Beruflich Dokumente
Kultur Dokumente
Confronting New Challenges and Sharing Knowledge, 11–13 September 2007, London, UK
Abstract
This paper addresses the influence of pipe-soil interaction on the design of pipelines susceptible to
lateral buckling and pipeline walking. The pipe-soil response is the largest uncertainty in the design of
such systems, and so generic guidance has been developed to guide the design process.
Force-displacement-response models were developed during Phase I of the SAFEBUCK joint-industry
project (JIP) based on large- and small-scale tests carried out by the SAFEBUCK JIP and project-specific
test data donated by JIP participants. These models are currently being applied by JIP participants on
a number of projects, to quantify the susceptibility to lateral buckling and pipeline walking, and design
safe and effective means to control these phenomena. However, of all the design parameters, the soil
response causes the greatest uncertainty in design because of the extreme sensitivity of design solu-
tions to the axial and lateral resistance imposed by the soil. Improving the understanding of pipe-soil
response provides the greatest scope for refining the design of such systems.
The purpose of this paper is to outline the significant influence that pipe-soil interaction has on the
pipeline design process and to highlight the way in which the inherent uncertainty in pipe-soil resist-
ance severely complicates pipeline design. The paper then describes the research, development and
model refinement that is ongoing to reduce the uncertainties.
133
2. Pipeline Structural Response axial friction case, µa = 0.10, is reduced by a factor of about 4
2.1 Pipeline installation and the pipeline does not reach full constraint. This illustrates
a real design case with upper and lower bound frictions based
Pipe-soil interaction influences the behaviour of the pipe
on upper and lower bound soil responses that correspond to
from the moment installation commences. The interaction
drained and undrained axial movement, respectively.
between the pipeline touchdown loads, combined with the
dynamics of the pipe catenary, and the seabed surface soil The compressive effective axial force in a pipeline therefore
defines the initial pipeline embedment. The remoulding of depends on the operating condition of the pipeline and
the soil that occurs during installation then influences the the axial friction. If the compressive force is large enough,
axial resistance, affecting the conditions at which the pipe then the pipeline may be susceptible to lateral buckling.
becomes constrained2. Meanwhile the lateral resistance af- Susceptibility to lateral buckling occurs when the com-
fects the tightness of route curves that the installation con- pressive force exceeds the ‘critical buckling force’, above
tractor can achieve. which the pipeline becomes unstable laterally. This lateral
instability is critically dependent upon the lateral soil re-
Furthermore, the time that passes between installation and
sistance (specifically the ‘lateral breakout resistance’ which
operation modifies the pipe-soil response as the soil is con-
is discussed later). Once lateral buckles have formed, the
solidated under the weight of the pipe. This consolidation
axial force drops significantly as pipe feeds axially into the
leads to a small increase in pipe embedment and changes
buckle, as shown in Figure 3.
the axial response associated with pipeline start-up. This is
particularly relevant for soft clays, where consolidation ef- In the example shown the pipeline has a totally different
fects can significantly increase the soil strength close to the behaviour at lower bound axial friction and upper bound
pipe, increasing the resistance to first movement. axial friction. The lower bound friction, µa = 0.10, means
that the pipeline will experience significantly greater end
2.2 First load: general pipeline expansion behaviour expansion and will be susceptible to pipe-walking. The up-
After installation, some residual effective lay tension remains per bound friction, µa = 0.58, means that the pipeline will
in a pipeline. This force is usually small in comparison with become fully constrained over some of its length so that
the forces that develop in operation and is not considered this section will not move axially, thus preventing walking.
further here. When the internal pressure and temperature However, the maximum effective force in the pipeline in-
increase to operating conditions the pipeline tends to ex- creases significantly, making it much more susceptible to
pand, but this expansion is resisted by the axial resistance lateral buckling.
between the pipe and the seabed. This restraint causes an
axial compressive force to develop in the pipeline. Low axial friction will increase the end expansion and axial
feed-in to lateral buckles, while high levels of axial friction
The ends of a pipeline are usually free to expand, so the will tend to reduce end expansion and feed-in, as illustrated
force at the ends is zero. However, as the cumulative axial in Figure 4.
resistance increases with distance from the pipe ends, the
force can increase to a condition of ‘full-constraint’, as il- Figure 4 highlights how the different levels of axial resist-
lustrated in Figure 1. ance fundamentally modify the pipeline expansion behav-
iour. At low axial friction, the pipe is fully mobilised and
The fully constrained effective force presented in Figure 1 expands from its centre, reaching a maximum displacement
is the maximum effective axial force that can occur in a in excess of 2.5m. At high axial friction, in the absence of
pipeline. This fully constrained force drives the axial expan- buckling, there is no displacement over the central section
sion and structural response. The effective force is made up (the pipe is fully constrained) and the pipe expands only at
of the (true) force in the pipe wall and the pressure induced the ends to reach a maximum displacement of about 1m.
axial force. Since pressure and temperature vary along the With lateral buckling the expansion occurs over the whole
pipeline length, the fully constrained force also varies along length, but the direction changes as feed-in occurs towards
the length as the pipeline cools (with heat loss to the envi- the buckles, as well as towards the ends of the pipe, reduc-
ronment). This is shown in Figure 1 by the slight fall in the ing the end expansion to about 0.7m.
fully restrained force along the pipeline.
