You are on page 1of 157

Department of Mechanical and Manufacturing Engineering

Aalborg University
Special Report No. 80

On lateral buckling of armouring wires in flexible pipes


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

Copyright @ 2012 Niels Hjen stergaard


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.

Niels Hjen stergaard


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

Extended abstracts presented at seminars and symposia


(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.

NKT-Flexibles documentation of laboratory test program


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.

Unbonded flexible pipe structures ....................................................................................... 2

1.2.

Failure by wire buckling in flexible pipes ............................................................................ 4

1.3.

Scope of work ........................................................................................................................ 8

2.

Experimental reconstruction of the lateral buckling failure mode ......................................... 11


2.1.

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

2.2.

Measurements and instrumentation ................................................................................. 16

3.

Summary of conducted experiments ......................................................................................... 17


3.1.

The initial trial test, 6 riser ............................................................................................... 17

3.2.

Test series I, 6 riser test pipe samples in G1 bending rig ................................................. 20

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.

Discussion of experimentally obtained results .................................................................. 24

4.

Theoretical work .................................................................................................................... 27

4.1.

Theoretical approach to armour wire equilibrium........................................................... 28

4.2.

Approach to elastic stability .............................................................................................. 31

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

Additional comparison of experimental and simulated results ............................................... 39


5.1.

Pipe designs and material properties ................................................................................ 40

5.2.

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

5.3.

Calculation of the linear pipe response .............................................................................. 41

5.4.

Load carrying ability ........................................................................................................... 43

5.5.

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

5.6.

Stresses and lateral wire contact tjecks .............................................................................. 47

5.7.

Flexural hysteresis .............................................................................................................. 49

5.8.

Implementation as design tool............................................................................................. 50

Concluding remarks ................................................................................................................... 53


6.1.

Novel contributions ............................................................................................................ 54

6.2.

Future research .................................................................................................................. 54

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.

In the following, a brief introduction to flexible pipe structures is included.

1.1. Unbonded flexible pipe structures


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].

1.2. Failure by wire buckling in flexible pipes


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.

The mechanism leading to lateral wire buckling is presently understood as follows:


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.

1.3. Scope of work


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.

Global analysis with varied bending radius,


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

Single wire analysis with


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

2. Experimental reconstruction of the lateral buckling failure mode


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).

2.1. Experimental setup


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+

Outer casing (simplified)


Bearing holder
Moving frame end

Static 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

Load cell (max. 200 kN)

Change of circumference of test pipe


sample in five equally spaced points

Extensometers

Pipe rotation in a number of discrete


points (for radius of curvature estimates)

Inclinometers

Twist of torsionally pipe end, B

Inclinometer

Pipe midpoint deflection (with respect to


overbar)

Extensometer

Stroke of compression hydraulics

Position sensor

Temperature of outer sheath

Temperature sensor

Table 1 Measurements taken during lateral buckling experiments.

16

2.2. Measurements and instrumentation


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

3. Summary of conducted experiments


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].

3.1. The initial trial test, 6 riser


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

Figure 16 Twist measured during the initial trial test of a 6 riser.

300
Upper measuring limit of instrument

Compressive stroke (mm)

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

Change of circumference (mm)

30

25

180 bending
cycles

Upper measuring limit of instrument


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.

3.2. Test series I, 6 riser test pipe samples in G1 bending rig


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 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

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

Pipe twist (deg)

Pipe twist (deg)

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

Compressive stroke (mm)

160

Pipe twist (deg)

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 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

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.

3.5. Discussion of experimentally obtained results


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.

4.1. Theoretical approach to armour wire equilibrium


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.

4.2. Approach to elastic stability


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

Wire angle, (rad)

pure bending
0.93

bending and tension, =0.001


loxodromic curve
geodesic curve, FD-algorithm

0.92

bending and tension, =0.002


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

Longitudinal compressive wire force, P (N)

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

Moment contribution from wire (Nm)

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.

Figure 29 Modes of deformation, simulated and experimentally reconstructed.

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

6" Riser, radially elastic pipe structures unless stated


1

Full pipe model, radially inelastic, =1/11 m

120

Full pipe model, =1/11 m1


Model simplification A, =0
Linear response
Offset linear response

100
Longitudinal force (kN)

Longitudinal compressive 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

5. Additional comparison of experimental and simulated results


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

5.1. Pipe designs and material properties


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

Outer layer of tensile


armour wires

High strength antibirdcaging tape *3

Outer sheath *4

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)
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).

