19 views

Uploaded by Gabriel Gonzalez

Thesis

Thesis

© All Rights Reserved

- Beam Design Lecture Notes
- Steel Column Design
- HDPE TechInfo English
- Steel design manual
- Buckling 04
- Sway Mode - EC3
- [0]DNVRPC201-platedstructurebuckling
- Loaded Timoshenko Beam
- Syllabus 17
- AISC vS EC
- CONTEC Design Guidelines Handbook
- 2002 Inelastic Seismic Response of Steel Bracing Members. Robert Tremblay
- EN 1993-1-5.pdf
- 3373056.pdf
- 1-s2.0-S0263823113002097-main (1)
- unit-8-Design-of-Beams.pdf
- Strength and Ductility of CFS Bolted Connections
- Perfiles a 60
- 40783-45459-1-SM(1)
- Face Sheet Wrinkling

You are on page 1of 157

Aalborg University

Special Report No. 80

Preprint

Industrial Ph.D.-Thesis

by

Niels Hjen stergaard

Department of Mechanical and Manufacturing Engineering, Aalborg University /

NKT-Flexibles, Department of Structural Development

Fibigerstrde 16, DK-9220 Aalborg East, Denmark

e-mail: Niels.HojenOstergaard@nktflexibles.com

This report, or parts of it, may be reproduced without the permission of the

author, provided that due reference is given. Questions and comments are

most welcome and may be directed to the author, preferably by e-mail.

Printed in Aalborg, February 2012.

ISBN 87-91464-35-8

Preface

This thesis has been submitted to the Faculty of Engineering, Science and Medicine at Aalborg

University in partial fulfillment of the requirements for the degree of Doctor of Philosophy in

Mechanical Engineering. The conducted work has been carried out at the Department of Mechanical

and Production Engineering at Aalborg University and at NKT-Flexibles in Brndby, Denmark, from

November 2008 to February 2012.

The work contained in the present thesis was carried out as part of an industrial PhD project involving

Aalborg University and NKT-Flexibles with financial support from the Danish Agency for Science,

Technology and Innovation. The project was supervised by Associate Professor, PhD Jens H.

Andreasen and Development Engineer, PhD Anders Lyckegaard, whom I thank for their support and

guidance.

I wish to thank my friend, PhD Alf Se-Knudsen for his support, optimistic mood and never failing

talent for expressing himself in differential equations rather than in words. Furthermore, a great

thanks to NKT-Flexibles Specialist Engineer Thomas Dettlaff for invaluable help and expertise in the

field of instrumentation, for ensuring that the numerous mechanical parts I constructed could be

assembled and for keeping me company watching a flexible pipe being bent repeatedly throughout

many a day and night. I would also like to thank the remaining crew involved in the experimental

work, especially Work Shop Technicians Jesper Nielsen and Tommy Harboe Friis for their work

related to dissections of failed flexible pipes.

I owe my gratitude to Lead Engineer Jan Rytter, who initiated and motivated the work with the lateral

buckling failure mode within NKT-Flexibles and to Pipe Design Engineers Lars Rude and Geir

Agustsson for their help with- and knowledge regarding well established design methods. I also wish

to thank Development Manager Niels Rishj and Lead Engineer Erik Bendiksen for encouraging the

present project and allowing the publication of obtained results.

Finally, I most of all wish to thank my family, Susi, Ronja and Malte for putting up with a boyfriend and

father who often while working on this project had his mind on wire mechanics rather than where it

should have been.

February, 2012

ii

iii

Abstract

The objective of the work documented in the present thesis has been by theoretical as well as

experimental means to study the physical behavior of the tensile armour layers of flexible pipes

related to a given failure mode involving lateral instability. This mode of failure, which is often

referred to as lateral wire buckling, is most common during flexible pipe laying in deep waters. In this

load scenario, a flexible pipe is subjected to compressive longitudinal loads and repeated bending

cycles.

Flexible pipes are usually designed as steel-polymer composite structures constituted by a number of

layers with different properties. Such structures have a wide range of applications in the offshore

industry. In the present work, the focus has mainly been on the mechanics of the two oppositely

wound layers of helical armour wires, since the investigated mode of failure occurs in these specific

pipe layers. Lateral wire buckling is a phenomenon characterized by the fact that it contrary to several

other flexible pipe failure modes is difficult to detect by visual inspection. However, it has on basis of

laboratory experiments been concluded, that lateral wire buckling causes a twist of a flexible pipe,

which can be quite severe. This twist occurs when the tensile armour layers no longer are torsionally

stable, which leads to a compression-twist coupling. Furthermore, lateral wire buckling is associated

with a shortening of a flexible pipe, while only small to moderate changes of circumference can be

detected.

As mentioned, lateral wire buckling of flexible pipe armouring layers can be reproduced

experimentally. In the present work, such experiments have been conducted by use of mechanical test

benches, in which a flexible pipe can be subjected to longitudinal compression and repeated bending.

It is, however, a widely accepted fact that results obtained by this experimental principle do not corelate to results obtained in the field, since failure during experiments occurs at lower compressive

load levels than encountered with offshore field conditions. The reason for this discrepancy is

unknown, but is possibly caused by differences in boundary conditions between the two scenarios.

However, experimental reconstruction of the lateral wire buckling failure mode remains a valuable

source of information related to failure in flexible pipes, since deformations and applied loads can be

measured, which is not the case in the field. Lateral wire buckling is characterized by large differences

in wire lay angles with respect to the initial helical state. However, the underlying mechanism does not

lead to failure of a flexible pipe, before repeated loading causes wire slippage towards geometrical

configurations in which the yield stress of the wire steel is exceeded.

In order to develop design methods for avoidance of failure in an unpressurized flexible pipe,

theoretical studies of armour wire mechanics have been conducted on basis of a formulation of the

mechanical equilibrium state of a beam embedded in a frictionless toroid. On this basis, the torsional

equilibrium state of all wires contained in the pipe wall is derived so that the compressive load

carrying ability can be calculated. The determination of this state rests on the assumption, that the

equilibrium states, which the wires will slip towards as cyclic bending is applied, coincide with the

equilibrium configurations determined directly, if friction is neglected. Effects due to friction and cyclic

loadings have partly been investigated without detection of significant impact.

The theoretical methods for analysis of armour wires that have been developed as part of the present

project are documented in four scientific journal papers. Furthermore, results obtained by these

methods are compared to experimental results in the present report as well as in two papers from

conference proceedings.

iv

Resume

Formlet med arbejdet, der er dokumenteret i nrvrende afhandling, har vret med svel teoretiske som

eksperimentelle metoder at studere den fysiske opfrsel af fleksible rrs trkarmering relateret til en given

svigtmekanisme involverende lateral instabilitet. Denne svigtmekanisme, der ofte benvnes lateral wire

buling, forekommer hyppigst under installation af fleksible rr p stor vanddybde. I dette belastningsscenarie

er et fleksibelt rr udsat for kompressive laster i lngderetningen og gentagne bjningscykler.

Fleksible rr er normalt designede som stl-polymer kompositte strukturer sammensat af en rkke lag med

forskellige egenskaber. Sdanne strukturer har et bredt spektrum af anvendelsesmuligheder i offshore

industrien. I det nrvrende arbejde er fokus hovedsageligt lagt p trkarmeringslagene, to lag af modsat

viklet helisk bndarmering, da den undersgte svigtmekanisme optrder i disse lag. Lateral wire buling er

som fnomen kendetegnet ved, at det i modstning til adskillige andre svigtmekanismer i fleksible rr kun

vanskeligt kan observeres ved visuel inspektion. Det er dog ved laboratorieeksperimenter, under hvilke

svigtmekanismen er blevet rekonstrueret under kontrollerede forhold, blevet fastslet, at lateral wire buling

forrsager et torsionelt vrid i fleksible rr, der kan vre ganske voldsomt. Dette er forrsaget af at

trkarmeringslagene ikke lngere stabiliserer hinanden torsionelt, sledes at en kobling mellem vrid og

forkortning opstr. Mens lateral wire buling er ledsaget af en forkortning af et fleksibelt rr, forekommer kun

sm til moderate ndringer i yderdiameter.

Som nvnt kan lateral wire buling i fleksible rrs trkarmeringslag reproduceres eksperimentelt. Indenfor

rammerne af det nrvrende projekt er dette foretaget under anvendelse af mekaniske testbnke, i hvilke et

fleksibelt rr udsttes for kompressiv last og gentagne bjningscykler. Det er et bredt accepteret faktum, at

dette eksperimentelle princip ikke rekonstruerer installationsbetingelserne p fyldestgrende vis, da svigt

forekommer ved lavere kompressive laster end observeret i felten. rsagen til denne diskrepans er ukendt,

men relaterer med stor sandsynlighed til forskelle i de randbetingelser, som et fleksibelt rr udsttes for i

felten og under laboratorieeksperimenter. P trods af dette forbliver laboratorieeksperimenter en vrdifuld

kilde mht. at studere svigt i fleksible rr, da deformationer og plagte belastninger kan mles, hvilket ikke er

tilfldet i felten. Lateral wire buling er karakteriseret ved store afvigelser i viklevinkel i forhold til den

oprindelige heliske tilstand. Der er dog frst tale om egentligt svigt, nr gentagne bjningscykler forrsager

at wirene slipper mod geometriske konfigurationer i hvilke flydespndingen er overskredet, da dette

muliggr dannelse af blivende plastiske deformationer.

For at vre i stand til at designe fleksible rr sledes at svigt ved lateral buling ikke forekommer, nr

et rr ikke et tryksat, er teoretiske studier i trkarmeringswires mekanik foretaget p grundlag af en

formulering for den mekaniske ligevgtstilstand for en bjlke p en friktionsls torusoverflade. P

grundlag af denne formulering er samtlige wires torsionelle ligevgtstilstand beskrevet, sledes at

den kompressive lastbreevne kan beregnes. Bestemmelse af denne tilstand hviler p en antagelse

om, at den ligevgtskonfiguration, som en armeringswire vil slippe mod efterhnden som

bjningscykler pfres, er sammenfaldende med den ligevgtskonfiguration, der bestemmes nr

friktion negligeres. Effekter forrsaget af friktion og cykliske belastninger har delvist vret berrt

uden at signifikant indflydelse kunne detekteres.

De teoretiske metoder til wireanalyse, der er udviklede som del af det udfrte arbejde, er

dokumenterede i fire videnskabelige artikler. Ydermere er sammenligninger med eksperimentelle

resultater dokumenteret i nrvrende rapport samt i to konferenceartikler.

vi

Dissertation

This dissertation is based on an introduction to the area of research and four papers submitted to

refereed scientific journals. Furthermore, two per-reviewed publications from conference proceedings

are included.

List of Publications

Journal Papers

1. Paper A

stergaard, N.H., Lyckegaard, A., Andreasen, J.H.

A method for prediction of the equilibrium state of a long and slender wire on a frictionless toroid

applied for analysis of flexible pipe structures

Engineering Structures, Vol. 34, pp. 391-399, 2012.

2. Paper B

stergaard, N.H., Lyckegaard, A., Andreasen, J.H.

Imperfection analysis of flexible pipe armour wires in compression and bending

Submitted to Applied Ocean Research, October 2011, under review.

3. Paper C

stergaard, N.H., Lyckegaard, A., Andreasen, J.H.

On modeling of lateral buckling failure in flexible pipe tensile armour layers

Submitted to Marine Structures, September 2011, under review.

4. Paper D

stergaard, N.H., Lyckegaard, A., Andreasen, J.H.

Simulation of frictional effects in models for calculation of the equilibrium state of flexible pipe

armouring wires in compression and bending

Rakenteiden Mekaniikka (Journal of Structural Mechanics), Vol. 44, No. 3, pp. 243-259, 2011,

Special Issue for the 24th Nordic Seminar on Computational Mechanics (NSCM-24).

Papers in proceedings

5. Paper E

stergaard, N.H., Lyckegaard, A., Andreasen, J.H.

On lateral buckling failure of armour wires in flexible pipes

Proceedings of the ASME 2011 30th International Conference on Ocean, Offshore and Arctic

Engineering, OMAE2011-49358.

6. Paper F

stergaard, N.H., Lyckegaard, A., Andreasen, J.H.

Simplified models for prediction of lateral buckling in flexible pipes armour wires

Accepted for presentation at the ASME2012 31th International Conference on Ocean, Offshore

and Arctic Engineering, OMAE2012-83080.

vii

(not included in the dissertation)

7. Extended Abstract A

stergaard, N.H., Lyckegaard, A., Andreasen, J.H.

Lateral buckling of the tensile armor layers of flexible pipes

12th Internal Symposium of the Danish Center for Applied Mechanics and Mathematics

(DCAMM), 2009.

8. Extended Abstract B

stergaard, N.H., Lyckegaard, A., Andreasen, J.H.

On lateral buckling of armour wires in flexible pipes

23rd Nordic Seminar on Computational Mechanics (NSCM-23), 2010.

9.

Extended Abstract C

stergaard, N.H., Lyckegaard, A., Andreasen, J.H.

Lateral buckling of the tensile armor layers of flexible pipes

13th Internal Symposium of the Danish Center for Applied Mechanics and Mathematics

(DCAMM), 2011.

The laboratory experiments conducted as part of the present project were documented in a number of

technical reports issued as NKT-Flexibles documents [45]-[71]. Usually each experiment was

conducted in accordance with a test procedure. Results were documented in a test report and a

separate dissection report was issued after disassembling the tested pipe structure.

These documents are not available to the public. However, a summary of the experimental results is

contained in chapter 3.

viii

Contents

Preface .................................................................................................................................................. i

Abstract ..............................................................................................................................................iii

Dissertation ........................................................................................................................................ vi

List of Publications ............................................................................................................................. vi

Contents ............................................................................................................................................ viii

1.

Introduction .................................................................................................................................. 1

1.1.

1.2.

1.3.

2.

2.1.

Experimental setup............................................................................................................. 12

2.2.

3.

3.1.

3.2.

3.3.

Test series II, 14 jumper test pipe samples in G2 bending rig ......................................... 21

3.4.

Test series III, 8 riser test pipe samples in G2 bending rig .............................................. 23

3.5.

4.

4.1.

4.2.

4.3.

Paper A ................................................................................................................................ 32

4.4.

Paper B ................................................................................................................................ 33

4.5.

Paper C ................................................................................................................................ 34

ix

5.

6.

4.6.

Paper D................................................................................................................................ 35

4.7.

Paper E ................................................................................................................................ 36

4.8.

Paper F ................................................................................................................................ 37

5.1.

5.2.

Discussion of the definition of lateral wire buckling limit state design .............................. 40

5.3.

5.4.

5.5.

Modes of deformation.......................................................................................................... 45

5.6.

5.7.

5.8.

6.1.

6.2.

References .......................................................................................................................................... 57

Internal NKT documents ................................................................................................................... 59

Appendix A: Paper A ... 61

Appendix B: Paper B ... 73

Appendix C: Paper C .... 87

Appendix D: Paper D .101

Appendix E: Paper E ......121

Appendix F: Paper F....133

1. Introduction

The present thesis deals with unbonded flexible pipes. This type of composite structures is usually

constituted by numerous steel and polymer layers. Flexible pipes have a wide range of applications in

the offshore industry related to drilling and extraction of oil and gas from subsea reservoirs. This type

of structure may be used as flowlines, running along the seabed, as conductors for water injection, or

as risers connecting a subsea reservoir to a floating platform at the sealevel. Flexible pipes are usually

designed in accordance with the API 17J code [1].

During the past few decades, the direction taken by the world economy has made development of oil

and gas fields at large water depths feasible. This process has introduced a need for flexible pipes

capable of resisting the extreme loads induced by hydrostatic pressure occurring at water depths

larger than 1000 meters. Furthermore, the need for experimentally validated design tools for

prediction of the numerous different failure mechanisms has grown tremendously. Due to this

development, the complex mechanical behavior of flexible pipes arising from interaction between the

different pipe layers has been subject of both industrial and academic research. Among the failure

modes, which have been investigated most intensively by theoretical and experimental means, are

collapse resistance and fatigue failure. Despite research is still ongoing and computational models and

experimental principles are continuously being improved, the methods for prediction of these failure

modes have been developed to a stage, at which they can be applied successfully for engineering

analysis.

The work presented in this thesis is related to lateral wire buckling in the tensile armour layers of

flexible pipes, see Figure 1. This is one of a few failure mechanisms, which cannot yet be predicted

with sufficient accuracy. Hence, no computational tool for design against lateral wire buckling has

been proposed. In order to obtain a torsionally stable design, the tensile armour layers in flexible pipes

are usually constituted by two layers of initially helically wound steel wires

wires with opposite lay

directions. The specific failure mode is usually encountered during pipe laying. In this scenario, the

flexible pipe is in a free hanging configuration from an installation vessel to the seabed. During pipe

laying, the pipe is exposed to longitudinal compression due to hydrostatic pressure, since the pipe

bore is empty during installation. Furthermore, the pipe will be exposed to repeated bending cycles

due to natural loads and movements of the installation vessel. The lateral buckling failure mode was

first encountered in 1997 and is described by Braga and Kaleff [2], who also were the first to

reproduce it by experimental means in the laboratory.

Figure 1 Severe state of lateral wire buckling in the inner layer of armouing wires, triggered experimentally by

laboratory testing of 6 flexible pipe.

The theoretical work documented in this thesis constitutes studies of armour wire mechanics

investigating the mechanisms leading to lateral wire buckling. Furthermore, the industrial cooperation

enabled investigation of lateral buckling failure by reconstructing the failure mechanism in the

laboratory under controlled conditions. In order to do so, mechanical test rigs constructed specifically

for this purpose were used.

The thesis is outlined as an introduction to a number of peer-review papers enclosed in Appendix A-F.

In the present first chapter, the structures and failure mechanisms, which are subject of the

presented research, will be introduced.

In the second chapter, the principle used for laboratory experiments will be presented

In the third chapter, a summary of the experiments conducted by means described in chapter 2

will be presented.

In the fourth chapter, the theoretical approach to wire mechanics is addressed before a

summary of the publications related to the present project is included.

In the fifth chapter, additional experimental and simulated results are included for the sake of

completeness.

In the sixth and final chapter, the thesis will be finalized by concluding remarks. Furthermore,

the contributions from the present research to the field of flexible pipe mechanics will be

summarized and directions of future research will be discussed.

Flexible pipes are in most known designs steel-polymer composite structures constituted by a number

of initially either helically wound or cylindrical layers, see Figure 2. This type of design is chosen in

order to obtain a structure which is flexible and capable of resisting large longitudinal loads as well as

external and internal pressure.

The size of a flexible pipe is usually measured in inches and refers to the inner diameter of the pipe

bore. The bore is surrounded by the carcass, which is constructed by helically wound steel profiles

with a high pitch angle. This enables the flexible design to be capable of resisting the internal pressure

in the bore as well as external hydrostatic pressure. The carcass is surrounded by a fluid barrier

denoted the inner liner which is an extruded polymeric layer. On the outside of this layer, a pressure

armour, also constructed of high pitched steel profiles, is applied in order to provide sufficient collapse

resistance against external pressure.

Although flexible pipes for deep-water applications have been designed with four layers of tensile

armour, see for example Secher et. al. [3], most designs contain two layers of tensile armour

constituted by oppositely wound helical steel wires. These layers ensure the structural integrity

against axisymmetric loads. The wires in flexible pipes are usually of rectangular cross-section in

which the wire height measured in the radial pipe direction is smaller than the wire width

corresponding to the dimension in the circumferential pipe direction. The tensile armour layers are

often surrounded by wound plastic tape layers. These do not contribute significantly to the structural

properties of the pipe, but prevent wear due to steel-against-steel contact. In order to prevent radial

deflections of the wires, a high-strength tape, often made of aramid reinforced plastics, is wound

helically above the tensile armour layers, see Figure 2. If radial deflections are not restricted,

longitudinal compression may lead to a wire buckling phenomenon denoted birdcaging. This failure

mode is characterized by large localized radial deflections, see Figure 3. The high-strength tape also to

some extend prevents rotations of the wires around the local axes, an undesired phenomenon usually

referred to a fishscaling.

The pipe interior is protected from seawater by an outer sheath, which like the inner liner is an

extruded polymeric layer. The space located between the inner liner and the outer sheath is usually

referred to as the pipe annulus.

1.Carcass

2.Inner liner

3.Pressure armour

4.Tensile armour

5.Tensile armour

6. Outer sheath

Figure 2, left: Example of the most common flexible pipe design, right: High-strength tape for prevention of

wire birdcaging (ABC-layer).

Since flexible pipes are designed as composite structures with numerous layers, the global structural

behavior is complex and has been investigated intensively in numerous publications. Due to the

unbonded nature of the structures, layers may to some extend slip internally. Furthermore, couplings

between longitudinal strain, radial strain, and pipe twist may arise due to the internal helically wound

components. In general, a flexible pipe is designed in such a manner that coupling effects are

minimized. As a consequence of this, the number of wires in the outer layer is larger than in the inner

layer of tensile armour. The first commercially available design tool for flexible pipe design, CAFLEX

documented in [4], enabled analysis of straight flexible pipes based on radial, torsional, and

longitudinal equilibrium of internal components. The mathematical model applied considered layers

either as constituted by helical elements or as isotropic thin shells. The underlying theory was

partially published in 1987 by Fret and Bournazel [5].

Due to wire slippage in bending, the flexural moment-curvature responses of flexible pipes exhibit

hysteresic behavior. This behavior was described in the CAFLEX theory manual [4] and investigated

further by, among others, Witz and Tan [6], Tan, Quiggin and Sheldrake [7], Kraincanic. and Kabadze

[8], Alfano et. al. [9]-[10] and Dastous [11]. For low values of curvature, frictional resistance prohibits

wire slippage, and the wires contribute to the cross sectional moment of inertia around the global pipe

axis. When a critical curvature value s is exceeded, wire slippage occurs, and the wires bend solely

around their local axes. This causes the wire contributions to the pipe bending stiffness to decrease as

wire slippage occurs. This behavior has often been idealized as bilinear as shown in Figure 3, in which

S1 denotes the pipe bending stiffness prior to wire slippage, and S2 the bending stiffness after slippage

occurs. The transition from S1 to S2, which in Figure 3 is idealized as sudden, is in reality smooth, since

all wires do not slip exactly at the same level of curvature. However, for most practical purposes the

idealization is sufficient to model the global flexural behavior in a fulfilling manner.

M

s

S1

S2

Figure 3, left: Birdcaging of armouring wires reproduced experimentally by NKT-Flexibles, right: Idealized

flexural behavior of flexible pipe.

The consequence of failure in a flexible pipe applied as riser for extraction of oil and gas from a subsea

reservoir may be oil spill, possibly with tremendous environmental consequences. However,

installation of a replacement riser, which possibly has to be manufactured first, leads to production

down time. Furthermore, shut down of an oil well causes loss of the reservoir pressure, which must be

reestablished before the production can be continued. Due to these reasons, both oil companies and

flexible pipe manufactures are extremely cautious with respect to modifying a design, which in the

field already has shown potential as proven concept. Due to these commercial issues, the offshore

business is dominated by a conservative approach to the design process. Therefore, it is in general

considered feasible to accept the significant costs related to laboratory experiments for validation of

existing designs rather than to attempt to redesign the product in new and innovative ways. Tests are

used to reconstruct failure modes in flexible pipes, and obtained results are used for validation or

calibration of theoretical models. Examples of how the limit state design related to different failure

mechanisms may be defined are summarized by Jiao [12].

Studies of failure in flexible pipes by wire buckling have to a very high extend been motivated by the

Brazilian Oil and Gas Company Petrobras. The main reason for this is that the Campus Bassin in the

Atlantic Ocean, where Petrobras operates, contains the known oil and gas fields, which are located

deepest. Due to the extreme hydrostatic pressure at these water depths, Petrobras have been the first

company to encounter wire buckling. Related failure modes impose a danger to the structural integrity

of a flexible pipe, when this is subjected to compressive loads. Such forces usually occur during pipe

laying, since the pipe in this scenario is empty. Two types of buckling have been encountered by

Petrobras and were described and reproduced experimentally by Braga and Kaleff [2].

It is noted, that both rigid and flexible pipes in compression may fail due to global instability. The term

lateral buckling has in the past often referred to the arising buckling phenomena. Theoretical methods

for prediction of such failure modes are well-described in the literature, see for example Vaz and Patel

[13]-[16]. Such mode of failures should, however, not be confused with the lateral wire buckling

failure mode, which refers to instability of internal components within the pipe wall.

The first type of wire buckling is the birdcaging mode, see Figure 3, which is characterized by large

localized radial deflections of the armouring wires. The mode was encountered for the first time in

1977. Detection of birdcaging led to that radial wire deflections in most known pipe designs for deep

waters were limited. This was done by introducing an extra pipelayer, namely, a high strength tape

(ABC-layer) wound on the outer side of the tensile armour layers. As this failure mode was

encountered, Petrobras initiated the so-called Deep Immersion Performance tests (DIP-tests). By these

tests, it is under full-scale offshore conditions examined, if a flexible pipe can be installed without

occurrence of armour wire buckling, see Figure 4.

During DIP-testing of flexible pipes with ABC-layers, lateral wire buckling was encountered in 1997. It

was observed, that buckling leads to disorder of the tensile armour wires in the circumferential pipe

direction, see Figure 5. The failure mode was discovered for flexible pipes subjected to compressive

loads and repeated bending cycles. Generally, the risk of lateral buckling is considered largest, when

the outer sheath of the pipe is damaged. This is due to the fact, that external pressure in the case of

flooded annulus to a much lower extend introduces contact stresses causing frictional effects to limit

wire slippage, than when the outer sheath is intact. Damaged sheaths are quite common during pipe

laying and wet annulus conditions must therefore be considered when designing flexible pipes against

lateral wire buckling. Due to this issue, DIP-tests are performed first with dry and afterwards flooded

annulus conditions.

Lateral wire buckling imposes a larger danger than birdcaging, since it often cannot be detected by

visual inspection of a test specimen without dissection. In order to study the failure mode under

controlled conditions, it was reconstructed experimentally in the laboratory. Braga and Kaleff [2],

developed the first principle for testing of flexible pipes in form of mechanical test benches. By this

experimental principle, a mounted flexible pipe could be subjected to repeated bending cycles.

Compression was applied by mounting a smaller flexible pipe inside the test sample pipe and

mounting those on the same base flange in one end of the setup. Furthermore, this end of the setup

was mounted in a manner enabling the test pipe sample to be torsionally free. As the inner pipe was

tensioned, the test sample pipe was compressed. Braga and Kaleff concluded that birdcaging could be

reproduced at compressive load levels corresponding well to the hydrostatic pressures causing failure

in the field. However, lateral wire buckling occurred at compressive load levels lower than detected

during field testing. The conclusion was furthermore drawn, that lateral buckling was associated with

shortening of the test sample and a pipe twist of the torsionally free pipe end. These observations

were confirmed during the experimental work included in the present project.

Similar experiments have been conducted by flexible pipe manufactures. Furthermore, the test

principle has been extended to include experiments conducted in mechanical bending frames mounted

inside pressure chambers. By this method, the stabilizing effect of hydrostatic pressure with dry

annulus conditions could be studied. Furthermore, compression could be applied directly to the pipe

end-cap without the necessity of a tension pipe in the bore of a test sample pipe. Experiments are

documented by Secher et. al. [3] and [17], Bectarte and Coutarel [18], and Tan et. al. [19]. The wire

deformation modes obtained experimentally in this project correspond well to the buckled wire

configurations, which were presented in [3], see Appendices E and G. Failure was during both

laboratory tests and DIP-tests detected in the inner layer of armouring wires. The reason for this is

deemed to be, that the inner layer of tensile armour experiences a larger compressive force than the

outer layer when the pipe is loaded. This is due to a minor radius of the layer, a lower number of wires

than in the outer layer, and complex interaction effects between the internal pipe components. The

outer layer may be prone to lateral buckling when the pipe is exposed to large externally applied

torsional moments. Such loads may occur during spooling of flexible pipes, but are not considered in

the present project.

While lateral wire buckling to a high extend has been investigated experimentally in the laboratory,

very few theoretical studies have been published. Custdio [20] formulated the equations of

equilibrium for the tensile armour wires analytically for a straight pipe. Buckling was considered a

bifurcation-problem formulated on basis of a radially elastic pipe wall. This effect was simulated by

use of elastic foundations. However, Custdio interpreted several of the obtained modes of wire

deformation as being unphysical and discarded those. Recently, further studies have been conducted

by Vaz and Ricci [13], using finite element modeling of a single armouring wire taking frictional effects

into calculation. Brack et. al. [21] also applied finite element methods for calculation of the load

carrying ability of an armouring wire. However, in none of the cases cyclic loadings were considered,

and obtained results were not compared to experimental data.

Installation vessel

Sea level

Flexible pipe in

free-hanging catenary

Touch-down

zone

Sea bed

Figure 4, left: Photo of flexible pipe touchdown zone during DIP-testing taken by ROV (underwater robot),

right: Principle sketch of flexible pipe during installation.

M1

M2

M3

M4

Figure 5 Modes of deformation in the inner layer of armouring wires detected during dissection of 6 flexible

pipe after it had been subjected to lateral buckling laboratory testing, M1: No sign of buckling, M2: Lateral wire

buckling detected as large localized gaps, M3: Lateral wire buckling detected as localized S-shaped mode of

deformation, possibly of a plastic nature M4: Severe state of lateral wire buckling, it is estimated that the yield

strength of the wires has been exceeded which has led to plastic deformations.

Considering a flexible pipe subjected to longitudinal compression, the wires may, if the external

pressure is insufficient to limit wire slippage, slip in the lateral direction. The wires may ultimately, as

bending is applied, reach an equilibrium state in which the load carrying ability is reduced with

respect to the initial helical configuration. This is visualized in Figure 6. 1-2. Reduced load carrying

ability may lead to a state in which the torsional equilibrium between the tensile armour layers can no

longer be maintained. The consequence of this may be, that compressive longitudinal pipe strain

couples to the pipe twist, which increases as bending cycles are applied. The twist occurs in the

direction at which the outer layer is wound. The reason for this is, that the magnitude of the

compressive load in this layer must decrease in order to maintain torsional equilibrium. This is

visualized in Figure 6.3.

1.

2.

3.

Papp

Papp

Papp

Figure 6 Schematic drawing of the lateral buckling failure mechanism in flexible pipe tensile armour layers, Blue

wires: outer layer of armour wires, Red wires: inner layer of armour wires, 1, compression and cyclic bending:

Compression and repeated bending cycles are applied to a flexible pipe, 2, nonlinear geometrical softening of

inner layer: As bending cycles are applied, the wires contained in the inner layer slip towards a configuration in

which the load carrying ability is reduced with respect to both the initial configuration of the layer and to the outer

layer of tensile armour, 3, torsional coupling between layers: This causes a pipe twist in order to maintain

torsional equilibrium, which leads to relaxation of the compressive loads in the outer layer of armouring wires. This

twist adds further straining to the inner layer of armouring wires.

pu

B

R=

Flexible pipe

End fitting

pb

pu

Seabed

Figure 7, left: Schematic drawing of a flexible pipe idealized as having constant bending radius during laboratory

testing, right: Schematic drawing of the touchdown zone of a flexible pipe during DIP-testing.

The coupling between the layers is mainly due to the fact that all layers are fixed in end-fittings in both

ends of the pipe. Therefore, all wires are subjected to the same strain and twist on the boundaries.

While this relaxes the outer layer, the inner layer is stressed further by the pipe twist. The inner layer

is loaded further, since the wires, which are already compressed, are twisted against the winding

direction. This causes further shortening of the wires and may at some point lead to plasticity. This

may cause permanent deformations and, ultimately, failure of the pipe structure.

The lacking one-to-one correspondence between results obtained from experimental lateral buckling

experiments in mechanical rigs and DIP-tests is a complex issue, for which no absolutely fulfilling

explanation has yet been given. No measurements of the pipe responses during pipe laying are

available, since it is a tremendous challenge to solve the problems related to instrumentation of a

flexible during a DIP-test. Information regarding the touchdown zone of a flexible pipe is therefore

only available from images transmitted by ROVs (small subsea robotsystems) during the tests. An

example of such an image is presented in Figure 4. However, the view of an ROV is quite limited, and it

is often difficult to identify possible 3D-effects. It is therefore difficult to know the exact circumstances

under which lateral buckling occurs with respect to seabed interaction and possibly out-of-plane

motion of the test pipe. The reasons for the lacking correspondence are therefore of a highly

speculative nature. However, the boundary conditions used for laboratory experiments may be

considered and compared to the installation scenario, see Figure 7. It is clear that the pipe end denoted

B in the experimental setup is torsionally fixed, while this end during DIP-tests may only be partially

fixed. This effect may be one reason for the lacking correspondence. Furthermore, the pipe may not

necessarily be subjected solely to in-plane bending. NKT-Flexibles has as the present project was

finalized never encountered lateral wire buckling in the field. Therefore, no exact knowledge regarding

in-field wire buckling has been established.

Another issue regarding the lacking one-to-one correspondence between the two test principles is the

difficulties related to determination of the limit compression causing failure during a DIP-test. In order

to determine the limit compressive load, the test had to be performed several times at varying water

depths until failure was detected. The necessary process would be extremely time consuming and

expensive. However, an example of such a campaign is documented by Secher et. al. [3] and [17].

Furthermore, for obvious commercial reasons, flexible pipe manufactures, oil and gas companies and

offshore contractors wish to validate the installability of a given flexible pipe. Therefore, the main

objective of a DIP-test is usually to conclude the test without failure rather than to determine limit

compressive loads. Furthermore, a flexible pipe may be installed at water depths at which the limit

compressive load is exceeded, if only a limited number of bending cycles insufficient to trigger failure

is applied. Secher et al. [3] and [17] claimed to have obtained correspondence between DIP-tests and

laboratory tests conducted in pressure chambers, however, using very short test pipe samples. In

order to draw conclusions regarding the validity of this issue, experiments with varied length of test

pipe sample must be conducted in order to examine the influence of test sample length on the

obtained results.

The experimental results obtained in the present work suggest, that buckling may develop in the

elastic regime. Furthermore, in some cases a tremendous number of bending cycles must be applied,

before the yielding limit of the wire steal is exceeded. These observations may alter the understanding

of the phenomenon. Presently, the only criteria for when lateral buckling is considered as failure

available are contained in certain flexible pipe specifications provided by oil and gas companies. The

API 17J code [1] only states that wire buckling should be considered, but does not prescribe by which

methods. This demonstrates the need for clarification of the difference between the distinct terms

lateral wire buckling and failure by lateral wire buckling. From a design perspective, the limit state

may be taken as the conditions, at which compressive loads cause torsional imbalance of the pipe

structure. However, yielding of the wires has not necessarily occurred in this state. Therefore, the limit

state may alternatively be considered as the conditions in which compressive loads and torsional

imbalance have led to permanent deformation of the pipe structure due to plastic behavior of the

wires. The limit state design issue is elaborated further in section 5.8.

The overall objective of the work presented in this thesis is to gain further insight in the mechanisms

leading to lateral wire buckling. This has been conducted by experimental means, but also by

theoretical studies in the field of wire mechanics. Key aspects of the present work are:

Experimental reconstruction of the lateral wire buckling failure mechanism in eight available

test sample pipes. This part of the project has included construction of the necessary

mechanical gear, execution of bending tests, and pipe dissections.

Derivation of the equations governing the equilibrium of an armouring wire within the wall of

a flexible pipe bent to a constant radius of curvature.

Investigation of lateral wire buckling of a single armouring wire with small imperfections.

These studies are based on full non-linear analyses for calculation of the wire equilibrium

paths.

Modeling of the mechanical behavior of both layers of armouring wires arising from lateral

wire buckling. Analyses will be based on multiple single wire analyses and proper

idealizations.

Estimation of how effects such as friction, boundary effects, and adjacent pipe layers influence

the load carrying ability.

The ultimate goal would be to deliver a set of experimentally validated design rules for flexible pipes.

These could then be implemented directly in the pipe design process and in the installation limitations

for a given flexible pipe design. However, due to lack of information regarding the exact conditions

encountered during pipe installation, the scope of work has been limited to prediction of the limit

compressive load causing lateral buckling during a laboratory experiment. The methods may enable

prediction of the outcome of a DIP-test. This may be addressed either on basis of calibration against

DIP-test results or on basis of future measurements taken during DIP-tests, which reveal the true

conditions encountered during pipe installation.

A visualization of the scope of work is presented in Figure 8.

friction, adjacent layers and dynamic loads

Global analysis of

armouring layers

Experimentally validated

Design tool

Simplified analysis

with adjacent layers

Single wire

imperfection analysis

DIP-test results /

measurements

friction and cyclic loads

Present project

Frictionless single

wire equilibrium

Laboratory

experiments

Preliminar understanding of

the lateral wire buckling problem

Figure 8 Schematic overview of the work contained in and related to the present project.

10

11

Laboratory experiments were in the present project executed in mechanical test rigs enabling

application of compression and repeated bending cycles. Two different test rigs were used. Both test

rigs were located in the laboratory facilities of NKT-Flexibles in Brndby, Denmark.

The first test rig was constructed for R&D purposes two years before this project was initiated by

NKT-Flexibles engineer Jan Rytter, see Figure 11. Four 6 riser pipes were tested using this rig, which

then was damaged attempting to test the first of two 8 riser pipe samples. The rig was afterwards

upgraded, partly before this project was initiated, but also as part of the present project, see Figure 12.

This allowed the tests of the two 8 pipe samples. Furthermore, two 14 jumpers were tested. After the

experiments had been completed, the pipe samples were dissected in order to detect if failure had

occurred in the tensile armour layers. The tests were conducted with special very short custom build

pipe end-fittings rather than end-fittings used for production pipes. In the pipe end-fittings the wires

were fixated by grouting and welding in a manner similar to the principle used in production endfittings. This principle was chosen in order to obtain a shorter end-fitting design. Furthermore, it

allowed end-fittings to be mounted in a manner, which was less complex than the standard end-fitting

mounting procedure.

In general, experimental studies should be designed taking appropriate statistical means into account.

A number of experiments should be carried out to test if the results are normally distributed to a

reasonable extend. Furthermore, the mean and variance should be calculated on basis of several

experiments to test how much the outcome of the experiment varies for fixed input. Statistical

methods appropriate for such analyses are well described in the literature, see for example

Montgomery [22].

a.

P

L

e

b.

c.

P

e

dy

M

P

dS

Figure 9 Schematic drawings of test setup, a) Test principle used for laboratory experiments, Pipe end A

longitudinally and torsionally fixed, Pipe end B longitudinally and torsionally free, b) Pipe geometry, c) Pipe

equilibrium.