For the low axial friction condition, the pipe is not suscep-
The gradual increase in effective axial force, from zero at the tible to buckling. The design challenge is then controlling
free ends to full constraint, is due entirely to the cumulative the extreme end expansion and its susceptibility to pipeline
axial restraint provided by the seabed. The slope of this line walking. For the high axial friction condition, the pipe is
is equal to the axial resistance (force) per unit length, which susceptible to buckling over most of its length, and the de-
is typically modelled as being ‘frictional’, thus proportional sign challenge is controlling the severity of the buckles and
to the pipeline weight, W’µa. This force profile is funda- pipeline walking. Somewhere between these two extremes,
mental to the pipeline response. a pipeline design solution must be found.
The important influence of a reduced level (or lower bound 2.3 Response in a buckle: first load
estimate) of axial friction on the effective force profile is il-
The attraction of the lateral buckling design solution is in
lustrated in Figure 2.
the reduction in compressive axial force that occurs in the
In Figure 2 the maximum effective force in the lower bound pipeline. However, the buckling displacement results in sig-
134
Figure 1: Effective axial force in a straight pipeline Figure 3: Effective axial force in a short pipeline with lateral
buckles
Figure 2: Effective axial force for a range of friction in a Figure 4: Displacement along a short pipeline with lateral
straight pipeline buckles
nificant bending moment at the crown of the buckle. This Figure 6 shows that low lateral resistance is extremely desir-
moment can be very high and stresses in excess of yield are able since it produces a much lower bending severity in the
normal (although these are relatively limited in axial extent). crown of the buckle. It is also beneficial from a buckle for-
Thus, under the lateral buckling philosophy, although the mation point of view (as discussed later). However, a quite
axial force is reduced, the bending stresses are significantly small increase in lateral resistance can significantly increase
increased. This can ultimately lead to a local buckling fail- the design-limit usage factor, so that uncertainty in lateral
ure with associated deformation of the pipe cross-section. resistance is likely to compromise the design (and must not
At start-up, the soil ahead of the buckle crown is heavily be underestimated).
remoulded as a berm of soil builds up in front of the lat- If the loads within the buckle are too high, then the de-
erally sweeping pipe (Figure 5). The lateral displacements sign must seek to reduce these. The severity of the buck-
involved are very significant; displacements of 10 or 20 di- ling problem is driven by how much pipe feed-in must be
ameters are typical (in absolute terms, displacement in the absorbed by the buckle. Decreasing the distance between
range 2 to 10m has been observed in operating pipelines). buckles reduces the feed-in and hence controls the severity.
The length of pipe over which the lateral displacement oc- The lateral-buckling design solution relies on the regular
curs is typically between 100 and 300m (the lateral scale is formation of lateral buckles so that the load in each does
exaggerated in Figure 5).
not exceed design limits (local buckling, strain capacity or
The lateral resistance during pipe movement governs the fatigue). This approach defines the maximum allowable
level of curvature and bending stress in
the pipe. Lateral soil resistance is usually Figure 5: Side-scan sonar image of a lateral buckle
the largest uncertainty in designing for
lateral buckling and unfortunately has a
significant influence on the design limit
states, as illustrated in Figure 6. This fig-
ure shows the effect of the lateral resist-
ance, expressed as a friction coefficienta,
on the limit state utilisation within the
crown of the buckle.
135
buckle spacing, or virtual anchor spacing (VAS)b , which is each parameter and there is significant uncertainty over the
typically thousands of metres. true magnitude of each. Lateral breakout resistance is gen-
erally the largest uncertainty.
In Figure 6, it is shown that reducing the distance between
buckles (equivalent to a reduction in the VAS from 2.5 to If lateral buckles are initiated at regular intervals along the
2.0 km) reduces the loading. A greater reduction in VAS pipeline, the loads are effectively shared between buckle
will reduce the loading further, but this can make it dif- sites. However, the shorter the spacing between buckle ini-
ficult to ensure that the buckles form reliably. If the spacing tiators, the lower the probability is of buckles forming at
cannot be reduced further, then the only other approach each site as desired. Therefore, selecting a suitable spacing is
is to reduce the lateral resistance. In Figure 6 this would often a difficult design compromise between these compet-
be required if the lateral resistance is expected to be much ing requirements.
higher than an equivalent friction coefficient of 1.0; this
If there is no strategy to form buckles at regular intervals
would mean adopting alternative buckle control methods
along a pipeline, they will form randomly and (generally)
to achieve system integrity.
less frequently than if an initiation strategy is employed.