5.3. Calculation of the linear pipe response


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

Layer share of torsional moment

M
P(i)
P(i+1)

Pipe twist

Pressure on inner side of layer


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

tan 2 hel F + P(i + 1)1 + P (i )1 = 0


2
r
2
r
2r 2

For isotropic sheaths the layer specific equations are given by

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'

Figure 32 Coordinate triad used for derivation of the CAFLEX-equations.

43

5.4. Load carrying ability


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, 10 deg. twist


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

* Analyzed only with the simplified methods, see Figure 34


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

Applied Compressive load (kN)

600

Experimental results, 8" riser

500

400

300

Slowly increasing, no failure


Calculated equilibrium path, 8" riser
Stable
Increasing, failure

200

Experimental results, 6" riser

Slowly increasing

Calculated equilibrium path, 6" riser


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)

Longitudinal compressive force (kN)

160

120
100
80
60

150

100

50

Inner layer of armouring wires


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

5.5. Modes of deformation


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

Figure 37 Modes of deformation obtained experimentally, 14 jumper.

5.6. Stresses and lateral wire contact tjecks


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

max throughout wire (MPa)

350
300
250
200
150

6" riser, L/L=-510-4


6" riser, L/L=-110-3

100

6" riser, L/L=-1.510-3


8" riser, L/L=-0.510-3

50

8" riser, L/L=-0.810-3


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 .

5.7. Flexural hysteresis


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

Bending moment (Nm)

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

Test II, LC 2, 210 kN

=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

Bending moment (Nm)

Bending moment (Nm)

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.

5.8. Implementation as design tool


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.

6.1. Novel contributions


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.

6.2. Future research


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]

CAFLEX Theory manual, IFP/SINTEF, 1991.

[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]

Bazant, Z.P. and Cedolin, L. : Stability of structures, World Scientific, 2010.

[26]

Terndrup Pedersen, P. : Equilibrium of offshore cables and pipelines during laying, International
Shipbuilding Progress, Vol. 22, pp. 399-408, 1975.

[27]

Reissner, E. : On finite deformations of space-curved beams, Journal of Applied Mathematics and


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]

BFLEX Theory Manual, Marintek SINTEF-group, 1999.

[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]

Weppenaar, N. and Kristiansen, M. : Present and Future Possibilities in Optical Condition


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.

Internal NKT documents


[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]

Lateral buckling dissection report, Pretest of 2876-NKT-904 14" jumper, 2876-DEV-4019.

[61]

Lateral buckling pretest report, NKT doc. No. 2876-DEV-3866.

[62]

Lateral buckling laboratory testing report, NKT doc. No. 2876-DOC-TST-220.

[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]

Qualification dossier lateral buckling, NKT doc. No. 8000-ENG-108.

[75]

Lateral buckling design methodology certification, Bureau Veritas doc. No. E&P11190Z-R2011-01 NKT Lateral buckling, Rev. Draft 2011 12 23.

[76]

Paper A issued as NKT technical note, NKT doc. No. 8098-DEV-6550.

[77]

Paper B issued as NKT technical note, NKT doc. No. 8098-DEV-6372.

[78]

Paper C issued as NKT technical note, NKT doc. No. 8098-DEV-6373.

[79]

Paper D issued as NKT technical note, NKT doc. No. 8098-DEV-6939.

[80]

Paper E issued as NKT technical note, NKT doc. No. 8098-DEV-6374.

[81]

Paper F issued as NKT technical note, NKT doc. No. 8098-DEV-6940.







 ] 
 


Engineering Structures 34 (2012) 391399

Contents lists available at SciVerse ScienceDirect

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/Aalborg University, Department of Mechanical and Production Engineering, Denmark


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

used wire configuration is the geodesic curve on a torus surface,


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

N.H. stergaard et al. / Engineering Structures 34 (2012) 391399

Nomenclature
r
R
Lpitch
/
/hel

j
t
n
b
h
u

a
ds
ds0
L
SL

minor torus radius


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

a solution form based on experience with geodesic and loxodromic


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.