12

However, industrial experiments conducted for validation of computational models applied in flexible

pipe design are extremely time consuming and expensive. The scale of the experiments conducted in

order to qualify flexible pipe design methods can with respect to execution time and cost often be

compared to, for example, fatigue tests of wind turbine blades, see for example Kensche [23]. Another

example of similar experiments are those for determination of pile-soil interaction effects used to

validate computational methods in civil engineering, see for example Imamura et. al. [24]. Therefore, it

is in general necessary to assume the outcome of such experiments to be reasonably deterministic.

This corresponds to neglecting effects caused by statistical variance of the experimental outcome.

Despite the fact that it from a statistical point of view is difficult to justify this approach to

experimental work, it is crucial to assume the outcome of the experiments deterministic, when

experiments like the present are carried out with limitations in cost and time. In the present project, at

least half of the conducted experiments would have been used only to obtain a rough measure for the

statistical variance, if a classical approach to design and analysis of experiments had been followed. It

was therefore chosen not to consider these issues when planning the tests.

Figure 10 Overview of upgraded test rig based on CAD model, 1) Test pipe sample, 2) Static frame end A

with hydraulics for tensioning of the inner pipe, 3) Moving frame end B, test pipe supported by system of

bearings allowing a test pipe twist, 4) Idealized bending device, 5) Supporting frame (idealized model).

The chosen test principle is to a wide extend similar to the experimental principle used by Braga and

Kaleff [2]. A test pipe sample with custom built test end-fittings was mounted in a mechanical bending

rig with boundary conditions shown in Figure 9. It was chosen to construct the test rig such that

bending was applied in the vertical plane, in order to avoid gravitational effects to cause out-of-plane

curvature. An example of a bending rig, in which the curvature is applied in the horizontal plane, is

described by Secher et al. [3] and [17]. It is noted, that even though similar experimental work has

already been conducted and documented in [2],[3], and [18], results are due to commercial issues not

available. Therefore, the present work has included experiments in order to obtain results which may

validate the developed computational models.

13

Figure 11 GI Lateral buckling test rig during the initial trial test of a 6 riser pipe sample, bending applied using

electrical drives.

Figure 12 G2 lateral buckling test rig during the test of the first 14 jumper, bending applied by a hydraulical

system.

14

The end-fittings were mounted on base-flanges connected to pinned H-formed frames. These were

above the test sample pipe connected by a bar, which could be shortened. This system constitutes the

bending device of the test rig, since shortening of the top bar causes the test pipe sample to bend

downwards as the H-formed frames rotate. The key difference between the initial and the upgraded

test rig is, that larger and more robust mechanical components were used for the upgrade. This

allowed larger pipes to be tested. Furthermore, the bending device in the initial setup was constituted

by a system of electrical screwdrives. This was replaced with a more robust hydraulic system in the

upgraded rig.

A maximum compressive load of 1000 kN could be applied in both bending rigs. In order to protect the

crew operating the rig in case of failure of the setup or rupture of the test pipe sample, a number of

two meter high concrete blocks shielded the position in which the computers used for datalogging and

control were placed. Furthermore, it could during the planning of a few experiments with extreme

load inputs not be ensured that the test pipe sample behaved in the desired manner. These

experiments were therefore conducted at times, when the test facilities and workshops were empty.

Base flange

Flange

Test end-fitting

Test pipe (flexible)

Test end-fitting

Tension pipe (flexible)

Base flange

Tension pipe endbody

Compression hydraulics

Threaded bar

v+

Bearing holder

Moving frame end

Figure 13 Sectional drawing of assembled test pipe sample, compression applied to test pipe sample when the

compression hydraulics moved in the direction marked v+, means for pipe centralization not included.

Figure 14, left: Assembled bearing arrangement for torsionallly free pipe end, right: Plastic centralizers for test

of the 8 risers and 14 jumpers constructed as two half-parts assembled by two pins and a band-it metal strip.

15

Figure 15, left: Unmounting of tension pipe with plastic centralizers during teardown after test of 14 jumper,

right: Dissected end-fitting from 6 riser pipe, wire fixated with weldings to a weld-ring.

Compression was applied by tensioning the inner pipe, which was centralized in the bore of the test

sample pipe by appropriate means. A sketch demonstrating this principle is shown in Figure 13.

During the execution of the tests of the 6 riser, polymeric insulation materials were used, since the

difference between the bore diameter of the test sample pipe and tension pipe was small. For the tests

of the remaining pipes, custom built plastic centralizers were used, see Figure 14 and Figure 15. The

torsionally free end of the test pipe sample was supported by a system of bearings allowing this end to

twist. The bearing system was constituted by two axial bearings transferring the longitudinal forces

from the tensionpipe to the surrounding structure for each direction of motion in the compression

hydraulics. Furthermore, two axial bearings enabling transferring of bending were applied. The

bearing arrangement is shown in assembled form in Figure 14.

Measurement

Compressive load on test pipe sample

Device

Load cell (max. 1000 kN)

Bending load

sample in five equally spaced points

Extensometers

points (for radius of curvature estimates)

Inclinometers

Inclinometer

overbar)

Extensometer

Position sensor

Temperature sensor

16

Several measurements were taken during the experiments, see Table 1. The loads applied for bending

and tensioning of the inner pipe were measured by load cells. Since the experimental principle

simulates flooded annulus conditions, by which external pressure acts on the inner liner, the

longitudinal load applied during experiments simulates hydrostatic pressure induced by water depth

acting on the liner area. The equivalent compressive load to be applied during a laboratory test is

therefore in terms of outer liner diameter Dliner, the water density , the water depth h, and

gravitational constant g

2

PREC ,wet = Dliner

hg

4

(1)

The system for compressing the test sample pipe was controlled directly by adjusting the hydraulic

station, until the desired compressive load was measured by the load cell mounted for this purpose.

The system used to bend the test pipe sample was in the G1 test rig controlled from a control box

placed directly behind the bending rig. This control box interfaced directly to the PLC controlling the

electrical drives. In the G2 test rig, in which bending was applied hydraulically, the bending system

control was programmed in a labview application. This could run parallel to the interface used for

datalogging on the computer integrated in the setup. The maximum radius of curvature was calculated

on basis of the response from two inclinometers placed on each side of the pipe midpoint. With the

rotation responses i+1 and i from transducers placed in Si+1 and Si, the pipe curvature is approximated

as a finite difference by

=

i +1 - i

S i +1 - S i

(2)

As described in section 1.2, lateral wire buckling is associated with a twist of end B, a shortening, but

only a small change of pipe circumference. These values were all measured during the experiments. All

instrumentation was by appropriate means connected to a digital interface that enabled the

measurements to be shown online during the experiments. Furthermore, the logged responses were

saved in an ascii-file for further studies and postprocessing. The radius of curvature of the pipe

midpoint was processed online and written to the computer interface during the experiments. This

allowed small adjustments necessary due to changes of the flexural behavior of the test pipe sample.

17

In this chapter, the experiments conducted by means described in section 2 are summarized. A brief

overview is presented in Table 2. The pipedesigns of the flexible pipe test samples used for the

experiments are contained in section 5.1. Initially, the very first lateral buckling experiment, referred

to as the initial trial test, will be described. The outcome and the knowledge gained by conducting this

experiment were crucial to the manner, by which the following experiments were designed.

Afterwards the three series of lateral buckling experiments will be described. The full documentation

of the measurements and test sequence from the experiments are described in the NKT-Flexibles

documents [45]-[71]. These were written as part of the present project.

In general, the tests were conducted in accordance with the following scheme:

Initially, a test of anti-birdcaging performance was conducted. During this part of the

experiment, the compressive load was applied to a straight test pipe sample. The objective was

to ensure, that the compressive loads applied during the actual lateral buckling test would not

cause birdcaging. Static compression was applied with a holding time of 20 minutes.

The test pipe sample was tensioned by a load of 50 kN and bent a number of times (usually

between 10 and 20) in order to ensure, that the wires contained in the tensile armour layers

were as close to the initial helical configuration as possible. Handling prior to the test may have

caused wire slippage.

Finally, the actual lateral buckling experiment was performed by applying a prescribed load

program. During most experiments, a maximum of 1200 cycles of repeated bending from

straight configuration was chosen as a reasonable estimate for a DIP-test. In similar

experimental studies, a total of 800 bending cycles was chosen, see [3].

The first lateral buckling experiment was conducted in the G1 bending rig shown in Figure 11 in April

2008 using a 6 riser sample. The purpose of this experiment was to test if the experimental principle

was capable of reproducing lateral buckling failure in a flexible pipe. Furthermore, it was desirable to

measure the pipe responses in order to study how a flexible pipe should be instrumented during

lateral buckling experiments. This experiment was necessary, since the only information regarding

how a flexible pipe would respond when lateral buckling failure occurs, was obtained by studying

publications from the OMAE, see [2],[3], and [18]. While instrumenting the test pipe sample prior to

the experiment, it was not expected that the pipe shortening, twist and radial expansion would be as

extreme as the pipe response values which were actually encountered. This caused several

transducers to be mounted in a manner, by which the measuring interval of the instrument was

exceeded. The unfortunate consequence was, that certain values of the pipe responses are unknown

after a given number of bending cycles had been applied. Furthermore, the inclinometer for

measurement of the pipe twist was mounted in a manner, which was incapable of sustaining the

severe twist encountered. This transducer was therefore torn off during the test. During tear-down of

the test pipe sample after the experiment had been concluded, it was determined that the twist of pipe

end B was approximately 90 degrees after unloading. The twist which was measured during the test is

shown in Figure 16. The compressive stroke and change of circumference are shown in Figure 17 and

Figure 18. These three responses constitute the primary mean for detection of lateral wire buckling.

18

15

Upper measuring limit of instrument

Instrument

torn off

10

Twist (deg)

48 bending

cycles

-5

Instrument

unmounted

-10

-15

500

1000

1500

2000

time (s)

2500

3000

3500

300

Upper measuring limit of instrument

250

48 bending

cycles

200

180 bending

cycles

Experiment stopped,

tests bends performed

150

100

50

1000

2000

3000

4000

5000 6000

time (s)

7000

8000

9000 10000

Figure 17 Compressive stroke measured during the initial trial test of a 6 riser.

19

OD1

OD2

OD3

OD4

OD5

30

25

180 bending

cycles

20

15

10

1000

2000

3000

4000 5000

time (s)

6000

7000

8000

9000

Figure 18 Change of circumference measured during the initial trial test of a 6 riser.

During dissection of the pipe sample, the following conclusions were drawn:

A severe remaining twist of approximately 90 degrees of pipe end B was measured. This was

the only sign of failure within the pipe wall, which could be observed by naked eye.

The anti-birdcaging tape of the test pipe sample had a highly irregular pitch. This may have

influenced the test result.

No signs of failure could be detected in the outer layer of tensile armour wires.

The inner layer of armoring wires was found in a severely buckled state, see Figure 1. It was

estimated, that plasticity in the wires had occurred during the test.

The radius of curvature was during the initial trial test estimated on basis of measured deflections.

These responses were logged by extensometers placed between the top-bar and the test pipe sample.

It was after the experiment found difficult to estimate the radius of curvature. By spline-interpolations

and regressions it was determined that at least 5 meter of bending radius had been applied during the

experiment, but the value at the pipe midpoint may have been lower. The difficulties were mainly

imposed by the fact, that estimating curvature included differentiating the deflection response twice,

which amplified noise in the responses significantly. The problem was during the remaining

experiments solved by replacing the extensometers with inclinometers mounted directly on the pipe

sample. The consequence is that the radius of curvature of pipe midpoint during the remaining tests

was estimated on basis of the rotation of the test pipe sample rather than on basis of deflections, see

equation (2).

As the electrical drives were turned on during the initial trial test, significant influence on the

responses from the remaining transducers was detected. Furthermore, a few transistor radios located

in the work shop began transmitting noise. The effect was found to be caused by the frequency

transformers in the electrical drives. In order to minimize noise in the logged measurements from the

remaining transducers, these were shielded electrically.

It was on basis of the detected behavior of the test pipe sample chosen to set a pipe twist of 45 degrees

as the stop criteria for future experiments.

20

Pipe ID

Experiment ID

Load

cycle

ID

Number of

bending

cycles

Applied

compression (kN)

Bending

radius (m)

*4

Result

obtained by

dissection *3

Pipe twist

before

unloading (deg)

Initial

trial test

0 *2

180 / 48*1

265

Failure

>90

Test

series 1:

204

265

11

Failure

45 (increasing)

800

80

11

No failure

<1 (stable)

II

392

210

11

Failure

45 (increasing)

1200

160

11

No failure

3 (slowly

increasing)

II

151

265

Failure

45 (slowly

increasing)

6000

277

18 to 24

No failure

<1 (stable)

II

1200

269

7 to 9

No failure

6.5

(increasing)

III

1200

411

9 to11

Failure

27

I*2

18

950

12

Failure

10 (rapidly

increasing)

II

1200

950

12

Failure

17 (increasing)

6 riser

L= 5m

Test

series 2:

14

jumper

L= 7.5m

Test

series 1:

1 *2

1200

700

12

Failure

27 (slowly

increasing)

8 riser

II

1200

300

12

No failure

<1

III

2400

400

12

No failure

15 (slowly

increasing)

L= 5m

* 1) On basis of the datalog it was after the test estimated that failure by lateral wire buckling occurred after 48

bending cycles,

*2) Experiment / load cycle was due to extreme load input conducted at an appropriate time when the test

facilities were empty (for example at night or started in the afternoons before holidays/weekends).

*3) Failure should in this context be interpreted as detection of lateral wire buckling during dissection of the test

pipe sample. The deformation patterns detected in the inner layer of armouring wires are included in section 5.5.

*4) Bending applied from straight configuration to prescribed value unless stated in the table.

Table 2 Summary of conducted experiments.

After the first lateral buckling experiment was conducted prior to initiation of the present project, the

G1 mechanical test rig was disassembled and moved to another workshop. As a consequence, the

experimental work was delayed. The present series of three 6 test pipe samples was tested and

dissected from October to December 2008 after approximately two months of bending rig reassembly.

Obtained results are shown in Figure 19.

The first experiment was constituted by a single load cycle, with the same compressive level as the

initial trial test but with the maximum bending radius of the pipe midpoint reduced to 11 m. The stop

21

50

80

Test I, LC 1, 265 kN, 11 m bending

Test II, LC 1, 80 kN, 11 m bending

Test II, LC 2, 210 kN, 11 m bending

Test III, LC 1, 160 kN, 11 m bending

Test III, LC 2, 265 kN, 8 m bending

30

Pipe twist (deg)

60

Compressive stroke (mm)

40

20

10

Test II, LC 1, 80 kN, 11 m bending

Test II, LC 2, 210 kN, 11 m bending

Test III, LC 1, 160 kN, 11 m bending

Test III, LC 2, 265 kN, 8 m bending

70

50

40

30

20

10

0

-10

200

400

600

800

Number of bending cycles

1000

1200

-10

200

400

600

800

Number of bending cycles

1000

1200

Figure 19, left: Pipe twist of the 6 risers applied bending cycles, right: Compressive stroke of the 6 risers.

criteria of 45 degrees was reached after 204 bending cycles. Failure was detected in the inner layer of

armoring wires during dissection. The modes of deformation were partly S-shaped, partly large gaps

also with an S-like shape (of the types M2 and M3, see Figure 5). The result showed, that reducing the

maximum bending radius of the pipe midpoint for a fixed compressive level had as consequence, that a

higher number of bending cycles had to be applied in order to trigger failure.

The second experiment in the present test series was constituted by two load cycles. The first load

cycle showed, that 800 bending cycles with 80 kN of compressive load could be applied without

causing failure to occur. This load cycle was conducted in order to ensure, that experiments could be

performed without leading to lateral buckling failure. The second load cycle was performed with a

compressive load of 210 kN but still with 11 m of bending radius. This caused failure in the test pipe

sample. Lateral buckling was during dissection of the test pipe sample detected as large gaps in the

inner layer of tensile armour wires.

The third experiment in the test series was also constituted by two load cycles. The first of those had

as objective to decrease the load with respect to the second experiment. A load level of 160 kN was

chosen. After 1200 bending cycles, the twist was approximately 3 degrees and progressing very

slowly. It was concluded, that if the experiment simulates installation, the limit compressive load of the

test pipe sample is between 160 and 210 kN for an 11 meter bending radius. The second load cycle

was performed with 265 kN compression and the maximum bending radius of the pipe midpoint

increased to 8 meters. This caused failure to occur. The progression in the twist response was during

this load cycle very rapid compared to what had been observed during previous experiments during

the last few bending cycles. The experiment was stopped with a twist of approximately 45 degrees, but

due to the very fast progression of the response, this was mainly due to coincidence. The last load cycle

reveled, that lateral wire buckling may lead to instability progressing at a quite high speed as cyclic

bending is applied.

3.3. Test series II, 14 jumper test pipe samples in G2 bending rig

After the first series of lateral buckling experiments, an 8 riser test pipe sample from the third series

of experiments was attempted tested in the original lateral buckling rig. However, this test pipe sample

had a bending stiffness so large, that it caused damage in the test rig due to misalignment of the screw

drives in the bending arrangement. After tear down of the test pipe sample without any sign of failure,

the test rig was upgraded as part of a NKT-Flexibles commercial project. The upgrade, including

assembly of mechanical parts and construction of a new tension pipe, was of approximately two

months duration.

22

The second experimental series using two 14 test pipe samples was started in June and July 2009

with the first test pipe sample. The test pipe sample used was not designed as a riser connecting a

reservoir at the seabed to a floating platform, but as a jumper connecting the top of a tower of rigid

steel risers to a floating platform. The load program, which is summarized in Table 2, was defined on

basis of client specifications as simulation of field conditions. Initially, 6000 bending cycles were

applied simulating shut down periods expected to occur in the pipe service-lifetime. For each 1200

bending cycles, the test was interrupted and the pipe sample was tensioned and bent repeatedly. This

procedure was followed in order to attempt to draw the armoring wires back towards the helical

configurations, before the experiment was continued. This simulates, that the test pipe is

repressurized and tensioned. No signs of failure could be detected during this part of the experiment.

After this load cycle, which was of a duration of a few days, two load cycles simulating the installation

process were conducted. The second of these load cycles led to severe pipe twist and shortening of the

test pipe sample, see Figure 20. However, the progression of these responses was, as cyclic bending

was applied, significantly slower than detected during the first test series, see Figure 19. During

dissection of the test pipe sample, lateral buckling was detected in the inner layer of tensile armour as

localized gaps of moderate size. However, it was unclear if plasticity had occurred in the wires.

The second lateral buckling experiment of the present series was conducted in December 2010. The

test objective was, motivated by the NKT-Flexibles qualification, to examine the response of a flexible

pipe to extreme loads. Therefore, a load scenario close to maximum rig capacity was chosen.

Performing 20 bending cycles with this load input and afterwards applying 1200 bending cycles with

decreased loads simulating service should constitute an experimental simulation of the long term

effects of extreme loads. During the first load cycle simulating extreme loads, the pipe twist increased

in a stable manner until 18 bending cycles had been applied. Then a very rapid twist occurred and the

experiment was stopped, see Figure 21. No reason for this mechanical behavior could be detected in

the measured responses or by visual inspection of the test pipe sample. It was chosen to continue the

experiment and load cycle II was initiated. The pipe sample was left for approximately 48 hours

between the two load cycles. Furthermore, tension and repeated bending was applied before load

cycle II was initiated. The twist increased in a more stable manner during the last part of the

experiment. During dissection of the test pipe sample it was determined, that gaps of moderate size

occurred throughout the inner layer of armouring wires without signs of localization. No other reasons

for the rapid twist during load cycle I could be found. It therefore seems, that some global mode of

buckling has been triggered by the extreme loads applied.

30

100

Logged response

LC 2 started

LC 3 started

25

80

Compressive stroke (mm)

20

Pipe twist (deg)

Logged response

LC 2 started

LC 3 started

90

15

10

70

60

50

40

30

20

10

-5

1000

2000

3000

4000

5000

6000

Number of bending cycles

7000

8000

9000

1000

2000

3000

4000

5000

6000

Number of bending cycles

Figure 20, left: Pipe twist of the first 14 jumper, right: Compressive stroke of the first 14 jumper.

7000

8000

9000

23

10

18

16

14

5

4

12

10

8

6

2

1

0

500

1000

1500

2000

2500 3000

time (s)

3500

4000

4500

5000

200

400

600

800

1000

number of bending cycles

1200

1400

Figure 21, left: Pipe twist of the second 14 jumper, load cycle I, right: Pipe twist of the second 14 jumper, load

cycle II.

200

30

180

25

160

20

15

10

140

120

Test I, LC 1, 700 kN, 12 m bending

Test II, LC 1, 300 kN, 12 m bending

Test II, LC 2, 400 kN, 12 m bending

100

80

60

40

Test II, LC 1, 300 kN, 12 m bending

Test II, LC 2, 400 kN, 12 m bending

500

1000

1500

Number of bending cycles

2000

20

0

500

2500

1000

1500

Number of bending cycles

2000

2500

Figure 22, left: Pipe twist of 8 risers, right: Compressive stroke of 8 risers.

3.4. Test series III, 8 riser test pipe samples in G2 bending rig

The third test series was conducted from February to March 2011 using two 8 riser test pipe samples

in the upgraded test rig shown in Figure 10. The first experiment was conducted with load inputs

corresponding to DIP-test conditions encountered for a pipe sample with the same pipedesign at 1500

meter of water depth in the Brazilian Campos Basin. Considering the pipe twist and compressive

stroke in Figure 22, which were measured during the experiment, the responses can be observed to

remain stable and increase slowly until a certain number of bending cycles have been applied. Then

the measured values can be observed to increase in a more progressive manner throughout a number

of bending cycles and afterwards soften and progress at lower speed as cyclic bending is applied. This

behavior differs from the responses measured during the first test series in which the responses

continuously progressed until the experiment was stopped, see section 3.2. During dissection of the

first 8 test pipe sample, lateral buckling was detected as large localized gaps in the inner layer of

tensile armour.

The second experiment of this test series had as scope to estimate the compressive limit depth leading

to lateral buckling for a fixed bending radius of 12 meters. Initially, 1200 bending cycles were

24

performed with a compressive load of 300 kN without signs of lateral buckling. During the second load

cycle, the compressive load was increased to 400 kN, which caused the pipe twist and compressive

stroke to increase in a manner similar to the first experiment. Since the responses had not reached a

constant level after 1200 bending cycles had been applied, it was chosen to apply additional 1200

bending cycles. Since this did not cause the responses to stabilize entirely, it was estimated, that lateral

buckling had occurred and the experiment was stopped. However, during dissection of the test pipe

sample, no signs of lateral buckling could be observed and only small variations in pitch length could

be detected in the inner layer of armouring wires.

All test pipe samples used during the experiments were tested until the pipe twist and compressive

stroke indicated that lateral buckling had occurred. During dissections it was determined, that this was

the cases except one, namely the second 8 test pipe sample. It is a very interesting result, that despite

the measured pipe responses exhibited a behavior, which for all remaining experiments was related to

lateral wire buckling, no sign of wire instability could be detected. This suggests that lateral buckling

for pipes with high strength armouring wires may develop solely in the elastic regime, which despite

causing a severe pipe twist and significant shortening in the pipe used during the experiment, did not

lead to failure.

Another highly interesting issue related to the experimental work is, that while the first 8 test pipe

sample failed during laboratory experiments with load input corresponding to DIP-testing, the actual

DIP-test did not fail. This experimental result supports the results obtained by Braga and Kaleff [2],

who concluded that lateral buckling occurred at lower load levels than in the field during laboratory

experiments.

It is also noticeable that all pipes showed signs of localizations in the deformation patterns

encountered in the inner layer of armouring wires except for the second 14 riser, in which a global

buckling mode characterized by gaps of moderate size was triggered throughout the test pipe sample

due to the extreme load input. The modes of deformation encountered during test series II and III

were of a less severe character than those detected during the first series of experiments. While the

inner layer of tensile armour wires in all 6 risers was severely deformed and possibly stressed

beyond the yielding level, this could not be concluded to be the case for the remaining test pipe

samples. Considering the global deformation responses of the pipes in Figure 16-Figure 22, twist and

shortening can also be observed to occur in a more rapid manner during the first test series than in the

following experiments. Furthermore, the measured global deformations did not for any experiments in

the first test series show signs of, that a deformed equilibrium state would be reached causing twist

and shortening to reach a constant stable level. While these responses also grew continuously during

the experiments involving the 14 jumper, however with a slower progression rate, responses seemed

to reach a level with lower rate of progression during experiments involving the 8 riser. This

difference in the measured pipe responses may be explained by the fact that the sudden and rapid

increase of progression encountered for the 6 riser is caused by yielding of the wires of basic grade

steel. This correlates with the fact, that the detected deformation patterns were of a more severe state

than encountered during the remaining experiments. Contrary, considering the 8 riser the lower

progression rate encountered after a large increase of shortening and twist may be explained by the

fact that the yield stress of the high-strength wires of this pipe has not been exceeded, and the wires

have reached an equilibrium state in a deformed configuration. The fact that the responses measured

during this test series continued to increase with a relatively slow rate of progression may be

explained by time-dependant effects such as creep in the polymeric layers and high-strength tape.

While the 6 risers can be concluded to have failed in the sense, that the pipe structure would not be

fit for field purposes, it is unclear if the deformation state of the armour wires detected during the

remaining tests would lead to failure, since repeated bending and tension may cause the wires to slip

back towards the initial helical state. If plasticity had not occurred in the wires, this is deemed likely,

25

but has due to the very limited tension capacity of the mechanical rigs not been examined

experimentally yet.

The conducted experiments have for flexible pipes subjected to the chosen boundary conditions

enabled rough estimates of the compressive load carrying ability of the 6 and 8 risers for fixed

bending radii, while the load carrying ability of the 14 riser is more unclear, since the experiments

were not conducted for fixed pipe curvature.

The deformation patterns observed during the dissections are contained in section 5.5 along with

further elaborations on how the limit state design against lateral buckling may be chosen.

26

27

4. Theoretical work

In this section, an overview of the theoretical work conducted in the present project will be presented.

Initially, elaborations regarding the choice of approach to wire mechanics will be included. Afterwards,

the content of the appended papers will be summarized one by one.

The mechanics of beams have been subject of research for several centuries. The well-known fourth

order differential equation usually referred to as the Bernoulli-Euler equation (3), which is derived

assuming deformations and rotations small, is probably among the theoretical results which to the

widest extend have been applied for engineering analysis

(3)

(4)

Equation (3) relates the fourth order derivative of deflection w as function of longitudinal position x to

the applied distributed load q with EI as measure for the flexural stiffness. The basic underlying

assumption, which is considered valid for long and slender members, is, that shear strains can be

neglected when calculating the deflection. As a consequence, cross sections plane before deformation

are assumed to remain plane in bending. For short members, this cannot be considered to be the case

and an extended formulation developed by Timoshenko can be applied in order to account for shear

effects. Large deflections and rotations were taken into calculation in an approach usually referred to

as the elastica equation, often by formulating the problem in rotations as function of wire arclength s

rather than in deflections, see Bazant and Cedolin [25]

in which P is the longitudinal load applied in a distance e, see Figure 9. The solution to equation (4) is

given by an elliptic integral of first kind. Similar means for formulation of the equilibrium equations

for offshore cables were followed by Terndrup-Pedersen, [26]. The extension of these approaches

necessary to describe the mechanical behavior of long and slender space curved beams has been

investigated for more than a hundred years. Despite of the fact that very few analytical solutions have

been obtained, a set of equations governing the equilibrium of space-curved beams is well-described

in the literature. Reissner [27] used the equations on the vectorial form

dP

+p=0

ds

dM

+ t p+m = 0

ds

(5)

in which P is the sectional beam load, M the sectional moment, p distributed external loads, m

distributed moments and t the beam tangent. It is unclear who first derived the equations, which were

addressed by Clebsch [28], but also contained in Loves book on theory of elasticity [29]. These

equations are in the present project solved in a mixed formulation including both geometry and force

variables for a beam which is partially constrained to a toroid, see section 4.1. This approach leads to a

nonlinear system of six differential equations. The equations of equilibrium for a long and slender

curved beam have been applied to a wide range of problems in numerous publications. Costello [30]

used the equations of equilibrium when addressing large deflections of helical springs. Spillers et. al.

[31] followed a similar approach when calculating stresses in a helical tape on a bent cylinder with the

underlying assumption, that the lay angle was constant. Stump and van der Heijden [32] and Vaz and

Patel [13]-[16] formulated the equations governing the structural behavior of offshore cables and

flexible pipes on basis of equilibrium equations similar to those given in equation (5). Furthermore,

similar equations were used to address the equilibrium of armouring wires in dynamically bent

flexible pipes by Leroy and Estrier [33]. This work constitutes one of very few known examples of

research in which transverse wire slippage is taken into calculation with frictional forces included as

transverse wire loads. However, an experience-based solution form was chosen as basis of the

analysis. Svik [34]-[35] developed a finite element model based on finite-strain continuum

28

mechanics addressing the same problem as Leroy and Estrier, with no restrictions regarding lay

angles and transverse slip. Furthermore, the global radius of curvature could be varied and boundary

effects close to end-fittings taken into calculation. The finite element model was afterwards applied as

basis for computer programs for analysis of armouring wires, see [36], however, assuming slip solely

to occur along curves with constant lay angles. Witz and Tan [6] followed a different approach based

on lineralized strains and potential wire energy, when addressing problems related to slippage along

bent helices. The determined critical curvature causing slippage was also derived by Kraincanic and

Kabadze [8] through a slightly different process. A question often addressed is which curve a wire

within the wall of a bent flexible pipe can be assumed to follow. While the assumption, that no

transverse slip will occur has been applied to a wide extend, Out and von Morgen [37] considered the

geodesic on a toroid as the equilibrium state in bending and tension focusing on constructing the

geodesic curve.

As discussed in section 1.2, the lateral wire buckling failure mechanism for a given pipe structure is

governed by compressive loads, repeated bending and often limited frictional resistance against wire

slippage due to a breached outer sheath. Wet annulus conditions are considered worst with respect to

lateral buckling (and other failure modes including pipe collapse). This is due to the fact, that there is

no longer a pressure difference across the outer sheath, and the normal force on the wires is therefore

limited. In order to develop mathematical models capable of describing the mechanism, these effects

should be included. Obviously, the first problem which should be addressed is the equilibrium state of

the armouring wires, which is encountered within the wall of a flexible pipe for given loads. The

geometrical wire equilibrium state is of a complex nature, since the pipe loads cause wire slippage in

the tangential and lateral wire directions. Furthermore, the radial deformations of the wires are to a

high extend restricted by adjacent pipe layers. The issue is further complicated by the fact that

validation of mathematical models for prediction of wire stresses and slips is quite difficult when a

high-strength tape imposes further restrictions of the radial deflections, as it is the case in flexible

pipes designed for deep waters. The reason for this is, that means for performing measurements below

the high-strength tape are still subject of industrial research. Cutting the tape to mount strain gauges

on the wires will change the mechanical behavior of the flexible pipe in an undesired manner. Ongoing

research suggests, that measurements will be available in a near future due to development of a

monitoring technology based on optical fibers embedded in the wires for measurement of strain and

curvature, see Weppenaar and Kristiansen [39] and Weppenaar et.al. [40].

In the present work, an analytical approach to the single wire equilibrium problem was chosen. The

following assumptions were chosen as basis for the model.

The problem is initially formulated without taking frictional effects into calculation. As a

consequence, the fatigue-like aspect of the problem is neglected by assuming, that the

physical wire equilibrium state, towards which convergence will occur as cyclic loads are

applied, equals the equilibrium state obtained directly without friction when the wire is

loaded.

The radius of curvature R of the modeled pipe will be considered constant. Hence, for the sake

of simplicity curvature variations are neglected.

Contact with adjacent layers will be considered by assuming the modeled wire embedded in a

cylinder bent into a toroid.

The wire dimensions will be assumed small compared to the minor torus radius, r, which is

convenient when formulating the constitutive relations

29

t n

tu

z

x

u (s)

t

b

Figure 23, left: An armouring wire modeled as a curve embedded in a toroid, right: Vector coordinate triads.

With these assumptions, the wire constitutes a curve (s) embedded in a toriod, which is

parameterized by an arclength coordinate u(s) along the torus centerline, and an angular coordinate

(s) along the circumferential direction, see Figure 23. The curve will be assumed parameterized by

arclength, s. A point on the torus surface is in Cartesian coordinates given by

,

(6)

A curve is constructed by establishing a relation between u(s) and (s). Assuming such a relation

established, a curvilinear coordinate triad can be attached to each point on the curve as a wire tangent

t, a normal n and a binormal b by means from basic differential geometry. The wire normal n is

assumed to coincide with the surface normal. As a consequence, the wire rotation around the local

tangent is governed by the underlying toroid. This is deemed to be a fair assumption for flexible pipes

with high-strength tape. In order to describe the curvature of the wire, this is split into a component in

the normal tn-plane, n, a geodesic component in the toriod tangent plane, g, and a torsional

component in the nb-plane, . A result from basic differential geometry known as the Darboux frame is

available for definition of the curvature component in terms of the triad vectors and their derivatives

in s

0

"

!#% &(

)

$

(

0

*

) "

* + !#%

0

$

(7)

Considering the vector triads in Figure 23 and Equation (7), it is noted that for a positive change of

wire arc length, a positive rotation around a local wire axis corresponds to a positive change of

curvature. Inserting proper definitions of the triad vectors in Equation (7), the following curvature

components are derived

( 01,-./
2 1
2

,-./

(8)

30

) 01,-./
2

.3(/

4

.

* 01,-./ 1
2
2

,-./

(9)

(10)

Alternatively, the definitions of curvature given by do Carmo [44] may be applied, which lead to the

same expressions, see Appendix A.

The equilibrium equations for long and slender curved beams have in the past served a wide range of

applications in applied mechanics. The equations were included in Loves famous book on theory of

elasticity, [29], and were given by Reissner [27], on vectorial form. It is desirable to express the

equilibrium equations with the local coordinate triad of the wire as basis

"

$

7

(

8

6 7

(

8

"

#

$ 97 " 9( # 98 $

:

:

7

(

( ( ) 8 97 ; " :

( 7 *8 9( ; #

8

) 7 *( 98 ; $ 0

<

"

$

?7

?(

?8

= 5 > " ?7

?(

?8

"

#

$

@7 " @( # @8 $ 8 # ( $

:

?7

?(

( ?( ) ?8 @7 ; " :

( ?7 *?8 8 @( ; #

(11)

(12)

?8

) ?7 *?( ( @8 ; $ 0

The following system of field equations governing the wire equilibrium can now be derived, if the

toroid is considered frictionless, causing distributed wire loads in the tangent plane to be neglectable

2

1

2

2

2 )

1

(13)

(14)

7

( ( ) 8

(15)

?(

( ?7 *?8 8

(17)

8

) 7 *(

(16)

(18)

31

in which equation (13) and (14) are derived on basis of the wire geometry in the toroid tangent plane,

equation (15) is derived on basis of the expression of the geodesic curvature given in equation

(9),

equation (16) governs the tangential force equilibrium, equation (17) the binormal force equilibrium

and equation (18) the normal moment equilibrium. This system of equations derived in Paper A form

the basis of the theoretical work conducted in the present project. They have been applied for single

wire analysis of perfect and imperfect wire geometries (Paper B), global models of the tensile armour

layers taking the torsional imbalance due to lateral wire buckling into calculation, (Paper C),

preliminar studies of frictional effects (Paper D) and, finally, simplifications of the global tensile

armour models (Paper F). Plasticity has not been simulated in the present work. However, this would

have been possible by applying appropriate constitutive relations.

The theory of elastic stability was originally developed in order to calculate the compressive load

carrying ability of slender structures, which failed due to instability prior to the expected load

capacity. The solution to the stability problem for initially straight beam-columns subjected to small

defections and rotations was originally studied by Leonard Euler, showing that stability could be

described mathematically as bifurcations. Hence, after a certain level of compressive loading, the

uniqueness of the solution to the equations of equilibrium was lost. The bifurcation loads and

corresponding mode shapes were obtained solving eigenvalueproblems formulated on basis of the

structural equilibrium of a deformed structure. While this approach led to quite accurate and

reasonable results for initially perfect structures, in which the deformations and rotations are small,

even slight violations of these basic assumptions could lead to enormous discrepancy between

measurements and calculation. This problem was addressed by Koiter, see for example [42], who

considered structural sensitivity to initial imperfections and showed that structural instability to a

wide extend was governed by the initial postbuckling behavior. The research in stability summarized

above is well described in the literature; see for example Bazant and Cedonlin [25], Brush and Almroth

[41], and Timoshenko and Gere [43].

In general terms, buckling is related to a deformed shape of a structure, usually characterized by large

deflections. This mechanical behavior may impose severe influence of the load carrying ability and

further loading of the structure may lead to large deflections, which at some point imposes

overstressing and leads to failure. The various types of behavior which are encountered in equilibrium

paths when considering a beam-column are shown in Figure 24. If rotations and deflections are

assumed small, buckling by bifurcation may be triggered, see Figure 24a. However, in physical

structures, rotations and deflections will often increase dramatically as buckling occurs causing

equilibrium to occur along a path as shown in Figure 24b. Obviously, instability is in many cases

related to the fact that the equilibrium equations have multiple solutions. In accordance with wellestablished energy methods, for example the principle of minimal potential energy, the geometrical

configuration, which will be encountered, is the deflected shape which minimizes the potential energy

of the structure. For the initially crooked beam in Figure 24, branch c is an example of an instable

primary path and branch d a stable secondary path.

In order to provide a strict definition of the terms stable and instable, the definition provided by

Liapunov, described by Bazant and Cedolin [25] pp. 176, may be adopted. In qualitative terms, this

states, that a system is stable if a small change of input parameters leads to a small change of response

or output. If this is not the case, the system is considered unstable. In Figure 24, branch c constitutes

an example of an unstable path, while branch d is an example of a stable path. This approach to elastic

stability has historically been applied to a wide range of problems and is well-described in the

literature.

However, if friction is present, a different approach must be followed since the system input

parameter causing instability is no longer the applied load, but the number of applied bending cycles.

The question of if the system is stable or not may then be addressed by considering the equilibrium

32

paths, which due to frictional energy dissipation exhibit a loop-like behavior. Obviously, if this loop is

closed after a few bending cycles, the system is stable and further cyclic loading can be assumed not to

cause instability. On the other hand, if the loop is open, wire slippage may occur towards a state, which

leads to lateral wire buckling. Further elaborations of instability, when friction is present, are

contained in Paper D.