Consequently, the pipeline designer has two ways of con- This random formation may produce an acceptable design.
trolling the design problem: However, the large uncertainty in pipe-soil resistance means
• Decrease the distance between buckles that the pattern of formation is extremely challenging to
predict. In all but the most benign cases, it is impossible to
• Decrease the lateral resistance.
demonstrate that inherent buckling will occur to an accept-
Both of these solutions lead to the adoption of buckle ini- able level of reliability. In these cases, an engineered buckle
tiation techniques (Section 2.5). initiation technique must be adopted.
2.4 Response in a buckle: cyclic behaviour – influence of 2.5.1 Buckle initiation techniques
berms A number of methods are available to initiate buckling at a
In operation, as the pipeline experiences fluctuations in controlled spacing; these include:
pressure and temperature, the pipe cycles back and forth • Snake-lay (Figure 8), where the pipe is laid with regular
across the same patch of seabed. Surface soil, swept ahead tight-radius route curves
of the pipe on each cycle, then builds up into berms at the
• Vertical upset (Figure 9), where rock or pipe sleepers are
extremes of pipe displacement. Subsequent consolidation
used to raise the pipe off the seabed
increases the strength of the soil berms after disturbance.
Pipe feeds axially into and out of the buckle with each cycle. • Local weight reduction (Figure 10), where distributed
On unloading, the pipeline attempts to return to the as-laid buoyancy or additional insulation coating is applied, or
position but is prevented from doing so by both axial and weight coating is reduced or removed.
lateral pipe-soil resistance. For the snake-lay method the lateral restraint is governed
by the lateral breakout resistance (discussed later), while the
Without soil berms, numerical modelling with lateral re-
OOS is defined by the tightest curvature for which that
sistance calculated using a single friction coefficient shows
installation contractor can install. This curvature is gov-
that buckles will grow in amplitude with each cycle. In real-
erned by two parameters: lateral breakout resistance and lay
ity, soil berms restrict the growth of the buckle so that cy-
tension. In reality, tighter than average design curvatures
clic displacements remain almost constant over a number of
usually occur due to installation dynamics, but these are
cycles. This increased restraint in the buckle causes higher
difficult to predict in advance of installation.
stress ranges in operation, leading to a higher level of fa-
tigue damage. This effect is illustrated by the example in The weight reduction and vertical upset solutions both
Figure 7. provide a vertical OOS to encourage buckle formation,
although this is perhaps easier to quantify in the vertical
This example clearly illustrates the influence of soil berms,
upset solution. Both techniques also benefit from the de-
although much uncertainty remains in predicting the level
sign by reducing the pipe weight or lifting the pipe off the
of restraint provided by berms. Currently, this uncertainty
seabed, reducing contact with the seabed and therefore de-
is addressed in design by assuming a range of berm resist-
creasing the lateral resistance across the buckle. This reduc-
ance to assess the sensitivity on allowable loading.
tion in loading means that fewer lateral buckles are needed
2.5 Buckle initiation to share the load. Therefore, buckles can be spaced further
Buckle initiation is governed by three parameters: (i) the apart, which increases the reliability of buckle formation.
effective compressive force in the pipeline, which is a func- The benefit that these techniques bring has resulted in their
tion of axial resistance; (ii) out-of-straightness (OOS) fea- use on a number of challenging deepwater projects (e.g.
tures; and (iii) lateral breakout resistance. Although only Harrison et al.3).
three parameters are involved, a number of factors feed into 2.5.2 Modelling issues
Special consideration is required in zones where the local ef-
b VAS is the distance between virtual anchors that form to each side of fective pipe weight changes due to loss of contact with the
the buckle, usually equivalent to the spacing between lateral buckles. seabed, for example at sleepers or at the start of a distributed
136
137
Figure 11:
Influence of
touchdown
load and
lateral
resistance on
buckle shape
Figure 12: Pipeline walking due to seabed slope for a range of Figure 13: Effective force distribution along a walking pipeline
axial friction coefficients restrained by an anchor
138
thermal transients can lead to buckle growth over a number buckle formation. This once again illustrates the need to
of load cycles. Lateral buckles are a significant influence on bound soil resistance; there is no conservative extreme.