Fig. 1. Example of application of helically wound steel wires as structural element


in flexible pipe.

jn
jg
s


Pt ; Pn ; Pb
Mt ; Mn ; Mb
pt ; pn ; pb
mt ; mn ; mb
DL
L
Dw
L

normal wire curvature


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

The Cartesian coordinates of each point on the torus surface is now


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

Fig. 2. Wire geometry.

393

N.H. stergaard et al. / Engineering Structures 34 (2012) 391399

form a, and assuming this curve parameterized by arclength s, a


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

Since it is desirable to let the tangent space be spanned by unit


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

t cos /tu sin /th

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 Eq. (5), the angle / between tu and t is used to relate tu and th to t.


In order for this definition to be consistent with the definition given
in Eq. (2), the following must hold

du cos /

ds
kxu k

2.2. Wire curvature components


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

jn cos2 /j1 sin2 /j2

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

in which the shape operators denoted S are defined as

This yields the following norms


2

kxu k 1 r j cos h

kxh k r 2

Sxu nu
9

@n
@u

Sxh nh

@n
@h

14

Applying the definition given in Eq. (2), the normal vector is

cosju cos h

6
7
n 4 sinju cos h 5

15

sin h

On basis of Eq. (14), the shape operators can be derived

3
j sinju cos h
6
7
Sxu 4 j cosju cos h 5
0
2

Fig. 3. Wire coordinate triads.

cosju sin h

6
7
Sxh 4 sinju sin h 5
cos h

16

17

394

N.H. stergaard et al. / Engineering Structures 34 (2012) 391399

Rearranging Eqs. (16) and (17) yields

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

The normal curvature is now by Eq. (12) given by

jn

j cos h
1
2
cos2 / sin /
1 rj cos h
r

20

Having determined the principle curvature components, do Carmo


[12], provides a definition of the geodesic torsion which eases
determination of an explicit expression

s j1 j2 cos / sin /

21

in which a negative sign is added in order to obtain correspondence


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

A definition of the geodesic curvature is also given by do Carmo. The


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

in which (jg)1 and (jg)2 are the geodesic curvature components,


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

4 sin h cos u 5  4 sin u 5


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

cos2 u sin h sin u sin h


sin h

R r cos h
R r cos h

A curve with constant u-coordinate possess no geodesic curvature,


so

jg 2 0

24

The total expression for the geodesic curvature is then given by

j sin h
d/
jg
cos /
1 r j cos h
ds


25

The obtained curvature expressions can be validated and shown to


correspond to the definitions given in Eq. (11).
This concludes the geometrical analysis.
2.3. Equations of equilibrium

Fig. 4. Curved beam equilibrium.

It is now desirable to apply the definition of the triad vectors and


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

Equivalent componentwise equations were derived by Love [9], and


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]

2.4. Constitutive relations

dP
p0
ds

Since the analysis aims to predict the equilibrium state of long


and slender wires with dimensions much smaller than both minor

dM
tPm0
ds

26

N.H. stergaard et al. / Engineering Structures 34 (2012) 391399

and major torus radii, it is fair to assume the following constitutive


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

In the present approach, it is assumed that the material frame of the


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.

2.6. Solution means and numerical implementation


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

In order to account for extensibility causing small axial strains, s


and s0 can be related by Cauchys definition of strain

ds 1 ds0

51

45

in which jn and s are given in terms of u, h and / by Eqs. (20) and


(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

with initial curvature components

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

in which j0n ; j0g and s0 denotes curvature components calculated on


basis of s0. The derived field equations can on this basis be rewritten
as

396

N.H. stergaard et al. / Engineering Structures 34 (2012) 391399

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

It is noted, that since the tangent geometry is governed by the two


first equations, the triad vectors are still of unit length. Substituting
the strain definition given in Eq. (51) into the field equations above
yields

Fig. 5. Computational wire model.

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

In order to demonstrate which results the presented method is


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

The modified equations of equilibrium correspond to the equations


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

The only remaining unknown term is Pn, which is given by

Pn



1
dM b

j0g Mt s0 Mn
1
ds0

64

The derivative of Mb in arc length is given by

dM b
d
EIb
ds0
ds0

j0n

ds0
ds

EIb

0
n

dj ds0
d ds0
EIb jn
ds0 ds
ds0 ds


65

in which the strain derivative in arc length is obtained from

 


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

Combining Eqs. (64) and (66), an expression for Pn can be obtained


on analytical form.