Mechanical systems, which can be described assuming rotations and deformations small, may be

analyzed with respect to instability on basis of eigenvalue problems arising from the formulation of

the equilibrium conditions in the deformed state. If rotations and deformations cannot be assumed

small, this must be accounted for when formulating the governing equations. In the present context,

stability considerations are conducted on basis of full non-linear analyses of perfect and imperfect

structures.

P

P

c.

b.

a. P = EI

2

L

d.

P=

EI

L2

Bifurcation point

Figure 24 Examples of equilibrium paths in (compressive load-displacement) diagrams for simply supported

beam-column, reproduced after Brush and Almroth [41], a.) Initially straight beam-column, small deflections

and rotations (Euler solution), b.) Initially straight beam-column with large rotations, c.) Initially slightly

crooked beam, primary unstable path, d.) Initially slightly crooked beam, secondary stable path.

4.3. Paper A

A method for prediction of the equilibrium state of a long and slender wire on a frictionless toroid

applied for analysis of flexible pipe structures

In the first paper, the system of governing equations (13)-(18) is derived on basis of the equilibrium of

space curved beams and the geometry of a wire modeled as a curve on a torus surface. Frictional

effects were neglected assuming, that the wire equilibrium state obtained after a significant number of

bending cycles applied is equal to the equilibrium state, which is obtained for a given load input if

friction is neglected. It is demonstrated, that the derivation of the curvature components can be

performed on basis of concepts from abstract differential geometry, see do Carmo [44]. This approach

corresponds in a mathematical sense to direct application of the Darboux equation (7) to the derived

local vector triad of the wire, but eases the derivation significantly. Furthermore, it is demonstrated

how the system of governing equations can be solved numerically by application of a commercially

available solver. A solution is obtained after having converted the differential equations from being

functions of deformed arclength to functions of undeformed arclength. This is conducted on basis of

the assumption that strains will remain sufficiently small to be described using Cauchys definition of

strain. Obtained results reveal, that the calculated equilibrium state of the wire approaches the

geodesic of a toroid (a curve possessing no geodesic curvature) when the simulated pipe is bent and

tensioned.

33

pure bending

0.93

loxodromic curve

geodesic curve, FD-algorithm

0.92

geodesic curve - [Out and von Morgen]

0.91

0.9

0.89

0.88

0

3

4

5

Wire arclength, s(m)

Figure 25, left A single armouring wire within the wall of a flexible pipe modeled as a curve embedded in a

torus surface, right: Wire lay angle for different configurations and loads.

4.4. Paper B

Imperfection analysis of flexible pipe armour wires in compression and bending

On basis of the established method for frictionless single wire analysis proposed in Paper A, the

mechanical behavior of a wire in compression and bending is addressed. In order to examine the

sensitivity of the wire to small initial imperfections, a small harmonic perturbation is added to the

initial geodesic curvature. The system of equations governing the static wire equilibrium is solved for

stepwise increased compressive loads with the solution of one load step as initial guess to the

following. By this method, the equilibrium path of the loaded end of the wire was studied. For perfect

wire structures, a linear equilibrium path (Branch A, Figure 26) and a path exhibiting significant

softening behavior (Branch B, Figure 26) were obtained. For small geometrical imperfections, which

were chosen in a manner, so the initial linear response was not influenced, another solution was

obtained. This solution softened at a load level below what was obtained for perfect structures

(Branch C, Figure 26). While equilibrium states along branch A only differs little from a bent helix, both

branch B and C differs significantly, which is interpreted as buckling. However, branch B is

symmetrical with respect to the pipe midpoint, which is not the case for solutions along branch C. The

effect of key parameters is examined and it is determined that decreasing the pitch length of the

modeled wire increases the load carrying ability significantly. On the other hand, it is worth noticing

that the applied pipe curvature seems to have very little effect on the load carrying ability. Finally, it is

demonstrated, that the number of pitches included in the computational model impose significant

influence on the load carrying ability and mode of deformation for low number of pitches. The

analyses may indicate that buckling by bifurcation has been encountered, since the uniqueness of the

solution vanishes after a certain load level even for perfect wire geometries. This, however, is not

demonstrated in an exact mathematical sense. For the imperfect structures analyzed in the present

context, buckling is usually considered likely to occur as limit points rather than bifurcations as

described by Koiter [42]. This corresponds well with the obtained results.

34

1400

A

1200

B

C

1000

D

m=0

m=1, =0.001

m=2, =0.001

800

m=5, =0.001

i

m=20, =0.001

i

600

m=1, =0.001

i

m=2, =0.001

i

m=5,i=0.001

400

m=20, =0.001

i

m=20, =0.01

i

m=20, =0.0001

200

m=20, =0.005

i

Primary path

3

Compressive pipe strain

6

4

x 10

Figure 26 Equilibrium paths for armouring wire in compression and bending (wire geometry shown in Figure

25), m denotes the number of harmonic terms taken into calculation in the applied imperfection, the magnitude

of the imperfection amplitudes (these are set equal to each other).

4.5. Paper C

On modeling of lateral buckling failure in flexible pipe tensile armour layers

On basis of the methods for single wire analysis proposed in paper A and B, a global model of the

armouring layers is proposed. It is shown that the outer layer of armouring wires can be modeled with

the assumption that wire slippage can only occur along curves with constant lay angles. This approach

leads to the same load carrying ability as an analysis in which full lateral slippage is allowed in both

layers. However, neglecting lateral slippage in the outer layer decreases the computational power

necessary to obtain a solution significantly. As softening occurs, the torsional balance of the pipe

structure can no longer be maintained. This leads to a pipe twist causing further straining of the wires

in the inner layer. A model for calculation of the torsional moment in the free end of the pipe for a

specified pipe twist and a given longitudinal pipe strain is established on basis of multiple single wire

analyses. Lateral wire contact is neglected in this context. On basis of this model, the torsional

equilibrium equation for all wires is solved with respect to pipe twist for fixed pipe strain values by

Newton-Raphson iterations. Results are compared to the equilibrium paths obtained for torsionally

fixed-fixed pipes. Significant difference in load carrying ability is detected between the two scenarios.

However, the difference is imposed solely by the forces in the outer layer. These are for a torsionally

fixed-free pipe free to relax, which is not the case for a torsionally fixed-fixed pipe. It is demonstrated,

that the model is capable of simulating modes of deformation, which locally correspond well to the

experimentally triggered modes. Radial elasticity of the pipe wall is accounted for by considering the

minor torus radius in the loaded configuration a function of the applied longitudinal pipe strain. This

approach is equivalent to assigning a Poissons ratio to the global pipe structure. It is demonstrated,

that little effect on the load carrying ability can be detected, when radial deformations are accounted

for in this manner.

35

M

A

1

R=

1

R=

Figure 27 Equilibrium paths of modeled layers of armouring wires, radially stiff structures, left: Torsionally

fixed-fixed pipe, right: Torsionally fixed-free pipe.

4.6. Paper D

Simulation of frictional effects in models for calculation of the equilibrium state of flexible pipe

armouring wires in compression and bending

In the fourth paper, frictional effects on a single wire within the wall of a compressed flexible pipe subjected

to cyclic bending are investigated. Friction was modeled by a regularized Coulomb-law and directed

oppositely to the slip velocity. This is implemented in the system of field equations as distributed transverse

and tangential loads. Frictional effects are addressed in a similar manner by Pfister [38]. The system of

equations is solved stepwise for a prescribed load history. However, for the sake of simplicity, inertia terms

were neglected in the system of equations solved. This was estimated to be fair in the case of cyclic bending

applied at very slow speed. The system is hereby analyzed in a quasistatic manner and the mathematical

formulation of the problem still solved as a boundary value problem.

Papp=2.0 kN

10

0.02

0.04

0.06

Pipe curvature, (1/m)

0.08

0.1

Figure 28, left: Calculated pipe strain as function of load step number from quasistatic frictional analysis, right:

Local moment contribution to global bending hysteresis.

36

The system of equations can only be solved for moderately small transitions from zero to full friction. As a

consequence, the number of bending cycles leading to instability in simulations cannot be expected to

correspond to experimental results. On basis of the present research it cannot be concluded that frictional

effects impose significant influence on neither mode of deformation nor load carrying ability. The model

exhibited a mechanical behavior which in a qualitative sense corresponds well to results published in related

publications with respect to bending hysteresis and transverse vs. binormal slip.

The present paper can due to the unresolved issues only be considered as a preliminary study in frictional

effects on tensile armour wires. Further research is needed in order to formulate the problem by more strict

methodical means.

4.7. Paper E

On lateral buckling failure of armour wires in flexible pipes

The fifth paper is a full length conference paper from proceedings of the OMAE (Offshore, Marine and

Arctic Engineering). It contains a summary of the work presented in Paper A-C, in which a simplified

approach to the single wire model is applied. Only the field equations governing the tangent geometry

are converted from deformed to undeformed arclength. The approach proposed in Paper C is followed

in order to establish a computational model of the armouring layers. This model is used to calculate

the frictionless equilibrium paths of the 6 riser, which is considered torsionally fixed-free. The

obtained load carrying ability is compared with the experimental results from test series I, which are

described in section 3.2. Neglecting the remaining layers of the flexible pipe, it is furthermore

attempted to study how the effect of limited wire slippage close to end-fittings influences the load

carrying ability. This is conducted by setting the model length shorter than the physical length of the

pipe. Furthermore, the maximum inner layer stresses are calculated. It is shown that while the

maximum pipe curvature has very little influence on the calculated load carrying ability of the wires, it

imposes significant influence on the calculated stresses. Finally, it is demonstrated that the modes of

deformation obtained by the model locally correspond well to the buckling modes, which are triggered

during the experiments, see Figure 29.

37

4.8. Paper F

Simplified models for prediction of lateral buckling in flexible pipes armour wires

The sixth paper is like paper E a full length conference paper from proceedings of the OMAE (accepted

for presentation in 2012). In this paper, simplified global models of flexible pipes are proposed in

order to demonstrate means, by which the work contained in Paper A-C and E can be converted to a

form, which is more appropriate for engineering analysis. In order to limit the computational power

necessary to perform lateral buckling analyses, only a single wire in the inner layer of armouring wires

is analyzed, and the torsional moment contribution is scaled in order to model the mechanical

behavior of the entire layer. Two different approaches were followed and denoted A and B.

In the first model simplification A, only the mechanics of the tensile armor layers is addressed. The

mechanical behavior of the inner layer of armourig wires was described as the scaled result from a

single wire analysis. The outer layer of armouring wires is modeled with the linear equations

contained in [4]. It is demonstrated, that while this approach leads to a poor representation of the

force response prior to buckling, the load carrying ability in the buckled state is approximated with

very good accuracy. This is deemed to be due to the fact, that the load carrying ability of the armouring

wires exhibit little dependency on the boundary conditions in the circumferential pipe direction. The

equilibrium paths of the full global pipe model and the present simplification are shown in Figure 30.

By the second model simplification B it is assumed, that the inner layer of armouring wires all behave

linearly in a (force-strain)-diagram until the load carrying ability determined by a single wire analysis

is reached. After this load level, the layer is assumed to soften completely. Results are shown in Figure

30. The main conclusion of the present analysis, despite the simple and rather crude approach chosen,

is that while adjacent pipe layers do not impose significant influence on the load carrying ability prior

to buckling, an influence can be detected as a slope of the equilibrium path in the buckled state.

Obviously, this improves the prediction of the experimental results, since the conservatism of the

calculated load carrying ability is limited. This is obviously related to the fact, that it with this

approach seems possible to let overstressing of the wires define the limit state design. Further results

and elaborations regarding the choice of limit state design are included in section 5.2.

140

1

120

Model simplification A, =0

Linear response

Offset linear response

100

Longitudinal force (kN)

200

150

100

80

60

40

Inner layer of armouring wires

Outer layer of armouring wires

ABC-layer

Outer sheath

Sum of layers

20

50

-20

0

0.2

0.4

0.6

0.8

Longitudinal compressive pipe strain

1

3

x 10

10

15

20

25

30

Pipe twist (deg)

35

40

45

50

Figure 30 Equilibrium paths of 6 riser pipe, left: Model simplification A compared to the full global pipe model

proposed in Paper C, only tensile armour layers are modeled, right: Model simplification B, postbuckling

response taken into calculation. Tensile armour layers, high-strength tape and outer sheath taken into

calculation.

38

39

In this chapter, experimentally obtained results are compared with results obtained from numerical

simulations. As described in section 4.7 and 4.8, experimental and simulated results are contained in

Paper E and F. However, for the sake of completeness, additional comparisons are included in the

present chapter.

In general, theoretical studies have been focused on the mechanical behavior of the tensile armour

layers. However, effects from high-strength tape layers and outer polymeric sheath are also

considered. The remaining pipe layers are in the present context neglected, since the contributions

from those to longitudinal pipe load and torsional moment are very small.

The following methods for analysis of flexible pipes have been established in the current project:

1.

2.

3.

A full global model of the armouring layers which for a prescribed longitudinal strain by

Newton-Raphson iterations can be used to determine the pipe twist of the free pipe end for

which the torsional equilibrium is fulfilled. The model was based on separate analyses of all

armour wires neglecting lateral wire contact. The method was proposed in Paper C and

numerically simulated results were compared to experimentally determined modes of

deformation and load carrying abilities in Papers E and F.

This model has been applied in order to analyze the armouring layers of the 6 and 8 risers.

Contributions from adjacent layers were neglected, since these effects were found to slow down

convergence significantly. The 14 jumper has not been analyzed using this model, since the

larger number of wires applied in this pipedesign turned out to slow down convergence to an

extend, by which the method was no longer well-posed to be used on desktop computers.

A simplified iterative model, denoted A, by which the force and moment contributions from the

inner layer was determined by scaling the results from a single wire analysis. The outer layer of

armouring wires was modeled using linear equations. A solution in pipe twist to the torsional

equilibrium of the free pipe end was for fixed values of longitudinal pipe strain obtained in a

manner equivalent to the methods described under 1).

This method was by analysis of the armouring layers of the 6 and 8 risers in paper F shown to

correspond very well to the full global pipe model in the postbuckled state, despite of the fact

that the response of the armouring layers prior to buckling was not estimated accurately. This

method is used to calculate the load carrying ability of the armouring layers of the 14 jumper in

the present chapter.

A simplified linear model, denoted B, was established. In this model it was assumed that the

inner layer of armouring wires exhibited a linear mechanical behavior based on equilibrium on

perfect helices until the load carrying ability of the layer was reached. This was assumed to

cause the load in the layer to soften completely to a constant level, which correlates well with

the results obtained by the models described above. The adjacent layers were modeled linearly.

This model was in Paper F used to estimate the postbuckling response of all three flexible pipes,

which had been tested experimentally. Despite of the fact that the model was coarse and timedependant effects in the polymer and tape layers are not taken into calculation, it was

demonstrated that those layers which prior to buckling generate insignificant contributions to

the load carrying ability and torsional moment, impose significant influence on the postbuckling

response.

In the following, results obtained with the various proposed methods will be presented.

40

The designs of the three flexible pipes investigated in the present project are contained in G1.

Pipe layer / Pipe Design

Inner layer of tensile

armour wires

armour wires

Outer sheath *4

Pitch length, Lpitch (m)

Pitch angle, hel (deg)

Wire size (height width) (mm)

Number of wires

Outer diameter (m)

Pitch length, Lpitch (m)

Pitch angle, hel (deg)

Wire size (height width) (mm)

Number of wires

Outer diameter (m)

Pitch length, Lpitch (m)

Pitch angle, hel (deg)

Tape size (height width) (mm)

Number of windings

Outer diameter (m)

Thickness (mm)

Number of inner layer pitches in

pipe sample (including end-fittings)

6 Riser *1

0.201

1.263

26.2

3 10

52

0.209

1.318

-26.2

3 10

54

0.212

0.075

83.5

1 60

1

0.225

6.0

8 Riser *2

0.276

1.474

30

5 12.5

54

0.289

1.525

-30.3

5 12.5

56

0.292

0.025

88.4

1.8 1.3

8

0.434

10.0

14 Jumper *2

0.442

2.247

31.5

4 15

70

0.452

2.345

-31.0

4 15

72

0.455

0.140

-84.4

1 60

2

0.477

10.0

3.96

3.39

3-34

*1) A basic grade steel used for wires with yield strength of approximately 650 MPa, elastic modulus 210 GPa,

Poisons ratio 0.3

*2) A high strength grade steel used for wires with yield strength of approximately 1350 MPa, elastic modulus

210 GPa, Poisons ratio 0.3

*3) Tape material properties chosen are: elastic modulus 27 GPa. Poisons ratio 0.4

*4) Sheath material properties chosen are: elastic modulus 400 MPa, Poisons ratio 0.4

Table 3 Pipe designs and material properties.

5.2. Discussion of the definition of lateral wire buckling limit state design

As discussed in section 1.2, lateral wire buckling may lead to failure in a flexible pipe structure as the

inner layer of tensile armour deforms to an extend by which the plastic limit of the wire steel is

exceeded. However, this may not always be the case as demonstrated experimentally by testing of the

second 8 riser. This flexible pipe twisted and shortened severely during the experiment, but no sign of

lateral wire buckling could be detected by dissection of the test pipe sample after the experiment had

been concluded, see section 3.4. Considering the computational models proposed in the present thesis,

it is clear that the load carried by the entire pipe structure may still increase in the postbuckled state.

Lateral wire buckling may therefore from a pipe design perspective be defined in two different

manners:

A. The conservative definition of the phenomenon is that softening occurs in the inner layer of

armouring wires due to repeated bending and longitudinal compression. Hence, lateral wire

buckling is governed mainly by the applied compression and occurs when compression causes

the pipe structure to become torsionally unstable. With this definition, lateral wire buckling

can be predicted with reasonable accuracy with computational models only including the

tensile armour wires.

41

B. A less conservative definition of lateral wire buckling taking the postbuckling response into

calculation may also be given. The limit state design can be defined as the state in which the

structure does no longer function in a satisfactory manner when subjected to in-serviceconditions due to plastic deformations in the inner layer of armouring wires. This definition

allows large deformations of the pipe structure and must be considered using models in which

high-strength tape and outer sheath are included.

Obviously, the two definitions are related to the way by which the allowed limit state used for pipe

design is chosen. Considering only the tensile armour layers, definition A will from a computational

point of view give a lower bound for the point at which lateral wire buckling needs not to be

considered in the design process.

Definition B of the limit state design has not been considered in detail by global analysis of the

armouring layers, since the computational effort related to loading the wires until yielding is quite

severe. However, this way of defining the limit state shows large potential, since it enables the

postbuckling response to be taken into account when calculating the load carrying ability.

The ultimate objective of computational modeling of the failure phenomenon would be to accurately

predict the deformation state of a flexible pipe subjected to compression and repeated bending as

function of the number of applied bending cycles. Preliminar studies of cyclic loads are contained in

Paper D, but are with the computational power of the computers usually applied for flexible pipe

designs not a very practical approach. Furthermore, as described in section 6.2, further research is

needed in order to obtain a consistent model capable of predicting the deformation state of a flexible

pipe as function of the number of applied bending cycles.

Linear compressive

pipe response

First yield

P

Postbuckled pipe response,

all layers

Load carrying ability of model

Equilibrium path, only

including only armour layers

armouring layers modeled

Torsional pipe stability can

no longer be maintained

Figure 31 Equilibrium paths calculated by computational models including 1) only the tensile armour layers

(marked in blue, related to definition A), 2) All pipe layers (marked in red, related to definition B).

In this section, the linear equations referred to as the CAFLEX-equations, which in the present work

are used to simplify and extend the global pipe model, are included for the sake of completeness. The

42

equations are widely applied for prediction of the force or deformation response of straight flexible

pipes, see [4] and [5]. Each layer is initially considered separately and is modeled either as helically

wound or isotropic cylindrical. In the present approach, the change of thickness of each layer will be

neglected, so each layer contributes to the global system of equations with three equations. The

parameters given in Table 4 can pair wise either be considered unknown or specified.

Force parameters

F

Layer of share of longitudinal force

Deformation parameters

Longitudinal pipe strain

L

M

P(i)

P(i+1)

Pipe twist

r

Change of radius

Pressure on outer side of layer

Table 4 Force and deformation parameters in the CAFLEX equations.

A number of global equations ensuring equilibrium and compatibility of the pipe structures may be

introduced. Along with equations for each layer, these constitute a linear system of equations which

easily can be solved. For helically wound layers the equations are given by

F

L

P (i ) P (i + 1) sin 2 hel

+

+

r cos 2 hel

r sin hel cos hel

=0

nwires ,i cos hel EA E 2

E

2

r

L

L

r tan hel F M = 0

1

t

t

2

r

2

r

2r 2

C E 3 1

E - 1

HI

: ; : ; 1

0

D F 2

F 2

I

rF r ri

r r

+ + P(i ) o P(i + 1) r = 0

EA

E t 2

E t 2

M

=0

GJ

L

in which ri and ro denotes inner and outer radius of the modeled layer and t the layer thickness. It is

noted that the equations are derived using a different coordinate system than in the present work, see

Figure 32. It is necessary to account for this difference in the formulation when combining the two

approaches to simplified global models for lateral buckling prediction. Furthermore, by modeling all

layers except the armour layer prone to buckling with these equations, bending terms are neglected. It

has in Paper F been demonstrated that this is reasonable for the outer layer of tensile armour.

b'

n'

t'

43

In this section, the calculated load carrying ability is compared with the experimentally measured

values in accordance with the definitions proposed in section G2. Considering the descriptions of the

experimental program included in chapter 3, it is clear that only rather coarse measures for the load

carrying ability are available. Furthermore, the two different definitions given in section G2 yield

different experimentally determined load carrying abilities.

For the 6 riser, experiment 1, LC I was concluded with approximately 80 kN of compression without

any sign of torsional instability in the tested pipe. However, experiment 3, LC I was concluded with

approximately 3 degrees of pipe twist and further twisting may have occurred if more than 1200

bending cycles had been applied. In accordance with definition A, the load carrying ability of the 6

riser is between 80 and 160 kN. However, if definition B is followed, the load carrying ability is

between 160 and 210 kN, since the first compressive load level did not cause severe pipe twist, which

was the case for the second compressive load level. Equivalent considerations lead to an

experimentally determined load carrying ability of the 8 riser between 300 and 400 kN in accordance

with definition A and between 400 and 700 kN in accordance with definition B. The 14 jumper was

not tested for varied compressive levels and fixed bending radius. Estimation of the load carrying

ability on basis of experimental results would therefore be of a highly speculative nature.

Pipe design

6 Riser, =1/11 m-1

8 Riser, =1/12 m-1

14 Jumper, =1/10 m-1

Pcr, calc

(kN)

Pcr, 20 deg. twist

Pcr, exp window

(kN)

(kN)

(kN)

100

110

118

A. 80-160 / B. 160-210

256

300

331

A. 300-400 / B. 400-700

157*

226

294

NA

Table 5 Load carrying ability, Pcr, calc load carrying ability based on full global model and simplified model A

based only on tensile armour layers, Pcr, 10 deg. twist - load carried by pipe structure at 10 degrees of twist obtained

with simplified model B including tensile armour layers, high-strength tape and outer sheath, Pcr, 20 deg. twist - load

carried by pipe structure at 20 degrees of twist obtained with simplified model B including tensile armour

layers, high-strength tape and outer sheath, Pcr, exp load carrying ability window determined experimentally.

The load carrying ability can now in accordance with definition A be calculated by the full global model

and simplified global model A by considering the determined equilibrium paths. Only the tensile

armour layers are included in the models. The results for the 6 and 8 pipe structures are contained in

paper E. The 14 jumper was due to the very large number of wires only analyzed using the simplified

model. The obtained equilibrium paths are shown in Figure 34. It is interesting to note, that a tensionbending coupling arises in the simplified model. The load carrying abilities are summarized in Table 5.

For the 6 riser, the load carrying ability is within the window determined experimentally in

accordance with definition A. For the 8 riser the simulated load carrying ability is lower than the

experimentally determined values. Hence, the load carrying ability is calculated conservatively.

The postbuckling response has only to full extend been taken into calculation by model simplification

B. This does not imply that this would not have been possible with the full global pipe model, but the

computational effort necessary to perform such analyses is quite severe, and has therefore not been

considered as part of the present work. Values will be presented for respectively 10 and 20 degrees of

pipe twist. In all cases, this increases the load carrying ability significantly. However, for the 6 and 8

risers, the load carrying ability can still be observed to be estimated conservatively. Considering

definition B of lateral wire buckling, stress calculations must be incorporated in the established

methods in order to determine when failure occurs. However, in order for this to be reasonable for the

simplified models, which are based on analyses of only a single wire in the inner layer of tensile

armour, the stress levels must be at least approximately the same in all wires. This is investigated

further in section G6 on basis of the full global pipe model.

44

The experimental results summarized in Table 2 and the load carrying ability calculated with model

simplification B are compared in Figure 33. Modeling the postbuckling response can be observed to

improve the calculated load carrying ability, despite the fact that predictions are still conservative with

respect to the experiments.

700

Slowly increasing, failure

600

500

400

300

Calculated equilibrium path, 8" riser

Stable

Increasing, failure

200

Slowly increasing

100

Stable

0

10

15

20

25

30

35

Twist of free end before unloading (deg)

40

45

Figure 33 Experimental results (contained in Table 2) compared with load carrying ability predictions from

model simplification B, comments to measurements in the figure are related to the progression of twist

response as the experiment was concluded.

14" Jumper, simplified pipe model, radially stiff pipe structures

300

180

250

140

200

Longitudinal force (kN)

160

120

100

80

60

150

100

50

Outer layer of armouring wires

ABC-layer

Outer sheath

Sum of layers

40

=0

=1/10 m-1

=0, reduced length

20

0.5

1.5

2

2.5

3

3.5

Longitudinal compressive pipe strain

-50

4.5

5

-4

8

10

Pipe twist (deg)

12

14

16

x 10

Figure 34, left: Equilibrium paths of 14 jumper calculated with simplified global pipe model A, only tensile

armour layers modeled, right: Equilibrium paths of 14 jumper calculated with simplified global pipe model B,

outer sheath and high-strength tape included in model.

45

In this section, the modes of deformation detected experimentally will be compared with the

deformation patterns simulated by the full global pipe model.

Test 0

Test I

Test II

Test III

Mode I

Mode II

Mode III

Figure 35 Modes of deformation from experiments and simulations (radially stiff analyses), 6 riser, Mode I:

R=11 m, pipe strain J 2.5 10N , Mode II: R=11 m, pipe strain J 1.0 10NO , 20 imperfection terms with

amplitudes of -0.001, Mode III: Equivalent to mode II, but with positive imperfection amplitudes.

46

Initially, considering the 6 riser the modes of deformation were at a local level shown to correspond

well to simulated results. However, the detected modes of deformation did not localize in the same

manner as detected experimentally, see Figure 35. The result from the initial trial test has deformed to

an extend which seems governed by plasticity. The remaining deformation patterns form localized

large gaps, S-shapes or a combination of the two deformation modes, by which the wires, despite the

formation of gaps, seem to sustain some S-shape. Three different modes of deformation could be

simulated. Mode I is a periodic S-shaped deformation mode, which corresponds well to the wire

configurations detected localized during Test I. The two other simulated modes of deformation are

characterized by large wire gaps localized in each end of the layer. Comparing these modes of

deformation with the modes encountered during dissection of pipes, gaps can be observed to localize

closer to the boundaries in the simulations, while occurring on each side of the pipe midpoint during

experiments. It seems likely, that this discrepancy occurs due to boundary effects in form of limited

wire slippage close to end-fittings.

Considering the modes of deformations of the 8 riser, the simulated modes of deformation were of a

more severe character than encountered during laboratory experiments, see Figure 36. The gaps in the

simulated deformation pattern are localized symmetrically around the pipe midpoint, while gaps only

occurred between the static frame end and the pipe midpoint during the experiment.

The 14 jumper was not analyzed using the full global pipe model. Experimentally triggered modes of

deformation are shown in Figure 37.

Test I

Test II

Figure 36 Modes of deformation obtained experimentally and simulated, 8 riser (radially stiff analysis),

R=12m, J 4.5 10NQ , 20 imperfection terms with amplitudes of -0.001.

47

Test I

Test II

The wire stresses can on basis of the determined equilibrium states within the pipe wall be calculated

for each step of compressive pipe strain applied in the global model, see Figure 38. In general, it can on

basis of the conducted full global analyses be concluded, that softening of the inner layer of tensile

armour occurs at lower load levels than yielding. This corresponds well to the observations made

during experiments, since the logged responses are interpreted in the sense that severe deformation of

a pipe structure may occur before failure due to yielding occurs. An example of the maximum cross

sectional normal wire stress in all mesh points is presented in Figure 38. The maximum normal wire

stress of all wires in the inner layer of tensile armour for the 6 and 8 riser is shown in Figure 39.

While the maximum normal wire stress can be observed to be approximately the same for all wires in

the 6 riser structure, this is clearly not the case for the 8 riser. This may be due to boundary effects

since a lower number of inner layer pitches are included in the 8 riser model than in the 6 pipe

structure. Further investigations are needed in order to draw conclusions regarding this issue.

However, it can be observed that a different global mode of deformation has been triggered by the

analysis of the 8 riser than the mode obtained by analysis of the 6 riser. This constitutes one possible

explanation for the differences in stress variation obtained by the two analyses.

48

n

t

1 =

Pt M n n Eh

+

+

A 2In

2

2 =

Pt M n n Eh

+

A 2I n

2

3 =

Pt M n n Eh

A 2In

2

4 =

Pt M n n Eh

+

A 2I n

2

max = max( 1 , 2 , 3 , 4 )

Figure 38 Maximum stress in wire cross-section, radially inelastic pipe structure, pipe strain R 7 10NQ ,

levels: blue : max < 100 MPa, green : 100 max <200, yellow : 200 max <300, red : 300 max.

Maximum wire stress

450

400

350

300

250

200

150

6" riser, L/L=-110-3

100

8" riser, L/L=-0.510-3

50

0

ini (rad)

Figure 39 Maximum stresses in wires as function of initial wire angle (related to position in circumferential

direction), 6 and 8 risers, radially inelastic global pipe models.

49

The full global pipe model is incapable of taking lateral wire contact into account. This effect may

cause interlocking effects which prevent shortening or twist, and is therefore estimated to increase the

load carrying ability. However, it is possible to perform computational checks of whether the wires are

in transverse contact or not during post processing of obtained results. An example of such a contact

check is shown in Figure 40. In general, no lateral wire contact occurs prior to softening of the inner

layers and is at larger compressive strains mainly caused by formation of gaps.

Figure 40 Lateral wire contact tjeck, contact violation marked in red, radially inelastic pipe structure, pipe strain

J 7 10NQ .

In section 2.2 it was demonstrated how the curvature of the pipe midpoint was estimated during the

experiments as a finite difference formulated on basis of measured rotations on each side of the

midpoint. It is interesting to observe, that the flexural hysteresic behavior of the test pipe samples can

be estimated by calculating the mechanical moment applied by the testrig at the midpoint as

? T U

in which the quantities are given in Figure 9. Gravitational effects have been neglected. The deflection

y of the pipe midpoint was measured during the experiment. However, it is noted that the bending

arrangement in the upgraded test rig is constructed in such a manner, that deflection measurements

by extensometers between the topbar and the test pipe sample will include a horizontal component.

This is caused by a change of the horizontal distance between the pipe midpoint and the extensometer

mounting point on the topbar, as bending is applied. This was not the case in the G1 test rig, in which

this distance remains constant. However, the influence on the calculated moment was estimated to be

below 3% and is therefore neglected in the present context.

The hysteresis loops which by this method are obtained on basis of experimentally measured data are

shown in Figure 41 and Figure 42. The following observations can be made:

The hysteresis loop does due to the presence of gravity on the test pipe sample not intersect

the origin of the (,M)-diagram like the idealized behavior shown in Figure 2. This causes the

hysteresis loop to have a negative offset on the vertical axis. The curvature values also reveal,

that the test pipe sample was not straightened completely due to gravitation.

Differences in compressive load level can be observed to impose very large influence on the

flexural behavior

The hysteresic behavior of the 14 jumper can for the experiment representing extreme load

conditions be observed to change for each bending cycle as the wires slip.

50

8000

Test I, LC 1, 700 kN

Test II, LC 1, 300 kN

Test II, LC 2, 400 kN

=1/12 m

6000

Bending moment (Nm)

x 10

10000

4000

2000

0

-2000

1

0

-1

Test I, LC 1, 265 kN

-4000

Test II, LC 1, 80 kN

-6000

-2

=1/11m

-8000

0.02

0.04

0.06

0.08

Curvature, , of pipe midpoint (1/m)

0.1

-3

0.12

0.01

0.02

0.03

0.04

0.05

0.06

0.07

Curvature, , of pipe midpoint (1/m)

0.08

0.09

Figure 41, Flexural hysteresis of pipe midpoint in curvature-moment diagram, left: 6 riser various load

cycles, right: 8 riser, all load cycles.

4

x 10

4

LC I, 277 kN compression

LC II, 269 kN compression

LC III, 411 kN compression

x 10

2

0

-1

-2

1

0

-1

-2

LC I

LC II

-3

-3

-4

0.04

0.06

0.08

0.1

0.12

Curvature, , of pipe midpoint (1/m)

0.14

0.16

-4

0.12

=1/6m

=1/8m

0.13

0.14

0.15

0.16

Curvature, , of pipe midpoint (1/m)

0.17

0.18

Figure 42, Flexural hysteresis of pipe midpoint in curvature-moment diagram, left: 14 jumper Experiment I,

right: 14 jumper Experiment II.

The need within NKT-Flexibles for a tool for engineering design against lateral wire buckling was, due

to a number of deep-water projects, urgent already as this project was initiated. It was therefore

chosen by NKT-Flexibles to implement the methods developed in the present project before this was

finalized. It was chosen to implement model simplification A (see Paper F), however, with highstrength tape and polymeric sheaths taken into calculation. The mechanical behavior of the inner layer

of armouring wires was described as the scaled results from a single wire analysis.

Remaining pipe layers included in the calculation were modeled using the linear responses given in

section 5.3. The torsional equilibrium equation of the free pipe end was solved numerically by Newton

Raphton iterations. Non-linear constitutive models appropriate to describe time-dependant effects

were applied in order to model the polymeric sheaths. These have not been investigated in the present

project. However, constitutive models were adopted from an already established tool for design

against birdcaging. The implementation of the method was programmed and documented by NKTFlexibles development engineer, Anders Lyckegaard, who also supervised the present PhD-project.

However, the implementation of the method was checked as part of the present project. As limit state

51

design criteria, definition B given in section 5.2 was chosen. Hence, the single wire model was loaded until

first yield occured. Analyses showed that this was conservative with respect to the results obtained by the

laboratory tests described in chapter 3 in the present thesis. In order to address the lacking one-to-one

correspondence between laboratory experiments and DIP-tests in the best possible manner, the model was

calibrated against available DIP-test results. This is, due to commercial issues, not included in the present

thesis. The developed design tool was named TFlex, see Figure 43.

Figure 43 The logo of the NKT-Flexibles TFlex tool for design of flexible pipes against lateral wire buckling.

The documentation of the TFlex code [71]-[73] and NKT-Flexibles test and analysis reports were

reviewed by Bureau Veritas as third part in order to obtain a certificate confirming the correctness of

the developed method. The review was conducted in autumn 2011. At the time, at which this project

was finalized, a draft revision of a design methodology certification report [74] had been issued, see

Figure 44. The draft section certifies the validity of the method for design of flexible pipes of 2 to 10

bore diameter for water depths not larger than 2000 m. The papers contained in the present thesis

were also reviewed by Bureau Veritas as part of the certification process. These were issued as NKTFlexibles documents [75]-[80].

Figure 44 Frontpage for Bureau Veritas draft revision of type approval certification report.

52

53

6. Concluding remarks

The lateral wire buckling failure mode has been reconstructed under controlled conditions in the

laboratory by use of mechanical test rigs simulating wet annulus conditions. Four 6, two 8 and three

14 pipes were tested by applying cyclic bending and longitudinal compression. Results confirmed the

widely accepted fact that lateral wire buckling occurs at lower load levels in the laboratory than

encountered with field conditions. The limit load carrying ability for pipes subjected to laboratory

testing could roughly be estimated for the 6 and 8 pipes. Furthermore, a global mode, which seemed

elastic, was detected during testing of a 14 pipe. It was also noticeable, that an 8 pipe exhibited a

global deformation behavior, which is usually related to lateral wire buckling, while dissection of the

pipe after unloading revealed no sign of failure at all.

A method for calculation of the equilibrium state of a beam embedded in a frictionless cylinder bent

into a toroid has been developed. The proposed method was used to calculate the equilibrium state of

an armouring wire within the wall of a flexible pipe of finite length bent to a constant radius of

curvature. The analysis was carried out assuming that the equilibrium state, which a wire will reach

after a significant number of bending cycles, equals the state obtained directly if friction is neglected.

With this assumption, it was shown, that the modeled wire approached the geodesic curve on a torus

surface in bending and tension.

This method was used for imperfection analysis of armoring wires within flexible pipes in bending and

compression. It was demonstrated, that significant softening occurred in the equilibrium path of the

modeled wire, and that this behavior corresponded to large change of wire lay angle. Furthermore,

imperfections caused changes of wire lay angle to localize, while perfect wire geometries exhibited a

deformation pattern which was symmetrical around the pipe midpoint. Surprisingly, it was found that

the bending radius of the modeled pipe had little influence on the load carrying ability. The initial lay

angle seemed to be the geometrical parameter which along with the dimensions of the wire cross

section governs the load level at which softening occurs.

On basis of multiple single wire analyses, a model of both armouring layers was obtained. This model

made it possible to calculate the equilibrium paths and load carrying ability taking the effect of the

torsional imbalance of the pipe structure due to buckling of the inner tensile armour layer into

calculation. A comparison of the the calculated load carrying ability with experimentally obtained

results reveals that the method, as expected, is conservative. The reason for this is most likely that

effects due to friction, limited wire slippage close to end-fittings and load carried by the remaining

pipe layers were not considered in this analysis. Modes of deformation obtained experimentally were

shown to correspond well to the simulated buckling modes at a local level.

On basis of the global model of the armouring layers, model simplifications have been proposed. These

have proven capable of predicting the load carrying ability with quite high accuracy, although the

representation of the response prior to buckling may be of poor precision. By this method it was

shown, that while the remaining pipe layers imposed very little influence on the load which leads to

softening, high-strength tape and outer sheath imposed some influence on the force response in the

postbuckled state. Along with end-fitting effects, this may improve the prediction of the result of a

laboratory test. However, further research is needed to improve the models of sheaths and antibirdcaging layer. These may experience large deformations and often exhibit time-dependant

constitutive behavior. These issues are not included in the present implementation of the model.

