shipping B.V. Page 25

GHS 10.50 BLUE MARLIN LO04

REFERENCE 0611266, STEP 4

1 MIDDLE C WBT1M.C SEA 0.00 0.0 0.000

1 MIDDLE P WBT1M.P SEA 0.00 0.0 0.000

1 MIDDLE S WBT1M.S SEA 0.00 0.0 0.000

1 OUTER WING P WBT1O.P SEA 0.00 0.0 0.000

1 OUTER WING S WBT1O.S SEA 0.00 0.0 0.000

2 MIDDLE P WBT2M.P SEA 432.87 17.4 1.143

2 MIDDLE S WBT2M.S SEA 756.07 30.6 2.060

2 OUTER WING P WBT2O.P SEA 0.00 0.0 0.000

2 OUTER WING S WBT2O.S SEA 0.00 0.0 0.000

3 MIDDLE C WBT3M.C SEA 0.00 0.0 0.000

3 MIDDLE P WBT3M.P SEA 0.00 0.0 0.000

3 MIDDLE S WBT3M.S SEA 0.00 0.0 0.000

3 OUTER WING P WBT3O.P SEA 0.00 0.0 0.000

3 OUTER WING S WBT3O.S SEA 2,546.81 80.4 10.846

4 MIDDLE C WBT4M.C SEA 2,321.52 100.0 22.750

4 MIDDLE P WBT4M.P SEA 2,487.61 100.0 22.750

4 MIDDLE S WBT4M.S SEA 2,467.71 100.0 22.750

4 OUTER WING P WBT4O.P SEA 210.73 6.5 1.050

4 OUTER WING S WBT4O.S SEA 2,447.06 75.0 10.099

5 MIDDLE C WBT5M.C SEA 0.00 0.0 0.000

5 MIDDLE P WBT5M.P SEA 0.00 0.0 0.000

5 MIDDLE S WBT5M.S SEA 0.00 0.0 0.000

5 OUTER WING P WBT5O.P SEA 2,464.04 75.2 10.126

5 OUTER WING S WBT5O.S SEA 3,276.98 100.0 26.510

6 MIDDLE C WBT6M.C SEA 1,588.29 100.0 22.750

6 MIDDLE P WBT6M.P SEA 1,732.38 100.0 22.750

6 MIDDLE S WBT6M.S SEA 1,711.63 100.0 22.750

6 OUTER WING P WBT6O.P SEA 0.00 0.0 0.000

6 OUTER WING S WBT6O.S SEA 2,126.90 100.0 26.510

7 OUTER WING P WBT7O.P SEA 0.00 0.0 0.000

7 OUTER WING S WBT7O.S SEA 1,859.96 100.0 24.070

1 LOWER C WBT1L.C SEA 1,116.96 100.0 26.510

1 LOWER P WBT1L.P SEA 1,214.41 100.0 26.510

1 LOWER S WBT1L.S SEA 1,214.41 100.0 26.510

2 LOWER C WBT2L.C SEA 1,325.95 100.0 26.510

2 LOWER P WBT2L.P SEA 1,307.69 100.0 26.510

2 LOWER S WBT2L.S SEA 1,302.22 100.0 26.510

3 LOWER C WBT3L.C SEA 1,327.37 100.0 26.510

3 LOWER P WBT3L.P SEA 1,304.22 100.0 26.510

3 LOWER S WBT3L.S SEA 1,309.48 100.0 26.510

4 LOWER C WBT4L.C SEA 0.00 0.0 0.000

4 LOWER P WBT4L.P SEA 0.00 0.0 0.000

4 LOWER S WBT4L.S SEA 0.00 0.0 0.000

5 LOWER C WBT5L.C SEA 0.00 0.0 0.000

5 LOWER P WBT5L.P SEA 1,309.43 100.0 26.510

5 LOWER S WBT5L.S SEA 0.00 0.0 0.000

6 LOWER C WBT6L.C SEA 0.00 0.0 0.000

6 LOWER P WBT6L.P SEA 0.00 0.0 0.000

6 LOWER S WBT6L.S SEA 0.00 0.0 0.000