68

It is desirable to test the equilibrium state of the wire in tension


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

in which m = 1 + rjcosh and B represents an integration constant.


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

N.H. stergaard et al. / Engineering Structures 34 (2012) 391399

The boundary value problems arising from the differential


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

P cos /Pt sin /Pb

76

Obtained results can be compared with equations determined by


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.

Firstly, the wire response in pure bending and tension will be


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

Wire angle, (rad)

Change of curvature, (1/m)

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)

3.3. Wire response in bending and tension

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

Wire arclength, s (m)


Fig. 6. Curvature components of loxodromic curve.

Wire arclength, s(m)


Fig. 7. Wire angle / in tension and bending.

398

N.H. stergaard et al. / Engineering Structures 34 (2012) 391399

Change of normal curvature, (1/m)

0.1

Change of curvature, (1/m)

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

Wire arclength, s(m)


Fig. 8. Changes of wire curvature components, pure bending.

Wire arclength, s(m)

30

Pt 103

20

Pn

Pb

Sectional wire forces (N)

0.1

Fig. 11. Change of normal curvature for a wire in bending and varying tensile pipe
strain.

0.15

Change of curvature, (1/m)

0.05

10
0
10
20
30

0.05
40

0.1
0

50

Wire arclength, s(m)

Wire arclength (m)

Fig. 9. Changes of wire curvature components, bending and tension.

Fig. 12. Sectional wire forces in arclength.

20

0.2

15

0.1

Distributed loads on wire

Geodesic curvature, g (1/m)

25

0
0.1
0.2
0

10
5
0
5
10

15

x 10

Pipe strain

10

Wire arclength, s(m)

mt (Nm/m)

20

pn103 (N/m)

25
0

Wire arclength (m)


Fig. 10. Change of geodesic curvature for a wire in bending and varying tensile pipe
strain.

Fig. 13. Distributed loads on wire in arclength.

N.H. stergaard et al. / Engineering Structures 34 (2012) 391399

The transverse loads, Pn and Pb can be observed to be small but


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

ture components for a wire resting on a surface, which is not of


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

Imperfection analysis of flexible pipe armour wires in compression and bending


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

engineering analysis have proven capable of predicting


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

stergaard, Lyckegaard and Andreasen / Applied Ocean Research 1 (2012) 111

of an empty pipe. In this scenario, lateral wire buckling


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],

and Leroy and Estrier, [12], investigated transverse slip


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-

stergaard, Lyckegaard and Andreasen / Applied Ocean Research 1 (2012) 111



 

 
 
  


  


  

Figure 2: Lateral buckling in the inner layer of tensile armour wires,


generated by NKT-Flexibles by laboratory testing

  
 


Figure 1: Flexible pipe design

ear phenomena, which may be related to buckling.


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,

Figure 3: Severe lateral buckling in the inner layer of tensile armour


wires, generated by NKT-Flexibles by laboratory testing

that the pipe radius of curvature is constant. With this


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)

in which r denotes the minor torus radius and the pipe


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)

stergaard, Lyckegaard and Andreasen / Applied Ocean Research 1 (2012) 111

Figure 4: Wire geometry

Figure 5: Wire coordinate triads

in which xu and x constitute a basis for the toroid tangent plane. These vectors are given by
xu =

x
u

x =

(3)

Demanding consistency between the tangent plane of


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)

In order to derive the curvature components of the wire,


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

in which n denotes normal curvature, g the geodesic


(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)

in which sectional wire force P, sectional wire moments


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

beam terms and assume the constitutive relations linear.