the walking response, which highlights the need to under-
stand this interaction. Currently the interaction between
2.8 Key influences of pipe-soil resistance
walking and buckling is assessed on a case-by-case basis by The preceding examples demonstrate the significant influ-
FEA in detailed design. ence that pipe-soil interaction has on pipeline design. The
axial pipe-soil resistance affects
The example in Figure 12 illustrates the influence of pipe-
• The maximum axial effective force in the pipeline
line length (or buckle spacing) and axial friction, µa, on
pipeline walking for a range of seabed slopes. In many cases, • The effective force available for reliable buckle initiation
as shown in Figure 12, a high level of axial resistance can • Pipe-end expansion and feed-in to lateral buckles
reduce walking significantly. Pipeline length (or section • Pipeline walking, including the rate of walking and the
length) is a significant parameter in the assessment of pipe- propensity to walk.
line walking, but axial resistance has a greater influence. If
the pipe weight is sufficiently high, or the fully constrained Similarly, lateral pipe-soil resistance affects
force sufficiently low, walking can stop altogether. • The lateral instability that is required for buckle initiation
• Route curve stability under axial tension
2.7 Control of pipe walking
When walking is considered unacceptable, it becomes nec- • The lateral buckle bending load at large displacements
essary to anchor the flowline. Hold-back suction anchors • The cyclic lateral buckle loading due to soil berms.
are commonly employed to control pipeline walking. In The influence of pipe-soil resistance on these various issues
most cases these anchors allow expansion towards the an- is often conflicting. For example, high lateral resistance is
chor but prevent the end of the pipe pulling back much beneficial for curve stability, but detrimental to the bending
beyond the installed position; this ensures that the effective loads developed in a buckle. High axial resistance is benefi-
compression in operation is not increased. cial for controlling feed-in to a buckle, but increases the sus-
However, as walking commences the anchor loads up, until ceptibility to buckling in the first place. Consequently, it is
after some tens of cycles, the anchor experiences the full shut- not possible to define a safe resistance. Instead, it is necessary
down tensile load. This restraining anchor load induces ten- to bound the pipe-soil resistance and ensure that the design
sile loads along the length of the pipeline. The force profile at is acceptable throughout the potential design envelope.
shutdown and the restraining anchor load is then defined by
the level of axial resistance, as shown in Figure 13. 3. Current Approach to Modelling Pipe-Soil
Interaction
In the example shown in Figure 13, walking is caused by
the presence of a steel catenary riser with a nominal tension 3.1 Soil resistance models
of 1MN at one end of the pipeline (at KP 0). At this tension The interaction between the pipe and the seabed is incor-
an anchor is required at the opposite end of the pipeline porated into the structural analysis of a pipeline – which is
with a capacity of 3.5MN. The effective tension within the usually finite element-based – by attaching ‘spring-slider’
pipeline exceeds 4.5MN at KP 3.0. This is a significant lev- elements at intervals along the pipe. This approach is analo-
el of tension, which introduces a further design challenge: gous the t-z and p-y load transfer methods of analysing pile
route curve stability. The concern is that this tension could response. The most basic pipe-soil elements provide a bilin-
be sufficient to cause instability of the pipeline – through ear elastic–perfectly plastic response in the axial and lateral
ratcheting lateral displacement – at the route curves. This directions. However, in order to capture the more advanced
instability can pull out the curve, allowing further pipe to effects of interaction, particularly the large displacement
walk, until the curvature is small enough to be stable. behaviour, it is necessary to introduce subroutines in which
the element response is modified to account for brittle brea-
The minimum stable radius of curvature is governed by the kout behaviour and cyclic berm growth.
minimum lateral resistance that the soil can provide. If the
lateral resistance is low, then this radius of curvature may be The most simple bilinear models involve a limiting value
so large as to compromise field architecture. This issue can of axial or lateral pipe-soil resistance, which is calculated
lead to major field layout changes and should be addressed using a simple friction law (Coulomb friction) linking
as early as possible in front-end engineering design. The in- the effective pipe weight to the maximum available resist-
stability of route curves has led in the past to some radical ance. A suitable axial and lateral friction coefficient can be
changes to field architecture to eliminate route curves or used successfully for some flowline design functions (e.g.
increase their radius. In some cases, the production facility simple stability calculations or end-expansion) and can be
has been relocated, in others the flowline has been split into employed in conceptual evaluation of lateral buckling, if
two straight sections with a mid-line tie-in, which has seri- treated with care. However, a single friction coefficient is
ous capital expenditure (CAPEX) implications. not appropriate in detailed numerical modelling design
for lateral buckling, particularly for large-amplitude lateral
Although a low level of lateral resistance is a disadvantage movement, where a frictional model represents an oversim-
for stability of route curves, it is an advantage for reliable plification of the behaviour.
139
Nevertheless, it is common to model many tens of kilome- contact area, which affects the axial resistance, and the pas-
tres of pipeline within a single FE analysis. It is therefore sive soil resistance against the pipe, which affects the lateral
necessary to keep the pipe-soil interaction models relatively break-out force. Heave of soil against the shoulders of the
simple. It is not feasible to model the soil domain around pipe raises the contact area above the value for a wished-in-
the pipe along the entire pipeline length, and so it remains place pipe at the same embedment. Figure 14 illustrates the
necessary to encapsulate the pipe-soil behaviour into the initial embedment of a pipeline.
response of a single node. Non-linear force displacement re-
3.3.2 Embedment uncertainties
sponses are therefore used to represent pipe-soil behaviour.