FWD LOWER WB P WBTFL.P SEA 0.00 0.0 0.000

07/15/08 11:44:48 Dockwise Shipping B.V. Page 26

GHS 10.50 BLUE MARLIN LO04

REFERENCE 0611266, STEP 4

FWD LOWER WB S WBTFL.S SEA 0.00 0.0 0.000

FWD UPPER WB P WBTFU.P SEA 0.00 0.0 0.000

FWD UPPER WB S WBTFU.S SEA 0.00 0.0 0.000

AFT WB P WBTA.P SEA 0.00 0.0 0.000

POOP AFT WB S WBTPA.S SEA 0.00 0.0 0.000

POOP FWD WB S WBTPF.S SEA 0.00 0.0 0.000

AFT L BALLAST S WBTALWL.S SEA 0.00 0.0 0.000

AFT PEAK C WBTAPC.C SEA 0.00 0.0 0.000

AFT POOP LOWER WB P WBTAPL.P SEA 0.00 0.0 0.000

AFT POOP UPPER WB P WBTAPU.P SEA 0.00 0.0 0.000

ENGINE RM TOP P WBTERTOP.P SEA 0.00 0.0 0.000

ENGINE RM TOP S WBTERTOP.S SEA 0.00 0.0 0.000

LOWER PEAK C WBTLOWPEAK.C SEA 0.00 0.0 0.000

HIGH PEAK C WBTHIGHPEAK.C SEA 0.00 0.0 0.000

WB OVRFL FWD WBTOVERFWD.C SEA 0.00 0.0 0.000

WATER BAL OFLOW WBTOVERF.S SEA 0.00 0.0 0.000

Total as loaded----------> 48,590.86

Soundings in m.---------------------------------------------------------------

07/15/08 11:44:48 Dockwise Shipping B.V. Page 27

GHS 10.50 BLUE MARLIN LO04

REFERENCE 0611266, STEP 4

SHELTERED CRITERION LONGITUDINAL STRENGTH SUMMARY at Heel = Stbd 0.07 deg.

Largest Bending Moment: -173,974 MT-m. at 125.700f (Sagging)

----------------m.------------MT-----%--------MT---------MT-m.----%------MT-m.

Fr.-5 3.500a 181.65 3 6,538.00 628.9 1 119,623.0

Fr.17 11.900f -378.00 6 -5,820.00 -511.6 1 -86,500.0

Fr.29 20.300f 544.66 4 13,275.00 -156.8 0 -186,087.7

Fr.47 34.700f 278.60 1 18,703.00 8,875.6 3 329,713.2

Fr.53 49.700f 79.53 1 14,682.00 11,039.7 2 597,778.0

Fr.60 67.199f -1,178.19 8 -14,768.00 6,417.6 1 664,318.4

Fr.63 74.700f -2,877.91 20 -14,124.00 -8,762.0 1 -709,742.7

Fr.73 99.700f -4,062.05 29 -14,124.00 -118,884.6 19 -625,742.0

Fr.83 124.700f -306.77 2 -15,272.00 -173,748.8 28 -622,378.4

Fr.93 149.700f 4,215.20 28 15,134.00 -126,942.0 24 -533,195.0

Fr.103 174.700f 3,015.24 55 5,491.00 -19,053.1 8 -237,181.0

Fr.131 194.300f 523.51 12 4,421.00 6,211.9 9 66,269.0

------------------------------------------------------------------------------

SHELTERED CRITERION SUMMARY

Largest Shear: 54.9% at 174.700f

Largest Bending Moment: -28.0% at 123.100f (Sagging)