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)

stergaard, Lyckegaard and Andreasen / Applied Ocean Research 1 (2012) 111

(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)

The chosen approach implies that the wire in the initial


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)

Radial expansion of the flexible pipe being modeled is


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)

The detected equilibrium paths are shown in Figure 7


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

stergaard, Lyckegaard and Andreasen / Applied Ocean Research 1 (2012) 111

150




Compressive force (N)



100
B

50

Solution, straight beam


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

Figure 6: Simplified wire geometry, clamped-clamped beam

1400
A

Longitudinal compressive wire force, Pu (N)

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

B, are obtained on basis of a perfect geometry, but with


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

is exceeded, the uniqueness of the solution is lost for the


perfect wire geometry.
The branch B is obtained for a perfect structure by a

x 10

stergaard, Lyckegaard and Andreasen / Applied Ocean Research 1 (2012) 111

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

deformation controlled analysis. Furthermore, it can be


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

stergaard, Lyckegaard and Andreasen / Applied Ocean Research 1 (2012) 111


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)

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)

Figure 12: Wire lay angle, , at L


L = 0.0001, R = 15m (prior to softening), all imperfections measured in m1

Figure 13: Wire lay angle, , at L


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.

3.2. Influence of model parameters


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

Finally, it will be examined how the number of


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.

stergaard, Lyckegaard and Andreasen / Applied Ocean Research 1 (2012) 111

R=15 m
1400

0.2
Longitudinal compressive wire force, Pu (N)

Change of normal curvature, n (1/m)

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)

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

equilibrium paths influenced by varied pipe cur-

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

stergaard, Lyckegaard and Andreasen / Applied Ocean Research 1 (2012) 111

0.55

2500

0.5

L=1.25 m (one pitch)


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

Wire lay angle, (rad)

Longitudinal compressive wire load, Pu (N)

3000

1000

0.45

L=1.25 m (one pitch)


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

initial wire lay angle, hel


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

it was found that increasing the pitch length decreased


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]

pression and Design Optimization for Prevention Proceedings


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

stergaard, Lyckegaard and Andreasen / Applied Ocean Research 1 (2012) 111


[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

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


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

wound steel wires ensure the structural integrity against


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

stergaard, Lyckegaard and Andreasen / Marine Structures 1 (2012) 112



 

 
 
  


  


  
  
 


Figure 1: Flexible pipe design

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],

and Tan et. al., [4].


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

stergaard, Lyckegaard and Andreasen / Marine Structures 1 (2012) 112

Figure 2: Localized lateral deformations in the inner layer of tensile armour wires, generated by NKT-Flexibles by laboratory testing



Figure 3: Birdcaging, localization of radial deformations, generated by


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]

and Tan, [13], and Kraincanic and Kabadze, [14], on


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

in a frictionless cylinder bent into a toroid subjected to


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.

stergaard, Lyckegaard and Andreasen / Marine Structures 1 (2012) 112

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

For a curve on a toroid, a curvilinear coordinate triad


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)

in which xu and x constitute the partial derivatives


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)

The wire curvature was divided into a transverse or


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)

The torsion of the wire was derived to be given by


!
1
cos

cos sin
(7)
=
1 + r cos r

Considering force equilibrium, the following equations


on vectorial form were given by Reissner, [20]
dP
+p=0
ds

dM
+tP+m=0
ds

(8)

in which sectional wire force is denoted P, sectional


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 [17] it was shown that unknown functions are given


in terms of known quantities after a solution is obtained.
However, since the arclength s is not equal to s0 in the

stergaard, Lyckegaard and Andreasen / Marine Structures 1 (2012) 112

Figure 5: Wire geometry

Figure 6: Wire coordinate triads

initial state in which the pipe is straight and the wire


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)

The wire forces in the longitudinal- and transverse pipe


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)

It is noted that the present approach to wire mechanics


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.

stergaard, Lyckegaard and Andreasen / Marine Structures 1 (2012) 112





 

2.3. Global pipe model


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.





Figure 7: Torsionally fixed-fixed flexible pipe in bending to a constant


radius of curvature


 



Figure 8: Torsionally fixed-free flexible pipe in bending to a constant


radius of curvature

The torsional moment in S = L can be calculated by


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

Mti cosi + Mbi sini

i=1

(Pit sini Pib cosi ) ri +

nX
sheets

Gi J i

i=1

In a torsionally fixed-free pipe, the equation T = 0


must be fulfilled in S = L. In order to establish torsional
equilibrium for a prescribed pipe strain L
L , a pipe twist

L must be determined such that T (L) = 0. This will be


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

Pit cosi + Pib sini +

nX
sheets
i=1

E i Ai

L
L

Results for pipes which are fixed-fixed and fixed-free in


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.