The current model and background for prediction of em-
These responses are usually presented as an equivalent fric-
bedment has already been published by the SAFEBUCK
tion coefficient – that is the maximum resistance divided by
JIP5. However, pipeline embedment is notoriously difficult
the submerged weight of the pipe. This coefficient is then
to predict. The values of as-laid embedment predicted using
updated automatically throughout the analysis to simulate
this and other models6 is generally found to be substan-
the underlying behaviour.
tially exceeded in practice. This discrepancy is because the
A friction coefficient is an unusual concept to apply to the touchdown load is actually cyclic in nature, due to dynamic
limiting resistance of a clay, which is usually characterised effects including vessel heave and hydrodynamic loading of
by an undrained strength. However, this is a convenient the pipe catenary. This increase in embedment is captured in
way of presenting the data for input into design using ana- design by multiplying the embedment found from a static
lytical or finite element based models and remains compa- embedment calculation by a ‘dynamic embedment factor’,
rable with historic approaches for pipe-soil interaction. This which is a function of the lay vessel dynamics, lay rate, lay
terminology should not, however, be taken to indicate that tension, pipeline configuration and environmental loading
lateral pipe behaviour is purely frictional; the limiting lat- during installation. However, these secondary variables are
eral and axial resistance is not solely dependent on the pipe not known prior to installation. Over-embedment, due to
weight, but is significantly influenced by the embedment the uncertain and variable influence of the laying process, is
and soil conditions. the main source of uncertainty in the prediction of pipeline
embedment and breakout resistance.
3.2 Interaction between axial and lateral resistance
Axial and lateral resistance should be treated as independ- The dynamic embedment factor carries significant uncer-
ent parameters in design for any given set of soil conditions. tainty. Comparison between field experience for soft clay
Since the axial resistance response is global – affecting long seabeds with shear strengths of 2 to 4kPa and predictions
sections of pipeline – the total axial resistance is negligibly from static analysis (based on the as-laid pipe weight mul-
influenced by the localised mobilisation of lateral resistance. tiplied by a suitable stress concentration factor at the TDP)
give a typical dynamic embedment factor in the order of
This approach can provide a low assessment of axial resist- 1.0 to 3.0. On stiff clay seabeds with shear strength exceed-
ance where axial and lateral interaction is possible locally, ing 100kPa, the factor can be in the order of 5.0 to 8.07.
at specific regions of a lateral buckle away from the crown. Further work is ongoing in this area, but evaluation has
Axial resistance involves shearing at the pipe-soil surface, been hampered by the limited availability of suitable data
whereas lateral resistance involves failure within the soil from installed pipelines. Simulation of the lay process is
ahead of the pipe. These mechanisms have minimal inter- possible in model scale testing, but the input parameters
action. However, in drained conditions, lateral loading on related to the motion of the vessel and hydrodynamic load-
a pipe will increase the effective stresses between the pipe ing of the lay catenary are difficult to assess.
and the soil, raising the available axial resistance. Therefore,
treating axial and lateral resistance as independent quanti- A further complication related to pipe embedment is the
definition of the soil surface, which forms the datum for
ties is likely to underestimate the axial resistance at specific
observed levels of embedment on installed pipelines. In situ
regions of the lateral buckle where drained axial and lateral
test data of shear strength often does not register an increase
displacement occurs simultaneously; this approach is usu-
in resistance at the observed soil surface level, even when
ally conservative.
using a sensitive T-bar penetrometer. This can either be be-
Caution is necessary with some FEA software that includes cause the T-bar has already penetrated the surface during
built-in soil elements. These often combine axial and lateral placement of the seabed frame, leading to some depth offset
resistance using an elliptical combined yield surface. This is between the visual observations and the in situ test data.
non-conservative as it results in reduced lateral resistance Alternatively, the visible surface layer may be so weak that its
when axial resistance is mobilised. In short, any combina- significance can be overlooked. In deep water, a very weak
tion of lateral and axial resistance causes a reduction of lim- layer of soil about 70 to 100mm thick is commonly found
iting resistance on both axes. at the surface and its presence is also commonly ignored.