Comparing the calculated equilibrium paths obtained by the full global model of the armouring layers,

it has been demonstrated that the load carrying ability is estimated conservatively. Furthermore, it has

been demonstrated, that the calculated predictions can be increased if the postbuckling response is

taken into account due to effects from adjacent pipe layers. Limited wire slippage close to end-fittings

may increase the load carrying ability of the computational model further.

Finally, it can be concluded that the objective of the present work formulated in section 1.3 has been

fulfilled in a satisfactory manner.

54

The work conducted in the present project served as basis for a tool for design of flexible pipes against

lateral wire buckling. The methods developed in the project were implemented by NKT Development

Engineer Anders Lyckegaard. The documentation of the method was reviewed by Bureau Veritas. At

the time at which the present project was finalized, a draft section for a type approval certificate

validating the correctness of the method had been issued. The design tool implementation is briefly

described in section 5.8.

The method for determination of the wire equilibrium proposed in paper A constitutes a novel

contribution to the field of wire mechanics. The issue related to this method, which has not been

addressed in this manner in related publications, is the determination of the equilibrium state by a

method, which is free of geometrical constraints except for the assumption, that the wire is embedded

in the toroid. The equilibrium state approached the geodesic when tension and bending was applied.

This has in the past often been assumed without arguments based on calculation.

In paper B it has been shown that the method can be applied for wire stability assessment on basis of a

full nonlinear analysis. It was furthermore demonstrated, that the modeled wires were sensitive to

small initial imperfections in compression. This application of the model proposed in paper A is also a

novel contribution, since tendon stability has not before been assessed on basis of analytical models

applied for full nonlinear analyses of perfect and imperfect structures. However, the studies in single

wire mechanics are insufficient for instability calculation of entire flexible pipes. Therefore, the global

model of the tensile armour layers proposed in Paper C constitutes an important contribution. It

enables both calculation of the load carrying ability and deformation patterns of all wires taking

torsional imbalance into calculation. Although boundary effects and ovalization demand further

research, the model is the first of its kind, since the torsional imbalance of the armouring layers due to

buckling has not been addressed in related publications.

The experimental work included in the present project is conducted by an experimental principle,

which to a wide extend has been applied for similar tests in the past. However, the obtained results

revealed, that lateral wire buckling may develop in the elastic regime. Despite causing deformations of

a flexible pipe, which usually are related to failure, no sign of failure could be observed during

dissection after unloading. This suggests that the limit state design may be taken as the state, at which

overstressing of the wires occurs, rather than the state at which the torsional balance can no longer be

maintained. This observation may alter the understanding of design against lateral wire buckling, if

taken into calculation in appropriate design tools. Model simplifications have been proposed as novel

contribution to the design process. These enable calculation of the load carrying ability of a given pipe

structure in a fast and efficient manner compared to the full global model of the tensile armour layers.

Methods for implementation of adjacent pipe layers and frictional effects have been investigated and

sketched. However, further research is needed in order to finalize those.

Among the issues which have not been addressed in the present work, is the lacking one-to-one

correspondence between laboratory experiments and full-scale field conditions. In order to investigate

this issue in a fulfilling manner, exact knowledge regarding the conditions encountered in the touchdown zone during pipe laying needs to be established. Especially the boundary conditions which the

pipe encounters in the touchdown zone are of major importance. Measurement of DIP-test conditions

may be conducted by instrumentation of the test pipe sample, either by strain-gauges or

accelerometers mounted in a manner so the gauges are protected from seawater, or by the optical

monitoring technology described in [39] and [40]. However, the development task related to such

measurements is considerable and has been deemed beyond the scope of the present project. The

method for calculation of the load carrying ability proposed in the present project, may be generalized

to predict the load carrying ability, if the exact boundary conditions are known. Presently, only very

55

few DIP-test results, which could be used for calibration of the computational model, are known.

Furthermore, no field failures have been encountered by NKT-Flexibles. The experiments conducted as

part of the present project have shown, that pipes may be installed at water depths corresponding to

compressive loads which could lead to lateral buckling, if only a limited number of bending cycles are

applied. This complicates the problem related to prediction of a DIP-tests result further.

Presently, it does not seem that laboratory experiments conducted inside pressure chambers

correspond better to field conditions than the principle applied in the present work. However,

laboratory experiments with longer test pipe samples than applied during the present experiments

would enable investigation of how boundary effects influences lateral wire buckling. Since it has been

demonstrated that lateral buckling may develop in the elastic regime, so no sign of failure can be

detected during dissection, the experimental results have shown that lateral buckling and failure by

lateral buckling can be considered as distinct terms in the flexible pipe design process. Further

research is needed in order to apply this very interesting observation as basis for design rules and

definition of the limit state design.

Regarding the theoretical work proposed in the present thesis, several issues could be addressed in

future research. First of all, the assumption, that the equilibrium state can be determined neglecting

friction, should be investigated further. Furthermore, the effect of limited wire slippage close to endfittings should be considered. In the present work, the effect has been considered by shortening the

computational model. However, this is strictly speaking not a correct method, and can only be used

crudely to estimate the effect. In order to investigate the interaction between friction and boundary

effects on the buckling load, further research following the approach used in the present work to

model frictional resistance is needed. Furthermore, measurement of the strains and curvature

components with optical monitoring technology would be both interesting and valuable. In bending a

tension, results were obtained which supported the assumptions made by Out and von Morgen [37]

regarding the geodesic curve as wire limit state. These results to some extend contradict the results

obtained by Leroy and Estrier [33], who included friction in the analysis and found cyclic bending to

cause slips which were centered around the geodesic. However, due to the experience-based solution

form, which was not justified, it is difficult to draw conclusions regarding the validity of these results.

Further research is needed in order to examine this issue with experimental and theoretical means.

Finally, the instability analyses in the present project are undertaken on basis of full-nonlinear

analyses of initially perfect and imperfect wire geometries. Future research should include further

investigations of the influence of the chosen imperfections on the obtained results. It would in this

context be both interesting and valuable to apply linearization techniques in order to investigate the

initial postbuckling behavior by analytical means.

While it in the present work has been demonstrated, that the adjacent pipe layers impose little

influence on the load carrying ability of the armouring layers, high-strength tape and polymeric

sheaths may have some influence on the postbuckling response. In the present approach, a method

based on linear responses of the remaining pipe layers has been applied neglecting time-dependent

effects in polymeric layers. Further research should include more detailed models of the adjacent pipe

layers including more sophisticated material models. Effects caused by ovalization of the pipe crosssection in bending have been neglected in the present context. These effects may, however, possibly

generate physical imperfections which should be accounted for in future research. Finally, an

improved method for modeling of radial deformations of the pipe wall taking variations of radial strain

into calculation should be considered.

If frictional effects are included when modeling the single armouring wires, the governing equations

for the wires could be formulated for varied radius of curvature throughout the length of the test pipe

sample. This would make it possible to model the global constitutive behavior in detail on basis of the

internal components. The arising method would obviously be extremely demanding to solve

numerically, but would on the other hand constitute a very valuable multipurpose tool for flexible pipe

analysis.

56

57

References

[1]

API 17J, Specification for Unbonded Flexible Pipes, American Petroleum Institute, 3rd edition,

2008.

[2]

Braga, M.P. and Kaleff, P. : Flexible pipe sensitivity to birdcaging and armor wire lateral buckling,

Proceedings of OMAE 2004, OMAE2004-51090.

[3]

Secher, P., Bectarte, F. and Felix-Henry, A. : Lateral Buckling of Armor Wires in Flexible Pipes:

reaching 3000 m Water Depth, Proceedings of OMAE 2011, OMAE2011-49447.

[4]

[5]

Fret, J.J. and Bournazel, C.L. : Calculation of stresses and slips in structural layers of unbonded

flexible pipes, Journal of Offshore Mechanics and Arctic Engineering, Vol. 109, pp. 263-269,

1987.

[6]

Witz, J.A. and Tan, Z. : On the Flexural Structural Behaviour of Flexible Pipes, Umbillicals and

Marine Cables, Marine Structures, Vol. 5, pp. 229-249, 1992.

[7]

Tan, Z., Quiggin, P. and Sheldrake, T. : Time Domain Simulation of the 3D Bending Hysteresis

Behavior of an Unbonded Flexible Riser, Journal of Offshore Mechanics and Arctic Engineering,

Vol. 131, 031301, 2009.

[8]

Kraincanic, I. and Kabadze, E. : Slip initiation and progression in helical armouring layers of

unbonded flexible pipes and its effect on pipe bending behavior, Journal of Strain Analysis, Vol.

36, No. 3, pp. 265-275, 2001.

[9]

Alfano, G., Bahtui, A. and Bahai, H. : Numerical derivation of constitutive models for unbonded

flexible risers, International Journal of Mechanical Sciences 51, pp. 295-304, 2009.

[10]

Alfano, G., Bahtui, A. and Bahai, H. : Numerical and Analytical Modeling of unbounded flexible

pipes, Journal of Offshore Mechanics and Arctic Engineering, Vol. 31, 2009.

[11]

Dastous, J.B. : Nonlinear Finite-Element Analysis of Stranded Conductors With Variable Bending

Stiffness Using the Tangent Stiffness Method, IEEE Transactions On Power Delivery, Vol. 20,

No.1, pp. 328-338, 2005.

[12]

Jiao, G. : Limit State Design for Flexible Pipes, Marine Structures, Vol. 5, pp. 431-454, 1992.

[13]

Vaz, M.A. and Rizzo, N.A.S. : A finite element model for flexible pipe armor wire instability, Marine

Structures, 2011.

[14]

Vaz, M.A. and Patel, M.H. : Post-buckling behavior of slender structures with a bi-linear bending

moment-curvature relationship, International Journal of Non-linear Mechanics, Vol. 42, p. 470483, 2007.

[15]

Vaz, M.A. and Patel, M.H. : Lateral buckling of bundled pipe systems, Marine Structures, Vol. 12 ,

pp. 21-40, 1999.

[16]

Vaz, M.A. and Patel, M.H. : Initial post-buckling of submerged slender vertical structures subjected

to distributed axial tension, Applied Ocean Research Vol. 20, pp. 325-335, 1998.

[17]

Secher, P., Bectarte, F. and Felix-Henry, A. : Qualification Testing Of Flexible Pipes For 3000m

Water Depth, Proceedings of OTC 2011, OTC-21490.

[18]

Bectarte, F. and Coutarel, A. : Instability of Tensile Armour Layers of Flexible Pipes under

External Pressure, Proceedings of OMAE 2004, OMAE2004-51352.

58

[19]

Tan, Z., Loper, C., Sheldrake, T. and Karabelas, G. : Behavior of Tensile Wires in Unbonded

Flexible Pipe under Compression and Design Optimization for Prevention, Proceedings of OMAE

2006, OMAE2006-92050.

[20]

Custdio, A.B. : Analytical model for instability assessment of flexible pipes armour, Doctoral

thesis, COPPE/UFRJ, 2005 (in Portuguese).

[21]

Brack. A., Troina, L.M.B. and Sousa, J.R.M. : Flexible Riser Resistance Against Combined Axial

Compression, Bending and Torsion in Ultra-Deep Water Depths, Proceedings of OMAE 2005,

OMAE2005-67404.

[22]

Montgomery, D.C. : Design and Analysis of Experiments, John Wiley & Sons 2009, ISBN: 978-0470-39882-1.

Kensche, C. W. : Fatigue of composites for wind turbines, International Journal of Fatigue, pp.

1363-1374, 2006.

[23]

[24]

Imamura, A., Hijikata, K., Tomii, Y., Nakai, S. and Hasegawa, M. : An experimental study on

nonlinear pile-soil interaction based on forces vibration tests of a single pipe and a pile group,

Eleventh World Conference on Earthquake Engineering, 1996, Paper No. 563, ISBN: 0 08

042822 3.

[25]

[26]

Terndrup Pedersen, P. : Equilibrium of offshore cables and pipelines during laying, International

Shipbuilding Progress, Vol. 22, pp. 399-408, 1975.

[27]

Physics (ZAMP), pp. 734-744, 1981.

[28]

Clebsch, A. : Theorie der Elasticitt Fester Krper, Druck und Verlag von B.G. Teubner, Leipzig,

1862.

Love, A.E.H. : A treatise on the Mathematical Theory of Elasticity, Dover Publications Inc., N.Y.,

1944.

[29]

[30]

Costello, G.A. : Large deflections of helical spring due to bending, Journal of the Engineering

Mechanics Division, Vol. 103, No. 3, pp. 481-487, 1977.

[31]

Spillers, W. R., Eich, E.D., Greenwood, A.N. and Easton, R., J. : A helical tape on cylinder subjected

to bending, Eng. Mech. Div. Proc. ASCE, Vol. 109, 1983.

[32]

Stump, D.M and van der Heijden, G.H.M. : Matched asymptotic expansions for bent and twisted

rods: applications for cable and pipeline laying, Journal of Engineering Mechanics, Vol. 38, pp.

13-31, 2000.

[33]

Leroy, J.M. and Estrier, P. : Calculation of stresses and slips in helical layers of dynamically bent

flexible pipes, Oil and Gas Science and Technology, REV. IFP, Vol. 56, No. 6, pp. 545-554, 2001.

[34]

Svik, S. : A finite element model for predicting stresses and slip in flexible pipe armouring

tendons, Computers and Structures. Vol. 46, No.2, pp. 219-230, 1993.

[35]

Svik, S. : On stresses and fatigue in flexible pipes, Doctoral Thesis, NTNU, Trondheim, 1992.

[36]

[37]

Out, J. M. M. and von Morgen, B. J. : Slippage of helical reinforcing on a bent cylinder, Engineering

Structures, Vol. 19, No. 6, pp. 507-515, 1997.

[38]

Pfister, J. : Elastic Multibody Systems with Frictional Contacts, Doctoral Thesis, University of

Stuttgart, 2006.

59

[39]

Monitoring of Flexible Pipes, Proceedings of OTC 2008, OTC-19427.

[40]

Weppenaar, N., Kosterev, A., Dong, L., Tomazy, D. and Tittel, F. : Fiberoptic Gas Monitoring of

Flexible Risers, Proceedings of OTC 2009, OTC-19901.

[41]

Brush, D.O. and Almroth, B.O. : Buckling of bars, plates and shells, Mcgraw-Hill, 1975.

[42]

Koiter, W.T. : Current Trends in the Theory of Buckling, Buckling of Structures, Symposium

Cambridge/USA, June 17-21, 1974, IUTAM, Springer-Verlag, pp. 1-16.

[43]

Timoshenko, S. and Geere, J.M. : Theory of Elastic Stability, McGraw-Hill, 2nd Ed. 1961.

[44]

do Carmo, M.P. : Differential Geometry of Curves and Surfaces, Prentice-Hall Inc., 1976.

[45]

Lateral buckling laboratory Testing Dissection Procedure, NKT doc. No. 2895-DOC-TST-501.

[46]

Test procedure for lateral buckling: Gomez 6 export riser, NKT doc. No. 8062-DEV-3018.

[47]

Test report for lateral buckling: Gomez 6 export riser, initial trial test, NKT. doc. No 8062-DEV3019.

[48]

Dissection report Lateral buckling test Gomez 6, initial trial, NKT doc. No. 8062-DEV-3161.

[49]

Radius of curvature estimation - Gomez 6" - initial trial, NKT doc. No. 8062-DEV-3019.

[50]

Lateral Buckling Laboratory Testing Procedure, 6" Gomez, Test I, NKT doc. No. 2895-DOC-TST502.

[51]

Lateral buckling laboratory testing report, Gomez Test I, NKT doc. No. 2895-DOC-TST-504.

[52]

Lateral buckling laboratory testing dissection report, Gomez Test I, NKT doc. No. 2895-DOC-TST503.

[53]

Lateral Buckling Laboratory Testing Procedure, 6" Gomez, Test II, NKT doc. No. 2895-DOC-TST515.

[54]

Lateral buckling laboratory testing report, Gomez Test II, NKT doc. No. 2895-DOC-TST-516.

[55]

Lateral buckling laboratory testing dissection report, Gomez Test II, NKT doc. No. 2895-DOC-TST514.

[56]

Lateral Buckling Laboratory Testing Procedure, 6" Gomez, Test III, NKT doc. No. 2895-DOC-TST518.

[57]

Lateral buckling laboratory testing report, Gomez Test III, NKT doc. No. 2895-DOC-TST-519.

[58]

Lateral buckling laboratory testing dissection report, Gomez Test III, NKT doc. No. 2895-DOCTST-517.

[59]

Lateral buckling laboratory testing procedure, NKT doc. No. 2876-DOC-TST-120 (14" jumper).

[60]

[61]

[62]

[63]

Lateral buckling test program for Block-15 - 906, NKT doc. No. 8098-DEV-5423.

60

[64]

Lateral buckling laboratory testing report, Block-15 906, NKT doc. No. 8098-DEV-5807.

[65]

Lateral buckling laboratory testing dissection report, Block-15 906, NKT doc. No. 8098-DEV5808.

[66]

Lateral buckling laboratory testing procedure, 8" DWD, NKT doc. No. 8098-DEV-5595.

[67]

Lateral buckling laboratory testing report, 8" DWD, TEST I, NKT doc. No. 8098-DEV-5791.

[68]

Lateral buckling laboratory testing dissection report, 8 DWD, TEST I, NKT doc. No. 8098-DEV6349.

[69]

Lateral buckling laboratory testing procedure, 8" DWD, TEST II, NKT doc. No. 8098-DEV-5716.

[70]

Lateral buckling laboratory testing report, 8" DWD, TEST II, NKT doc. No. 8098-DEV-5792.

[71]

Lateral buckling laboratory testing dissection report, 8 DWD, TEST II, NKT doc. No. 8098-DEV6350.

[72]

A simplified model for the axial compression properties of a flexible pipe, NKT doc. No.

8098-DEV-5943.

[73]

A model for a beam defined on a frictionless torus, NKT doc. No. 8098-DEV-5820.

[74]

[75]

Lateral buckling design methodology certification, Bureau Veritas doc. No. E&P11190Z-R2011-01 NKT Lateral buckling, Rev. Draft 2011 12 23.

[76]

[77]

[78]

[79]

[80]

[81]

]

Engineering Structures

journal homepage: www.elsevier.com/locate/engstruct

A method for prediction of the equilibrium state of a long and slender wire

on a frictionless toroid applied for analysis of flexible pipe structures

Niels Hjen stergaard a,, Anders Lyckegaard b, Jens H. Andreasen c

a

NKT-Flexibles, Priorparken 480, Brndby, Denmark

c

Department of Mechanical and Production Engineering, Aalborg University, Denmark

b

a r t i c l e

i n f o

Article history:

Received 11 April 2011

Revised 5 September 2011

Accepted 3 October 2011

Available online 9 November 2011

Keywords:

Mechanics of flexible pipes

Curved beams

Armour wire equilibrium

a b s t r a c t

This paper concerns the behavior of a helical wire on a frictionless cylindrical surface subjected to bending, such that the wire in the deformed state constitutes a curve on a toroid. In order to determine the

equilibrium in the loaded state, a sixth order system of differential equations based on curved beam equilibrium and concepts from differential geometry will be presented. On this basis, a method for analysis of

curved beams on frictionless toroids is established. Helically wound steel wires are widely used in flexible pipes, which have various applications in the offshore industry, in order to ensure the structural

integrity against axial loads. The research presented in this paper constitutes a contribution to the field

of wire mechanics, since it enables calculation of the frictionless wire equilibrium state within the wall of

a flexible pipe.

2011 Elsevier Ltd. All rights reserved.

1. Theory

In the present paper, the mechanics of a helically wound wire

modeled as a curved beam is investigated. The wire will be considered embedded in a frictionless cylindrical surface, which is bent

into a toroid. Helically wound wires are widely used in steelpolymercomposite structures such as flexible pipes and umbilicals.

Such structures are usually unbonded and often used in the offshore

industry for transport of liquid or gas. The helical windings constituting the tensile armour of these structures have as primary function to ensure the structural integrity against longitudinal loads,

see Fig. 1. In most known designs two layers of armour wires are applied and designed such that axial strain and twist do not, or to a very

low extend, couple. When a pipe structure with helical windings as

structural element within the pipe wall is subjected to bending and

longitudinal loads, the wires will slip towards an equilibrium state in

which the lay angle may not be constant, which is the case in the initial unloaded state. However, since friction on the wires limits sliding, multiple load cycles must be applied in order for the wires to slip

towards the limit equilibrium curve. Established models of this

mechanical behavior have mainly been based on prescribed geometrical wire configurations. It is often assumed either, that the lay

angle remains constant in bending, which for constant radius of curvature yields a loxodromic curve on a torus surface. Another widely

Corresponding author. Tel.: +45 41908457.

E-mail addresses: Niels.HojenOstergaard@nktflexibles.com (N.H. stergaard),

Lyckegaard@nktflexibles.com, Anders.Lyckegaard@nktflexibles.com (A. Lyckegaard),

jha@m-tech.aau.dk (J.H. Andreasen).

0141-0296/$ - see front matter 2011 Elsevier Ltd. All rights reserved.

doi:10.1016/j.engstruct.2011.10.012

which may be applied assuming, that a wire in bending and tension

will seek equilibrium by taking the path which yields the shortest

possible distance on the torus surface.

The mechanics of helically wound structural elements have

been the subject of research, among others by Costello [1], considering large deflections of helical springs by applying curved beam

equilibrium equations. Spillers et al. [2], used a similar approach

for structures in which radial deformations are restricted in order

to calculate stresses in a helically wound tape on a bent cylinder

surface. However, the lay angle was assumed constant throughout

the tape.

Calculation of stresses and slips in armour wires has been investigated extensively in research published throughout the past few

decades. Fret and Bournazel [3], presented methods capable of

predicting the radial and axial deformations and axial twists of

straight flexible pipes and umbilicals subjected to axial loads, also

on basis of the assumption of constant lay angles. Witz and Tan [4]

and Kraincanic and Kabadze [5] investigated slip initiation and

progression along curves with constant lay angles in armouring

layers subjected to bending and tension. Svik [6], proposed a

computational model based on a finite element formulation for

prediction of slips and stresses in armour wires. This approach

was used to calculate the fatigue lifetime of the wires. The geometry of the wires, when assuming slip to occur towards the geodesic, was investigated by Out and von Morgen [7], by considering

the solution to the EulerLagrange equation applied to a torus

geometry. Leroy and Estrier [8], modeled the mechanical behavior

of the armouring wires in repeated bending with friction. However,

392

Nomenclature

r

R

Lpitch

/

/hel

j

t

n

b

h

u

a

ds

ds0

L

SL

major torus radius

pitch length

wire angle

helical lay angle

pipe curvature, j 1R

local unit wire tangent

local unit wire normal

local unit wire binormal

angular torus coordinate

arclength torus coordinate

wire geometry as relation in u and h

wire arclength increment,loaded state

wire arclength increment,unloaded state

total length of toroid/cylinder

total length of wire

curves was applied. The method proposed in the present paper

aims to determine the geometrical wire configuration, which actually satisfies the equations of equilibrium for a curved beam

embedded in a torus surface. This topic has only to a very low extend been subject of research. Further investigations are therefore

needed in order to obtain a method which with no geometrical

constraints is capable of predicting this equilibrium state. In order

to fulfill these requirements, curved beam equilibrium as presented by Love [9], and Reissner [10], will be considered. Furthermore, concepts from differential geometry will be applied in order

to describe the wire geometry. The derived system of equations

was briefly presented on a simplified form by stergaard et al.

[11], applied to a stability problem occurring in flexible pipes.

The full derivation of the equations and results obtained from single wire analyses will be included in this paper.

In order to obtain a prediction of the limit state, which the wires

slip towards, frictional loads will be neglected. Furthermore, only

cylindrical structures bent to a constant curvature, j 1R, will be

considered. In the present approach, lateral wire interaction will

be neglected. The validity of this assumption can be tested by modeling of all wires within the pipe wall, but inclusion of frictional effects in the formulation may be necessary in order to ensure, that

analyzing all wires separately is reasonable.

in flexible pipe.

jn

jg

s

Pt ; Pn ; Pb

Mt ; Mn ; Mb

pt ; pn ; pb

mt ; mn ; mb

DL

L

Dw

L

geodesic wire curvature

wire torsion

tangential wire strain

wire sectional forces

wire sectional moments

wire distributed loads

wire distributed moments

pipe strain

pipe twist

elastic modulus

shear modulus

wire moments of inertia

wire cross sectional area

E

G

In ; I b ; J

A

2. Methods

A consistent system of differential equations for prediction of

the wire equilibrium state will now be derived by considering a

single armour wire. For the sake of simplicity, the radius of curvature for the centroid of the underlying surface will in the loaded

state be considered constant. The initial stress-free wire state will

be assumed helical.

2.1. Surface parameterization and tangent geometry

Having assumed the radius of curvature R j1 constant, the wire

will in the loaded state constitute a curve on a toroid. Initially, a

parameterization of the torus surface must be established. An arclength coordinate, u, along the torus centerline, and an angular

coordinate, h, along the minor torus radius, will be chosen as

parameterization, see Fig. 2. The torus equation is then

2 1

6

xu; h 4

3

cos h cosju j1

7

1

j r cos h sinju 5

r sin h

j r

given by Eq. (1) for a set of (u, h)-coordinates. On the other hand, a

curve on the torus surface is constructed by specifying a relation between u and h. Assuming that such a relation is given on explicit

393

triad of unit vectors can be attached to each point on the curve

da

ds

xu xh

kxu xh k

btn

in which the vectors xu and xh are the surface derivatives with respect to the torus coordinates, which also span the tangent space

of the torus surface, see Fig. 3

xu

@x

@u

xh

@x

@h

vectors, the following definition of the basis will be applied in the

following

tu

xu

kxu k

th

xh

kxh k

The definition given in Eq. (4) enables the following alternative definition of the tangent of the curve, which the wire constitutes

da

du

dh

xu

xh

ds

ds

ds

cos /tu sin /th

This corresponds to stating

dh sin /

ds kxh k

These equations govern the kinematics of the wire. The norms kxu k

and kxh k are determined by calculating xu and xh as defined in Eq.

(3) on basis of the torus geometry given in Eq. (1)

6

xu 4

Since kinematic equations for the wire have now been derived,

curvature components must be determined. Having followed an

approach by which a curvilinear coordinate triad is attached to

each point on the wire, it is desirable to express curvature components in terms of a geodesic and a normal component, jg and jn,

respectively, along with a component representing the geodesic

torsion, s. A result from basic differential geometry known as the

Darboux frame, based on the FrenetSerret differential formulas,

is available for relating the triad vectors to their first order derivatives with respect to arclength in terms of curvature components.

The Darboux frame is defined as

2 3 2

0

t

d6 7 6

n

4 5 4 jn

ds

jg

b

jn

jg

s 0

32 3

t

76 7

54 n 5

b

10

Assuming the unit triad and corresponding arclength derivatives given, this corresponds to the following definitions of the curvature

components

dt

dn

t

ds

ds

db

dt

jg t b

ds

ds

dn

db

s b n

ds

ds

jn n

In order for this definition to be consistent with the definition given

in Eq. (2), the following must hold

du cos /

ds

kxu k

j j1 r cos h sinju

1

j j r cos h cosju 7

5

r cosju sin h

6

7

xh 4 r sinju sin h 5

2

11

The curvature components can on basis of this definition be calculate in terms of u, h and / along with appropriate derivatives of

those. This direct approach was followed both by Svik [6], and

Leroy and Estrier [8]. However, direct application of Eq. (11) to

the coordinate triad given in Eq. (2) leads to very long expressions

which are difficult to handle. Therefore, concepts from abstract differential geometry will be applied. The approach has been shown in

a mathematical sense to be similar to calculating the dot products

and derivatives given in Eq. (11), but eases the task significantly.

Initially, it is noted that the normal curvature component can be

determined on basis of Eulers formula

12

In this equation, j1 and j2 denotes the principle curvature components of the torus surface. These are by definition given by

Sxu j1 xu

Sxh j2 xh

13

r cos h

2

kxu k 1 r j cos h

kxh k r 2

Sxu nu

9

@n

@u

Sxh nh

@n

@h

14

cosju cos h

6

7

n 4 sinju cos h 5

15

sin h

3

j sinju cos h

6

7

Sxu 4 j cosju cos h 5

0

2

cosju sin h

6

7

Sxh 4 sinju sin h 5

cos h

16

17

394

Sxu

j cos h

xu

1 rj cos h

1

Sxh xh

r

18

Reviewing Eq. (14), it is now clear that the principle curvature components of the torus surface are given by

j1

j cos h

1 rj cos h

j2

1

r

19

jn

j cos h

1

2

cos2 / sin /

1 rj cos h

r

20

[12], provides a definition of the geodesic torsion which eases

determination of an explicit expression

s j1 j2 cos / sin /

21

with the sign convention given in Eq. (10). Substituting appropriate

values into this definition yields the following expression

j cos h

1

cos / sin /

1 r j cos h r

22

definition is derived for a curve a(s) parameterized by arc length

embedded in a surface in which the angle / is given equivalent to

the definition applied in Eq. (12)

jg

d/

jg 1 cos / jg 2 sin /

ds

23

respectively of a curve defined by a constant h-coordinate and of

a curve defined by a constant u-coordinate. (jg)1 is obtained directly from the definition given in Eq. (11)

jg

3

2

3 2

cos u

sin h cos u

1

7

6

7 6

R r cos h

0

cos h

in which the scalar factors and the first vectorial term represents

the change rate of the curve tangent in arclength and the second

term represents a vector in the tangent plane normal to this tangent. Performing the scalar product the following expression is

obtained

2

jg 1

sin h

R r cos h

R r cos h

so

jg 2 0

24

j sin h

d/

jg

cos /

1 r j cos h

ds

25

correspond to the definitions given in Eq. (11).

This concludes the geometrical analysis.

2.3. Equations of equilibrium

their derivatives with respect to arclength to obtain a set of

component wise equilibrium equations, see Fig. 4. The internal

and external force and moment-vectors in Eq. (26) are within this

framework given by

P Pt t Pn n Pb b

M Mt t Mn n Mb b

p pt t pn n pb b

m mt t mn n mb b

With the current definition of an orthonormal basis, the equations

of equilibrium can be written as

dP

dt

dn

db

dPt

dPn

dPb

p Pt Pn

Pb

t

n

b

ds

ds

ds

ds

ds

ds

ds

pt t pn n pb b 0

27

dM

dt

dn

db

dM t

dMn

m t P Mt Mn

Mb

t

n

ds

ds

ds

ds

ds

ds

dM b

b

mt t mn n mb b Pb n Pn b 0

ds

28

Applying Eq. (10), the derivatives in Eq. (28) can be calculated, so

the componentwise equations of equilibrium are

dPt

jn Pn jg Pb pt 0

ds

dPn

jn Pt sPb pn 0

ds

dPb

jg Pt sPn pb 0

ds

dM t

jn Mn jg M b mt 0

ds

dM n

jn Mt sM b Pb mn 0

ds

dM b

jg Mt sM n Pn mb 0

ds

29

30

31

32

33

34

the derivation of those is mainly included here in order to ensure,

that the signs correspond to the definition given in Eq. (10).

The equilibrium equations for a curved beam segment is derived on vectorial form by Reissner [10]

dP

p0

ds

and slender wires with dimensions much smaller than both minor

dM

tPm0

ds

26

relations valid

Pt EA

35

M t GJ Ds

36

M b EIb Djn

37

M n EIn Djg

38

in which the changes of curvature are given with respect to the initial helical state in which the structure is unloaded

ini

Ds s sini Djn jn jini

n Djg jg jg

sin /hel

r

sin /hel cos /hel

ini

s

r

jini

g 0

jini

n

The angle /hel is the initial wire lay angle defined in terms of the

minor torus radius and the pitch length Lpitch

39

modeled wire does not rotate with respect to the Darboux frame,

since the tangential wire rotation is constrained such that it is governed solely by the underlying toroid. Since the geometry, equilibrium and constitutive relations for a wire have now been derived on

basis of assumptions which provides sufficient precision for most

engineering purposes, means for obtaining a solution for the wire

equilibrium state can now be considered.

2.5. System of field equations

Considering the wire kinematics given in Eq. (6), rearranging

the expression for the geodesic curvature in Eq. (25) and the equilibrium in the tangent plane in Pt, Pb and Mn, the following sixth order system of equations is derived

du

cos /

ds 1 r j cos h

dh sin /

ds

r

d/

j sin h

cos / jg

ds

1 r j cos h

dPt

jn Pn jg Pb

ds

dPb

jg Pt sPn

ds

dMn

jn Mt sM b Pb

ds

40

dMb

djn

EIb

ds

ds

42

43

44

47

djn

ds

can be obtained by analytical means. Having obtained a solution to the differential equations, unknown quantities, pn and mt

are given in terms of known functions. The external loads are geometrically specified by the two remaining equations of equilibrium governing the normal force and torsional equilibrium of

the wire

48

49

t

Unknown terms dPdsn and dM

can be derived analytically. The physical

ds

explanation for the presence of the moment reaction mt, which is

assumed induced by adjacent pipe layers is, that it constrains the

modeled wire to the toroid such that the rotation around the local

tangent is governed solely by the underlying surface. This is equivalent to assuming that the material coordinatesystem and the

Darboux frame do not rotate relative to each other.

The derived system of field equations is solved as a boundary

value problem of sixth order, which can be solved with respect

to six boundary conditions. The analytical solution to the equilibrium equations for a curved beam segment is only available in a

few simple cases. It is therefore desirable to establish a numerical

method for solving the obtained system of field equations, rather

than to attempt to obtain an analytical solution. The system will

be solved by use of a commercially available solver known as

bvp4c in the MATLAB programming environment. The solver

applies the Lobatto-IIIa method based on finite differences and is

capable of performing re-meshing until convergence has occurred.

However, the differential equations have been derived as function

of arc length, s, in the deformed state, which is unknown. Considering the modeled wire inextensible, s is equal to the arclength in

the initial helical state, s0, which is given by the well-known

relation

41

s0

rh

sin/hel

50

and s0 can be related by Cauchys definition of strain

ds 1 ds0

51

45

(22). Since the torus surface is considered frictionless, the distributed loads in the tangent plane pt and pb can be considered zero.

Furthermore, the distributed moments mn and mb will be considered zero. This assumption corresponds to, that the twist around

the wire tangent is geometrically governed by the underlying toroid

and that local tangential twist, usually denoted fish-scaling in flexible pipes armour wires, is neglected. Pn can be obtained from the

equilibrium equation governing the binormal moment equilibrium

in Eq. (34)

dMb

Pn

jg M t sMn

ds

can be determined by

dP n

jn Pt sP b

ds

dM t

jn Mn jg M b

mt

ds

2pr

Lpitch

dM b

ds

pn

tan/hel

in which

395

46

Considering the derived curvature components, these can be rewritten on the form

jn n

jg t

sb

dt

dt ds0

ds0

n

j0n

ds

ds0 ds

ds

db

db ds0

ds0

t

j0g

ds

ds0 ds

ds

dn

dn ds0

ds0

b

s0

ds

ds0 ds

ds

basis of s0. The derived field equations can on this basis be rewritten

as

396

du ds0

cos /

ds0 ds

1 rj cos h

52

dh ds0 sin /

ds0 ds

r

53

d/ ds0

j sin h

cos / jg

ds0 ds

1 rj cos h

54

dP t ds0

jn Pn jg Pb

ds0 ds

55

dP b ds0

jg Pt sP n

ds0 ds

56

dM n ds0

jn M t sMb Pb

ds0 ds

57

first equations, the triad vectors are still of unit length. Substituting

the strain definition given in Eq. (51) into the field equations above

yields

du

cos /

1

ds0

1 r j cos h

58

3. Results

dh

sin /

1

ds0

r

59

d/

j sin h

cos / j0g

1

ds0

1 r j cos h

60

dP t

j0n Pn j0g Pb

ds0

61

dP b

j0g Pt s0 P n

ds0

62

dM n

j0n Mt s0 Mb Pb 1

ds0

63

capable of providing, five 1 m long pitches of wire within a pipe

wall will be modeled, see Fig. 5. The wire cross section will be

set to 5 10 mm. The wire will in the initial state be assumed to

constitute a helix of radius 0.2 m. This helix will restrained against

radial deformation be bent to a radius of curvature, R = 10 m. Since

the wire now rests on a torus surface, the equilibrium state can be

determined. Geometrical boundary conditions will be applied

corresponding to end-fittings, which are usually applied in flexible

pipe or umbilical designs. This corresponds to fixing the displacements in both ends of the wire and setting the wire angle equal

to the pitch angle, /hel. Generalized longitudinal loads are applied

as axial strain DLL and twist DLw

which were obtained by Reissner [10], if the shear strains which

were taken into calculation in this publication, are neglected. While

j0g is proportional to Mn obtained from the modified system of field

equations, the remaining curvature components must be modified

in the constitutive relations in order for those to be consistent

P t EA

ds0

sini

ds

ds0

EIb j0n

jini

ds

M t GJDs GJ

M b EIb Djn

s0

Pn

1

dM b

j0g Mt s0 Mn

1

ds0

64

dM b

d

EIb

ds0

ds0

j0n

ds0

ds

EIb

0

n

dj ds0

d ds0

EIb jn

ds0 ds

ds0 ds

65

d ds0

d

1

d

1

ds0 1 2

ds0 ds

ds0 1

dP t

1

1

j0n Pn j0g P b

2

ds0 EA1

EA1 2

u0 0 h0 hAini / 0 /hel

DL

Dw

uSL L 1

hSL hBini

L /SL /hel

L

L

on analytical form.

68

and bending against the geodesic as reference curve. A geodesic

curve on a surface is per definition the curve which minimizes

the arc length between two specified points. It can be shown mathematically, that this corresponds to a curve possessing zero geodesic curvature.

The geodesic can be obtained as the solution to the following

two field equations with boundary conditions on h, in which the

jg-term is set to zero

dh sin /

ds

r

d/

j sin h

cos /

ds

1 rj cos h

69

70

Alternatively, the geodesic can be constructed on basis of the following derivative, which is given in [13]

p

dh m m2 B2

du

rB

66

67

71

Applying the following boundary conditions h0 hAini and

hS hBini , the wire angle, /, can be determined by

tan /

dh

r

du 1 r j cos h

72

397

equations which specifies the geodesic are solved by the same

BVP-solver as applied to the full system of equations. For the sake

of simplicity, the geodesic is determined assuming s = s0.

3.1. Curvature components comparison

In order to validate the derived curvature components, these

will be compared with results obtained by Leroy and Estrier [8].