stergaard, Lyckegaard and Andreasen / Marine Structures 1 (2012) 112

2.4. Buckling analysis of tensile armour wires


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

stergaard, Lyckegaard and Andreasen / Marine Structures 1 (2012) 112

Longitudinal compressive wire force (N)

1500


1000

500

 

perfect wire geometry, secondary path



imperfect wire geometry, amp=0.001


perfect wire geometry, primary path




imperfect wire geometry, amp=0.001


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

Table 1: 6 FLEXIBLE PIPE DESIGN

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

Table 2: Material parameters

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

the outer layer transversely fixed, since a transversely


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

stergaard, Lyckegaard and Andreasen / Marine Structures 1 (2012) 112


200
inner layer
outer layer
sum of layers

180
50
Longitudinal compressive force (kN)

Longitudinal compressive force in inner layer (kN)

60

40

30

20
imperfect wire geometry, 1...20=0.001

10

imperfect wire geometry, 1...20=0.001

160
140
120
100
80
60
40
20

perfect wire geometry


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

model. Mode A and C are found for all imperfection


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

Longitudinal compressive force (kN)

Longitudinal compressive force (kN)

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

to buckling modes presented in [2]-[3] and [19].


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

stergaard, Lyckegaard and Andreasen / Marine Structures 1 (2012) 112


150

110

Compressive longitudinal force (kN)

Longitudinal compressive force (kN)

100

100

50

radially elastic structure, torsionally fixedfree pipe


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

imperfect wire geometry, 1...20=0.001,


radially stiff

50

imperfect wire geometry, 1...20=0.001,


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

Figure 17: Detected buckling modes for L


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

prised solely by the higher contribution from the outer


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-

lation, the fixed-free model may yield a limit load which


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

stergaard, Lyckegaard and Andreasen / Marine Structures 1 (2012) 112

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

a manner that lateral buckling will never occur. It is yet


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

is by no means a trivial task. The presented methods


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

stergaard, Lyckegaard and Andreasen / Marine Structures 1 (2012) 112

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]

and slip in flexible pipe armouring tendons. Computers and


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









  
          !"" #$%% &


%$#

Rakenteiden Mekaniikka (Journal of Structural Mechanics)


Vol. 44, No 3, 2011, pp. 243 259

Simulation of frictional effects in models for calculation of


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.



 

 
 
  


  


  
  
 


Figure 1. Flexible pipe design.

Figure 2. Buckling mode, triggered experimentally


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


 
 


    
   



 

Figure 3. Principle drawing, flexible pipe during installation.

a significant number of bending cycles when friction is present, is reached instantaneously


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

 








 

 





Figure 4. Wire geometry.

Figure 5. Wire coordinate triads.

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)

in which hel is the initial wire lay angle.


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

Papp = Pt cos + Pb sin

(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

D(s)i = x(s)i x(s)i1

(17)

The slip speed can on this basis be calculated as


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)

Applied longitudinal wire load

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

Figure 7. Load history, applied longitudinal


wire force.

500

1000
1500
load step number

2000

2500

Figure 8. Load history, curvature applied as


harmonic response.

With these assumptions, the frictional loads can be calculated as


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

Figure 9. Modes of deformation, red curve:


Papp = 3.0 kN after 20 bending cycles,
blue curve: Papp = 3.25 kN after 20 bending cycles.
P

Figure 10. Modes of deformation, red curve:


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)

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

Figure 11. Equilibrium path loop, Papp =


2.0kN , s = SL /2.

2.2 2.15 2.1 2.05 2 1.95 1.9 1.85 1.8 1.75


3
Pipe strain
x 10

1.45
3

x 10

Figure 12. Equilibrium path loop, Papp = 3.0


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

Figure 13. Pipe strains.

6
8
10
12
14
Number of applied bending cycles

16

18

20

Figure 14. Change of pipe strain.

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

Equilibrium paths, loaded end of wire (Papp prescribed)


3500

0.5

2500

Pipe strain

Compressive longitudinal wire load (N)

=0.15
=0.10
=0.05

bending cause equilibrum point to cycle along path

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

Figure 15. Equilibrium paths of loaded end of


wire, analyses with and without friction.

500

1000
1500
Load step number

2000

2500

Figure 16. Influence of variation of the frictional


coefficient on the pipe strain response.