SHELTERED criterion LONGITUDINAL STRENGTH at 0.07 degrees STBD HEEL

GHS 10.50

07/15/08 11:44:48

160

4

140

120

100

2

80

60

40

General

BLUE MARLIN

20

0

Scale

-20

-40 1

3

-60

-80

2

-100

-120

-140

4

-160

Page 28

LO04

0 20 40 60 80 100 120 140 160 180 200

Aft <-- Location (M.) --> Fwd

07/15/08 11:44:48 Dockwise Shipping B.V. Page 29

GHS 10.50 BLUE MARLIN LO04

REFERENCE 0611266, STEP 4

Body @ 5.000f Body @ 30.000f Body @ 62.200f Body @ 90.000f Body @ 120.000f Body @ 175.000f

78

74

77 71

70

79 80 79 80 14 13 15 11 10 12 8 7 9

48 72 83 83

76 39 38 40 34 33 35 29 28 30

48 49 68 69

41 63 36 37 31 32

62 64 42 60 59 61 57 56 58

70 82

79 17 14 11 8 5 2

72 83

76

44 39 34 29 24 20 68 81

66 63 60 57 54 51

82

79 16 13 10 7 4 1

76 19

43 38 33 28 81

65 62 59 56 53 50

84

71

82

80 18 15 12 9 6 3 69

83

76

45 40 35 30 25 21 81

67 64 61 58 55 52

Plan View

77

78 48 46 41 36 31 26 22

72 17

44

66 63

39

14 60

34

11 57

29

8 54

24

5 2

20

51 70

68

79

76 65

43 16 62

38

13 33

10

59 28

56

7 53

4 19 50

1 83 81

82

84

80

67

45

18 64

40

15 61

35

12 58

30

9 55

25

6 52

21

3 69

71

75

49 47 42 37 32 27 23

73

74

1 WBT1U.C.........6% 18 WBT6U.S.......100% 35 WBT4M.S.......100% 52 WBT1L.S.......100% 69 WBTFL.S.........0%

2 WBT1U.P.........0% 19 WBT1M.C.........0% 36 WBT4O.P.........6% 53 WBT2L.C.......100% 70 WBTFU.P.........0%

3 WBT1U.S.........0% 20 WBT1M.P.........0% 37 WBT4O.S........75% 54 WBT2L.P.......100% 71 WBTFU.S.........0%

4 WBT2U.C.........0% 21 WBT1M.S.........0% 38 WBT5M.C.........0% 55 WBT2L.S.......100% 72 WBTA.P..........0%

5 WBT2U.P.......100% 22 WBT1O.P.........0% 39 WBT5M.P.........0% 56 WBT3L.C.......100% 73 WBTPA.S.........0%

6 WBT2U.S.......100% 23 WBT1O.S.........0% 40 WBT5M.S.........0% 57 WBT3L.P.......100% 74 WBTPF.S.........0%

7 WBT3U.C.......100% 24 WBT2M.P........17% 41 WBT5O.P........75% 58 WBT3L.S.......100% 75 WBTALWL.S.......0%

8 WBT3U.P.......100% 25 WBT2M.S........31% 42 WBT5O.S.......100% 59 WBT4L.C.........0% 76 WBTAPC.C........0%

9 WBT3U.S.......100% 26 WBT2O.P.........0% 43 WBT6M.C.......100% 60 WBT4L.P.........0% 77 WBTAPL.P........0%

10 WBT4U.C.........0% 27 WBT2O.S.........0% 44 WBT6M.P.......100% 61 WBT4L.S.........0% 78 WBTAPU.P........0%

11 WBT4U.P.........0% 28 WBT3M.C.........0% 45 WBT6M.S.......100% 62 WBT5L.C.........0% 79 WBTERTOP.P......0%

12 WBT4U.S.........0% 29 WBT3M.P.........0% 46 WBT6O.P.........0% 63 WBT5L.P.......100% 80 WBTERTOP.S......0%

13 WBT5U.C........97% 30 WBT3M.S.........0% 47 WBT6O.S.......100% 64 WBT5L.S.........0% 81 WBTLOWPEAK.C....0%