In this paper the following curvature components were used

3

dg dh

jg sin h2r2 g R2 g2 g 3 rRg

dh

ds

2

dh

jn r Rg cos hg 2

ds

2

dh

s Rg

ds

73

74

Table 1

Straight pipe in tension and torsion.

DL

L

0.001

0.001

0.001

0

p

180

p

180

2

ds

r 2 R2 g2 g 2

dh

dx

g

dh

r

g 1 cos h

R

in which u = xR. Considering a loxodromic curve with pitch length,

Lpitch = 1 m, on a toroid with minor radius r = 0.2 m and major radius

R = 10 m, changes of curvature with respect to the initial helical

configuration as presented on Fig. 6 are obtained. The two sets of

curvature components can be observed to correspond to each other.

3.2. A wire on a tensioned cylindrical surface

Considering a 5 10 mm wire of 1 m pitch length on a cylindrical surface (j = 0) of radius r = 0.2 m subjected to tension, the load

in the axial direction of the cylinder is given by

76

Fret and Bournazel [3], who derived the axial force to be given by

2.535

8.585

13.655

77

Results are presented in Table 1. Results can be observed to correspond well, since the method derived in this paper predicts no

transverse slip in tension and with zero pipe curvature, and the

equations used for references are derived for perfect helices.

considered. In order to study the geometrical configuration

obtained by the analysis, the wire angle / will be considered, see

Fig. 7. The wire angle can be observed to approach the angle corresponding to the geodesic curve, when tensioned. It is noted that

significant change of angle occurs with respect to the initial helical

angle. Considering the changes of curvature with respect to the initial helical state for the two equilibrium configurations, see Figs. 8

and 9, it is clear that while the normal curvature increases from a

state in pure bending to a state with both bending and tension, the

geodesic curvature decreases significantly.

It will be investigated further, if this tendency holds. Considering multiple values of tensile axial strain, the results presented in

Fig. 10 are obtained. It can be observed, that the geodesic curvature

decreases significantly when the wire is tensioned further. The

developed method has therefore this far exhibited the expected

behavior since the geodesic curvature vanishes in bending when

the wire tension is increased. This corresponds to that the wire

slips towards a limit curve possessing no geodesic curvature,

namely, the geodesic curve. As basis of comparison regarding the

magnitude of curvature, the normal curvature is shown in an

equivalent manner in Fig. 11.

Considering the sectional force responses of the wire in tension

DL

0:001 on Fig. 12, the tangential wire load in the wire can be

L

observed to be large, but with very small variations in arclength.

pure bending

bending and tension,=0.001

loxodromic curve

geodesic curve, FDalgorithm

bending and tension,=0.002

geodesic curve ref. [13]

0.93

ref.[8]

g

n ref.[8]

ref.[8]

0.02

0.04

0.92

2.537

8.587

13.642

F

DL

Dw

cos2 /hel

r sin /hel cos /hel

cos /hel EA

L

L

0.02

F(kN)

0.04

P(kN)

75

in which a minus is added to jg and s in order to obtain correspondence between the sign conventions, which can be derived

comparing the applied Darboux frames. The unknown terms are

given by

0.06

Dw

L

0.91

0.9

0.89

0.06

0.88

0.5

1.5

Fig. 6. Curvature components of loxodromic curve.

Fig. 7. Wire angle / in tension and bending.

398

0.1

0.2

n

0.08

0.06

0.04

0.02

0

0.1

0

0.1

0.2

0

1

0.02

x 10

0.04

0.06

0

Pipe strain 4

Fig. 8. Changes of wire curvature components, pure bending.

30

Pt 103

20

Pn

Pb

0.1

Fig. 11. Change of normal curvature for a wire in bending and varying tensile pipe

strain.

0.15

0.05

10

0

10

20

30

0.05

40

0.1

0

50

20

0.2

15

0.1

25

0

0.1

0.2

0

10

5

0

5

10

15

x 10

Pipe strain

10

mt (Nm/m)

20

pn103 (N/m)

25

0

Fig. 10. Change of geodesic curvature for a wire in bending and varying tensile pipe

strain.

with significantly larger variations in arclength. On this basis it

can be concluded that it remains fair to neglect shear strains in

the constitutive relations, since this assumption depends on small

shear forces.

The distributed normal load pn and the distributed moment mt,

which are needed to have the two remaining equations of equilibrium fulfilled, are shown on Fig. 13. pn can be observed to be quite

large but as expected directed outwards with respect to the pipe

center. The physical explanation for the presence of mt is, as mentioned in Section 2.5, that this moment constrains the wire to the

surface in a manner, such that no rotation occurs between the

material frame and the Darboux frame. This is considered physically meaningful, since adjacent layers limits tangential wire

rotations.

In a physical pipe structure, radial expansion of the underlying

toroid is allowed, which may have large influence on the magnitude of pn. It was demonstrated briefly in [11] how the effect of a

radially elastic pipe structure on the wire responses can be taken

into calculation by considering the minor torus radius r a function

of the applied longitudinal strain. This aspect, however, should be

investigated further, since it may impose large impact on the stiffness of the armouring wires being modeled by the presented

method.

4. Conclusions

On basis of curved beam equilibrium and concepts from differential geometry a consistent sixth order system of differential

equations describing the mechanical behavior of a long and slender

wire on a frictionless toroid surface has been derived. Solving this

system as boundary value problem, a method for prediction of the

frictionless wire equilibrium state is demonstrated. This method is

based only on the geometrical assumption, that the tangential wire

rotation is governing solely by the underlying surface. A method

for derivation of the wire curvature components based on concepts

from abstract differential geometry has been presented. Despite

the focus in the presented approach has been solely on constant

radius of curvature of the centroid of the supporting surface, the

approach followed in this paper may ease the derivation of curva-

399

constant radius of curvature. Results show, that bending and tension tend to eliminate the geodesic curvature component of the

wire. This indicate, that the curve, which possesses no geodesic

curvature, namely, the geodesic on a torus surface, constitutes

the asymptotic limit state of the wire, when the pipe is tensioned

or bent further. However, further research is needed in order include frictional effects in the present formulation in order to examine, which influence this may impose on the obtained results.

The presented method may have a wide range of applications

such as analysis of transverse stability as shown in [11] in the field

of wire mechanics which constitutes an important discipline when

designing flexible pipes. Furthermore, the presented method may,

if frictional effects and radial pipe equilibrium are included, serve

as basis for computation of how cyclic loadings effects the mechanical behavior of the wires.

References

[1] Costello GA. Large deflections of helical spring due to bending. J Eng Mech Div

Proc ASCE 1977;103:112433.

[2] Spillers WR, Eich ED, Greenwood AN, Easton R. A helical tape on cylinder

subjected to bending. J Eng Mech Div Proc ASCE 1983;109:112433.

[3] Fret JJ, Bournazel CL. Calculation of stresses and slips in structural layers of

unbonded flexible pipes. J Offshore Mech Arctic Eng 1987;109.

[4] Witz JA, Tan Z. On the flexural structural behaviour of flexible pipes,

umbillicals and marine cables. Mar Struct 1992;5.

[5] Kraincanic I, Kabadze E. Slip initiation and progression in helical armouring

layers of unbonded flexible pipes and its effect on pipe bending behavior. J

Strain Anal 2001;36:3.

[6] Svik S. A finite element model for predicting stresses and slip in flexible pipe

armouring tendons. Comput Struct 1993;46:2.

[7] Out JMM, von Morgen BJ. Slippage of helical reinforcing on a bent cylinder. Eng

Struct 1997;19(6):50715.

[8] Leroy JM, Estrier P. Calculation of stresses and slips in helical layers of

dynamically bent flexible pipes. Oil and Gas Science and Technology. REV. IFP

2001;56(6):54554.

[9] Love AEH. A treatise on the Mathematical Theory of Elasticity. New

York: Dover Publications, Inc.; 1944.

[10] Reissner E. On finite deformations of space-curved beams. J Appl Math Phys

(ZAMP) 1981;32.

[11] stergaard NH, Lyckegaard A, Andreasen J. On lateral buckling failure of

armour wires in flexible pipes. OMAE2011-49358. In: Proceedings of OMAE,

2011.

[12] doCarmo MP. Differential geometry of curves and surfaces, 2008.

[13] CAFLEX Theory manual. IFP/SINTEF, 1991.

]

Applied Ocean

Research

Applied Ocean Research 1 (2012) 111

Niels Hjen stergaard

NKT-Flexibles / Aalborg University, Department of Mechanical and Production Engineering, Denmark

E-mail: Niels.HojenOstergaard@nktflexibles.com

Anders Lyckegaard

NKT-Flexibles, Priorparken 480, Brndby, Denmark

E-mail: Anders.Lyckegaard@nktflexibles.com

Jens H. Andreasen

Department of Mechanical and Production Engineering, Aalborg University, Denmark

E-mail: jha@m-tech.aau.dk

Abstract

The work presented in this paper is motivated by a specific failure mode known as lateral wire buckling occurring

in the tensile armour layers of flexible pipes. The tensile armour is usually constituted by two layers of initially

helically wound steel wires with opposite lay directions. During pipe laying in ultra deep waters, a flexible pipe

experiences repeated bending cycles and longitudinal compression. These loading conditions are known to impose

a danger to the structural integrity of the armouring layers, if the compressive load on the pipe exceeds the total

maximum compressive load carrying ability of the wires. This may cause the wires to buckle in the circumferential

pipe direction, when these are restrained against radial deformations by adjacent layers.

In the present paper, a single armouring wire modeled as a long and slender curved beam embedded in a frictionless

cylinder bent into a toroid will be studied in order to gain further understanding of this failure mode. In order to study

the compressive behavior, both perfect beams as well as beams with small geometrical imperfections are studied. The

mathematical formulation of the problem is based on curved beam equilibrium and allows large deflections to be

taken into calculation.

1. Introduction

Marine structures such as flexible pipes and umbilicals are usually unbounded steel-polymer composites.

Such structures are often used in the offshore industry

during field development at large water depths. In most

known pipe designs, two layers of helically wound steel

wires are applied in order to ensure the structural integrity with respect to longitudinal loads, see Figure 1.

The mechanical behavior of these armouring tendons

subjected to longitudinal loads and bending have been

widely investigated in academic as well as industrial research in the past few decades. Established methods for

numerous failure modes. However, a few failure modes

are still of such complexity, that they remain a motivation of research. Among those are lateral wire buckling

of the tensile armour layers, see Figure 2 and 3. This

failure mode is known mainly to occur during pipe laying in deep-waters. In this scenario, the flexible pipe

is in a free-hanging position from an installation vessel

to the seabed, which leads to repeated bending cycles

due to wave loads and vessel movements. Furthermore,

the pipe is empty during installation, such that compression occurs due to hydrostatic pressure on the end cap

is known to impose a danger to the structural integrity,

especially when the outer sheath (a polymeric layer) is

damaged. This implies, that the pipe annulus is flooded,

such that the armouring wires are surrounded by water.

In this situation, external hydrostatic pressure does not

induce sufficient friction to prohibit wire slippage.

The problem was first described by Braga and Kaleff,

[1], who reproduced the failure mode experimentally

in a mechanical test bench. Further experimental research is described in [2], [3] and [4]. However, a

fulfilling mathematical investigation of the underlying

mechanisms has not yet been conducted on analytical

basis, despite attempts to capture the physics of the failure mode by application of finite element analysis have

been published, see [5] and [6].

The mathematical approach followed in the present

paper is essentially based on the mechanics of curved

beams, which have already been investigated in numerous publications. The equations of equilibrium for a

three-dimensional slender curved beam segment were

derived by Love, [7], on component form. On basis of a

vectorial formulation, these were investigated further by

Reissner, [8], who by means of variational calculus derived the mathematics governing the rotational field and

on basis of a finite-strain approach showed how extensibility of the modeled beam effected the equations of

equilibrium. Similar equations were used by Costello,

[9], investigating the equilibrium state of a bent helical

spring with large deformations, and by Spillers et. al.,

[10], addressing the problem of a tape on a bent cylinder. Stump and van der Heijden, [11], studied the equilibrium state of cables modeled as curved beams. Furthermore, Leroy and Estrier, [12], based their analysis of

armouring wires in flexible pipes subjected to dynamic

loads on similar equations. However, the underlying assumption regarding a prescribed solution form based on

experience with geodesic and loxodromic curves, see

do Carmo, [13], prevents this approach from being applicable when describing the mechanical behavior of a

curved beam in compression and bending. The detailed

mechanical behavior of the armouring wires of flexible

pipes subjected to bending and tension have been investigated in various publications. Svik, [16], showed

how slip could be predicted by a finite element model

formulated on basis of finite strain continuum mechanics. Witz and Tan, [14], and Kraincanic and Kabadze,

[15], investigated wire slip mechanisms by analytical

means for flexible pipes in bending to a constant curvature. However, all these approaches were based on

the underlying assumption, that the wire lay angle remains constant in bending. Out and Von Morgen, [17],

towards the geodesic on a toroid.

The approaches summarized above are all capable of

providing results which correspond well to the physical behavior of flexible pipes in bending and tension.

However, none of them are free of geometrical constraints so that the limit equilibrium state of an armouring wire reached after a significant number of bending cycles can be determined. A sixth order system of

equations specifically derived for this purpose was proposed by stergaard et. al., [18], in which it was shown

that bending and tension caused slippage towards the

geodesic configuration. The mathematical formulation

of this problem was based on the geometry of a long

and slender beam embedded in a frictionless cylinder

bent into a toroid and curved beam equilibrium as given

in [7] and [8]. Furthermore, it was assumed that if frictional effects are neglected, the beam will when loaded

instantly reach the equilibrium state, which when friction is present is reached after a significant number of

bending cycles. This assumption may be justified, when

the outer sheath is damaged, since friction induced by

external pressure does not limit wire slippage.

The proposed method was applied to the lateral wire

buckling problem in [3] on a simplified form. However, further research is needed in order to investigate

the possibilities arising from this method as a tool for

stability analysis.

In the present paper, the mechanical behavior of a single tensile armour wire within the wall of a flexible pipe

subjected to longitudinal compression and bending will

be investigated. Analyses will be conducted by applying loads in small steps so that the equilibrium path of

the wire can be studied. Furthermore, the influence of

key-model parameters and small geometrical imperfections on the load carrying ability and deformation mode

of the wire will be examined.

The failure mode addressed in the present paper has

traditionally been characterized as a buckling problem.

The theory of elastic stability for non-linear systems

was originally formulated by Koiter, who considered

the initial postbuckling behavior of structures and

showed, that the mechanical behavior of nonlinear

systems in compression may be sensitive to small initial

imperfections. The established framework is well

described in the literature, see [19] and [20]. Equilibrium paths of non-linear systems in force-displacement

diagrams may exhibit limit point behavior, softening

or bifurcation. These types of behavior may be related

to loss of structural stability. In the present work, a

full non-linear analysis of perfect and imperfect wire

geometries will be used to detect possible nonlin-

generated by NKT-Flexibles by laboratory testing

Buckling due to small geometrical imperfections

has been addressed among others by Tvergaard and

Needlemann, [21], and Taylor and Gan, [22]. In both

cases the imperfections were chosen in a manner so

these were representing the buckling modes attempted

triggered. In more complex cases, like the present, the

imperfections cannot be chosen directly on basis of

the system physics. The imperfections will therefore

be chosen in such a manner, that these are capable of

triggering a wide range of deformation modes. It is

noted, that bifurcation points cannot be detected by

the present approach, since the structural behavior is

determined on basis of a full non-linear analysis.

2. Theory

2.1. Wire equilibrium state with transverse slip

In this section, the method for determination of the

wire equilibrium state presented in [18] is summarized.

In order to obtain the limit equilibrium state of an armour wire, friction on the wires will be neglected. It

is assumed that the wire being modeled will reach the

frictionless configuration after a significant number of

bending cycles. Furthermore, the analysis of a single

tensile armour wire will be based on the assumption,

wires, generated by NKT-Flexibles by laboratory testing

assumption, an armour wire can be modeled as constituting a curve on a toroid.

The toroid will be parameterized by an arclength coordinate u along the torus centerline and an angular coordinate, , along the minor torus radius, see Figure 4.

Specifying a relation between u and , a curve can be

constructed. The relation between a point on the surface

in torus- and in Cartesian coordinates is given by

1

+ r cos cos (u)

1

x(u, ) =

+

r

cos

sin (u)

r sin

(1)

curvature related to the major torus radius by R = 1 .

Embedding a curve in the surface, a coordinate triad of

unit length can be attached to each point on the curve

in form of a tangent vector, t, a normal vector, n and a

binormal vector, b given by

t=

d

ds

n=

xu x

kxu x k

b=tn

(2)

in which xu and x constitute a basis for the toroid tangent plane. These vectors are given by

xu =

x

u

x =

(3)

the wire and the underlying toroid, the following differential equations governing the wire tangent geometry

were derived in [18]

du cos

=

ds

kxu k

d cos

=

ds

kx k

(4)

the Darboux frame, a basic result from differential geometry, can be applied

t 0

n g t

d

n = n 0

n

(5)

ds b

b

0

g

(or transverse) curvature and the wire torsion. In [18]

it was shown how curvature components consistent with

the definition given in Eq. 5 could be derived on algebraic form. Considering force equilibrium, the following equations were given by Reissner, [8], on vectorial

form

dP

+p=0

ds

dM

+tP+m=0

ds

(6)

M, distributed external loads p and distributed external

moments m can be written with the curvilinear coordinate triad as basis. This leads to a system of equations

similar to those derived by Love, [7]. Furthermore, assuming that the wire dimensions are small with respect

to minor and major torus radii, it is fair to neglect curved

Assembling the unknowns, the following sixth order

system of coupled differential equations is derived

du

ds

d

ds

d

ds

dPt

ds

dPb

ds

dMn

ds

cos

1 + r cos

sin

=

r

sin

=

cos + g

1 + r cos

(7)

(8)

(9)

= n Pn g Pb

(10)

= g Pt Pn

(11)

= n Mt + Mb + Pb

(12)

Unknown functions are given in terms of known quantities after a solution has been obtained.

In order to obtain a solution by numerical means, it

is desirable to convert the system of field equations

from deformed arclength s to a system in undeformed

arclength s0 . Assuming that axial strains are small,

Cauchys definition of strain applies. Substituting this

into the field equations ensures that a solution is obtained on a regular mesh. A solution to the wire equilibrium is obtained using a commercially available BVPsolver. The model is, since generalized loads are applied

as longitudinal strain L

L and pipe twist L , deformation

controlled. Applying boundary conditions corresponding to the mechanics of an armour wire within the wall

of a flexible pipe, the following can be stated

A

u(0) = 0 (0) = ini

(0) = hel

u(S L ) = L 1 +

L

L

B

(S L ) = ini

(13)

L

L

(14)

(S L ) = hel

in which S L denotes the total wire arclength and L the

total pipe length.

2.2. Nonlinear imperfection analysis

Nonlinear systems may be sensitive to small initial

imperfection, which may trigger buckling. In the following it will be examined if small initial imperfections

cause a mechanical behavior which may be related to

instability of the wire being modeled. The imperfection

will be added directly to the geodesic curvature and is

chosen in such a manner that it may represent a wide

range of physically possible deformation modes

g = g (g,ini g,imper )

(15)

g,ini = 0

(16)

g,imper =

m

X

i=1

i sin

is

L

(17)

state when Mn = 0 possesses a small geodesic curvature. In the following, the influence of the number of

terms taken into account in the sine-series included in

equation 15 and the magnitude of the chosen amplitudes

will be examined. The equilibrium path of the loaded

end of the wire will be considered in order to detect wire

instability. Having obtained a solution to the wire equilibrium state, the wire forces in the tu and t -directions

are given by

Pu

P

=

=

=

=

Pt t tu + Pb b tu

Pt cos + Pb sin

Pt t t Pb b t

Pt sin Pb cos

(18)

(19)

not included in an exact manner in the proposed model.

However, means for estimation of the effect of a pipe

wall with a non-zero Poissons ratio are included in [3]

by an approach equivalent to the methods used in [12]

and [15].

3. Results

3.1. Wire imperfection analysis

In order to examine if buckling of a simple structure can be triggered by the present approach, a plane

straight clamped-clamped steel beam of 3 10mm rectangular cross section and length 1.5m will be analyzed

using the suggested method. Youngs modulus for steel

will be set to E = 210GPa. The analysis is performed

by reformulating the angular torus coordinate as an arclength coordinate w = r along the minor torus radius

and setting r = R = . The Euler buckling load, PE for

a clamped-clamped beam is given by the well-known

formula

PE =

42 EIn

L2

(20)

for a perfect structure, branch A, and an imperfect structure, branch B, with initial imperfection given in Equation 15 with m = 1 and 1 = 0.001m1 . While branch

A can be observed to be linear, branch B exhibits softening behavior with the Euler load as horizontal asymptote. While the present approach is incapable of predicting bifurcation points, it can be concluded that the

imperfect wire exhibits non-linear geometrical softening. This compressive behavior of a beam, in which

large rotations and displacements are allowable, is welldescribed in the literature, see [19].

Now modeling an initially helical steel wire subjected

to bending, four pitches of 3 10mm steel wire with

A

ini

= 0, L pitch = 1.25m and r = 0.1m will be analyzed,

see Figure 6. Poissons ratio will be set to = 0.3 and

the material assumed isotropic. Solutions can now be

obtained by solving the formulated nonlinear boundary

value problem with stepwise increased compressive longitudinal strain. The boundary conditions are given in

equation 13 and 14 with

L = 0. Setting the bending radius, R = 15m, and calculating the sectional force in the

loaded end of the wire as function of the applied pipe

strain, the equilibrium path can be determined. This

curve will be used for detection of behavior which may

be related to buckling.

The influence of the chosen set of possible imperfections in equation 15 will now be examined, see figure

9. The path denoted A on which the longitudinal wire

load remains a linear function of L

by reL , is detected

placing the boundary condition u(S L ) = L 1 + L

L in

equation 14 with a force boundary condition Pu (S L ) =

Papp = cos Pt + sin Pb . The branches denoted B-E

are calculated using the deformation controlled model.

Significant non-linear geometrical softening can be observed. A measure for the maximum compressive load

which can be carried by the wire can be obtained on

basis of the equilibrium paths and the asymptotic load,

which can be considered an estimated value of the buckling load of the wire. It is noted, that the branches, A and

150

100

B

50

Euler Load

0

0.5

1.5

2

2.5

3

Compressive strain

3.5

4.5

5

4

x 10

Figure 7: ( L

L , Pu )-equilibrium paths for clamped-clamped beam in compression, see Figure 6, A: perfect initial geometry, B: imperfect initial

geometry

1400

A

1200

B

C

1000

800

m=0

m=1,i=0.001

m=2,i=0.001

m=5,i=0.001

m=20,i=0.001

600

m=1,i=0.001

m=2,i=0.001

m=5,i=0.001

400

m=20,i=0.001

m=20,i=0.01

m=20,i=0.0001

200

Figure 8: Model of armour wire within the wall of a flexible pipe subjected to bending and longitudinal loads

two different set of boundary conditions. Hence, when

the point, at which geometrical softening starts to occur,

m=20,i=0.005

Primary path

0

3

Compressive pipe strain

Figure 9: ( L

L , Pu )-equilibrium paths of modeled wire for R = 15m, see

Figure 8. A: perfect wire geometry, primary path, B: perfect wire geometry, secondary path, C: equilibrium path, small initial imperfections, D-E:

equilibrium paths, large initial imperfections, all imperfections measured

in m1

perfect wire geometry.

The branch B is obtained for a perfect structure by a

x 10

4

Figure 10: Wire equilibrium states (prior to nonlinear softening), R = 15m, L

L = 1.05 10 , blue curve: bent helix ( = hel in bending),

red curve: Wire equilibrium state, m = 20, 1...20 = 0.001m1 , green curve: Wire equilibrium state, m = 20, 1...20 = 0.001m1

3

Figure 11: Wire equilibrium states (nonlinear force-deformation regime), R = 15m, L

L = 10 , blue curve: bent helix ( = hel in bending),

1

red curve: Wire equilibrium state, m = 20, 1...20 = 0.001m , green curve: Wire equilibrium state, m = 20, 1...20 = 0.001m1

observed that branch B is obtained if the amplitudes of

the imperfection is chosen very small. Increasing the

imperfection amplitudes, a different branch, denoted C,

is obtained. However, it can be concluded that the number of terms taken into account has little, or no influence on the determined equilibrium path. The branches

D and E are obtained if the imperfection amplitudes are

chosen in such a manner, that the wire behavior in the

linear regime on the equilibrium path is influenced. In

the following, only imperfections which prior to buckling have negligible influence on the determined equi-

librium paths will be considered. The detected equilibrium paths are of a shape which correspond well to

tendon behavior detected by finite element analysis in

[5]. The force in pure bending can be observed to be

negative, hence, if the arc length of the torus centerline

is considered constant, a bending-compression coupling

occurs. To give an impression of the magnitude of the

chosen imperfections, the change of normal curvature,

n , with respect to the initial helical state for a wire in

pure bending is shown in Figure 14. The chosen amplitudes are observed to be relatively small.

In order to study the effect of the chosen imperfections

0.54

m=0

hel

0.5

m=1,i=0.001

m=0,i=0.001

0.52

0.49

m=20,i=0.001

m=1,i=0.001

m=2,i=0.001

0.47

m=5,i=0.001

m=20,i=0.001

0.46

m=20,i=0.001

m=1,i=0.001

0.5

Wire lay angle, (rad)

0.48

m=5,i=0.001

m=2,i=0.001

m=5,i=0.001

m=2,i=0.001

m=1,i=0.001

m=2,i=0.001

m=5,i=0.001

m=20,i=0.001

0.48

Primary path

0.46

A

0.45

0.44

0.44

3

4

Wire arclenght, s(m)

0.42

3

4

Wire arclength, s(m)

L = 0.0001, R = 15m (prior to softening), all imperfections measured in m1

L = 0.0005, R = 15m (nonlinear

force-deformation regime), all imperfections measured in m1

on the wire geometry, the wire lay angle will be considered, see figure 12 and 13. It is interesting to note,

that switching the sign of the imperfection amplitudes

causes the changes in wire lay angle to localize opposite, see the curves denoted C + and C corresponding

respectively to positive and negative amplitudes.

is different for different values of . The compressive wire force in pure bending ( L

L = 0) can

be observed to increase, when the curvature is increased. Hence, the pipe curvature does not effect the load in the nonlinear regime after softening

has occurred, but is again detected as a bendingcompression coupling.

3. A and B in equation 13 and 14 representing where

in the pipe crosssection the wire starts and ends has

little detectable influence on the calculated wire

equilibrium paths.

A parameter study can now be conducted in order

to determine, how the model parameters influence the

maximum compressive load, which can be carried by

the wire by considering the equilibrium paths. All necessary analyses will be conducted for a perfect and an

imperfect wire structure with m = 20 (20 terms) and all

corresponding amplitudes set to -0.001. Results are presented in figure 15-16. The following conclusions can

be drawn:

1. The pitch length proportional to the initial wire lay

angle has major influence on the equilibrium path,

see Figure 15 in which L pitch is set to 1.0, 1.25 and

1.5 m. The point, at which softening occurs, is detected at higher compressive loads when the pitch

length is decreased. A wire with a short pitch has

therefore a larger load carrying ability than a wire

with a large pitch.

2. The pipe curvature, = R1 , seems to have little

influence on the point at which softening occurs,

see Figure 16, in which imperfect structures with

set to 0, 20m1 , 15m1 and10m1 are compared.

However, the behavior detected prior to softening

pitches being modeled effects the wire deformation

modes and the maximum compressive load, which can

be carried by the wire. Analyzing an imperfect wire

(m = 20, 1...20 = 0.001m1 ) with geometry as specified above, but varying the number of pitches included

in the model, the equilibrium paths shown in figure 18

are obtained. The number of pitches included in the

model are set to 1, 2, 4, 6, 8 and 10. The corresponding wire lay angles are presented in figure 19. It can be

observed that the number of modeled pitches and localization of variation in lay angle remains approximately

equivalent until the model is shortened sufficiently for

influencing the mode of deformation. This becomes the

case when less than four wire pitches are included in the

model, which can be observed to cause different modes

of deformation. A higher number of pitches can be observed all to deform in a similar manner in one end of

the wire. The remaining path of the wire exhibits a harmonic pattern in the remaining part of the solution array.

R=15 m

1400

0.2

Longitudinal compressive wire force, Pu (N)

A

0.15

0.1

0.05

0.05

1200

1000

A

B

800

600

400

Lpitch=1.0 m

200

0.1

Lpitch=1.25 m

Lpitch=1.5 m

0.15

0

3

4

Wire arclength, s(m)

Figure 14: Change of normal wire curvature n with respect to the initial

1

, the magnitude of the initial

helical state, state of pure bending to = 15m

helical normal curvature is 2.02m1

0.5

1.5

2

2.5

3

3.5

Compressive pipe strain

4.5

5

4

x 10

Figure 15: ( L

L , Pu )-wire equilibrium paths for varied pitch lengths, A:

perfect wire geometry, B: imperfect wire geometry, state of bending to

1

= 15m

and compression

R=15

1100

Longitudinal compressive wire force, Pu (N)

1200

1000

800

600

400

200

=0

=(20 m)

800

700

600

A=B=0

500

A=B=/2

A=B=

A=B=3/2

300

0

=(10 m)

Figure 16:

vature

=(15 m)1

1

200

0

A

900

400

1000

0.1

0.2

( L

L , Pu )-wire

0.3

0.4

0.5

0.6

0.7

Compressive pipe strain

0.8

0.9

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Compressive pipe strain

0.8

0.9

1

3

x 10

x 10

4. Conclusions

Based on a mathematical formulation of the mechanical behavior of a long and slender beam embedded in

a frictionless cylindrical surface bent into a toroid, the

equilibrium state of a single armouring wire within the

wall of a flexible pipe can be determined. The model

allows curvature components and deformations to be

large, but is in the present approach restricted to the

small strain regime. On this basis, the behavior of a

tensile armour wire in compression and bending has

Figure 17: ( L

L , Pu )-wire equilibrium paths for varied pipe bending radii

(for the sake of simplicity only imperfect wire structures are considered),

A: perfect wire geometry, B: imperfect wire geometry

been examined, including both initially perfect and imperfect wire geometries. Significant nonlinear softening

has been observed in the calculated force-displacement

responses. It has been demonstrated, that when a certain

load level is exceeded, the uniqueness of the solution to

the equilibrium state of an initially perfect wire geometry is lost. The effect of various imperfections was found

to have noticeable effects on the obtained solution.

The effect of key model parameters was examined. The

input parameter which had the largest influence on the

obtained solution, was the pitch length of the wire, since

10

0.55

2500

0.5

L=2.5 m (two pitches)

L=3.75 m (three pitches)

L=5 m (four pitches)

L=7.5 m (six pitches)

L=10 m (eight pitches)

L=12.5 m (ten pitches)

2000

1500

3000

1000

0.45

L=2.5 m (two pitches)

L=3.75 m (three pitches)

L=5 m (four pitches)

L=7.5 m (six pitches)

L=10 m (eight pitches)

L=12.5 m (ten pitches)

0.4

0.35

500

0

0.2

0.4

0.6

0.8

1

1.2

1.4

Compressive pipe strain

1.6

1.8

x 10

Figure 18: ( L

L , Pu )-wire equilibrium paths detected when the number of

modeled pitches is varied, imperfect wire geometry

the level of load at which softening occurred in the force

responses. Furthermore, it was found, that modeling

less than four pitches of wire, the mode of deformation found in the nonlinear geometrical regime differed

radically from the mode of deformation obtained when

modeling more pitches of wire.

In the past few years, the occurrence of armour

wire buckling in the circumferential pipe direction has

demonstrated a need for methods, which may serve as

basis when analyzing the compressive behavior of tensile armour layers in flexible pipes. The present work

may constitute a valuable contribution in the field of

wire mechanics, since it enables the determination of

the wire equilibrium state in compression and bending

when frictional effects are neglected due to a damaged

outer sheath. The presented method may constitute a

valuable basis for derivation of methods for simulation

of the effect of cyclic bending if methods for inclusion

of frictional effects are developed. In order to classify

the buckling phenomenon and investigate the possible

existence of buckling by bifurcation, further research is

needed.

[1] Braga, M.P. and Kaleff, P. (2004): Flexible Pipe Sensitivity to

Birdcaging and Armor Wire Lateral Buckling. Proceedings of

OMAE, 2004

[2] Bectarte,F. and Coutarel,A. (2004): Instability of Tensile Armour Layers of Flexible Pipes under External Pressure. Proceedings of OMAE, 2004

[3] stergaard, N.H., Lyckegaard, A. and Andreasen, J. (2011): On

lateral buckling failure of armour wires in flexible pipes. Proceedings of OMAE, 2011

[4] Tan, Z., Loper, C., Sheldrake, T. and Karabelas, G. (2004): Behavior of Tensile Wires in Unbonded Flexible Pipe under Com-

6

8

Wire arclength, s (m)

10

12

14

Figure 19: Wire lay angles, , detected when the number of modeled

pitches is varied, imperfect wire geometry

[5]

[6]

[7]

[8]

[9]

[10]

[11]

[12]

[13]

[14]

[15]

[16]

[17]

of OMAE, 2006

Brack. A., Troina, L.M.B. and Sousa, J.R.M. (2005): Flexible

Riser Resistance Against Combined Axial Compression, Bending and Torsion in Ultra-Deep Water Depths. Proceedings of

OMAE, 2005

Vaz, M.A. and Rizzo, N.A.S. (2011): A finite element model for

flexible pipe armor wire instability Proceedings of OMAE, 2005

Love, A.E.H. (1944): A treatise on the Mathematical Theory of

Elasticity. Dover Publications, Inc., N.Y

Reissner, E. (1981): On finite deformations of space-curved

beams. Journal of Applied Mathematics and Physics (ZAMP),

Vol. 32

Costello, G.A. (1977): Large deflections of helical spring due to

bending. J. Eng. Mech. Div. Proc. ASCE,vol. 103, 1124-33

Spillers, W. R., Eich, E.D., Greenwood, A.N. and Easton, R.

(1983): A helical tape on cylinder subjected to bending. J. Eng.

Mech. Div. Proc. ASCE,109 (1983), 1124-33

Stump, D.M and van der Heijden, G.H.M. (2000): Matched

asymptotic expansions for bent and twisted rods: applications

fro cable and pipeline laying Journal of Engineering Mechanics,

38(1), 13 31, 07 2000

Leroy, J.M. and Estrier, P. (2001): Calculation of stresses and

slips in helical layers of dynamically bent flexible pipes. Oil and

Gas Science and Technology, REV. IFP, Vol. 56, No. 6, pp. 545554, 2001

do Carmo, M P. (2008): Differential Geometry of Curves and

Surfaces.

Witz, J.A. and Tan, Z. (1992): On the Flexural Structural Behaviour of Flexible Pipes, Umbillicals and Marine Cables. Marine Structures, Vol. 5

Kraincanic I. and Kabadze E.. (2001): Slip initiation and progression in helical armouring layers of unbonded flexible pipes

and its effect on pipe bending behavior. Journal of Strain Analysis, Vol. 36, No. 3

Svik, S. (1993): A finite element model for predicting stresses

and slip in flexible pipe armouring tendons. Computers and

Structures. Vol. 46, No.2

Out, J. M. M. and von Morgen, B. J. (1997): Slippage of helical

reinforcing on a bent cylinder. Engineering Structures, Vol. 19,

No. 6, pp. 507-515

[18] stergaard, N.H., Lyckegaard, A. and Andreasen, J. (2011): A

method for prediction of the equilibrium state of a long and slender wire on a frictionless toroid applied for analysis of flexible

pipe structures. Engineering Structures, 2011

[19] Brush, D.O. and Almroth, B.O. (1975): Buckling of bars, plates

and shells Mcgraw-Hill

[20] Bazant, Z.P. and Cedolin, L. (2010): Stability of structures

World Scientific

[21] Tvergaard, V. and Needleman, A. (1981): On Localized Thermal Track Buckling. Int. J. Mech. Sci. Vol. 23, No. 10, pp. 577587

[22] Taylor, N. and Gan A.B. (1986): Submarine Pipeline Buckling

- Imperfection Studies. Thin-Walled Structures 4 (1986) p. 295323

[23] Feret, J.J. and Bournazel, C.L. (1987): Calculation of stresses

and slips in structural layers of unbonded flexible pipes. Journal

of Offshore Mechanics and Arctic Engineering, Vol. 109

[24] CAFLEX Theory manual. IFP/SINTEF, 1991

11

Marine Structures

Marine Structures 1 (2012) 112

Niels Hjen stergaard

NKT-Flexibles / Aalborg University, Department of Mechanical and Production Engineering, Denmark

E-mail: Niels.HojenOstergaard@nktflexibles.com

Anders Lyckegaard

NKT-Flexibles, Priorparken 480, Brndby, Denmark

E-mail: Anders.Lyckegaard@nktflexibles.com

Jens H. Andreasen

Department of Mechanical and Production Engineering, Aalborg University, Denmark

E-mail: jha@m-tech.aau.dk

Abstract

In the present paper, a mathematical model which is capable of representing the physics of lateral buckling failure

in the tensile armour layers of flexible pipes is introduced. Flexible pipes are unbounded composite steel-polymer

structures, which are known to be prone to lateral wire buckling when exposed to repeated bending cycles and longitudinal compression, which mainly occurs during pipe laying in ultra-deep waters. On basis of multiple single wire

analyses, the mechanical behavior of both layers of tensile armour wires can be determined. Since failure in one layer

destabilizes the torsional equilibrium which is usually maintained between the layers, lateral wire buckling is often

associated with a severe pipe twist. This behavior is discussed and modeled. Results are compared to a pipe model,

in which failure is assumed not to cause twist. The buckling modes of the tensile armour wires can be obtained by the

presented method.

1. Introduction

Unbounded Flexible pipe structures are widely used

as risers or flowlines for development of subsea oil

or gas reservoirs in deep waters. In order to obtain a

structural design capable of resisting large curvature,

internal and external pressure and longitudinal loads,

flexible pipes are constructed as composite structures

comprised by a number of layers with different mechanical properties, see Figure 1. The pipe bore denoted the

carcass is constructed in such a manner that it retains

flexibility while capable of resisting pressure. On the

outside of this layer, a helically wound pressure armour

comprised by steel profiles is applied. Two polymer

liners constitute fluid barriers which ensure that the

pipe is tight. Furthermore, two layers of helically

axial loads. These tensile armour wires are known to be

prone to instability in longitudinal compression.

Two types of wire buckling failure modes have been

observed in the field, see [1]:

1. Birdcaging, by which the armouring wires become

radially unstable due to compressive loads, see

Figure 3. The birdcaging failure mode is in most

known pipedesigns prevented by a layer of high

strength tape wound around the wires which restricts the radial deformations.