252

It is a well-described phenomenon that stick-slip effects in the tensile armour layers


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)

Wire lay angle, (rad)

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

Figure 17. Comparison of deflection modes,


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

transverse slip (mm)

transverse slip (mm)

10

15

0.5

1.5
2
tangential slip (mm)

2.5

Figure 19. Tangential wire slip vs. transverse


wire slip for Papp = 3.0 kN.

0.5

1.5
2
tangential slip (mm)

2.5

3.5

Figure 20. Tangential wire slip vs. transverse


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

[18] 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. 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

Appendix: Derivation of equations governing the wire equilibrium state


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)

Assuming the curve parametrized by arclength s, a local curvilinear coordinate triad of


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)

in which P denotes sectional force, M sectional moments, p distributed loads and m


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)

Equation (35) can now be rewritten on the form


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)

in which the crossproduct t P is given by


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

ON LATERAL BUCKLING FAILURE OF ARMOUR WIRES IN 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
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.

ditions and obtained results can be predicted by engineering analysis.


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

FIGURE 1. Birdcaging failure mode in flexible pipe, generated by


NKT-Flexibles by laboratory testing

FIGURE 3. Lateral buckling in the inner layer of tensile armour


wires, generated by NKT-Flexibles by laboratory testing

main in loxodromic configuration (having constant pitch angle).


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

Coutarel, [2]. Since the results obtained by experiments are not


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)

in which the surface derivatives are given by

xu =

x
x
xv =
u
v

(4)

With the definitions given in equation 3 and 4, a wire tangent


vector can be defined as
t = cos tu + sin tv

in which denotes the wire angle with respect to tu . In order


for this definition to be consistent with the definition given in
equation 2, the following vectorial equation must hold

FIGURE 4. Wire geometry

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)

This corresponds to stating


du
cos
=
ds0
kxuk

dv
sin
=
ds0
kxv k

(7)

Equation 11 relates the arc length derivatives to the wire angle


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

A curve can be constructed by relating u and v. Assuming the


curve parameterized by undeformed arc length s0 , a curvilinear
coordinate triad of unit length given by

t=

(5)

(8)

The length of the tangent vector for a curve parameterized by s


is given by

(2)

t t = (1 + )2

(9)

The following definition of the tangent vector fulfills this requirement

can be attached to each point on the curve. Now considering the


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

This corresponds to stating

Curvature of loxodrome (constant pitch)


0.06
g Leroy and Estrier

sin
dv
= (1 + )
ds
kxv k

(11)

kxv k=r

Leroy and Estrier


n

Calculating xu and xv as defined in equation 4, the following


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

Since the governing equations for the wire tangent geometry


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

The internal and external force and moment-vectors in equations


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

FIGURE 5. Curvature components of loxodromic curve

The chosen sign convention can be shown to correspond to, that


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)

Since the surface is considered frictionless, the external forces pb


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

loxodromic curve on a torus surface, the derived system can be


simplified. Setting dds = 0 corresponding to constant pitch angle,
the expression for the geodesic curvature reduces to

equations linear. These are given by


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)

The corresponding boundary conditions are


u(0) = 0 v(0) = vAini

in which remaining unknown functions are given in terms of


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)

With dds = 0 all curvature components are only functions of v and


. 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 solutions to the presented boundary value problems are in


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

1.2 Flexible pipe model


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

Mti cos i + Mbi sin i (Pti sin i Pbi cos i ) ri

i=1
nsheets

Gi J i

i=1

=0
L

in which forces P and moments M are calculated with respect to


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

FIGURE 6. Wire coordinate triads

Pui +

nsheets

Pti cos i + Pbi sin i +

pipe will be taken as fixed-free corresponding to the conducted


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

Torsional equilibrium can be establish by solving equation 41


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

FIGURE 7. Flexible pipe model

(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

TABLE 1. 6 FLEXIBLE PIPE DESIGN

Inner layer

Outer layer

OD(m)

0.2012

0.209

L pitch (m)

1.263

1.318

Wire size (mm)

3 10

3 10

Number of wires

52

54

Axial compressive load (kN)

150

100

50
R=11 m

imperfection can be added directly to the geodesic curvature as

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

FIGURE 8. (Load-strain)-curve for different bending radii, R

with m = 20 and amp = 0.001.