However, any visual assessment of pipeline embedment
3.3 Pipeline embedment from post-installation surveys will include this weak layer.
3.3.1 Definition of pipeline embedment
Pipe embedment is defined as the depth of penetration of 3.4 Axial pipe-soil interaction
the invert (bottom of pipe) relative to the undisturbed sea- 3.4.1 Definition of axial resistance
bed. Pipeline embedment therefore influences the pipe-soil Axial pipe-soil resistance is often modelled using a simple
140
141
ing apparatus operating at usual effective stress levels. The dif- Breakout lateral resistance: A significant peak in lateral re-
ference can be critical to pipeline design, as demonstrated by sistance occurs when the pipe first moves laterally (point 1
the preceding examples. Friction coefficients as high as unity in Figure 17). The size of this peak depends on the level of
(corresponding to friction angles in the range 40–50º) have initial embedment (Figure 18). Breakout resistance is criti-
been measured for soft deepwater clays at low stress levels. cal to the initiation of buckles and is also associated with
During fast axial pipe movement, the apparent friction lateral stability at route curves, when a pipeline can become
coefficient that should be used in the ‘beta’ effective stress unstable under tensile load.
model may increase or decrease depending on whether neg- Suction release: The breakout force includes the effect of suc-
ative or positive excess pore pressure is generated by shear- tion between the pipe and soil, which is released (point 2 in
ing at the pipe-soil interface; this tendency depends on the Figure 17) as a crack forms between the back of the pipe and
current over-consolidation ratio (OCR) of the soil. the soil. Suction release has little influence on subsequent
Since undrained (fast) and drained (slow) responses gen- behaviour. For very slow loading, this crack may form prior
erally provide very different values of axial resistance, it is to the full mobilisation of breakout resistance. In this case,
important to consider the interaction between the available the breakout load will be smaller and no suction release will
resistance and the resulting pipe velocity. An analysis based be evident. However, the initial breakout of the pipeline is
on drained parameters that yields high pipe velocities is usually a fast event, and some level of tension is sustained at
meaningless, and vice versa. Limited field data is available the rear of the pipe, leading to a two-way failure mechanism
on pipe velocities during operation, but one analysis exam- involving soil ahead of and behind the pipe.
ple is illustrated in Figure 16, which shows some typical Residual lateral resistance: Once the pipe has started to
pipe velocities during a shutdown and restart cycle. move, the residual lateral resistance drops or rises to a
The velocities shown here are generally quite low and only steady-state value at large displacements (point 3 in Figure
exceed 0.2mm/s at locations close to the pipeline ends (KP 0 17) as the pipe pushes a berm of surface soil across the sea-
and KP 7.5) and then for quite short durations. The impli- bed. The residual resistance during the first loading cycle
cation is that the axial response is likely to remain drained, controls the lateral displacement at which the first buckle
but this is by no means certain. stabilises, defining the initial shape of the lateral buckle and
the peak bending stress in the pipe.
The important transition from undrained to drained condi-
tions is not well understood. While dissipation of the excess Cyclic lateral resistance: As the pipe undergoes cycles of
pore pressure from laying is expected to occur relatively shutdown and restart during operation, the pipe cycles
quickly, it does not prevent undrained conditions from be- back and forth across the same patch of seabed. Surface soil,
ing re-established under conditions of rapid or large pipe swept ahead of the pipe on each cycle (points 5 and 8 in
displacements. However, re-consolidation after axial move- Figure 17), builds up into berms at the extremes of the pipe
ment is likely to be quicker than after embedment, since displacement (points 6 and 9 in Figure 17). These berms
only a small region of soil at the pipe surface is sheared offer significant resistance to pipe movement and define the
during axial motion. This topic is an area of ongoing re- shape of the buckle in operation. Suction release also oc-
search under Phase II of the SAFEBUCK JIP to improve curs as the pipe pulls away from each berm (points 4 and
the fundamental understanding and better define the most 7 in Figure 17). The release usually occurs as a crack forms
appropriate axial resistance models and test methods. between the back of the pipe and the soil, but during steady
cycling a lump of soft remoulded soil can adhere to the rear
3.5 Lateral pipe-soil interaction
of the pipe, so a more gentle suction release occurs at some
3.5.1 Definition of lateral resistance distance from the berm. Inclusion of the suction release
While lateral breakout loads have been the subject of much behaviour in the pipeline analysis has little influence on
research and published papers, until recently there has been resulting response. However, the mechanism of crumbling
little guidance on modelling lateral resistance for the large soil from the fixed berms falling back into the trench creates
cyclic displacements of typically 10 or 20 diameters experi- a bowl-shaped trench. The shape of this trench means that
enced in lateral buckling. the cyclic residual resistance is initially low as the pipe slides
The current models and background for prediction of brea- down one side of the trench, and then gradually increases as
kout and residual resistance have already been published by the pipe slides horizontally and then rises up the other side
the SAFEBUCK JIP5. The simplified response is presented of the trench before meeting the berm.
as an equivalent friction coefficient, H/V, where H and V are 3.5.2 Lateral resistance uncertainties
the horizontal and vertical load on the pipe per unit length,
Many of the tests carried out to date have evaluated the
respectively.
lateral resistance of quite light pipes, representing wet-in-
A simplified force-displacement lateral-friction response, sulated or gas pipelines; uncertainty remains over the long
based on model tests, is illustrated in Figure 17. term cyclic response of heavier pipelines, such as water-
filled lines or pipe-in-pipe systems.