2. Lateral buckling, by which the wires become transversely unstable below the high-strength tape, see

Figure 2

The lateral buckling failure mode is from a design perspective of a more complex nature than buckling by

birdcaging. Both failure modes can be observed as localized wire buckling, but while birdcaging can be detected by visual inspection without dissection of the

failed pipe, lateral buckling is significantly harder to detect without precise measurements.

Lateral buckling of tensile armour wires of flexible

pipes are known mainly to occur during pipe laying in

deep-waters, see Figure 4.T1. In this scenario, the flexible pipe is in a free-hanging position from an installation vessel to the seabed. Furthermore, the pipe is often

empty during installation. The conditions, which are

known to cause lateral buckling in a flexible pipe are

1. Axial compression due to hydrostatic pressure on

the end cap of an empty pipe

2. Repeated bending cycles due to waves, current and

vessel movements

3. Often breached outer sheath, so the pipe annulus is

flooded, so hydrostatic pressure does not introduce

contact stresses which limits wire slip

The wire buckling failure modes were described by

Braga and Kaleff, [2]. A mechanical test principle for

reconstruction of the failure mode in the laboratory was

described. However, it seems to be a widely accepted

fact that results obtained experimentally in the laboratory by use of mechanical test rigs do not correspond

to observations made in the field, since failure occurs at

lower load levels in mechanical test rigs. The test principle was developed further by use of hyperbaric chambers by Bectarte and Coutarel, [1], Bectarte et. al.,[3],

Lateral buckling of armour wires is presently understood as follows:

1. Due to the manner by which a flexible pipe is designed to be torsionally stable, the number of wires

in the outer layer is larger than in the inner layer.

The complex interaction between the layers causes

compressive loads to be larger in the wires of the

inner layer than in the wires contained in the outer

layer of tensile armour.

2. As cyclic bending is applied, the frictional restraining against wire movement vanishes and the wires

slip transversely from a state at which the wire lay

angles are constant in bending (a bent helix) towards an equilibrium state. In compression this

state may for the inner layer of armour wires correspond to a decreasing ability to carry load compared to the load carrying ability occuring in the

outer layer.

3. This causes torsion and longitudinal loads to couple and the flexible pipe will experience a twist,

which may be very severe.

4. This twist causes further compressive straining of

the armouring wires in the inner layer, which may

lead to plasticity, large gaps or more severe modes

of buckling. This may cause deformations within

the pipe wall to an extend at which the flexible

pipe will not function in service in a satisfactory

manner.

The lateral buckling failure mode has been studied by

finite element analysis, see [5] and [16]. However, a

fulfilling mathematical investigation of the underlying

physics has not yet been published.

The mechanics and interaction of both isotropic and

helically wound layers in a straight flexible pipe

subjected to axisymmetric loads were investigated by

Feret and Bournazel, [7], leading to a linear system

of equation for prediction of radial and longitudinal

pipe strain and pipe twist. This method was implemented in the computer program CAFLEX, [8], and

has proven capable of providing accurate predictions

of the response of a given flexible pipe subjected to

known axisymmetric loads. Furthermore, the effect

of cyclic bending, which leads to hysteresis in the

global constitutive moment-curvature relation, was

investigated. Numerous examples of refinement of

the theoretical description of the global constitutive

behavior of flexible pipes have been conducted, see

[9]-[12].

The effect of bending was investigated further by Witz

Figure 2: Localized lateral deformations in the inner layer of tensile armour wires, generated by NKT-Flexibles by laboratory testing

NKT-Flexibles by laboratory testing

Figure 4: Flexible pipe configurations, T.1: Pipe laying scenario, T.2: Idealized scenario, used during laboratory testing, see [2],[1] and [19]

basis of armour wire mechanics with the assumption

that the wire lay angle remains constant in bending.

Svik, [15], used a finite element approach to predict

slippage along a curve with constant lay angle. Further

research was conducted by Leroy and Estrier, [16],

taking frictional effects and transverse wire slippage in

cyclic bending into calculation on basis of a prescribed

solution form based on experience with geodesic curves

and bent helices. While the approaches summarized

above are based on assumptions, which can be justified

for a flexible pipe subjected to bending and tension, this

is not the case in bending and compression, since the

prescribed wire geometries are incapable of predicting

the geometrical wire configurations detected, when

lateral buckling occurs due to either neglected or

prescribed transverse wire slippage. stergaard et.

al. derived a system of equations for prediction of

the wire equilibrium state if friction is neglected, see

[17], and demonstrated how this method could serve

as basis for buckling analysis of flexible pipe armour

wires, see [18]. The proposed method was based on

the equilibrium of a long and slender beam embedded

longitudinal straining and twist. The analysis of this

type of frictionless structure is based on the assumption

that the wire when loaded reaches an equilibrium state

which is equal to the state obtained after a significant

number of bending cycles when friction is present.

As buckling occurred in an analyzed wire, the (longitudinal force-pipe strain)-relation exhibited nonlinear

geometrical softening in such a manner, that the

maximum compressive longitudinal load, which can be

carried by a single wire, could be estimated. Furthermore, a preliminar implementation of a global model

for modeling of lateral buckling failure was described

in [19] based on a simplified version of the method

for single wire analysis. In the present paper, it will

therefore be described how lateral buckling failure in a

flexible pipe can be modeled on basis of multiple single

wire analyses coupled in torsion in one single point.

This enables a large large number of local single wire

models to be coupled to the global deformations of a

flexible pipe.

2. Theory

In order to construct a global model of a flexible pipe,

the means which are needed for modelling of a flexible

pipe will initially be introducted. Afterwards, it will be

desribed how multible single wire analyses form a basis

for a global pipe model.

2.1. Wire equilibrium state with transverse slip

Initially, a brief summary of the method derived for

single wire analysis in [17] is given. A single tensile

armour wire within the wall of a flexible pipe is modeled. Assuming the pipe curvature = R1 constant, the

wire can be assumed to constitute a curve on a torus surface with minor radius r, see Figure 5, parameterized

by the coordinates u and . A point on the torus surface with specified (u, )-coordinates has the cartesian

coordinates

1

+ r cos cos (u) 1

1

(1)

x(u, ) =

+ r cos sin (u)

r sin

constituted by a tangent t, a normal n and a binormal

vector b is given by

t=

d

ds

n=

xu x

kxu x k

b=tn

(2)

of the surface equation with respect to the torus coordinates. These vectors span the tangent plane of the

toroid. s denotes the wire arclength. It can be observed,

that the normal vector of the wire is chosen equal to the

surface normal, since the wire is modeled as embedded

in a toroid.

The differential equations governing the wire geometry

in the tangent plane are given by

du cos

=

ds

kxu k

d cos

=

ds

kx k

(3)

geodesic component g and a normal component n .

cos

1

cos2 sin2

1 + r cos

r!

d

sin

cos +

1 + r cos

ds

(4)

(5)

(6)

!

1

cos

cos sin

(7)

=

1 + r cos r

on vectorial form were given by Reissner, [20]

dP

+p=0

ds

dM

+tP+m=0

ds

(8)

wire moment denoted M, distributed external load denoted p and distributed external moment denoted m.

The equations of equilibrium may be written with the

(tnb)-frame as basis

dPt

n Pn + g Pb + pt

ds

dPn

+ n Pt Pb + pn

ds

dPb

g Pt + Pn + pb

ds

dMt

n Mn + g Mb + mt

ds

=0

(9)

=0

(10)

=0

(11)

=0

(12)

dMn

+ n Mt Mb Pb + mn = 0

ds

dMb

g Mt + Mn + Pn + mb = 0

ds

(13)

(14)

Assuming the wire dimensions small with respect to minor and major torus radii, it is fair to neglect curved

beam terms and assume the constitutive relations linear

Pt

Mt

Mb

Mn

= EA

= GJ = GJ( 0 )

= EIb n = EIb (n n,0 )

= EIn g = EIn (g g,0 )

(15)

in which denotes tangential wire strain. Now assembling the unknowns, the following system of equations

is derived

du

ds

d

ds

d

ds

dPt

ds

dPb

ds

dMn

ds

cos

1 + r cos

sin

=

r

sin

=

cos + g

1 + r cos

=

(16)

(17)

(18)

= n Pn g Pb

(19)

= g Pt Pn

(20)

= n Mt + Mb + Pb

(21)

in terms of known quantities after a solution is obtained.

However, since the arclength s is not equal to s0 in the

helical, due to wire strains, these will be assumed sufficiently small to calculate using Cauchys definition of

strain. Since s is unknown prior to the analysis, but ds0

is easily calculated by well-known relations for helices,

ds is rewritten in terms of ds0 and . A solution can be

obtained using a commercially available solver (bvp4c

in MATLAB). Boundary conditions corresponding to

flexible pipe end fittings will be applied in s = 0 and

s = S L in which S L denotes the total wire arclength and

L the total pipe length.

A

u(0) = 0 (0) = ini

(0) = hel

u(S L ) = L 1 +

L

L

L

L

(S L ) = hel

(S L ) = ini

(22)

(23)

directions are given by the transformation

Pu

P

=

=

=

=

Pt t tu + Pb b tu

Pt cos + Pb sin

Pt t t Pb b t

Pt sin Pb cos

(24)

(25)

is based on the assumption, that assuming that no friction acts on the wire has the effect, that the equilibrium

state determined equals the state, which slips will occur

towards in cyclic loading if friction is included and will

be obtained after a significant number of load cycles.

2.2. Wire equilibrium state, slip along a bent helix

The derived system of field equations can be modified so the equilibrium state of a wire can be determined

with the conventional assumption, that the wire lay angle remains constant. Slip in this model may only occur

tangentially along a bent helix. For a wire with constant

, the term d

ds vanishes. On this basis the system of field

equations can be reduced to

du

cos

=

(26)

ds

1 + r cos

d

sin

=

(27)

ds

r

d

= 0

(28)

ds

dPt

= n Pn g Pb

(29)

ds

The corresponding boundary conditions are

A

u(0) = 0 (0) = ini

!

L

B

(S L ) = ini

u(S L ) = L 1 +

L

L

(30)

(31)

All observations made during test execution have supported the assumption, that transverse wire slip is very

limited in the outer layer of tensile armour, see [2] and

[19]. This simplified system of equations may therefore

be used in order to model the mechanical behavior of

the outer layer of armour wires.

A layer of armouring wires has this far been assumed

radially inelastic. In order to take effects of a radially

elastic pipe wall into calculation, the minor torus radius

will be set as function of the applied compressive strain.

The following relation specifying the deformed radius

rd was derived in [18] and [19].

!

!

ka L

r

r = 1

r

(32)

rd = 1 +

r

kr L

in which ka and kr denotes respectively an axial and an

radial linear pipe stiffness.

In this section, it will be described how a global pipe

model can be constructed on basis of multiple single

wire analyses. The local single wire models described

in section 2.1 and 2.2 will be used to model all wires

within the pipe wall seperately. Compability between

all wires in both layers must be ensured by determining

the equilibrium of all wires in a configuration in which

the entire pipe is in equilibrium when subjected to gen

eralized loads, , L

L and L . This is done by considering

the equilibrium in the end points of the pipe in which all

wires are subjected to equal deformations, since these

are mounted in pipe end fittings, corresponding to presribed wire rotation and displacement. Two global set of

flexible pipe boundary conditions will be studied:

1. A flexible pipe of finite length L which bent to a

constant radius of curvature is fixed against twist

in S = 0 and S = L, see Figure 7. In this model

of the flexible pipe the two layers of armour wires

are torsionally uncoupled. Compability between

the pipe layers in point A and B is in this scenario

ensured by prescribed boundary conditions on all

wires corresponding to

L = 0.

2. A flexible pipe of finite length L which bent to a

constant radius of curvature is fixed against twist

in S = 0 and free to twist in S = L, see Figure 8.

Failure by lateral buckling in one layer of armouring wires will with these boundary conditions lead

to a pipe twist of the torsionally free end, B. Hence,

compressive pipe loads are in this model torsionally coupled to the pipe torsion. In the following,

the global torsional pipe equilibrium will only be

demanded fulfilled in the loaded end of the pipe,

S = L, which is free in torsion.

radius of curvature

radius of curvature

summing up the moment contributions from all wires

!

L

T

,

=

L L

nX

nX

wires

sheets

i

(Mui Pi ri ) +

Mu,sheets

=

i=1

(33)

i=1

nX

wires

i=1

nX

sheets

Gi J i

i=1

must be fulfilled in S = L. In order to establish torsional

equilibrium for a prescribed pipe strain L

L , a pipe twist

performed by Newton-Raphton iterations on all wires in

the flexible pipe. The total longitudinal load in the pipe

structure can be calculated in an equivalent manner by

summing up the force contributions from all wires

Pa

=

=

nX

wires

i=1

nX

wires

i=1

Piu +

nX

sheets

Piu,sheets

(34)

i=1

nX

sheets

i=1

E i Ai

L

L

torsion can now be obtained and compared.

In the present approach, transverse wire contact is neglected. The validity of this assumption may be tested

after a solution for all wires has been obtained.

The equilibrium path detected in a (Longitudinal

load, pipe strain)-diagram of a single armour wire with

a design as shown on Figure 10 is sensitive to small initial imperfections, see Figure 9. This is shown in [18].

Significant sorftening behavior of the equilibrium path

can be observed to occur. The imperfections were added

by assuming, that the wire in the initial state possessed

a small geodesic curvature component given by

m

is

Mn X

+

i sin

(35)

g =

EIn i=1

L

It was concluded that the number of terms taken into

calculation in the sine series m had virtually no influence on the obtained result. However, it was detected,

that the chosen signs of 1...m while having very little

influence on the equilibrium path, caused the buckling

phenomena to localize opposite.

3. Results

3.1. Flexible pipe response in bending and compression

In this section, the equilibrium state of armouring layers in flexible pipes subjected to compression and bending will be analyzed. The two sets of global boundary conditions shown in Figure 7, torsionally fixed-fixed

pipe, and Figure 8, torsionally fixed-free pipe, will be

analyzed and compared.

In the following, a flexible pipe with an armour layer

design given in Table 1 is analyzed. The length of the

pipe being modelled will be set to five times the inner

layer pitch length.

In the presented example, the force- and moment contributions from the polymeric sheaths are neglected for

the sake of simplicity, since these are made of a polymer

material with insignificant stiffness contributions. However, it is important to note that this may not always be

the case when analyzing flexible pipes.

The layers of armouring wires will initially be assumed radially stiff. In order to compare the mechanical behavior of the armouring layers in the non-linear

regime, in which buckling may occur, with the linear

compressive behavior, which is usually predicted, the

equations contained in [7] for analysis of straight flexible pipes subjected to axisymmetric loads are applied.

Initially, only the inner layer of armouring wires will be

analyzed. An entire torsionally fixed-fixed layer of armouring wires, which is free to slip transversely as described in section 2.1 will be modeled. The equilibrium

path of the layer can be calculated by applying compression in steps to all wires in the layer and calculating the

longitudinal load by Equation 34, see Figure 9. It is initially noted that the layer of wires exhibits a behavior

equivalent to what was observed for a single armouring wire, see Figure 9. Hence, the buckling analysis

performed for a single armouring wire can now be performed for an entire layer.

By equivalent methods the Equilibrium paths of a torsionally fixed-fixed pipe can be determined, see Figure

12. The inner layer of armouring wires, which for a

pipe subjected to longitudinal compression is prone to

lateral buckling, will be modeled by means described

in section 2.1, so transverse slip is allowed. The outer

layer of armouring wires will be described as described

in section 2.2. Hence, the wire lay angle remains constant when the pipe is loaded. The force in the inner

layer can be observed to exhibit significant softening

behavior, while the force in the outer layer can be observed to increase linearly, since this layer can not slip

transversely and thereby exhibit buckling behavior.

Now considering the pipe torsionally fixed-free with

the inner and outer layer of armour wires modeled in a

manner equivalent to what was applied during analysis

of a torsionally fixed-fixed pipe, the forces in the pipe

layers are shown on Figure 13. The computational time

necessary to perform this analysis is significantly larger

than for a fixed-fixed pipe, since the torsional equilibrium must be fullfilled by solving equation 33 iteratively

for all wires. The consequence of the softening behavior is, that applying compressive loads at levels larger

than the maximum value determined by the analysis,

will lead to transverse wire buckling in the inner layer

of armouring wires, which causes a severe pipe twist,

see Figure 16.

The effect of a radially elastic pipe wall can now be

estimated on basis of equation 32. By methods established on basis of equations derived in [8], the ratio between linear longitudinal and radial stiffness kkar is calculated to 2.52 which corresponds well to what has been

measured during a compression test of a pipe with the

given design. The test setup was described in [19]. Considering the equilibrium paths shown on Figure 15, the

radial elasticity taken into calculation can be observed

to have major influence on the longitudinal pipe stiffness, which prior to buckling corresponds well to the expected linear behavior. However, it can be observed that

the compressive load, which can be carried by the pipe

structure, only to a very low extend is effected by radial

deformations. The corresponding pipe twist is shown

on Figure 16. The behavior of the armouring layers in a

twist-longitudinal force diagram can be observed not to

be effected by radial elasticity.

In Figure 14 the equilibrium path for a radially stiff

1500

1000

500

perfect wire geometry, primary path

0

0.5

1.5

2

2.5

3

3.5

Longitudinal compressive pipe strain

4.5

5

4

x 10

Figure 9: ( L

L , Pu )-equilibrium paths of analyzed armour wires, see Figure 10 (pipe modeled torsionally fixed-fixed), all imperfections measured

in m1

Figure 10: Model of armour wire within the wall of a flexible pipe subjected to bending and longitudinal loads

Outer diameter(m)

L pitch (m)

Size (mm)

Number of windings

Inner layer

0.2012

1.263

3 10

52

Outer layer

0.209

1.318

3 10

54

High-strength tape

0.2117

0.075

1 60

1

Wire steel

High-strength tape

Youngs Modulus

210 GPa

27 GPa

torsionally fixed-free pipe in which transverse slip is allowed in both layers is presented. It can be concluded,

that the maximum load, which can be carried by the

structure, only to a very low extend is influenced by that

transverse slip is allowed in both layers of armouring

wires. However, the longitudinal stiffness can be concluded to be effected, since the strains are larger than in

results obtained with a transversely fixed outer layer, see

Figure 13. Since this is the case, it is desirable to model

Poissons ratio

0.3

0.3

free outer layer more than doubles the computational

time.

3.2. Buckling modes

The obtained buckling modes in the inner layer of armouring wires are presented on Figure 17. Mode A and

B are obtained for a torsionally fixed-fixed model, while

Mode C and D are obtained for a torsionally fixed-free

200

inner layer

outer layer

sum of layers

180

50

Longitudinal compressive force (kN)

60

40

30

20

imperfect wire geometry, 1...20=0.001

10

160

140

120

100

80

60

40

20

0

0.5

1.5

2

2.5

3

3.5

Longitudinal compressive pipe strain

4.5

Figure 11: ( L

L , Pa )-equilibrium paths of inner layer of armouring wires,

analysis comprized of mulitiple single wire analyses, torsionally fixedfixed boundary conditions on layer, all imperfections measured in m1

110

100

100

90

90

80

80

70

60

50

40

30

20

inner layer

outer layer

sum of layers

10

0

0.5

1.5

2

2.5

3

3.5

Longitudinal compressive pipe strain

0.5

1.5

2

2.5

3

3.5

Longitudinal compressive pipe strain

50

40

30

20

inner layer

outer layer

sum of layers

x 10

amplitudes equal to 0.001 and mode B and D are found

for all amplitudes equal to -0.001. Localization of wire

gaps can be observed in opposite ends of the modeled

layer. The assumption stating that no transverse wire

contact occurs is in all cases valid.

A closer examination of buckling modes is presented

on Figure 18. Mode E is a closer examination of the

gaps shown in Figure 17. Mode F and G corresponds

to severely buckled states obtained respectively for a

torsionally fixed-fixed and a fixed-free model for lonL

gitudinal pipe strains L

L = 0.01 and L = 0.0225.

While assuming that no transverse wire contact occurs

remains fair for mode E, the assumption is violated for

mode F and G. The deformation modes correspond well

5

4

x 10

60

Figure 13: ( L

L , Pa )-equilibrium paths of analyzed flexible pipe (torsionally fixed-free, see Figure 8),transverse slip taken into calculation in inner

layer, neglected in outer layer

4.5

70

10

4.5

Figure 12: ( L

L , Pa )-equilibrium paths of analyzed flexible pipe (torsionally fixed-fixed, see Figure 7), transverse slip taken into calculation in

inner layer, neglected in outer layer

5

4

x 10

0.5

1.5

2

2.5

3

3.5

Longitudinal compressive pipe strain

4.5

5

4

x 10

Figure 14: ( L

L , Pa )-equilibrium paths of flexible pipe (torsionally fixedfree, see Figure 8), transverse slip is taken into calculation in both layers

of armouring wires

4. Discussion

The main contribution to wire mechanics contained

in this paper is constituted by the global pipe model,

which torsionally fixed-free exhibits a behavior similar

to what can be observed during laboratory experiments,

see [2], [3] and [19].

The presented results clearly show, that the compressive

load, which can be carried by a flexible pipe structure to

a very high extend is dependent on the boundary conditions applied in the global model. However, it can

also be observed, that the significant higher load, which

can be carried by a torsionally fixed-fixed pipe, is com-

10

150

110

100

100

50

radially elastic structure, torsionally fixedfixed

offset linear compressive response

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Longitudinal compressive pipe strain

0.8

0.9

80

70

60

radially stiff

50

radially stiff

imperfect wire geometry, 1...20=0.001,

radially elastic

40

1

x 10

Figure 15: ( L

L , Pa )-equilibrium paths of flexible pipe, radially stiff and

radially elastic structures

A

90

2

Pipe twist, /L

5

3

x 10

Figure 16: Pipe twist of torsionally free pipe end, all imperfections measured in m1 , radially stiff and radially elastic structures

C

L =-0.001, R = 11m, in inner layer of armouring wires, A: torsionally fixed-fixed pipe, 1...20 =0.001, B:

torsionally fixed-fixed pipe, 1...20 =-0.001, C: torsionally fixed-free pipe, 1...20 =0.001, D: torsionally fixed-free pipe, 1...20 =-0.001

layer of armour wires, which in the presented approach

is locked. The forces in the inner layer of armouring

wires exhibits buckling behavior at exactly the same

load in both models. While the torsionally fixed-fixed

model obviously will generate a non-conservative estimate of the maximum compressive load causing wire

buckling failure in the pipe structure during pipe instal-

is conservative. A preliminar comparison of experimental and modeling results contained in [19] showed that

this may be the case. In order to clarify this issue, further research is needed.

The presented methods can only be used to calculate the

maximum compressive load in endurance-like scenarios. Hence, flexible pipes can only be designed in such

11

Figure 18: Closer examination of buckling modes, R = 11m, E: Localized buckling detected as large wire gaps, F: Periodic deformation mode

L

found at large pipe strain ( L

L = 0.01) for torsionally fixed-fixed pipe, G: Periodic deformation mode found at large pipe strain ( L = 0.025) for

torsionally fixed-free pipe

to be determined, if the approach adds too much conservatism to the design process or not.

Obviously, the buckling modes despite localizing in

a manner corresponding to the physics of a flexible

pipe, may be influenced by friction. Further research is

needed in order to determine, if the presented approach

is capable of representing this behavior or must be extended to include frictional effects in a more stringent

manner. This, however, is from a computational point of

view highly undesirable, since inclusion of friction and

simulation of stick-slip effects on all wires couple to the

global constitutive relation of a flexible pipe. The constitutive relations of flexible pipes are still subject of research, see among others [9]-[12], and proposing a constitutive law of a flexible pipe based on the true physical

state of all armouring wires in bending and compression

may therefore constitute a valuable basis for a simplified approach in order estimate how physical phenomena, which are not included in the presented research,

effects the limit compressive load.

5. Conclusions

Based on a mathematical model for determination of

the wire limit equilibrium state within the wall of a flexible pipe, a global model of a flexible pipe subjected

to bending to a constant radius of curvature and compressive longitudinal loads has been proposed on basis of multiple single wire analyses. On this basis, the

limit compressive load of a pipe structure, in which the

pipe interior is assumed flooded, can be calculated along

with deformation modes, which correspond well to preliminar experimental results. The wire models have

proven sensitive to initial imperfections, which cause

the gaps in the inner layer of armouring wires to localize

when lateral buckling occurs. The presented approach,

in which a flexible pipe is modeled torsionally fixedfree, may however on the shown form serve as basis

of conservative estimates of the maximum compressive

load, which can be carried by the pipe structure. Experimental validation and assessment of how friction effects

the triggered buckling modes have not been presented in

this paper. Further research is needed in order to draw

final conclusions regarding how design rules and methods for flexible pipe design can be formulated on basis

of the proposed model, due to the conservatism included

in the present approach.

[16]

[17]

[18]

[19]

[20]

Structures. Vol. 46, No.2

Leroy, J.M. and Estrier, P. (2001): Calculation of stresses and

slips in helical layers of dynamically bent flexible pipes. Oil and

Gas Science and Technology, REV. IFP, Vol. 56, No. 6, pp. 545554, 2001

stergaard, N.H., Lyckegaard, A. and Andreasen, J. (2011): A

method for prediction of the equilibrium state of a long and slender wire on a frictionless toroid applied for analysis of flexible

pipe structures. Submitted to Engineering Structures, April 11th ,

2011

stergaard, N.H., Lyckegaard, A. and Andreasen, J. (2011): Imperfection analysis of flexible pipe armour wires. Submitted to

Applied Ocean Research, July, 2011

stergaard, N.H., Lyckegaard, A. and Andreasen, J. (2011): On

lateral buckling failure of armour wires in flexible pipes. Proceedings of OMAE, 2011

Reissner, E. (1981): On finite deformations of space-curved

beams. Journal of Applied Mathematics and Physics (ZAMP),

Vol. 32

1211

[1] Bectarte,F. and Coutarel,A. (2004): Instability of Tensile Armour Layers of Flexible Pipes under External Pressure. Proceedings of OMAE, 2004

[2] Braga, M.P. and Kaleff, P. (2004): Flexible Pipe Sensitivity to

Birdcaging and Armor Wire Lateral Buckling. Proceedings of

OMAE, 2004

[3] Secher, P., Bectarte, F. and Felix-Henry, A. (2011): Lateral

Buckling of Armor Wires in Flexible Pipes: reaching 3000 m

Water Depth. Proceedings of OMAE, 2011

[4] Tan, Z., Loper, C., Sheldrake, T. and Karabelas, G. (2004): Behavior of Tensile Wires in Unbonded Flexible Pipe under Compression and Design Optimization for Prevention Proceedings

of OMAE, 2006

[5] Brack. A., Troina, L.M.B. and Sousa, J.R.M. (2005): Flexible

Riser Resistance Against Combined Axial Compression, Bending and Torsion in Ultra-Deep Water Depths. Proceedings of

OMAE, 2005

[6] Vaz, M.A. and Rizzo, N.A.S. (2011): A finite element model for

flexible pipe armor wire instability Proceedings of OMAE, 2005

[7] Feret, J.J. and Bournazel, C.L. (1987): Calculation of stresses

and slips in structural layers of unbonded flexible pipes. Journal

of Offshore Mechanics and Arctic Engineering, Vol. 109

[8] CAFLEX Theory manual. IFP/SINTEF, 1991

[9] McIver, D.B. (1995): A method of modeling the detailed component and overall structural behaviour of flexible pipe sections.

Engineering Structures, Vol. 17, No. 4, pp. 254-266

[10] Custodio, A.B. and Vaz, M.A. (2002): A nonlinear formulation

for the axisymmetric response of umbilical cables and flexible

pipes. Applied Ocean Research 24, 21-29

[11] Alfano, G., Bahtui, A. and Bahai, H. (2009): Numerical derivation of constitutive models for unbonded flexible risers. International Journal of Mechanical Sciences 51,295-304

[12] Vaz. M.A. and Patel, M.H. (2007): Post-buckling behaviour of

slender structures with a bi-linear bending moment-curvature relationship. International Journal of Non-linear Mechanics, Vol.

42, p.470-483

[13] Witz, J.A. and Tan, Z. (1992): On the Flexural Structural Behaviour of Flexible Pipes, Umbillicals and Marine Cables. Marine Structures, Vol. 5

[14] Kraincanic I. and Kabadze E.. (2001): Slip initiation and progression in helical armouring layers of unbonded flexible pipes

and its effect on pipe bending behavior. Journal of Strain Analysis, Vol. 36, No. 3

[15] Svik, S. (1993): A finite element model for predicting stresses

12

!""
#$%%
&

%$#

Vol. 44, No 3, 2011, pp. 243 259

the equilibrium state of flexible pipe armouring wires in

compression and bending

Niels Hjen stergaard, Anders Lyckegaard and Jens H. Andreasen

Summary. The motivation for the work presented in this paper is a specific failure mode

known as lateral wire buckling occurring in the tensile armour layers of unbounded flexible

pipes. Such structures are steel-polymer composites with a wide range of applications in the

offshore industry. The tensile armour layers are usually constituted by two layers of oppositely

wound steel wires. These may become laterally unstable when a flexible pipe is exposed to

repeated bending cycles and longitudinal compression.

In order to model the mechanical behavior of the armouring wires within the pipe wall,

a formulation based on the equilibrium of a curved beam embedded in an initially cylindrical

surface bent into a toriod is applied. In the present work, the response of a single armouring wire

subjected to compression and cyclic bending will be studied, in order to detect lateral buckling

of the wire. Frictional effects are included as distributed tangential and transverse loads based

on a simple regularized Coulomb model.

Key words: curved beam equilibrium, wire mechanics, friction, flexible pipes, lateral buckling of

armour wires

Introduction

Unbounded flexible pipes are steel-polymer composite structures with a wide range of

applications in the offshore industry. A flexible pipe structure is usually constituted

by numerous layers with different properties, see Figure 1. The pipe bore, denoted the

carcass, is constituted by helically wound profiles surrounded by a pressure armour. These

layers ensure the structural integrity against external and internal pressure. The pressure

armour is surrounded by a polymeric liner, which like the external pipe sheath, is a fluid

barrier layer. The space between liner and outer sheath is usually denoted the pipe

annulus. In the pipe annulus, the tensile armour layers are located, usually constituted

by two layers of oppositely wound steel wires. Usually, the total number of wires is

80 150. These layers ensure the structural integrity against longitudinal and torsional

loads. The tensile armour layers are in flexible pipes for deep-water applications usually

surrounded by a high strength tape in order to prevent radial deflections. Flexible pipes

are usually designed in accordance with the specifications given in the API17J-standard,

[1].

In the present paper, only the mechanics of the tensile armour wires are addressed.

During pipe laying, the flexible pipe is in a free-hanging configuration from an installation

vessel to the seabed, see Figure 3. Furthermore, the pipe is empty, in order to ease

the installation process, and hydrostatic pressure on the end cap causes longitudinal

243

compression. Due to vessel movements, wave loads and current the flexible pipe is also

exposed to repeated bending cycles. This is known possibly to lead to lateral wire buckling

failure, especially, if the outer sheath of the pipe is breached such that the pipe annulus

is flooded. This leads to, that external pressure no longer induce sufficient frictional

resistance to prohibit wire slippage. The failure mode was first described by Braga and

Kaleff, [2], who reproduced it experimentally in the laboratory. Further experimental

investigations were conducted in [3]-[6].

In repeated bending the wires within the pipe wall may slip towards a configuration

in which the wire lay angle is not constant, like in the initial helical configuration. For a

pipe subjected to longitudinal compression, the geometrical configurations of the wires obtained after a significant number of bending cycles, may be associated with wire buckling

within the pipe wall leading to a reduced load carrying ability of the pipe structure.

The mechanics of armouring wires in flexible pipes have been subject of both academic

and industrial research in the past few decades. Feret and Bournazel, [7], derived expressions for prediction of the global response of straight flexible pipes on basis of analysis of

internal components. The methods were implemented in a computerprogram, see [8], in

which the armouring tendons were described as perfect helices. The global behavior of

flexible pipes has been investigated further in numerous publications, see [9]-[12].

Witz and Tan, [13], and Kraincanic and Kabadze, [14], considered progression of wire

slippage along curves with constant lay angles for flexible pipes in bending. Svik, [16],

addressed the same problem, but based his analysis on a finite element formulation based

on finite-strain continuum mechanics. Out and von Morgen, [15], considered wire slip

towards the geodesic of a toriod in bending. Leroy and Estrier, [17], simulated wire

slippage due to cyclic bending based on curved beam equilibrium with frictional effects

taken into calculation. However, a prescribed experience-based solution form was applied.

and simulated.

The approaches to wire mechanics are obviously all incapable of predicting transverse

wire slippage for a flexible subjected to bending and compression. A method for calculation of the equilibrium state of a wire which was free of geometrical constraints was

proposed in [6] and elaborated further in [18]. The problem of a curved beam embedded in

a frictionless toroid was addressed, assuming that the wire equilibrium state reached after

244

when the wire is loaded, if friction is neglected. The proposed method was applied to the

lateral buckling problem, see [6], in which it was shown that the method was capable of

representing the buckling modes of deformation. An example of simulated and experimentally triggered buckling modes is presented in Figure 2. The governing equations were

formulated analytically, but solved by numerical means. However, frictional effects were

not investigated. Since transverse wire stabilization due to friction may possibly increase

the buckling load and shorten the modes of deformation, frictional effects are included in

the present approach. For the sake of simplicity, only a single wire within the wall of a

flexible pipe subjected to compression and repeated bending cycles will be analyzed, since

frictional effects on all wires and arising couplings to the global constitutive behavior of

flexible pipes demands severe computational power.

In the present approach, the focus has mainly been on adding some frictional stabilization in order to study, how this influences the wire responses, rather than to model an

exact physical behavior with a frictional law based on experimentally obtained parameters. Despite results may not correlate well with experimental measurements, the chosen

theoretical approach and the obtained results is of a very interesting nature due to, that

very little research in the mechanics of wire slippage is available.

Single wire mechanics

System of field equations

In this section, the system of field equations governing the wire equilibrium state is presented. The system of equations was derived in [6] and [18], but for the sake of completeness the derivation is summarized in the Appendix. The wire geometry is shown in

Figure 4 with a curvilinear (tnb) coordinate frame. Before proceeding, the assumptions

on which the present formulation is based are summarized

The wire will in the initial configuration be assumed to constitute a geodesic on a

cylinder, hence, a helix.

245

The pipe will be assumed bent to a constant radius of curvature. Hence, a wire

constitutes a curve on a cylindrical surface, which is bent into a toroid with major

radius R = 1/.

A wire will be modeled as a long and slender curved beam of rectangular cross

section. The dimensions of the cross sections are assumed small compared to both

minor and major torus radii.

Wire friction will be modeled using Coulombs law. Hence, the frictional load is

assumed speed independent.

Wire inertia terms are neglected, since these are estimated small compared to stiffness related terms. A similar approach was followed in [17].

The wires in the inner layer of tensile armour have responses which in terms of stable/unstable behavior are sufficiently equivalent to only consider a single armouring

wire.

Frictional effects will be accounted for by applying transverse loads. However, since inertia

terms are small, second order terms related to the wire slip acceleration in the equilibrium

equations will be neglected. Furthermore, applying Coulomb friction, the transverse wire

loads constituting frictional effects are governed only by the normal wire load and the

frictional coefficient. The direction of the frictional loads will be determined on basis of

the previous load step.

Six differential equations in torus coordinates u and , wire lay angle , tangential

wire force Pt , Shear force in the binormal wire direction Pb and normal moment Mn as

functions of wire arclength s is derived.

cos

du

=

ds

1 + r cos

d

sin

=

ds

r

246

(1)

(2)

d

ds

dPt

ds

dPb

ds

dMn

ds

sin

cos + g

1 + r cos

(3)

= n Pn g Pb p t

(4)

= g Pt Pn pb

(5)

= n Mt + Mb + Pb

(6)

The system is derived on basis of Kirchhoffs equations for curved beam equilibrium

given on vectorial form by Reissner, [19], and concepts from differential geometry for

mathematical description of curves on surfaces.

In order to discretize the system on a known regular mesh, the unknown arclength s

in the deformed state is converted to initial helical arclength s0 . Assuming strains small,

this can be done by applying Cauchys definition of strain, , which is given by

ds

= (1 + )

ds0

(7)

The initial arclength is given by the well-known relation valid for a helix

s0 =

r

sin(hel )

(8)

The system will be solved with respect to boundary conditions corresponding to the

physics of a wire within the wall of a flexible pipe

u(0) = 0

A

(0) = ini

(0) = hel

B

(SL ) = ini

(SL ) = hel

(9)

(10)

in which Papp is the external load on the wire in the longitudinal pipe direction and SL

B

A

is the total arclength of the wire. ini

and ini

denotes the circumferential wire angles in

both end of the pipe.

Wire stability in dynamic bending

Stability problems have to a wide extend been investigated and are well-described in the

literature. In general, compressive loads are known possibly to cause the equilibrium

equations of a given structure to be fulfilled in buckled geometrical configurations associated with large deflections and rotations. The corresponding equilibrium paths in

force-displacement diagrams may exhibit softening, bifurcation or limit point behavior.

Neglecting friction on the wires, a classical stability approach to the lateral wire buckling

problem was followed in [6]. In the present approach, simulation of frictional loads encaptures an additional physical effect, namely, that cyclic loads must be applied in order

for the wire to slip. A different definition of stability must therefore be considered. Considering the equilibrium paths of a point on the modeled wire, these will, except for the

points s = 0 and s = SL , exhibit a loop-like behavior (examples of such loops are given

247

in Figure 11 and 12). If the wire when subjected to cyclic loads converges towards an

equilibrium state in which this loop is closed, the wire will in the following be considered

stable. If this is the case, the pipe strain obtained by the analysis after each bending cycle

has been completed, will be constant after a number of bending cycles have been applied.

On the other hand, if the (load-strain) loops are not closed, the slip with respect to the

initial configuration will increase for each bending cycle. This may lead to, that the yield

strength of the wire steel is exceeded and failure occurs due to formation of plastic hinges.

In [6] it was demonstrated that buckling could be triggered by adding a small harmonic

response to the initial helical geodesic curvature, such that g can be determined by

m

is

Mn X

(11)

i sin

+

g =

EIn i=1

L

In the following, the imperfection will be calculated by setting 1...20 = 0.001 in accordance with [6].