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.

FIGURE 9. Localized wire buckling

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

Axial compressive load (kN)

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

FIGURE 12. Lateral buckling test rig

FIGURE 10. Pipe twist for different bending radii, R

Pipe sample twist, free end (degrees)

Maximum stress in test pipe (MPa)

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

FIGURE 11. Maximum wire stress

50

100

150
200
250
number of bending cycles

300

350

400

FIGURE 13. Twist responses measured during laboratory testing

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

FIGURE 14. Detected buckling modes

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

Axial compressive load (kN)

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

FIGURE 15. Model-experiments comparison

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.

between results. In Fig. 15 the calculated axial load carried by


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

SIMPLIFIED APPROACHES TO MODELING OF LATERAL WIRE BUCKLING IN THE


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

lead to a severe pipe twist relaxing the outer layer of armouring


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

level, see Figure 3.


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

FIGURE 3. Detected buckling modes, simulated and reconstructed by laboratory experiments

arclength coordinate u measured along the torus centerline and a


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].

tions of equilibrium for a long and slender rod given on vectorial


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)

The wire equilibrium was derived on basis of Kirchhoffs equa3

c 2012 by ASME
Copyright

FIGURE 4. Model of armour wire within the wall of a flexible pipe subjected to bending and longitudinal loads

FIGURE 5. Equilibrium paths, a.) only tensile armour layers modeled,


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

Mti, j cos i, j + Mbi, j sin i, j

i=1

(Pti, j sin i, j Pbi, j cos i, j ) r j

2.2

Global model of flexible pipe armouring layers


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)

The equation was solved with respect to L by Newton-Raphson


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

FIGURE 6. Flexible pipe model

=0

nlayers nwires ( j)

(Mui, j Pi, j r j )

i=1

j=1

Pui, j

i=1

Pti, j cos i, j + Pbi, j sin i, j

(9)

i=1

c 2012 by ASME
Copyright

tensile armour wires of basic grade steel with a yield strength of


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.

The inner layer of armouring wires were concluded to exhibit


nonlinear softening behavior to an almost constant level of load
in force-strain diagrams.
2.3

Model simplification A, calculation of the armour


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

1. While the equilibrium paths of the armouring wires in the


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.

In this section, results obtained by numerical solution of the


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

8" Riser, =1/12 m1,


=0.001 except radially elastic global model,
1...20

=0.003, Full pipe model unless stated

6" Riser, radially elastic pipe structures unless stated

1...20

300

Full pipe model, radially inelastic, =1/11 m


Full pipe model, =1/11 m1
Model simplification A, =0
Linear response
Offset linear response

250
Longitudinal compressive force (kN)

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

pitch angle hel (deg)

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

regime without leading to failure, since the yield strength does


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

Endfitting effects modeled as lengthreduction, simplified model, =0

Longitudinal compressive force (kN)

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 9. Estimation of the effect of limited wire slippage close to endfittings

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

Pipe twist (deg)

35
30

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

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

FIGURE 12. Compressive stroke measured during laboratory experiments

FIGURE 11. Pipe twist measured during laboratory experiments

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

Model simplification B: postbuckling response


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

6" Riser,Test II, LC 2


Gauge 1
Gauge 2
Gauge 3
Gauge 4
Gauge 5

4.5
4
Change of circumference (mm)

5
Change of circumference (mm)

8" Riser,Test II, LC 2

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

FIGURE 13. Change of circumference, 6 riser

500

1000
1500
number of bending cycles

2000

FIGURE 14. Change of circumference, 8 riser

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

No Failure ( < 1deg)

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

No Failure ( < 1deg)

8 riser

Test 2

II

400

12

2400

No Failure ( 15deg)

equilibrium equations relating axial force, torsional moment and


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

Longitudinal force (kN)

Longitudinal force (kN)

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

FIGURE 15. Postbuckling response, 6 riser, model simplification B


(outer sheath and ABC-tape included)

10

20
30
Pipe twist (deg)

40

50

FIGURE 16. Postbuckling response, 8 riser, model simplification B


(outer sheath and ABC-tape included)

periments in mechanical rigs than during pipe installation in the


field.

2. Boundary effects may increase the load carrying ability,


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.

Comparing the postbuckling force response in Figure 15 and 16


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