14 WBT5U.P........29% 31 WBT3O.P.........0% 48 WBT7O.P.........0% 65 WBT6L.C.........0% 82 WBTHIGHPEAK.C...0%

15 WBT5U.S........30% 32 WBT3O.S........80% 49 WBT7O.S.......100% 66 WBT6L.P.........0% 83 WBTOVERFWD.C....0%

16 WBT6U.C.......100% 33 WBT4M.C.......100% 50 WBT1L.C.......100% 67 WBT6L.S.........0% 84 WBTOVERF.S......0%

APPENDIX 11

Stress plots of the analysis step entire SPAR on vessel

with 4 skid beams by SPAR model loading

147

1. Stress components on Hull side shell on port side along the skid beam # 1

148

3. stress components on longitudinal bulkhead under skid beam #3

149

APPENDIX 12

Stress plots of the analysis step entire SPAR on vessel

with 3 skid beams by SPAR model loading

150

1. Stress components on Hull side shell on port side along the skid beam # 1

151

3. stress components on longitudinal bulkhead under skid beam #3

152

APPENDIX 13

Stress plots of the analysis step entire SPAR on vessel

with 2 skid beams by SPAR model loading

153

1. Stress components on Hull side shell on port side along the skid beam # 1

154

3. stress components on longitudinal bulkhead under skid beam #3

155

APPENDIX 14

Stress plots of the analysis step final stowage position

of SPAR on vessel with 2 skid beams

156

1. Stress components on Hull side shell on port side along the skid beam # 1

157

3. stress components on longitudinal bulkhead under skid beam #3

158

BIBLIOGRAPHY

[1] Subrata K. Chakrabarti, Handbook of Offshore Engineering. Volume 1, Elsevier, First Edition 2005

th

major_projects/perdido/about/spar.html. Last access 5 May 2009

th

[3] Home page of Sea Engineering http://www.sea-engr.com/images/spar_ops.gif. Last access 5

May 2009.

th

[4] Home page of Technip offshore http://www.technip.com/english/pdf/SparProg.JPG. Last access 5

May 2009.

th

spar017.jpg. Last access 5 May 2009

th

[6] Home page of Technip offshore http://www.technip.com/english/pdf/technip.pdf. Last access 5

May 2009.

[7] E.C.Tupper, Introduction to Naval Architecture, Butterworth Heinemann, Third Edition 1996

[8] K.J.Rawson & E.C.Tupper, Basic Ship Theory, Volume 1 - Hydrostatics and Strength,

Butterworth Heinemann, Fifth Edition 2001

th

[10] Home page of Wikipedia, http://en.wikipedia.org/wiki/Semi-submersible, Last access 10

June 2009

th

courses.d/IFEM.d/, Last access 12 June 2009.

th

en_us/products/velocity/femap/index.shtml, Last access 12 June 2009

th

[14] FEMAP version 9.3 documentation, last access 12 June 2009.

[15] Hearn E.J, Mechanics of Materials Vol 1, ButterWorth Heinmann, Third Edition 1997