Frictional forces

The wire loads in the toroid tangent plane, pt and pb , will be defined such that they

constitute frictional resistance. In order to do so, the problem will be defined and solved

stepwise for a prescribed load history. First, the wire will be loaded longitudinally. Afterwards cyclic bending will be simulated. An example of such a definition of loads is

presented in Figure 7 and 8. Since the mass of the wire is small and assuming bending to

be applied slowly, inertia terms can be neglected. Coulomb friction is for a given speed v

defined as

v

(12)

pfric pn

kvk

(13)

The slip speed can be observed only to provide the direction of the frictional force. However, the formulation given in equation (12) in inconvenient for implementation in numerical solvers. A regularization will therefore be applied by assuming a transition, z,

between zero frictional force for v = 0 to full frictional force at v = z

v

(14)

v < z : pfric = p(v)

kvk

v

v z : pfric = pn

(15)

kvk

in which z is the length of the transition zone and p(v) is a polynomial of second order

determined on basis of the conditions

dp(z)

=0

(16)

dv

The slip D is calculated with respect to the previous load step, see Figure 6. For the load

step i and the curvature fixed to = i the slip is given by

p(0) = 0

p(z) = pn

(17)

v(s)i =

Di

t

248

(18)

Figure 6. Definition of wire slip, which is calculated with respect to the deformed underlying layer.

0.1

0.09

0.08

Pipe curvature, (m1)

500

1000

1500

2000

0.07

0.06

0.05

0.04

0.03

0.02

2500

0.01

3000

0

0

500

1000

1500

load step number

2000

2500

wire force.

500

1000

1500

load step number

2000

2500

harmonic response.

pt,i = pfric,i ti

pb,i = pfric,i bi

(19)

(20)

The normal load is obtained from equation 42 governing the normal force equilibrium

pn =

dPn

n Pt + Pb

ds

(21)

In the present approach, radial elasticity of the pipe wall being modeled, is not taken into

account in an exact manner. Since the effect is crusial to the magnitude of the frictional

forces, the minor torus radius will be assumed a function of the applied load. Neglecting

ovalization due to bending, the minor torus radius is given by

r

ka L

r = 1

r

(22)

rd = 1 +

r

kr L

in which ka and kr constitute respectively an axial and a radial spring coefficient of the

pipe being modeled. This approach is equivalent to the methods described in [7] and are

based on the equilibrium of perfect helices. The consequence of calculating the normal

249

load in this manner is, that pn is estimated in a fair manner prior to buckling, while the

value of pn may be inaccurate after occurrence of instability, since buckling leads to large

changes of wire lay angle.

Since the problem is solved numerically by a commercially available BVP-solver, the

length of the slip transition zone, z , must be chosen in such a manner, that convergence

can still be obtained. All time steps will be set to t = 1 s. With this assumption,

the slip speed is related to the slip by Di = vi [1 s]. By numerical experiments it was

determined, that a solution could not be obtained for very short values of z. A transition

length of z = 0.005 was the smallest value, for which the analysis could be performed with

reasonable precision. An obvious consequence of this choice, is that the wires when loaded

may not experience full friction, since the slip speed does not cause, that the length of the

transition zone is exceeded. In the present analysis, stick effects are therefore simulated

in a manner, so the wire has a small speed.

Results

An armouring wire with rectangular cross section within the wall of a flexible pipe will

be modeled on basis of the following geometrical input

r = 0.2762m

Lpitch = 1.474 m

width = 12.5 mm

hel = 30 deg

ka

= 1.9

height = 5 mm

kr

(23)

(24)

The wire steel will be considered isotropic with elastic modulus E = 210 GPa and Poissons ratio = 0.3. Five analyses will be conducted for compressive wire loads, 2.0 kN,2.5

kN, 2.75 kN, 3.0 kN and 3.5 kN. 20 bending cycles from = 1/1000 m1 configuration

(almost straight) to = 1/11 m1 will be simulated. The frictional coefficient will be set

to = 0.1, which corresponds well to values chosen in [9] and [17]. The length of the

frictional transition zone will be set to z = 0.005 m.

Initially, the geometry obtained at the last load step of the simulation will be considered for the compressive load levels 3.0 kN and 3.5 kN, see Figure 9. While the

configuration of the wire obtained for the first load level can be observed not to differ

significantly from the initial helical shape, the wire configuration obtained for the second

load level can be observed to have changed. Obviously, the conclusion can be drawn, that

the second load level has caused the wire to exhibit buckling behavior. However, it is

not possible solely on this basis to determine, if the first load level considered is stable or

not. The wire geometry with maximum curvature for the last simulated bending cycle is

shown in Figure 10.

The average pipe strain will now be considered. This is given by

u(SL ) u(0)

L

=

L

L

(25)

In Figure 11 and 12 examples of the loops formed by the equilibrium paths due to cyclic

bending are presented. Since it is difficult to draw conclusions regarding stability of these

loops, this will be studied on basis of the pipe strain.

In Figure 13 the pipe strain for all analyzed load levels are plotted as functions of

the load steps number. Yet, it is still difficult to draw conclusions regarding stability of

a specific load level. Furthermore, it is on this basis not possible to draw conclusions

regarding if buckling will occur if further bending cycles are applied. Therefore, the

250

Papp = 3.0 kN after 20 bending cycles,

blue curve: Papp = 3.25 kN after 20 bending cycles.

P

Papp = 3.25 kN, maximum bending,

blue curve: Papp = 3.25 kN, maximum

bending, first cycles.

=2.0 kN

app

=3.0 kN

app

2065

2070

2470

2075

Longitudinal load (N)

2480

2080

2085

2090

2095

2100

2500

2510

2520

2105

2110

2490

2530

1.7

1.65

1.6

Pipe strain

1.55

1.5

2.0kN , s = SL /2.

3

Pipe strain

x 10

1.45

3

x 10

kN, s = SL /2.

change of strain after each bending cycle has been concluded with respect to the strain

obtained after the first bending cycle will be considered, see Figure 14. The slope of these

curves can now be taken as basis for consideration of if the wire will remain in a stable

configuration, or if instability may occur after a larger number of bending cycles. The

magnitude of slope for the analyses with Papp set to 2.0 kN and 2.5 kN is decreasing

while this value for the remaining analyses is increasing. The conclusion can therefore be

drawn, that the wire for the two first load levels seem to converge against a closed loop in

(force-strain)-diagrams, while the geometry of the equilibria obtained with the remaining

load levels do not converge towards closed loop behavior. The limit compressive load for

the wire can on basis of this method be estimated to lie between 2.5 and 2.75 kN.

It is interesting to compare this measure for the maximum load carrying ability of

a single wire with the limit load obtained from a frictionless analysis of both layers of

armouring wires by methods proposed in [6]. Calculating the compressive load per wire,

an equilibrium path as shown on Figure 15 is obtained. A limit load of 2.33 kN is

251

0.001

0.5

0.002

0.003

1.5

Change of strain, i1

Pipe strain

0.004

0.005

0.006

Papp=2.0 kN

0.007

0.008

Papp=2.75 kN

=2.5 kN

2

2.5

3

P

Papp=2.5 kN

0.009

Papp=2.75 kN

=3.0 kN

app

500

1000

1500

Load step number

=3.0 kN

=3.25 kN

app

5

0

app

4.5

Papp=3.25 kN

=2.0 kN

app

3.5

app

0.01

x 10

2000

6

8

10

12

14

Number of applied bending cycles

16

18

20

obtained as maximum load carrying ability by the analysis. This is slightly less than the

value determined on basis of the present analysis.

In order to compare the wire mode of deformation associated with instability obtained

by the present method with the buckling mode determined with no friction, these are

shown in Figure 17. It is noted that the frictionless buckling mode is calculated with a

deformation controlled model and that direct comparison of the magnitude of the two

responses is not possible. Furthermore, the two responses do not represent the same load

level, since this cannot be ensured due to significant differences in the chosen means for

controlling the model. However, it can be concluded that the two deformation modes

have approximately the same shape. Hence, inclusion of friction in the model can not be

concluded to have changed buckling modes significantly. In order to investigate the effect

of the frictional coefficient, three analyses with Papp = 2.75 kN and frictional coefficients

0.05, 0.1 and 0.15 were carried out. The results are available in Figure 16. As expected,

the strain rate increases after a lower number of applied bending cycles if the frictional

coefficient is decreased.

P

3500

0.5

2500

Pipe strain

=0.15

=0.10

=0.05

3000

2000

1

1500

=2.75 kN

app

x 10

=1/12 m

along path

1.5

=0 along path

2.5

1000

500

Frictionless model

Model including friction

Frictionless limit load

bending

extension

coupling

3.5

0

0.5

1.5

Pipe strain

2.5

3

3

x 10

wire, analyses with and without friction.

500

1000

1500

Load step number

2000

2500

coefficient on the pipe strain response.

252

cause hysteresic flexural behavior in flexible pipes, see [11] and [13]. In order to investigate, how the present approach to frictional effects on armouring wires corresponds to

the descriptions given in other publications, the moment-curvature relation is studied in

Figure 18. The total local wire moment M is calculated on vectorial form as the sum of

the moments around the wire tnb-directions. The obtained moment can afterwards be

projected onto the z-axis on basis of a unit vector k in this direction.

M = Mt t + Mn n + Mb b

Mz = M k

(26)

The behavior detected by the present approach corresponds well to the expected hysteresic

flexural behavior despite only the contribution from a single wire is considered. However,

it is noted that a force term should be added in equation 26 if the total contribution from

the analyzed wire to the global pipe moment is desired, see [6].

Papp=2.0 kN

10

0.65

0.6

Moment contribution from wire (Nm)

0.55

0.5

0.45

0.4

0.35

After 1 bending cycle

After 10 bending cycles

After 15 bending cycles

After 18 bending cycles

Frictionless equilibrium, L/L=0.001

Frictionless equilibrium, L/L=0.002

0.3

0.25

0.2

0.15

0.02

0.04

0.06

Pipe curvature, (1/m)

0.08

0.1

wire lay angle, results obtained with and without inclusion of frictional effects.

Figure 18. Wire contribution from local moments to global pipe hysteresic flexural behavior for Papp = 2.04 kN, s = SL /2.

For small wire slips, it is reasonable to calculate slippage in terms of a tangential and a

transverse components by projecting the wire slip for load step i onto the initial tangent,

t0 , and initial binormal, b0 for fixed curvature

Dt = kD t0 k

Db = kD b0 k

The two slip components are plotted versus each other, see Figure 19 and 20. Similar

results are presented in [17] and [21].

Conclusions

On basis of an established model for determination of the equilibrium state of an armoring

wire within the wall of a flexible pipe, means for inclusion of frictional effects have been

presented. Solutions are obtained as the solution to a boundary value problem solved

for each step in a predefined load history. Friction has been modeled as tangential and

transverse distributed wire loads with magnitudes based on a regularized Coulomb law

and the normal distributed wire load. The directions of the frictional loads have been

calculated on basis of the wire slippage with respect to the the previous load step. Despite

the choice of slip speed transition is arguable, the proposed method has proven capable

253

Papp=3.0 kN

Papp=3.25 kN

25

20

10

15

0.5

1.5

2

tangential slip (mm)

2.5

wire slip for Papp = 3.0 kN.

0.5

1.5

2

tangential slip (mm)

2.5

3.5

wire slip for Papp = 3.25 kN.

of limiting the wire slippage in dynamic loading and representing key-effects which are

known to be caused by friction.

The proposed method was applied to a specified pipe design and the stability of a

single wire subjected to prescribed cyclic loads was examined. It was found, that when

simulating only a limited number of bending cycles, an estimation of if the wire would

remain in a stable configuration could be found by considering the change of strain obtained after each bending cycle with respect to the strain found after the first bending

cycle. The sloop of the obtained curves may serve as basis for stability considerations,

since they reveal if wire slippage converges towards a stable configuration or not. The

buckling modes determined were approximately of the same shape as buckling modes

found if friction was neglected. The load carrying ability was slightly larger than the

limit load determined when friction was neglected.

With larger computational power, than used for conducting the present analyses, the

proposed method may be used to model all wires within the wall of a flexible pipe. However, coupling the stick-slip effects to the global flexural pipe constitutive relations, which

due to friction is known to exhibit hysteresic behavior leading to variations of the radius

of curvature, is a task which calls for further research. Due to the assumption, that the

global curvature is constant, this cannot be conducted, without extending the present

formulation. Furthermore, it is desirable to investigate means for implementation of a

shorter transition zone and possibly a frictional law based on measured parameters. Inclusion of such means in the analysis are likely to limit wire slippage further and represent

the modeled physics in a more accurate manner. In order to do so, further research and

severe computational power is needed. However, very little research in slip mechanics

allowing transverse slips limited by friction has been conducted, and the present research

may therefore serve as a valuable basis for further research.

References

[1] API Spec 17J, Specification for Unbonded Flexible Pipe American Petroleum Institute,

2nd edition, 1999

[2] Braga, M.P. and Kaleff, P. Flexible Pipe Sensitivity to Birdcaging and Armor Wire

Lateral Buckling. Proceedings of OMAE, 2004

254

[3] Bectarte,F. and Coutarel,A. Instability of Tensile Armour Layers of Flexible Pipes

under External Pressure. Proceedings of OMAE, 2004

[4] Secher, P., Bectarte, F. and Felix-Henry, A. Lateral Buckling of Armor Wires in Flexible Pipes: reaching 3000 m Water Depth. Proceedings of OMAE, 2011

[5] Tan, Z., Loper, C., Sheldrake, T. and Karabelas, G. Behavior of Tensile Wires in

Unbonded Flexible Pipe under Compression and Design Optimization for Prevention

Proceedings of OMAE, 2006

[6] stergaard, N.H., Lyckegaard, A. and Andreasen, J. On lateral buckling failure of

armour wires in flexible pipes. Proceedings of OMAE, 2011

[7] Feret, J.J. and Bournazel, C.L. Calculation of stresses and slips in structural layers of

unbonded flexible pipes. Journal of Offshore Mechanics and Arctic Engineering, Vol.

109, 1987

[8] CAFLEX Theory manual. IFP/SINTEF, 1991

[9] McIver, D.B. A method of modeling the detailed component and overall structural

behaviour of flexible pipe sections. Engineering Structures, Vol. 17, No. 4, pp. 254-266,

1995

[10] Custodio, A.B. and Vaz, M.A. A nonlinear formulation for the axisymmetric response

of umbilical cables and flexible pipes. Applied Ocean Research 24, 21-29, 2002

[11] Alfano, G., Bahtui, A. and Bahai, H. Numerical derivation of constitutive models

for unbonded flexible risers. International Journal of Mechanical Sciences 51,295-304,

2009

[12] Vaz. M.A. and Patel, M.H. Post-buckling behaviour of slender structures with a bilinear bending moment-curvature relationship. International Journal of Non-linear Mechanics, Vol. 42, p.470-483, 2007

[13] Witz, J.A. and Tan, Z. On the Flexural Structural Behaviour of Flexible Pipes, Umbillicals and Marine Cables. Marine Structures, Vol. 5, 1992

[14] Kraincanic I. and Kabadze E. Slip initiation and progression in helical armouring

layers of unbonded flexible pipes and its effect on pipe bending behavior. Journal of

Strain Analysis, Vol. 36, No. 3, 2001

[15] Out, J. M. M. and von Morgen, B. J. Slippage of helical reinforcing on a bent cylinder.

Engineering Structures, Vol. 19, No. 6, pp. 507-515, 1997

[16] Svik, S. A finite element model for predicting stresses and slip in flexible pipe armouring tendons. Computers and Structures. Vol. 46, No.2, 1993

[17] Leroy, J.M. and Estrier, P. Calculation of stresses and slips in helical layers of dynamically bent flexible pipes. Oil and Gas Science and Technology, REV. IFP, Vol. 56,

No. 6, pp. 545-554, 2001

255

the equilibrium state of a long and slender wire on a frictionless toroid applied for

analysis of flexible pipe structures. Submitted to Engineering Structures, accepted for

publication

[19] Reissner, E. On finite deformations of space-curved beams. Journal of Applied Mathematics and Physics (ZAMP), Vol. 32, 1981

[20] Love, A.E.H. A treatise on the Mathematical Theory of Elasticity. Dover Publications,

Inc., N.Y, 1944

[21] Brack. A., Troina, L.M.B. and Sousa, J.R.M. Flexible Riser Resistance Against Combined Axial Compression, Bending and Torsion in Ultra-Deep Water Depths. Proceedings of OMAE, 2005

Niels Hjen stergaard, Anders Lyckegaard

NKT-Flexibles / Aalborg University, Department of Mechanical and Production Engineering

Priorparken 480, DK-2605 Broendby

Niels.HojenOstergaard@nktflexibles.com, Anders.Lyckegaard@nktflexibles.com

Jens H. Andreasen

Department of Mechanical and Production Engineering, Aalborg University

Pontoppidanstrde 103, DK-9000 Aalborg

jha@m-tech.aau.dk

In this section, the methods used for determination of the equilibrium state of an armouring wire within the wall of a flexible pipe are described.

Geometry

A point on the toroid given by a set of (u, )-coordinates is in cartesian coordinates given

by

1

+ r cos cos

(u) 1

1

+ r cos sin (u)

(27)

x(u, ) =

r sin

A curve is defined by specifying a relation in (u, )-coordinates. Assuming that such a

relation is given, the following norms are defined

xu =

x

u

x =

(28)

orthonormal vectors, tangent t, normal n and binormal b, see Figure 5, can be attached

to the curve

t=

d

du

d

= xu

+ x

ds

ds

ds

n=

xu x

kxu x k

b=tn

(29)

In equation (29), the wire normal has been defined equal to the surface normal. Hereby, it

is assumed that adjacent pipe layers are sufficiently stiff to prohibit the wire from rotating

256

freely around the local tangent. Hence, the rotation around t is geometrically governed

by the underlying toroid.

Adressing the definition of the wire tangent geometry, an alternative definition can be

based on the following vectors spanning the tangent space of the toroid

t = cos tu + sin t

(30)

in which tu and t , which span the toroid tangent space, are given by

tu =

xu

kxu k

t =

x

kx k

(31)

In order for this definition to be consistent with equation (29), the following two differential

equations must hold

du

cos

cos

=

=

ds

kxu k

1 + r cos

d

sin

sin

=

=

ds

kx k

r

(32)

These equations govern the wire geometry in the surface tangent plane.

Transformation formulaes

Having defined two orthonormal frames, (t, n, b) and (tu , t , n), see Figure 5, it is desirable

to relate those by a transformation formula

t

cos sin 0

tu

n = 0

0

1 t

(33)

b

sin cos 0

n

Furthermore, considering the (t, n, b)-frame, it is desirable to relate the triad vectors to

their derivatives in arclength. Defining a normal curvature component, n (curvature

in the (t, n)-plane), a geodesic curvature component, g (curvature in the (t, b)-plane)

and a wire torsion component, (in the (n, b)-plane), this transformation, known as the

Darboux frame, is given by

t

t

0

n g

d

n

n = n 0

(34)

ds

g

0

b

b

It is noted, that the transformation contained in equation (34) implies that a positive

rotation about a given triad axis corresponds to a positive change of curvature for a

positive change of arclength. This is sufficient to specify the signs in the constitutive

relations for the wire.

Equilibrium equations

The equations of equilibrium for a curved Bernoulli-Euler beam segment were formulated

by Kirchhoff and included in Loves book on theory of elasticity, [20]. On vectorial form,

the equilibrium equations were given by Reissner, [19]

dP

+p=0

ds

dM

+tP+m=0

ds

257

(35)

distributed moments. These may on components form be written as

P

M

p

m

=

=

=

=

P t t + Pn n + Pb b

Mt t + Mn n + Mb b

pt t + pn n + pb b

mt t + mn n + mb b

(36)

dP

+p=

ds

dt

dn

db

dPt

dPn

dPb

Pt + Pn

+ Pb

+t

+n

+b

+

ds

ds

ds

ds

ds

ds

pt t + pn n + pb b = 0

dM

+m+tP=

ds

dt

dn

db

dMt

dMn

dMb

+ Mb

+t

+n

+b

+

Mt + Mn

ds

ds

ds

ds

ds

ds

m t t + m n n + m b b Pb n + Pn b = 0

(37)

(38)

(39)

(40)

t P = t (Pt t + Pn n + Pb b) = Pt t t + Pn t n + Pb t b = Pn b Pb n

The equations of equilibrium can now be written on the following form

dn

db

dPt

+ Pn t

+ Pb t

+ pt

ds

ds

ds

dPn

dt

db

+ Pt n

+ Pb n

+ pn

ds

ds

ds

dPb

dt

dn

+ Pt b

+ Pn b

+ pb

ds

ds

ds

dn

db

dMt

+ Mn t

+ Mb t

+ mt

ds

ds

ds

dMn

dt

db

+ Mt n

+ Mb n

Pb + m n

ds

ds

ds

dMb

dt

dn

+ Mt b

+ Mn b

+ Pn + m b

ds

ds

ds

=0

(41)

=0

(42)

=0

(43)

=0

(44)

=0

(45)

=0

(46)

Applying the transformation given in equation 34, the following expressions are derived

dt

dn

= t

ds

ds

dt

db

= b

g = t

ds

ds

dn

db

=b

= n

ds

ds

n = n

258

(47)

The wire curvature components can now be calculated on basis of the chosen geometry

cos

1

n =

cos2 sin2

r

1 + r cos

sin

d

cos +

g =

1 + r cos

ds

1

cos

cos sin

=

1 + r cos r

(48)

(49)

(50)

Constitutive relations

In order to relate the changes of curvature with respect to the initial helical wire state

( = 0) to sectional wire moments, the constitutive relations will be assumed linear. This

is a reasonable assumption if the wire cross sectional dimensions are small compared to

the minor torus radius, which is the case when modeling a flexible pipe. Furthermore, it

will be assumed that the wire strains, , are small, so Cauchys definition of strain applies.

The following constitutive relations can then be assumed valid

Pt = EA

Mb = EIb n

Mt = GJ

Mn = EIn g

Similar constitutive relations have to a wide extend been applied when investigating the

mechanics of armouring wires, see [8, 13, 17].

Field equations

In order to determine the geometry of the wire which on basis of the chosen constitutive

relations satisfy the equations of equilibrium, a sixth order system of first order differential

equations can be derived by considering the following:

Equation (32) governing the wire geometry in the toroid tangent plane provides two

differential equations in u and .

The definition of the geodesic curvature, equation (49), provides one differential

equation in .

The equilibrium equations in tangential force, equation (41), in binormal force, equation (43) and normal moment, equation (45), provides three differential equations

in Pt , Pb and Mn .

This yields the system of six first order differential equations (1-6).

259

! "#

"$%&%' #"()"$ "

$$%" #

# #

!"#$$%&'()*"+*,-$*./01*2344*53,-*6',$!'7,&"'78*9"'+$!$'#$*"'*:#$7';*:++)-"!$*7'%*.!#,&#*1'(&'$$!&'(*

:0.12344*

<='$*4>?2@;*2344;*A",,$!%7B;*C-$*D$,-$!87'%)

:0.12344?@>358

Niels H. stergaard

NKT-Flexibles /

Aalborg University, Department of Mechanical

and Production Engineering

Denmark

Email: Niels.HojenOstergaard@nktflexibles.com

Anders Lyckeggaard

NKT-Flexibles

Jens H. Andreasen

Aalborg University,

Department of Mechanical and Production Engineering

ABSTRACT

This paper introduces the concept of lateral buckling of tensile armour wires in flexible pipes as a failure mode. This phenomenon is governed by large deflections and is therefore highly

non-linear. A model for prediction of the wire equilibrium state

within the pipe wall based on force equilibrium in curved beams

and curvature expressions derived from differential geometry is

presented.

On this basis, a model of the global equilibrium state of the armour layers in flexible pipes is proposed. Furthermore, it is

demonstrated how this model can be used for lateral buckling

prediction. Obtained results are compared with experiments.

However, a number of failure modes are still subject of academic

and industrial research. Among those are lateral buckling of flexible pipe tensile armour layers, which usually are designed as two

layers of helically wound steel wires with opposite lay directions.

The structural function of the layers is mainly to ensure the integrity of the structure against axial and torsional loads. In order

to prevent large radial wire deflections caused by axial compression, which leads to an instability failure mode usually denoted

birdcaging, see Fig. 1, most risers for deep waters are designed

with a high strength tape wound around the wires.

The lateral buckling failure mode has been observed to occur

during installation of flexible pipes in ultra deep waters. In the

installation scenario, flexible pipes are exposed to axial compression due to hydrostatic pressure on the end cap of an empty pipe,

and repeated bending cycles due to vessel movements, waves and

current, see Fig. 2. Furthermore, wet annulus conditions (corresponding to a damaged outer sheath) are known to increase the

risk of lateral buckling failure, since external pressure does not

introduce contact stresses in the wires, which would enable friction to limit wire slippage. The failure mode is governed by very

large lateral deflections of the armour wires, see Fig. 3

In order to investigate the physics of the lateral buckling failure

mode, it was reproduced experimentally by Braga and Kaleff,

[1], in a mechanical test rig. However, it was concluded that failure occured at lower load levels than experienced in the field, and

presently this seems like a widely accepted fact. Further experimental studies were conducted and presented by Bectarte and

INTRODUCTION

Flexible riser pipes are widely used in the offshore industry

for oil and gas extraction from subsea reservoirs at water depths

so large, that it is not possible or feasible to place a traditional

jacket supported oil rig on top of the reservoir. In this case, flexible risers may connect a floating platform to a subsea reservoir.

In order to obtain a structural design which provides sufficient

structural integrity against external and internal pressures, axial

loads and large deflections, flexible pipes are usually designed

as unbonded steel-polymer composite structures comprised by a

number of layers with different mechanical properties and structural functions. Due to the extreme loading conditions a flexible

pipe may experience both during installation and in operation,

multiple failure modes have been identified. Most failure modes

can today be reconstructed experimentally under controlled con1

c 2011 by ASME

Copyright

NKT-Flexibles by laboratory testing

wires, generated by NKT-Flexibles by laboratory testing

Svik [6] presented a wire model, in which slip is assumed to

occur along a loxodromic curve, for prediction of fatigue lifetime. This research can, however, since transverse slide is neglected, not be used for lateral buckling prediction. Leroy and

Estrier, [5], presented research in which transverse wire movements were modeled. However, only force equilibrium in tension

and bending was considered. Furthermore, a prescribed experience based solution-form was chosen, which cannot be expected

to hold in compression. Brack et. al. [7] demonstrated that the

non-linear buckling load of an armour wire could be calculated

by a finite-element model. However, in order to study the coupling effects leading to failure of an armour layer, an analytical

model taken transverse equilibrium into account needs to be developed.

The presented research aims to determine the limit state of the

tensile armour wires after many bending cycles have been applied. It is noted, that while this approach is reasonable when

modeling wire buckling, it is just one of many approaches to

wire mechanics, and other approaches may provide better results

with respect to other failure modes.

FIGURE 2. Touch-down zone of flexible pipe during DIP-testing simulating the installation scenario

publicly available, a series of experiments are conducted and presented in this paper.

The physical mechanism that leads to lateral buckling failure is

presently understood as loss of load carrying capacity of the inner layer of tensile armour wires due to buckling. This causes

bending and compression to couple to the pipe torsion, which

leads to a severe pipe twist in the pitch direction of the outer

layer of amour wires. Due to this twist, the stresses and deformations in the inner layer increases dramatically, which leads to

plastic deformation of the layer. Eventually, the pipe structure

will therefore be permanently deformed and will in most cases

not have the sufficient structural integrity to function in operation.

The mechanics of armour wires have been subject of research

for several decades, and numerous examples of research in how

to calculate stresses and slips are available. Feret and Bournazel,

[3], developed a model for prediction of flexible pipe responses

due to axisymmetric loads. Witz and Tan, [4], suggested a model

of armour layers in bending, however, assuming the wires to re-

1 METHODS

1.1 Single wire mechanics

The geometry, equilibrium and constitutive relations for a

single tensile armour wire subjected to axial loads and bending will be considered in this section. The radius of curvature

will, for the sake of simplicity, be assumed constant. A single

armour wire can therefore be considered as constituting a curve

on a torus surface.

Friction will, in order to determine the limit equilibrium state

of a wire subjected to given loads, be neglected. Slip towards

this limit state will occur, since the pipe annulus is considered

flooded, so friction does not restrain the wires to a loxodromic

2

c 2011 by ASME

Copyright

surface

tu =

xu

xv

tv =

kxu k

kxv k

(3)

xu =

x

x

xv =

u

v

(4)

vector can be defined as

t = cos tu + sin tv

for this definition to be consistent with the definition given in

equation 2, the following vectorial equation must hold

d

du

dv

= xu

+ xv

ds0

ds0

ds0

= cos tu + sin tv

configuration.

While the curvature components of a single tensile armour wire

in the analysis are allowed to be large, the axial strain of the wire

is assumed sufficiently small to determine using Cauchys definition of strain. It will furthermore be assumed that the curvature

components can be determined on basis of the geometry of an

inextensible curve due to small axial strains.

A parameterization of the torus by an arc length coordinate u

along the torus centerline and an angular coordinate v is chosen,

see figure 4. The torus surface is then, for pipe curvature = R1

and radius r given by

x(u, v) =

1

+ r cosv cos

(

u)

1

+ r cosv sin ( u)

r sin v

d

ds0

n=

xu xv

kxu xv k

b = tn

(6)

du

cos

=

ds0

kxuk

dv

sin

=

ds0

kxv k

(7)

and surface geometry. The effect of that the arc length s in the

loaded wire state does not correspond to the arc length in the initial helical state s0 will now be taken into calculation by modification of the tangent length. Considering the initial wire geometry parameterized by undeformed arc length s0 , the axial wire

strain can be related to s by

(1)

ds = (1 + )ds0

curve parameterized by undeformed arc length s0 , a curvilinear

coordinate triad of unit length given by

t=

(5)

(8)

is given by

(2)

t t = (1 + )2

(9)

geometry of the wire in the tangent plane, an alternative definition of the unit tangent vector is as a linear combination of the

following two unit vectors spanning the tangent plane of the torus

t = (1 + ) cos tu + (1 + ) sin tv

3

(10)

c 2011 by ASME

Copyright

0.06

g Leroy and Estrier

sin

dv

= (1 + )

ds

kxv k

(11)

kxv k=r

n

norms can be determined

kxu k=1 + r cosv

g derived

0.04

Change of curvature, (1/m)

du

cos

= (1 + )

ds

kxuk

(12)

n derived

Leroy and Estrier

derived

0.02

0.02

0.04

have now been derived, curvature components of the wire must

be calculated. Applying the well-known Darboux frame, the

triad vectors and their first order derivatives in arc length can

be related by the curvature components

t

0 n g

t

d

n = n 0 n

ds0

g 0

b

b

0.06

dM

+tP+m= 0

ds

1

1.5

wire arclength, s(m)

2.5

17 are given by

(13)

P = Pt t + Pnn + Pbb

M = Mt t + Mnn + Mb b

p = pt t + pnn + pb b

m = mt t + mnn + mbb

(18)

Applying equation 13, the equilibrium equations can be

rewritten on component form

dPt

n Pn + g Pb + pt = 0

ds

dPn

+ n Pt Pb + pn = 0

ds

dPb

g Pt + Pn + pb = 0

ds

dMt

nMn + gMb + mt = 0

ds

(14)

(15)

(16)

The obtained curvature components can be compared with components derived by Leroy and Estrier, [5], chosing a loxodromic

curve on a torus surface as reference curve, see Fig. 5. The

curvature components can be observed to be of the same magnitude and differences in signs can be observed to correspond

to different sign conventions. Since the wire geometry has now

been considered, a set of equilibrium equations must be derived.

The equilibrium equations of a curved beam are given by Reissner, [8], on vectorial form

dP

+p = 0

ds

0.5

a positive rotation about a given axis corresponds to a positive

change of curvature for a positive change of arc length, and is

furthermore consistent with respect to the sign conventions chosen in [5] and [6]. Furthermore, this choice of sign convention

secures that moments and changes of curvature have the same

sign, which is desirable when formulating constitutive relations,

see equation 25. Applying this definition to a coordinate frame

defined as specified in equation 2, the following wire curvature

components can be derived

1

cos v

n =

cos2 sin2

1 + r cos v

r

d

sin v

g =

cos +

1 + r cos v

ds

cos v

1

=

cos sin

1 + r cos v r

dMn

+ nMt Mb Pb + mn = 0

ds

dMb

gMt + Mn + Pn + mb = 0

ds

(19)

(20)

(21)

(22)

(23)

(24)

and pt in the tangent plane and the distributed moments mb and

mn can be considered zero.

Assuming the wire dimensions small both with respect to major

and minor torus radii, it is fair to neglect curved beam terms in

the cross sectional constants and assume the wire constitutive

(17)

4

c 2011 by ASME

Copyright

simplified. Setting dds = 0 corresponding to constant pitch angle,

the expression for the geodesic curvature reduces to

Pt = EA

(25)

Mt = GJ

Mb = EIb n

g =

Mn = EIn g

cos

1 + r cos v

sin

= (1 + )

r

sin v

=

cos + g

1 r cos v

= (1 + )

du

ds

dv

ds

d

ds

dPt

ds

(26)

(27)

(28)

= n Pn g Pb

(29)

= g Pt Pn

(30)

= n Mt + Mb + Pb

(31)

L

v(S) = vBini

L

u(S) = L 1 +

L

L

(S) = hel

cos

1 + r cos v

sin

= (1 + )

r

= (1 + )

(35)

(36)

=0

(37)

= n Pn g Pb

(38)

u(0) = 0 v(0) = vAini

known quantities after a solution is found. The system can be

solved with respect to boundary conditions corresponding to the

mechanical behavior of flexible pipe end fittings, in which the

wires are fixed in displacement and rotation

u(0) = 0 v(0) = vAini (0) = hel

(34)

. Relating changes of curvature with sectional wire moments by

applying the constitutive equations, the binormal sectional wire

force is now determined by the normal moment equilibrium. The

governing equations can then be reduced to

Now considering the equations governing the tangent wire geometry in equation 11, rearranging the obtained definition of the

geodesic curvature and considering equilibrium in the tangent

plane, the following consistent sixth order system is obtained

du

ds

dv

ds

d

ds

dPt

ds

dPb

ds

dMn

ds

sin v

cos

1 + r cosv

L

u(S) = L 1 +

L

v(S) = vBini

(39)

(40)

the following determined using a Matlab build-in solver by the

Lobatto-IIIa method.

It is noted, that since the wire is considered embedded in a torus

surface, the rotation around the wire tangent is governed only

by the underlying surface. In a physical pipe structure, the wire

may under some circumstances to some extend be allowed to rotate in a slightly different manner causing a phenomenon which

is usually denoted fishscaling. This leads to a model which

has a larger stiffness than the wire being modeled. The effect is,

however, due to contact effects from adjacent layers and in accordance with observations made during conducted experiments

deemed negligible in the present context.

(32)

(33)

in which S denotes the total wire arc length, hel the initial helical

wire lay angle and vAini and vBini the specified v-coordinate of the

wire, respectively, for s = 0 and s = S. Furthermore, L

L denotes

the pipe strain and L the pipe twist, which correspond to the

applied generalized loads.

A system of equations for prediction of the wire equilibrium state

has now been derived allowing for large transverse slips.

Modeling an armour wire with the conventional assumption, that

the wire angle remains constant such that the wire constitutes a

A model for prediction of the flexible pipe torsional response

to compression and bending can now, if transverse wire contact

is neglected, be constructed on basis of multiple single wire analyses. The global torsional boundary conditions of the analyzed

5

c 2011 by ASME

Copyright

nwires

i=1

nsheets

Gi J i

i=1

=0

L

the u and v-directions given on Fig. 6. Once the geometrical configuration satisfying equation 41 has been established, the axial

loads carried by the pipe structure can be calculated as

nwires

Pa =

i=1

nwires

Pui +

nsheets

lateral buckling laboratory experiments, see Fig. 7.

While the inner layer of tensile armour will be considered free

to seek equilibrium transversely, the outer layer will be assumed

locked in loxodromic configuration. The underlying assumption,

that no transverse slip occurs in the outer layer of armour wires,

is supported by observations made during execution of experiments, by which no transverse slip or change of lay angles could

be observed in the outer layer of armour wires, even when severe

failure was detected in the inner layer of tensile armour.

In order for the free end of the flexible pipe to be in equilibrium, a pipe twist L must be applied such that the following

global equation of equilibrium is satisfied

nsheets

i=1

i=1

i

Mu,sheets

=

nsheets

E i Ai

i=1

L

L

for given generalized load inputs, pipe bending radius R and axial pipe strain L

L by Newton-Raphton iterations.

The presented model does not take radial deformations due to

axisymmetric loadings into calculation. It will, in order to study

the effect of a radially elastic pipe wall be assumed, that the radial expansion in the model equals the radial expansion obtained

by axisymmetric analysis be means described in [3]. While this

assumption can be justified for loadings which do not cause the

inner layer of tensile armour to fail by lateral buckling, it does

not provide sufficient precision after failure has occurred. The

main reason for this is, that the applied methods are based on

axisymmetric loading of perfect helices, and the wire geometry

may change dramatically as the wires buckle. From axisymmetric analysis it is known that

(Mui Pvi ri ) +

(42)

i=1

i=1

nwires

i

Pu,sheets

Pa = ka

r

L

= kr

L

r

(43)

in which ka and kr denotes the axial and radial stiffness of a flexible pipe modeled with linear global properties. Defining the

radial strain and rearranging equation 43 yields

ka L

r

rd = 1 +

r = 1

r

r

kr L

(44)

The radius in the torus model can therefore be considered a function of the applied axial loading, while initial curvature components are calculated on basis of the radius in the unloaded state.

Since the system of governing equations for the wires in the inner layer is non-linear, an imperfection must be added to the geometry, in order to trigger stability phenomena if present. The

(41)

6

c 2011 by ASME

Copyright

Inner layer

Outer layer

OD(m)

0.2012

0.209

L pitch (m)

1.263

1.318

3 10

3 10

Number of wires

52

54

150

100

50

R=11 m

R=16 m

R=11 m, radially elastic

g,pert = g + i sin

i=1

i s

L

(45)

0.2

0.4

0.6

Axial compressive pipe strain

0.8

1

3

x 10

2 RESULTS

A 6 flexible pipe with tensile armour properties given in

table 1 will be modeled. Effects from other pipe layers are neglected. It is noted, that polymer sheaths, insulation and highstrength tape layers of flexible pipes may in some cases contribute significantly to the torsional- and compressive pipe stiffness. However, for the present pipe design, this is not the case.

The wires are made of steel with elastic modulus 2.1 105 MPa,

yield stress 765MPa and are considered isotropic. In order to

study the structural behavior in compression, the (load-strain)curve of the loaded end of the pipe will be presented for various

model parameters, see Fig. 8. Two different bending radii will be

studied for radially stiff pipe structures. Furthermore, a radially

elastic pipe structure will be analyzed by applying equation 44.