Four stages of lateral pipe-soil interaction should be con-
sidered in design, (i) breakout resistance (ii) suction release The key to successful modelling of the cyclic behaviour is
(iii) residual resistance and (iv) cyclic lateral resistance. to capture the reaction force from the static berms. Once
142
the static berms are formed on the first cycle, they quickly
become established and provide a significant resistance to
increasing cyclic amplitude of the pipe, although the pipe is
likely to encroach into the berm slightly during the first few
cycles, until the berm gains sufficient strength.
This behaviour is confirmed by the soil berms established
on operating pipelines where lateral buckles have occurred.
Figure 19 shows a typical cross profile at the crown of a
lateral buckle, where the as-installed and maximum excur-
Figure 17: Schematic of lateral force-displacement response
sion positions are inferred from survey data. For compari-
son, Figure 20 shows the pair of soil berms established by
a laterally sweeping pipe section during a centrifuge model
test. This test was part of a programme of lateral pipe-soil
interaction tests conducted at The University of Western
Australia (UWA) as part of SAFEBUCK Phase II9. By ana-
lysing images captured during this test programme using
particle image velocimetry (PIV)10, the soil deformation
mechanisms during lateral sweeping could be identified.
Figure 18: Breakout response based on embedment level
Figure 21a shows the displacement field as a pipe travels
horizontally on soft clay. The growing berm ahead of the
pipe primarily deforms by shear along the base, with some
soil being pushed into the berm close to the pipe invert.
To illustrate the significance of the berm resistance, the
berm resistance recorded during these tests after 50 cycles
of fixed amplitude was typically five times higher than the
residual lateral resistance during the first cycle.
Models for pipe-soil interaction that are used to simulate
large-amplitude cyclic motion should therefore capture the
Figure 19: Typical cross profile at crown of lateral buckle on effect of soil berms, which tend to constrain the buckle and
an operated pipeline can increase the loading at the apex of the buckle. The in-
143
crease in lateral resistance at the soil berm is a function of: This model for berm behaviour mimics the mechanisms
• The cumulative lateral pipe displacement, defined by the seen during experiments in a simple form. The model can
number and size of previous sweeps, and the vertical em- be used to assist in the interpretation of physical tests and
bedment during these sweeps: these distances govern the allow these to be generalised to arbitrary sequences of lateral
volume of soil pushed out from the trench and added to movement, providing values of berm resistance to be used
the berms; the size of the berm then governs the resist- in FEA. A future development would be to incorporate this
ance required to push through it model directly into the FEA, allowing berm resistance to
evolve automatically during the analysis.
• The proportion of berm resistance mobilised: as the pipe
moves into the berm, the resistance rises approximately 3.6 Advances in pipe-soil modelling methods
linearly until the full berm strength is mobilised The current pipe-soil response models are characterised by
• The rate of displacement into the berm, hence wheth- a piecewise linear response defined by equivalent friction
er the response tends towards undrained or drained coefficients (Figure 17). These provide a practical way to
behaviour capture the non-linear response demonstrated by pipe-soil
• The remoulded strength of the clay within the berm, de- interaction testing. However, this approach can be cumber-
fined by repeated cyclic loading and consolidation of the some, requiring the definition of a non-linear response for
berm. each pipeline configuration to be inserted into a user-sub-
routine that can update the equivalent friction coefficients
The amplitude of the cyclic sweeps contributes to how
for each section of the pipeline, as contact loads and dis-
quickly the berm is established during the initial cycles.
placements change throughout the analysis. These routines
However, during sweeps of constant amplitude, the berm
must be incorporated into the FEA in such a way that re-
resistance approaches a limiting value, as after the first few
versals of direction or non-constant pipe touchdown loads
sweeps, any fresh soil added to the berm on a given sweep
are handled competently.
falls back into the trenchc.