159

List of Figures

2.1 SPAR with production platform [3]

2.2 Progression of Spars (Technip Offshore) [4]

3.1 MV Blue Marlin

4.1 Typical web frame of Blue Marlin - original vessel

4.2 Typical web frame of Blue Marlin - modified vessel

4.3 Blue Marlin in submerged condition

4.4 General Arrangement of Future SPAR

4.5 Roll-off operation of a heavy-lift on SPMT

4.6 Skidding of a SPAR

4.7 Skidding by uniform loading using hydraulic cylinders

4.8 Stowage plan-1.7m off centre

4.9 Stowage plan-8.3m off centre

4.10 Locations of the skid beams

5.1 Typical finite element geometries in one through three dimensions.

5.2 Examples of primitive structural elements.

5.3 FEMAP environment

5.4 Rod Element

5.5 Beam Element

5.6 Plate Element

5.7 Solid Element

6.1 Rectangular coordinate system

6.2 FE model of Blue Marlin and coordinate system axes

6.3 FE model showing mesh of the Keel of Blue Marlin.

6.4 FE model of Double hull bottom structure of Blue Marlin

6.5 FE model details of bottom structure

6.6 FE model of typical Blue Marlin web frame with stiffeners

6.7 FE model of the vessel deck

6.8 FE model of the bow section

6.9 FE model of stern section showing the profile of the stern

6.10 FE model of stern section without deck showing the engine room section

6.11 Spring elements under the vessel model

6.12 API tool for implementing draft and hydrostatic pressure in the model

6.13 Hydrostatic pressure applied to the FE model

6.14 Closer look at the hydrostatic pressure applied

6.15 API tool for updating the mass elements in the model

6.16 Mass elements in the FE model of Blue Marlin

6.17 Skid beams on vessel model

6.18 Typical web frame details of the physical vessel

6.19 Typical web frame details of the FE model

6.20 FE model of SPAR and coordinate system axes

6.21 FE model of Hard tank section of the SPAR

6.22 FE model of Soft tank section of the SPAR

6.23 FE model of Heave plate section of the SPAR

6.24 FE model of the truss section of the SPAR

6.25 Line only FE model assembly of hard tank and truss

6.26 FE model of the SPAR without supports and outfittings

6.27 FE model of the hard tank supports

6.28 FE model of truss supports

6.29 FE model of SPAR with supports for 4 skid beam load-out

6.30 FE model of SPAR with supports for 3 skid beam load-out

6.31 FE model of SPAR with supports for 2 skid beam load-out

7.1 Uniform loading on skid beams

7.2 Constraint in bow section of the vessel

7.3 Constraints in stern section of vessel

7.4 Constraints in the spring elements

7.5 Draft level and hydrostatic pressure for the step hard tank on the vessel

7.6 Location of supports for 4 skid beams load-out

7.7 Location of supports for 3 skid beams load-out

7.8 Location of supports for 2 skid beams load-out

7.9 FE model of analysis step hard tank on vessel with 4 skid beams by uniform loading

160

7.10 Scaled total deformation of the analysis step uniform loading with 4 skid beams

7.11 FE model of analysis step hard tank on vessel with 3 skid beams by uniform loading

7.12 Scaled total deformation of the analysis step uniform loading with 3 skid beams

7.13 FE model of analysis step hard tank on vessel with 2 skid beams by uniform loading