All responses exhibit significant softening behavior, which is interpreted as limit point buckling. Considering the obtained equilibrium state of the wires, the added imperfection can be observed to cause wire gaps to localize in one end of the analyzed

pipe, see Fig. 9. Both the pipe curvature and the effect of radial expansion can be observed to have very limited influence on

the buckling load in the analyzed flexible pipe. However, the approach by which radial expansion is taken into calculation, can

for obvious reasons not be considered exact and can therefore

only be considered as an estimate.

Considering the twist angle of the free end of the flexible

pipe, this can be observed also exhibit limit point behavior, see

Fig. 10. The physical interpretation of this result is, that when

the pipe is subjected to loads equal to or larger than the limit

point buckling load, it will cause a severe twist. Calculating the

stresses in all wires in the equilibrium state, the maximum stress

in the armour layers can be determined, see Fig. 11.

3 MODEL-EXPERIMENT COMPARISON

In order to reconstruct the lateral buckling failure mode in

the laboratory, experiments were carried out on three 5 meter

long 6 pipe samples with armour layer design given in Tab. 1 in

mechanical test rigs, see Fig. 12. The test setup was quite similar to the one applied by Braga and Kaleff, [1]. The pipe samples were mounted with geometrical boundary conditions corresponding to the ones shown in Fig. 7. Compression was applied

by mounting a smaller flexible pipe inside the test pipe. Subjecting this to tension caused a compressive reaction in the test

sample. Cyclic bending from neutral position to a specific maximum pipe curvature was applied by rotating the pinned frames

on which the pipe endfittings were mounted. In one case, a large

number of bending cycles were applied without sign of failure

in the test sample. After the initial test cycle had been concluded, the test pipe was tensioned in cyclic bending in order

7

c 2011 by ASME

Copyright

110

100

90

80

70

60

R=11 m

R=16 m

R=11 m, radially elastic

50

0.5

0.5

1

Twist of free pipe end (deg)

1.5

350

300

250

200

150

100

50

0

Test Series I

Test Series II

Test Series III

60

400

0.2

0.4

0.6

Axial compressive pipe strain

40

30

20

10

R=11 m

R=16 m

0

50

0.8

1

3

x 10

50

100

150

200

250

number of bending cycles

300

350

400

to straighten the wires within the pipe wall, and the test pipe was

finally subjected to compressive loads larger than during the initial test cycle. This caused failure by lateral buckling in the test

pipe.

The key difference between results obtained from the model

and experiments is, that while modeled wires fail immediately

when critical loads are applied, cyclic bending must be applied

in the experiments to overcome frictional effects before failure

occurs. The measured twist of the free end of the test sample is

presented in Fig. 13. It is during laboratory testing observed, that

the pipe curvature has large influence on the number of bending

cycles which must be applied in order to trigger failure by lateral

buckling.

In Fig. 14 buckling mode shapes determined experimentally and

by modeling are compared. Buckling mode A. corresponds to

large wire gaps, while buckling mode B. corresponds to a geometrical state with large deviations from the initial helical angle

but with small gaps. It is noted, that while gaps in mode A. localize similar in test and model results, mode B. is detected as

localized buckling during experiments, but as a periodic solution

throughout the pipe length in the model. However, this mode has

in the current case occurred in the model after the yield strength

has been exceeded in the wires. This may explain the difference

8

c 2011 by ASME

Copyright

R=11 m

L=2.526 m

L=5 m

L=1.263 m

Test Result:No Failure

Test Result:No Failure

possibly not infinite lifetime

Test Result:Failure

250

200

150

100

50

0.5

1.5

2

2.5

3

3.5

Axial compressive pipe strain

4.5

5

4

x 10

c 2011 by ASME

Copyright

cluded in the model, only the limit load for which lateral buckling

will never occur can be determined.

The friction may along with end-fitting effects cause, that the

wires in each end of the analyzed pipe will never experience slip,

since they are locked in their positions. The impact of this effect

on the results can be estimated by setting the length of the mathematical model lower than the physical length of the test pipe.

Results show that this effect may impose severe impact on the

buckling load.

Future research will therefore include a method for determination of the length of the slip-free zones in each end of the test

sample. Furthermore, additional experiments will be conducted

in order to validate the obtained model.

the pipe structure is compared to tests results for a fixed bending

radius of 11m. Two test results can be observed to be, that lateral

buckling was not triggered in the test pipe, in which infinite lifetime may not be guaranteed in one case. Furthermore, two tests

can be observed to have failed by lateral buckling. These are represented by the two red dotted lines in the figure. If the model

length is set equal to the physical length of the test sample, the

model can be observed to generate a conservative estimate for

the limit load. However, it is a well-known fact that end-fitting

effects on the wires combined with friction causes the wires in

zones close to end-fittings to be restrained in their positions during bending. The effect of non-slip zones can be estimated by

setting the model length shorter than the physical length of the

pipe sample, since both slip-free zones in each end of the pipe

sample will deform rigidly. Assuming that only respectively one

and two pitches are free to slide, the buckling load can be calculated. While decreasing the model length to two pitches can

be observed to have moderate influence on the buckling load, the

model of one pitch length exhibits a dramatically larger buckling

load than the original model. A model length this short cannot be

justified on physical grounds on basis of the research presented in

this paper, but is included in the comparison in order to demonstrate that the difference between the two models with modified

length is significant. The physical reason for this is, that if less

than two pitches are free to slip transversely, the obtained buckling mode is influenced severely, so a different mode shape is

determined.

REFERENCES

[1] Braga, M.P. and Kaleff, P. (2004): Flexible Pipe Sensitivity

to Birdcaging and Armor Wire Lateral Buckling. Proceedings of OMAE, 2004

[2] Bectarte,F. and Coutarel,A. (2004): Instability of Tensile

Armour Layers of Flexible Pipes under External Pressure.

Proceedings of OMAE, 2004

[3] Feret, J.J. and Bournazel, C.L. (1987): Calculation of

stresses and slips in structural layers of unbonded flexible

pipes. Journal of Offshore Mechanics and Arctic Engineering, Vol. 109

[4] Witz, J.A. and Tan, Z. (1992): On the Flexural Structural

Behaviour of Flexible Pipes, Umbillicals and Marine Cables. Marine Structures, Vol. 5

[5] Leroy, J.M. and Estrier, P. (2001): Calculation of stresses

and slips in helical layers of dynamically bent flexible

pipes. Oil and Gas Science and Technology, REV. IFP, Vol.

56, No. 6, pp. 545-554, 2001

[6] Svik, S. (1993): A finite element model for predicting

stresses and slip in flexible pipe armouring tendons. Computers and Structures. Vol. 46, No.2

[7] Brack. A., Troina, L.M.B. and Sousa, J.R.M. (2005): Flexible Riser Resistance Against Combined Axial Compression, Bending and Torsion in Ultra-Deep Water Depths.

Proceedings of OMAE, 2005

[8] Reissner, E. (1981): On finite deformations of spacecurved beams. Journal of Applied Mathematics and Physics

(ZAMP), Vol. 32

4 CONCLUSIONS

In order to develop a method, which can predict lateral buckling of the tensile armour wires in flexible pipes, theoretical and

experimental studies have been conducted.

A mathematical single wire model based on equilibrium of

curved beams and curvature expressions derived on basis of differential geometry has been presented. Since the wire is assumed to rest on a frictionless surface, the equilibrium state of

the wire is reached immediately when loads are applied, while

cyclic loadings must be applied to a physical pipe structure in

order to overcome frictional effects so the equilibrium state is

reached.

On basis of the single wire model, a mathematical model of an

entire flexible pipe can be obtained by multiple single wire analyses, if transverse contact between the wires is neglected. This

model can be used to determine the torsional equilibrium state

for a flexible pipe subjected to given compressive and bending

loads.

This model exhibits behavior quite similar to the observations

made during experiments and can be applied in order to obtain

a conservative estimate of the limit buckling load, which can be

carried by the pipe structure. However, since friction is not in10

c 2011 by ASME

Copyright

!"#"$ %&!

!!"

'! '

(

Proceedings of the 31th International Conference on Ocean, Offshore and Arctic Engineering

OMAE2012

June 10-15, 2012, Rio de Janeiro, Brazil

OMAE2012-83080

TENSILE ARMOUR OF FLEXIBLE PIPES

Niels H. stergaard

NKT-Flexibles /

Aalborg University, Department of Mechanical

and Production Engineering

Denmark

Email: Niels.HojenOstergaard@nktflexibles.com

Anders Lyckeggaard

NKT-Flexibles

Jens H. Andreasen

Aalborg University,

Department of Mechanical and Production Engineering

ABSTRACT

In the present paper, simplifications of methods developed

for modeling of lateral wire buckling in the tensile armour layers

of flexible pipes are proposed. Lateral wire buckling may occur

during pipe laying in ultra-deep waters. In this scenario a flexible pipe is subjected to repeated bending and axial compression

due to hydrostatic pressure on the end cap of an empty pipe. If

the outer sheath is breached, these loads may cause wire slippage

towards states in which the load carrying ability is reduced and

wire buckling in the circumferential pipe direction occurs. This

leads to characteristic deformation patterns, which may compromise the structural integrity of the entire pipe structure. On the

other hand, these loads may cause overstressing of the wires, if

the outer sheath is intact.

Simplifications of established models for calculation of the load

carrying ability are in the present context proposed in a manner,

by which the effect of adjacent pipe layers on the postbuckled

response can be estimated. The simplifications enables significant reduction of the computational time, which is necessary to

calculate the load carrying ability of a given pipe structure.

structures are application as riser pipes for oil and gas extraction

from subsea reservoirs at large water depths. The complex mechanical behavior of flexible pipes has for the past few decades

been subject of industrial and academic research in order to develop methods capable of predicting the loads, which may lead

to the numerous failure modes, that have been identified. Most

failure modes can be prevented using well established methods

for engineering analysis.

Among the many layers constituting a flexible pipe are the tensile armour layers, which ensure the structural integrity against

longitudinal loads. The tensile armour layers are in most known

pipedesigns constituted by two layers of helically wound steel

wires. During pipe laying in deep-waters, a flexible pipe is in

a free-hanging configuration from an installation vessel to the

seabed, see Figure 1 and 2. In this scenario, the flexible pipe

is empty, such that hydrostatic pressure on the end cap causes

longitudinal compression. Furthermore, the flexible pipe is subjected to repeated bending cycles due to waves, current and vessel movements. These loads are known possibly to cause a failure mode referred to as lateral wire buckling, by which the armouring wires become unstable in the circumferential direction,

which leads to large deformations confined within the pipe wall

with respect to the initial helical state. Instability is known to occur in the inner layer of armouring wires, since the compressive

loads in this layer is larger than in the outer layer of armouring

wires. Furthermore, instability is known to cause, that the torsional equilibrium between the armouring layers can no longer

be maintained in the twist free pipe configuration, which may

1 Introduction

Unbonded flexible pipes are composite steel-polymer structures capable of withstanding large tensile loads, bending and

external as well as internal pressure. Flexible pipe structures are

usually designed in accordance with the specifications given in

the API 17J-code, [1]. Among the numerous applications of such

1

c 2012 by ASME

Copyright

FIGURE 1. Flexible pipe during Deep Immersion Performance tests simulating pipe laying

wires and stressing the inner layer further.

The failure mechanism was first described by Braga and

Kaleff, [2], who reconstructed the lateral buckling failure mechanism by experimental means in the laboratory in a mechanical test rig developed specifically for this purpose. Further experimental research was presented in [3] and [4], in which the

test principle was developed further including experiments conducted in pressure champers. In [5] experiments conducted by

use of a test principle similar to the one applied by Braga and

Kaleff were conducted. Furthermore, the obtained results were

compared to theoretical predictions of the load carrying ability.

These were based on a global pipe model of the armouring layers

which for a prescribed compressive strain was capable of calculating the pipe twist rate necessary to fulfill the torsional equilibrium. The model was based on multiple single wire analyses

including all wires contained within the pipe wall using a method

developed for prediction of the equilibrium state of a single armouring wire embedded in a frictionless toroid proposed in [6].

Hence, the aspect that repeated bending must be applied in order

for the wires to overcome frictional resistance and slip towards

buckled configurations was neglected and it was assumed, that

the buckled state of the layer is reached immediately, when the

wires are loaded. Some studies of frictional effects limiting wire

slippage are contained in [7]. Additional theoretical studies of

lateral wire buckling were conducted by Vaz and Rizzo [8] and

Brack et. al. [9].

Results revealed that the method was conservative with respect to

calculation of the load carrying ability of a given pipe structure.

It was furthermore shown by calculation, that boundary effect

may increase the load carrying ability, but, that the length of the

sections close to the boundaries, which should not be included

in the model, was unknown. Experimentally triggered and simulated buckling modes were shown to correspond well at a local

FIGURE 2. Schematic drawing of the principle which is usually used during pipe laying

In the present paper, it will be demonstrated, that the load carrying ability of the tensile armour layers may be assessed with high

accuracy on basis of a simplified model based on the scaled result

from a single wire analysis. The computational method, which is

developed for this specific purpose will be denoted model simplification A. This approach eases the computational effort necessary to calculate the load carrying ability significantly.

In order to estimate the effect of anti-birdcaging tape and outer

sheath, a second computational method, denoted model simplification B is introduced. This model is based on a bilinear response in the tensile armour layer prone to buckling and linear

responses of the remaining layers. The bilinear characteristic

is obtained on basis of scaled results from single wire analysis

and must therefore be applied externally in the established algorithm. However, the main advantage is that the governing system

of equations on this basis are linear and therefore can be solved

with little computational effort.

In the present paper, only modeling of the behavior related to lateral buckling triggered by laboratory testing is investigated. The

mechanical behavior during pipe laying is of a more complex nature. Investigations of model calibration against data from Deep

Immersion Performance tests are not considered in the present

paper.

2 Theory

2.1 Local model for determination of the frictionless

equilibrium of armouring wires

Initially, a single wire within the wall of a flexible pipe assumed bent to a constant curvature will be modeled as constituting a curve on a frictionless toroid, see Figure 4. The initial

helical configuration will be assumed stress-free. The toroid with

major radius R and minor radius r will be parameterized by an

2

c 2012 by ASME

Copyright

circumferential angular coordinate . To each point on the curve,

a curvilinear coordinate frame constituted by a tangent t, a normal n chosen equal to the surface normal and a binormal b can

be attached by methods from basic differential geometry.

Denoting the wire arclenght s, the wire lay angle , force

components P and moment components M, the following system of equations for calculation of the wire equilibrium state was

derived in [5].

form by Reissner, [10]. The normal curvature n , the geodesic

curvature g and the torsion was determined on basis of the

local coordinate triad. Furthermore, small initial imperfections

were added to the geodesic curvature in order to trigger instability related behavior such that g was given by

g =

du

ds

d

ds

d

ds

dPt

ds

dPb

ds

dMn

ds

cos

1 + r cos

sin

=

r

sin

=

cos + g

1 + r cos

=

m

Mn

i s

+ i sin

EIn i=1

L

(7)

(1)

in which m = 20 and i = 0.001 were used as parameters governing the imperfection. The system was solved by a commercially available BVP-solver with respect to boundary conditions

chosen in accordance with flexible pipe end-fittings and general

ized pipe loads in form of pipe strain L

L and pipe twist rate L .

Means for estimation of effects caused by radial elasticity were

presented and found not to influence the load carrying ability significantly. In order to model the outer layer of armouring wires,

in which failure does not occur, the pitch angle of the wires in this

layer was assumed constant, which reduced the governing equations to a system of fourth order. This enables determination of

the equilibrium state of the wires within the pipe wall.

(2)

(3)

= n Pn g Pb

(4)

= g Pt Pn

(5)

= n Mt + Mb + Pb

(6)

c 2012 by ASME

Copyright

FIGURE 4. Model of armour wire within the wall of a flexible pipe subjected to bending and longitudinal loads

b.) tensile armour layers, anti-birdcaging tape and outer sheath modeled, c.)

load carrying ability of model including only armouring layers

nlayers nwires ( j)

j=1

i=1

2.2

Considering the boundary conditions which a flexible pipe

is subjected to during laboratory testing in mechanical rigs, see

Figure 10,the following equation was shown in [5] to govern the

torsional equilibrium in point B, see Figure 6

nlayers nwires ( j)

Pu =

j=1

L

,

L L

nlayers nwires ( j)

j=1

(8)

iterations stepwise for all wires within the pipe wall for prescribed L

L in order to calculate the equilibrium paths of the armouring layers. The wires were modeled neglecting lateral contact by solving the system of equations 1-6 for all wires iteratively in order to have equation 8 fulfilled. In order to reduce

computational time, the outer layer of armouring wires, which

during laboratory testing is not prone to buckling, was modeled

with constant angle of pitch, see [5]. The remaining pipe layers

were neglected since these were found not to influence the mechanical behavior prior to buckling significantly. After obtaining

a solution, the load carried by the pipe structure can be calculated

by

=0

nlayers nwires ( j)

(Mui, j Pi, j r j )

i=1

j=1

Pui, j

i=1

(9)

i=1

c 2012 by ASME

Copyright

approximately 650 MPa, while the 8 riser was designed with

armour wires of high-strength steel with yield strength of 1350

MPa. The high-strength tape will be modeled with a module of

elasticity of 27 GPa.

Results obtained by experimental means will now be compared

with results obtained from simulations.

nonlinear softening behavior to an almost constant level of load

in force-strain diagrams.

2.3

layer load carrying ability

If only the load carrying ability of a flexible pipe structure is

desired, the model can be simplified significantly. The simplification will be based on the following observations:

3.1 Simulations

inner layer (modeled free to slip transversely) are different

prior to buckling, all wires seem to soften to the same approximately constant load level in the postbuckled configuration.

2. While the individual wires in the outer layer are influenced

by terms related to bending of the pipe crosssection, these

vanish if all wire forces are summed, since curvature terms

are harmonic function of . Therefore, the outer layer of

armouring wires can for prescribed generalized loads be described using the CAFLEX equations, [11] related to the

work presented in [12], for a layer of perfect helices within

the wall of a straight pipe.

full and simplified global pipe model will be presented. These

analyses are performed without stiffness contributions from the

outer sheath and anti-birdcaging layer, since these can be neglected for low strain and twist levels.

Results obtained by the full global pipe model described in [5]

will be compared with equilibrium paths calculated with model

simplication A. The equilibrium path of the 6 riser is shown in

Figure 7. Good correspondence with respect to the load carrying

ability can be observed to occur for the full global pipe model

and the simplified model.

The equilibrium path of the 8 riser is shown in Figure 8. The

load carrying ability can also in this case be estimated to correspond well. Results are in these analyses presented without

anti-birdcaging layers.

On this basis, the load carrying ability of the 6 structure is determined to be 100 kN of compressive load. The 8 structure can

be observed to have a load carrying ability equal to 256 kN of

compressive load. It is noted, that loadings above this level will

cause softening behavior of the inner layer of armouring wires.

This, however, causes a pipe twist, but does not necessarily lead

to failure. This effect is studied further in section 3.3.

The twist-rate vs. the longitudinal load was for a global model

of the 6 pipe structure contained in [5]. In order to examine

the effect of boundary effects, the wires will be assumed locked

until a certain distance to the end-fitting is exceeded. This effect

is simulated by setting the model length shorter than the physical length of the pipe sample. The exact length of the sections

in which wire slippage is limited is unknown. A value of half a

pitch was estimated in [3]. Despite the true value may be higher,

this value will be used for shortening the model. For the sake of

simplicity, the analyses will be carried out using the simplified

model with zero pipe curvature, since bending impose little influence on the load carrying ability. Results are presented in Figure 9. The boundary effects can on this basis only be observed

to impose limited effect on the load carrying ability. However,

boundary effects may not be limited to half a pitch length in each

end of the pipe. It was demonstrated in [5] that if longer zones

are considered slip free, significant influence on the load carrying

ability is imposed, since the mode of wire deformation is influenced.

It will now be assumed, that the mechanical behavior of the inner layer of armouring wires can be modeled as the scaled result

from a single wire analysis performed on basis of equation 16. The outer layer of tensile armour will be modeled using the

CAFLEX equations. Neglecting the effect of the outer sheath and

anti-birdcaging tape and denoting contributions from the outer

layer of tensile armour with index 2, equation 8 simplifies to

(Mt1 cos 1 + Mb1 sin 1 )n1wires

((Pt1 sin 1 Pb1 cos 1 ) r1 )n1wires + M2 = 0

(10)

This equation can also be solved in an equivalent manner to equation 8. This approach constitutes the first simplification, which

will be presented in the present work. The second simplification

is described in section 3.3 on basis of the results presented in

section 3. It is noted, that despite force and torsional moment

contributions from outer sheath and anti-birdcaging layers are

neglected in the present approach in order to ensure that obtained

results can be compared with results obtained by the full global

pipe model presented in [5], these can in principle be taken into

calculation by adding appropriate contributions to Equation 10.

3 Results

In the following, a 6 and an 8 riser are studied, see table

1. Both samples were 5 meter long. The polymeric sheaths will

be modeled with elastic modulus 400 MPa. The 6 riser had

5

c 2012 by ASME

Copyright

=0.001 except radially elastic global model,

1...20

1...20

300

Full pipe model, =1/11 m1

Model simplification A, =0

Linear response

Offset linear response

250

Longitudinal compressive force (kN)

200

150

100

50

200

150

100

Inner layer, radially inelastic

Outer layer, radially inelastic

Sum of layers, Radially inelastic

Inner layer, radially elastic

Outer layer, radially elastic

Sum of layers, radially elastic

Offset linear response

Model simplification A, sum of layers,

radially elastic

Analysis terminated

50

0

50

100

150

200

0.2

0.4

0.6

0.8

Longitudinal compressive pipe strain

0.2

0.4

0.6

0.8

1

1.2

1.4

Longitudinal compressive pipe strain

1.6

1.8

3

x 10

x 10

FIGURE 8. Equilibrium paths obtained for 8 riser with various methods, contributions from separate layers shown (only tensile armour layers

modeled)

FIGURE 7. Equilibrium paths obtained for 6 riser with full global pipe

model and model simplification A (only tensile armour layers modeled)

TABLE 1. Flexible pipe designs, abbreviations) TA1, inner layer of tensile armour, TA2, outer layer of tensile armour, ABC, anti-birdcaging tape

3.2

6,TA1

6,TA2

6,ABC

6,sheath

8,TA1

8,TA2

8,ABC

8,sheath

Outer diameter(m)

0.2012

0.209

0.2117

0.2253

0.2762

0.289

0.292

0.4336

L pitch (m)

1.263

1.318

0.075

1.474

1.525

0.025

26.2

26.2

83.5

30

30.3

88.4

Thickness (mm)

1.8

10

Number of windings

52

54

54

56

Laboratory experiments

not seem to be exceeded during the experiments. The reason for

the rapid twists occurring during experiments with the 6 risers,

is deemed to be that the wire yield strength has been exceeded

causing formation of plastic hinges in the wires. The responses

of the 8 riser can be observed to increase after a number of

bending cycles until a certain level is reached and the progression rate decreases. It is likely, that this mechanical behavior

occurs, since a stable equilibrium condition has been obtained in

a deformed state, and that the lower progression of change rate in

the pipe responses throughout the remaining experiment occurs

due to time-dependent effects in the polymeric layers and antibirdcaging tape.

Secondly, the 8 riser experiment denoted Test 1, LC I, has been

DIP-tested with loading conditions corresponding to the load input to the experiment. However, while no sign of lateral wire

Mechanical test rigs constructed specifically for experimental reconstruction of the lateral wire buckling failure mode was

used to test three 6 and two 8 pipe samples , see Figure 10. A

more detailed description of the test principle is included in [5].

It is widely accepted that lateral wire buckling is associated with

shortening, twist and change of circumference of a flexible pipe,

see Figure 11, 12, 13 and 14. The test programs are described in

table 1, while a summery of the results obtained is contained in

table 2.

Two very interesting observations were made during testing of

the 8 risers. Firstly, despite the experiment denoted Test II, LC

2 exhibit behavior indicating that failure had occurred in the test

sample, no signs of lateral wire buckling could be detected during dissection. Since the 8 riser has wires of high-strength steel,

this suggests that lateral buckling may developed in the elastic

6

c 2012 by ASME

Copyright

300

8" Riser, full length

8" Riser, reduced length

6" Riser, full length

6" Riser, reduced length

250

200

150

100

50

0.5

1.5

2

2.5

3

3.5

Longitudinal compressive pipe strain

4.5

4

x 10

FIGURE 10. Mechanical test rig applied for lateral wire buckling laboratory experiments

200

50

6" Riser, Test I, LC 1, 265 kN, 11 m bending

6" Riser, Test II, LC 1, 80 kN, 11 m bending

6" Riser, Test II, LC 2, 210 kN, 11 m bending

6" Riser, Test III, LC 1, 160 kN, 11 m bending

6" Riser, Test III, LC 2, 265 kN, 8 m bending

8" Riser, Test II, LC 2, 400 kN, 12 m bending

8" Riser, Test II, LC 1, 300 kN, 12 m bending

8" Riser, Test I, LC 1, 700 kN, 12 m bending

40

35

30

6" Riser, Test II, LC 1, 80 kN, 11 m bending

6" Riser, Test II, LC 2, 210 kN, 11 m bending

6" Riser, Test III, LC 1, 160 kN, 11 m bending

6" Riser, Test III, LC 2, 265 kN, 8 m bending

8" Riser, Test II, LC 2, 400 kN, 12 m bending

8" Riser, Test II, LC 1, 300 kN, 12 m bending

8" Riser, Test I, LC 1, 700 kN, 12 m bending

150

Compressive stroke (mm)

45

25

20

15

100

50

10

5

0

50

0

500

1000

1500

Number of bending cycles

2000

2500

500

1000

1500

Number of bending cycles

2000

2500

buckling could be detected by dissection after the DIP-test, failure occured during the laboratory experiment. This supports the

observation made by Braga and Kaleff, [2], that failure occurs at

lower compressive load levels during laboratory experiments in

mechanical rigs than during pipe installation in the field.

ering the pipe equilibrium paths in Figure 7 and 8, it can be concluded, that after buckling occurs, the load carried by the structures softens to a level, which with very good accuracy can be

assumed constant. This load is determined on basis of methods

described in section 2.2 and 2.3. It will now be assumed that the

mechanical behavior of the layer prone to buckling in a forcedisplacement diagram exhibits a bilinear trend. The mechanical

behavior prior to buckling is then based on the expected linear

compressive pipe response. The postbuckling behavior is based

on a constant force in the inner layer. The remaining layers will

be modeled using the CAFLEX equations, [11], however, neglecting the change of thickness in the layers, this yields three

3.3

for large twists with adjacent layers

Further simplifications can now be made in order to study

the postbuckling response taking both layers of tensile armour,

anti-birdcaging layer and outer sheath into calculation. Consid7

c 2012 by ASME

Copyright

Gauge 1

Gauge 2

Gauge 3

Gauge 4

Gauge 5

4.5

4

Change of circumference (mm)

5

Change of circumference (mm)

Gauge 1

Gauge 2

Gauge 3

Gauge 4

Gauge 5

3.5

3

2.5

2

1.5

1

0.5

0

50

100

150

200

250

Number of bending cycles

300

350

500

1000

1500

number of bending cycles

2000

TABLE 2. Test results, abbreviations) LC: load cycle, P: applied compressive load, n: number of bending cycles applied during experiment

Pipe ID

Test ID

LC ID

P(kN)

R(m)

Results

6 riser

Test 1

265

11

204

Failure ( 45deg)

6 riser

Test 2

80

11

800

6 riser

Test 2

II

200

11

392

No failure ( 45deg)

6 riser

Test 3

160

11

1200

No failure ( 3deg)

6 riser

Test 3

II

265

151

Failure ( 45deg)

8 riser

Test 1

700

12

1200

Failure ( 27deg)

8 riser

Test 2

300

12

1200

8 riser

Test 2

II

400

12

2400

No Failure ( 15deg)

pressure to axial strain, twist rate and radial strain for each pipe

layer. Furthermore, one global equilibrium equation governing

the torsional balance of the pipe is applied. This yields 13 linear equations with 13 unknowns. After buckling occurs, the load

carried by the inner layer is prescribed, and the dimension of

the system of equations is reduced with one. Obtained results

are presented in Figure 15 and 16 on basis of an externally input load carrying ability for the tensile armour layers of 100 kN

for the 6 structure and 256 kN for the 8 structure. The antibirdcaging layer and the outer sheath can be observed to contribute sufficiently to the torsional balance to impose influence

on the postbuckling response. The obtained results were com-

pared to iteratively solved simplified models for established moment equilibrium and were found to correspond well. It is noted

that while friction is neglected in the local wire model enabling

computation of the wire limit states, friction is in the present simplification assumed to cause all pipe layers in the cross section to

experience the same twist and strain.

4 Discussion

Considering the equilibrium paths calculated with the various presented methods, the following can be observed:

1. If only the armouring layers are modeled, the load carrying

ability of a given pipe structure is estimated conservatively.

8

c 2012 by ASME

Copyright

140

500

Inner layer of armouring wires

Outer layer of armouring wires

ABClayer

Outer sheath

Sum of layers

450

120

80

400

Inner layer of armouring wires

Outer layer of armouring wires

ABClayer

Outer sheath

Sum of layers

100

60

40

350

300

250

200

150

100

20

50

0

10

20

30

Pipe twist (deg)

40

50

(outer sheath and ABC-tape included)

10

20

30

Pipe twist (deg)

40

50

(outer sheath and ABC-tape included)

field.

since these may reduce the effective wire slip length.

3. While the outer sheath and anti-birdcaging tape do only impose little influence on the load carrying ability prior to

buckling, the torsional and longitudinal contributions from

these layers imposes significant influence when considering

the postbuckled response.

5 Conclusions

In the present paper, two simplifications of global pipe models for modeling of lateral buckling of flexible pipe armour wires

have been proposed. While the first simplification leads to a poor

representation of prebuckled response the postbuckled response

is estimated with good accuracy. The second method has been

applied in order to represent the global behavior of the pipe structure both before and after softening of the inner layer of tensile

armour due to lateral wire buckling. However, the load carrying ability of the layer prone to buckling must be established by

other means prior to the analysis. Compared to experimentally

obtained results, the present approach limits the conservatism of

the original model for lateral buckling prediction, which served

as basis for the present simplifications. The accuracy of the prediction may be increased further by inclusion of a better measure

of boundary effects.

with the experimental results in Table 1, it is clear that the the

load carrying ability is still estimated conservatively despite allowing that a pipe twist occurs. This discrepancy may be explained by boundary effects, which are not included. The value

of half a pitch which served as basis for calculation of the load

carrying ability shown in Figure 9 may be underestimated. Furthermore, lateral wire contact, which would limit the pipe twist

and increase the load carrying ability, is not included in the

present approach.

Both simplified approaches proposed in the present paper are

to a wide extend based on the CAFLEX equations [11]. The

equations governing the helical equilibrium are based on the assumption that the only small changes occur in the wire lay angle.

This assumption will at some point be violated in the present approach. Therefore, future research should have as objective to

implement a model of the helical layers capable of representing

large deflections and rotations for description of the unbuckled

layer, possibly by incremental means. The experiments with the

8 riser denoted Test I, LC I has been DIP-tested at the same load

level, but while the laboratory experiment caused failure by lateral wire buckling, this was not the case for the DIP-tests. This

supports the observation made by Braga and Kaleff, [2], that failure occurs at lower compressive load levels during laboratory ex-

REFERENCES

[1] API Spec 17J, Specification for Unbonded Flexible Pipes.

American Petroleum Institute, 3rd edition, 2008

[2] Braga, M.P. and Kaleff, P. Flexible Pipe Sensitivity to Birdcaging and Armor Wire Lateral Buckling. Proceedings of

OMAE, 2004

[3] Secher, P., Bectarte, F. and Felix-Henry, A. Lateral Buckling of Armor Wires in Flexible Pipes: reaching 3000 m

Water Depth. Proceedings of OMAE, 2011

9

c 2012 by ASME

Copyright

[4] Tan, Z., Loper, C., Sheldrake, T. and Karabelas, G. Behavior of Tensile Wires in Unbonded Flexible Pipe under Compression and Design Optimization for Prevention. Proceedings of OMAE, 2006

[5] stergaard, N.H., Lyckegaard, A. and Andreasen, J. On lateral buckling failure of armour wires in flexible pipes. Proceedings of OMAE, 2011

[6] stergaard, N.H., Lyckegaard, A. and Andreasen, J. A

method for prediction of the equilibrium state of a long and

slender wire on a frictionless toroid applied for analysis of

flexible pipe structures. Engineering Structures, Vol. 34, pp.

391-399,2012

[7] stergaard, N.H., Lyckegaard, A. and Andreasen, J. Simulation of frictional effects in models for calculation of the

equilibrium state of flexible pipe armouring wires in compression and bending. Rakenteiden Mekaniikka (Journal of

Structural Mechanics), Vol. 44, 2011, No. 3, 2011

[8] Vaz, M.A. and Rizzo, N.A.S. A finite element model for

flexible pipe armor wire instability Marine Structures, 2011

[9] Brack, A., Troina, L.M.B. and Sousa, J.R.M. Flexible Riser

Resistance Against Combined Axial Compression, Bending

and Torsion in Ultra-Deep Water Depths. Proceedings of

OMAE, 2005

[10] Reissner, E. On finite deformations of space-curved beams.

Journal of Applied Mathematics and Physics (ZAMP), Vol.

32, 1981

[11] CAFLEX Theory manual. IFP/SINTEF, 1991

[12] Feret, J.J. and Bournazel, C.L. Calculation of stresses and

slips in structural layers of unbonded flexible pipes. Journal

of Offshore Mechanics and Arctic Engineering, Vol. 109,

1987

10

c 2012 by ASME

Copyright

- Beam Design Lecture NotesUploaded byyyanan1118
- Steel Column DesignUploaded bymr740111
- HDPE TechInfo EnglishUploaded byMohammad Danish
- Steel design manualUploaded bymuggur
- Buckling 04Uploaded bythanzawtun1981
- Sway Mode - EC3Uploaded bytiagofreire
- [0]DNVRPC201-platedstructurebucklingUploaded by금동민
- Loaded Timoshenko BeamUploaded byap scattolin
- Syllabus 17Uploaded byTran Nguyen Khiem
- AISC vS ECUploaded byMohamed Adel Hassan
- CONTEC Design Guidelines HandbookUploaded byFarhad Khan
- 2002 Inelastic Seismic Response of Steel Bracing Members. Robert TremblayUploaded byCristian Camilo Martinez Jimenez
- EN 1993-1-5.pdfUploaded byГоран Милојевић
- 3373056.pdfUploaded byAnonymous bqV7MbFSu
- 1-s2.0-S0263823113002097-main (1)Uploaded byVasil Georgiev Georgiev
- unit-8-Design-of-Beams.pdfUploaded byveenu
- Strength and Ductility of CFS Bolted ConnectionsUploaded byFerhan Öztürk
- Perfiles a 60Uploaded byEDWINCIVILES
- 40783-45459-1-SM(1)Uploaded byShalonda Blevins
- Face Sheet WrinklingUploaded byManoj Kumar
- 3 BucklingUploaded byganesh_withucad
- Application of spline finite strip method in stability analyses of arbirarily shaped plates.pdfUploaded byAravind Bhashyam
- 7189-85-3.pdfUploaded byChris Powell-Williams
- TCM23dk5072_04Uploaded bymrnaeem
- URBeams1.pptxUploaded byDileep Lamba
- keimena_nomoi_04Uploaded bySokos
- Ultimate Flexural Strength of Prestressed Concrete Beams Validation and Model Error EvaluationUploaded byPCPC
- column_buckling.pdfUploaded by700spymaster007
- Engineering Ex Perv 00000 i 00010Uploaded byJulio García Martínez
- Standard Proctor Compaction TestUploaded byDilanka S Gunasinha

- Heinbockel - Tensor Calculus - PartUploaded byapi-3835172
- The Non Linear Field TheoryUploaded byGabriel Gonzalez
- MATH459 Project1 SolutionUploaded byGabriel Gonzalez
- A Finite Element Under CompressionUploaded byGabriel Gonzalez
- A Helical Tape on Cylinder Subjected to BendingUploaded byGabriel Gonzalez
- Technology Global WorldUploaded byGabriel Gonzalez
- Koiter, W.J. - On the Stability of Elastic EquilibriumUploaded byGabriel Gonzalez
- KANT, I., Principles of PoliticsUploaded byGabriel Gonzalez
- VIV - GabrielUploaded byGabriel Gonzalez
- Photo ElasticityUploaded byShreyas Mg

- main group chemistryUploaded bySurender Malik
- Hughes Kettner Vortex 80 Schematic DiagramUploaded byshinyredapple
- Biodiesel 2010Uploaded byJorge Alejandro DelaVega Lozano
- HTP-COPD, Chickenpox.doc.docxUploaded byArrjhey Doge
- Roi Econometric ModelingUploaded byAser A. Abdullah
- Remove Istockphoto WatermarkUploaded byapi-26979077
- The Lawgiver by Herman WoukUploaded bySimon and Schuster
- controlsystems.pdfUploaded bySirak Masresha
- 16524_structureunionenumtypedefUploaded bySatyam Singh
- 06 Process CostingUploaded byJohn Glenn Lambayon
- An Innovative Approach to Investigate VaUploaded bySergio Alan Martinez
- Essay - Ways of Seeing by John BergerUploaded byKarin Lobo
- SyllabusUploaded byalphayo
- PHOTOJOURNALISM HANDOUT.docxUploaded byAriel De La Cruz
- Batuan Sedimen Karbonat.pdfUploaded byToberMardain
- Nursing Admin- Organizational Theories and StructuresUploaded bycharmdosz
- Customer Relationship Management Project ReportUploaded byJAI SINGH
- Application Note Component Selection and PCB LayoutUploaded byheocuibap
- Chemistry Past Paper Ch1.7Uploaded byRaymond Chan
- Third Party Administrators and Health Insurance in IndiaUploaded byPiyush Trivedi
- Manual Hidrolavadora Db-383539 Mca. HotsyUploaded byLigreza
- ManualUploaded byjjAHEL
- Victorian EraUploaded byAru Saberfang D'crestfallen
- BS EN 00026-2015Uploaded byVag Katsikopoulos
- smedaUploaded byRakshit Lobin
- History of Fiat CurrencyUploaded byAlfieSHung
- Imo Solas Ui Sc Pdf1581Uploaded byDodol
- U113072 Anshuk Pradhan IKEAUploaded bySiddharth Sourav Padhee
- lazomercader.pdfUploaded byM Lapuz
- CH10AUploaded bynikowawa