7.14 Scaled total deformation of the analysis step uniform loading with 2 skid beams

7.15 Location of highest equivalent stresses in uniform loading case analysis.

7.16 Rod elements connecting nodes on supports and skid beams

7.17 Constraints of the SPAR supports under the heave plates and soft tank

7.18 FE model of analysis step hard tank on vessel with 4 skid beams by SPAR model loading

7.19 Scaled total deformation of the analysis step hard tank on vessel with 4 skid beams

7.20 Plot of vessel model translation in Z-axis for the step hard tank on vessel with 4 skid beams

7.21 FE model of analysis step hard tank on vessel with 3 skid beams by SPAR model loading

7.22 Scaled total deformation of the analysis step hard tank on vessel with 3 skid beams

7.23 Plot of vessel model translation in Z-axis for the step hard tank on vessel with 3 skid beams

7.24 FE model of analysis step hard tank on vessel with 2 skid beams by SPAR model loading

7.25 Scaled total deformation of the analysis step hard tank on vessel with 2 skid beams

7.26 Plot of vessel model translation in Z-axis for the step hard tank on vessel with 2 skid beams

7.27 Axial forces in rod elements on skid beam #1 – hard tank on vessel with 4 skid beams

7.28 Axial forces in rod elements on skid beam #2 – hard tank on vessel with 4 skid beams

7.29 Axial forces in rod elements on skid beam #3 – hard tank on vessel with 4 skid beams

7.30 Axial forces in rod elements on skid beam #4 – hard tank on vessel with 4 skid beams

7.31 Axial forces in rod elements on skid beam #1 - hard tank on vessel with 3 skid beams

7.32 Axial forces in rod elements on skid beam #2 - hard tank on vessel with 3 skid beams

7.33 Axial forces in rod elements on skid beam #3 - hard tank on vessel with 3 skid beams

7.34 Axial forces in rod elements on skid beam #2 - hard tank on vessel with 2 skid beams

7.35 Axial forces in rod elements on skid beam #3 - hard tank on vessel with 2 skid beams

7.36 Location of highest equivalent stresses in hard tank on vessel by SPAR model loading case analysis.

7.37 Final position of SPAR hard tank supports on skid beams for entire SPAR on vessel

7.38 Final position of SPAR truss supports on skid beams for entire SPAR on vessel

7.39 Constraints in supports under heave plates for the step entire SPAR on vessel

7.40 FE model of analysis step entire SPAR on vessel with 4 skid beams.

7.41 Total deformation of the analysis step entire spar on vessel with 4 skid beams

7.42 Axial forces on the rod elements in the analysis step entire spar on vessel with 4 skid beams

7.43 FE model of analysis step entire SPAR on vessel with 3 skid beams.

7.44 Total deformation of the analysis step entire spar on vessel with 3 skid beams

7.45 Axial forces on the rod elements in the analysis step entire spar on vessel with 3 skid beams

7.46 FE model of analysis step entire SPAR on vessel with 2 skid beams.

7.47 Total deformation of the analysis step entire spar on vessel with 2 skid beams

7.48 Axial forces on the rod elements in the analysis step entire spar on vessel with 2 skid beams

7.49 Location of highest equivalent stresses in entire SPAR on vessel by SPAR model loading case

analysis.

7.50 Final stowage position of SPAR hard tank supports on skid beams

7.51 Final stowage position of SPAR truss supports on skid beams

7.52 FE model of analysis step final stowage position of the SPAR with 2 skid beams.

7.53 Total deformation of the analysis step final stowage position of SPAR with 2 skid beams

7.54 Axial forces on the rod elements in the analysis step final stowage position of SPAR with 2 skid

beams

List of tables

3.1 Principal characteristics of Blue Marlin

4.1 Weight of SPAR components

4.2 Weight distribution of SPAR

6.1 Summary of FE model of the vessel

6.2 Properties of FE model of keel

6.3 Properties of FE model of bottom structure

6.4 Properties of FE model of bulkhead

6.5 Properties of FE model of deck

6.6 Properties of FE model of bow section

6.7 Properties of FE model of stern section

161

6.8 Properties of spring element under the vessel model

6.9 Properties of Mass element in vessel model

6.10 Comparison of FE model with the actual vessel.

6.11 Comparison of FE model with the actual vessel

6.12 Properties of FE model of hard tank

6.13 Properties of FE model of soft tank

6.14 Properties of FE model of Heave plates

6.15 Properties of FE model of Truss section

6.16 Properties of FE model of SPAR supports

6.17 Properties of FE model of SPAR with outfittings for 4 skid beam load-out

6.18 Comparison of FE model of SPAR with the design

7.1 Ballast condition for step hard tank on the vessel.

7.2 Load calculation for the uniform loading with 4 skid beams load-out

7.3 Load calculation for the uniform loading with 3 skid beams load-out

7.4 Load calculation for the uniform loading with 2 skid beams load-out

7.5 Summary of load calculation for all 3 steps of uniform loading

7.6 Summary of results of uniform loading cases

7.7 Highest equivalent stresses of the analysis step hard tank on vessel with 4 skid beams by uniform

loading

7.8 Highest equivalent stresses of the analysis step hard tank on vessel with 3 skid beams by uniform

loading

7.9 Highest equivalent stresses of the analysis step hard tank on vessel with 2 skid beams by uniform

loading

7.10 Properties of rod elements

7.11 Summary of results of hard tank on vessel by SPAR model loading cases

7.12 Load on skid beams in step hard tank on vessel with 4 skid beams

7.13 Load on skid beams in step hard tank on vessel with 3 skid beams

7.14 Load on skid beams in step hard tank on vessel with 2 skid beams

7.15 Summary of load on skid beams in step hard tank on vessel by SPAR model loading

7.16 Highest equivalent stresses of the analysis step hard tank on vessel with 4 skid beams by SPAR

. model loading

7.17 Highest equivalent stresses of the analysis step hard tank on vessel with 3 skid beams by SPAR

model loading

7.18 Highest equivalent stresses of the analysis step hard tank on vessel with 2 skid beams by SPAR

model loading

7.19 Ballast condition for the step entire SPAR on vessel

7.20 Summary of results of entire SPAR on vessel by SPAR model loading cases

7.21 Highest equivalent stresses of the analysis step entire SPAR on vessel with 4 skid beams

7.22 Highest equivalent stresses of the analysis step entire SPAR on vessel with 3 skid beams

7.23 Highest equivalent stresses of the analysis step entire SPAR on vessel with 2 skid beams

7.24 Highest equivalent stresses of the analysis step final stowage position of SPAR with 2 skid beams

162

- Wooden Floor JoistЗагружено:Joseph Cloyd L. Lamberte
- 20.195e-10.98Загружено:richard anish
- Welcome to International Journal of Engineering Research and Development (IJERD)Загружено:IJERD
- Consistency of Gait Characteristics as Determined From Acceleration Data Collected at Different Trunk LocationsЗагружено:Eduardo Vieira
- Stoica FreaЗагружено:AugustoZebadúa
- 2015_AppliedOptics (1)Загружено:Venkatesan Thanigai Arasu
- Calculating Dead Loads on Studded TrussesЗагружено:Thon Marvine Dionisio Urbano
- Rexel Tutorial EngЗагружено:Simon Petrovčič
- 3 BandwidthЗагружено:Tathagata Ghosh
- Interpreting Spectral AnalysisЗагружено:siriuslot
- DK342X_C006.pdfЗагружено:Bekraoui Keltoum
- a krauss - vfs 2014 - egg and spectral momentsЗагружено:api-261272633
- FFT FAQЗагружено:Muhammad Subekti
- Dattorro - Effect Design Part 3 - Oscillators: Sinusoidal and PseudonoiseЗагружено:Jon Dattorro
- Web FormingЗагружено:Kumar Swami
- Lifetime Load AnalysisЗагружено:phantom29
- On Albersheim's Detection EquationЗагружено:Sunwoo Kong
- TFA - Cohen.pdfЗагружено:anon020202
- PLESUMS_On Teaching Structure Systems_1423936Загружено:mulkeenm
- vibration DNVЗагружено:ANTON NEAGU
- Evalutation of Seismic Energy in Structures_ Uang and BerteroЗагружено:Manu Vega
- 423F44BF40CFB116D500184392743824Загружено:Akhlaq Hussain
- Lecture 02Загружено:chinchou
- TECHOS DE MADERAЗагружено:Josué Azurín Rendich
- ba_cpr_d_gbЗагружено:ajayprakash
- Hangar Design ReferenceЗагружено:Stephen James
- Aguiar-Conraria L 2011 - The Continuous Wavelet Transform a PrimerЗагружено:erik
- 07.docxЗагружено:locus
- Sigma Delta ExactЗагружено:zydscience
- Taipei Being a Coastal City the Problems Present AreЗагружено:Vatsal Chabhaiya

- Local Scour PileЗагружено:oe07m007
- jetstreamv5_enuЗагружено:Cristián Marcelo Abarca Ilabaca
- koei JIS.pdfЗагружено:Timo
- staad tech.pdfЗагружено:oe07m007
- Advance Soil MechanicsЗагружено:lakshmi1011
- Eindrapport Van WijngaardenЗагружено:bob monkton
- Dynamics and Joint Design for Offshore StrcutresЗагружено:oe07m007
- Offshore Hydro MechanicsЗагружено:rezbahal
- GrillЗагружено:oe07m007
- Inplace Analysis Workshop-GeniEЗагружено:oe07m007
- Metal Guide DesignЗагружено:oe07m007
- Guideline for metal structres.pdfЗагружено:oe07m007
- Groutand HertzЗагружено:oe07m007
- Dreel RiggЗагружено:oe07m007
- 1 Drilling RigЗагружено:daburto2
- Blast Resistant Design With Structural SteelЗагружено:Javier Flores
- Axial Pile design.pdfЗагружено:oe07m007
- Maheswaran_Jeron.pdfЗагружено:Rohan Karande
- pojaЗагружено:oe07m007
- Kitchen detailsЗагружено:oe07m007
- 4_Mudmat[1]Загружено:oe07m007
- Structural AnalysisЗагружено:oe07m007
- 09_Types of analysis.pdfЗагружено:Nazar Shafiq
- Deterministic Fatigue AnalysisЗагружено:oe07m007
- Is 1893 PART1 Exp.examplesЗагружено:jonnyprem
- v1-3-25Загружено:oe07m007

- thyristorЗагружено:Ahmed Elhadidy
- Copper(I) Alkoxides. Synthesis ReactionsЗагружено:Mike Burns
- Foundation HU Lec 2 - CopyЗагружено:zeru3261172
- Effect of Protective Clothing and Fatigue on Functional Balance of Firefighters 2165 7556.S2 004Загружено:kirana_patrolina
- Teaching Reading Using Quantum TeachingЗагружено:M I Andi Wahyono
- IM200-10 (LC Mag IOM).pdfЗагружено:Lazzarus Az Gunawan
- mc-ty-parts-2009-1.pdfЗагружено:Andrew Chen
- Ultrasound Quiz 1.Docx 20 05 13Загружено:Osman Gone Rana
- 6D. Science & Technology (212)Загружено:princeforjesus
- lmdqЗагружено:Mohd Azamuddin Mohd Saifullizan
- Chapter 23 Functional GroupsЗагружено:Yudi Permana
- uutrrЗагружено:Mahmoud Hegazy
- Mass Transfer CoefficientsЗагружено:juandiegoCO
- s 7102049Загружено:Dana Iulia Moraru
- DALSA Boa Manual en.pdfЗагружено:Vicente Ramon Balle
- Biomechanics of the Knee During Closed Kinetic Chain and Open Kinetic Chain ExercisesЗагружено:AndresLeivaMckendrick
- Glycol Dehydration UnitЗагружено:arispriyatmono
- VOR-ADFЗагружено:Erlendur Petersen
- UNIT-1 BITSЗагружено:REVANTH KUMAR K
- MathЗагружено:samim
- anna univ mechatronics questionЗагружено:Stephanie Johnson
- Folk.uio.No Yurig FYS3410 Solutions3410-TmpЗагружено:Meshal Alawein
- EeeЗагружено:Dharmateja Kadem
- Centrifugal Pumps for Molten Sulphur & h2so4 ProductionЗагружено:Samir Ganguli
- Sapo 34Загружено:Yiğit Ilgaz
- Mathcad - Elastica Viga Carga TrapecialЗагружено:lucho0314
- 30755d1304962898 Mumbai University 2nd Sem Engineering Drawing Ed Exam Download Previous Years Question Papers Mumbai University 2nd Sem. Engineering Drawing Ed Sample Paper 1Загружено:Abhishek Kulkarni
- D 4436 – 84 R98--Загружено:darkobogd75
- ASSIGNMENT (CH-DETERMINANTS).pdfЗагружено:Manas Ranjan Jena
- MTH301 Solved Finalterm Papers Solved MCQsЗагружено:

## Гораздо больше, чем просто документы.

Откройте для себя все, что может предложить Scribd, включая книги и аудиокниги от крупных издательств.

Отменить можно в любой момент.