An alternative to the equivalent friction coefficient ap-
FEA models should include the berm resistance based on
proach is to use yield envelopes in vertical and horizontal
the actual vertical load, V, exerted by the pipe caused by the
load space that bound the allowable combinations of load
formation of soil berms at lateral buckles. This includes lat-
for a given pipe embedment. Yield envelopes are well es-
eral buckles that occur on the seabed (on-bottom) or those
tablished as an approach to describe the bearing capacity
that occur at sleepers or distributed buoyancy sections. In
of shallow foundations under combined load. By invoking
every case, soil berms provide lateral restraint and strongly
plasticity theory and an appropriate flow rule, yield enve-
influence the cyclic response of the pipe. Berm resistance is
lopes can be used to create a model for the general load-dis-
generally incorporated into FEA by modifying the equiva-
placement response of a pipe or foundation12, 13, 14, 15. These
lent lateral coefficient as the pipe approaches the berms, ac-
are known as ‘force-resultant’ plasticity models.
cording to a user routine. In recent projects, the magnitude
of the berm strength and the rate at which it is mobilised These models are similar to the work-hardening plasticity
as the pipe approaches the berm has been estimated from models used to describe the behaviour of individual ele-
model tests. Currently, a range of values of berm strength ments of soil – such as Cam clay – and are able to capture
is employed in design to address the sensitivity to berm the brittle behaviour of over-penetrated pipes, but with a
strength during cyclic loading. smooth response rather than the abrupt changes in stiffness
given by the piece-wise linear simplification illustrated in
Prediction of berm resistance is not well understood and
Figure 17.
much work is ongoing in this area. To provide a more sys-
tematic basis for predicting berm resistance, a methodology A force-resultant plasticity model for small displacement be-
that links the erosion of material from the trench base to haviour of a pipe on carbonate sand is described by Zhang
the growth of the berm, and therefore the berm strength, et al.13 To capture the behaviour needed for lateral buck-
has been proposed11. ling design, this type of model must be extended to include
the large displacement effects described previously – for
This approach operates by considering conservation of vol-
example, using the berm modelling approach described in
ume between the soil berms and the trench between the
Section 3.5.2, in which the active berm area is an additional
berms, which is being ‘ploughed’ by the sweeping pipe.
hardening parameter.
The thickness of material ploughed by the pipe during
each sweep, tplough (Figure 21b) is calculated from the pipe The SAFEBUCK JIP is supporting ongoing research into
weight. From tplough, the growth of the ‘active’ berm ahead force-resultant plasticity models for pipe-soil interaction.
of the sweeping pipe can be calculated. The volume of ma- The aim is to develop a force-resultant model for pipe-soil
terial ploughed during each cycle is added to the fixed berm interaction that captures the necessary effects for use in lat-
at the end of each sweep, and the total berm volume is used eral buckling design. This research is currently focused on
to calculate the berm resistance. identifying the basic mechanisms involved – using model
tests and theoretical analysis – then distilling this behaviour
c The trench described here is one which is created by cyclic pipe dis- into appropriate extensions to existing force-resultant plas-
placement and is not the result of a pipeline trenching operation. ticity models.
144
4. Soil Data and Model Testing bed frame – to measure axial and lateral pipe-soil resistance
4.1 Soil properties required for design in situ is being developed by Fugro (SMARTPIPE). This
programme is ongoing and no test data is currently in the
The pipe-soil force-displacement responses used in the de-
public domain.
sign of a flowline should cover the full range of anticipated
shear strength and submerged unit weight at the site. The 4.2 Laboratory model test programmes
lower bound soil strength profile results in greater em- Lateral buckling and pipeline walking behaviour is extreme-
bedment, which leads to high lateral breakout resistance, ly sensitive to pipe-soil interaction, and there is much uncer-
whereas the upper bound strength profile results in shallow tainty associated with models for pipe-soil interaction in soft
embedment and a reduced breakout load. clay. In addition, the basic phenomena involved are not fully
For lateral buckling and pipe walking, it is important to understood, and detailed numerical modelling is not yet ca-
bound behaviour. Upper and lower bound values of soil pable of fully representing the response. This has led to the
resistance are both important. It is therefore essential that a need for project-specific model test programmes. These are
detailed soil investigation is conducted, providing detailed often undertaken at full-scale (or half-scale), which requires
measurements of soil strength, particularly in the upper 1m more time than project time scales normally allow.
of soil. Uncertain soil conditions results in a large range The aim of the SAFEBUCK pipe-soil interaction test pro-
between upper and lower bound behaviour, which will in- gramme was, therefore, to provide generic guidance for
crease the design challenge and potential mitigation costs; future projects and improve current understanding. Data
in some cases, it may preclude the ability to demonstrate a from these tests have been augmented by project-specific
robust design solution. tests, donated by JIP participants, to create a database that
It is therefore important to define a range of soil parameters now spans a wide range of pipe diameters, weights and
that includes realistic upper and lower bound values with soil conditions; this database is the source on which the
depth. It is also important not to assume a single value or an SAFEBUCK models published previously5 are based.
unrealistically narrow band of response, as this could lead However, much uncertainty remains and more testing is
to an unsafe design. ongoing. Many recent projects have carried out small- and
Key soils data required for design includes: large-scale tests in specialist laboratories. These tests have
concentrated on the low shear strength and low-permeabil-
• Shear strength profiles measured in situ to about 1.0m
ity soft clays that dominate the deepwater regions of the
depth, preferably using a T-bar penetrometer
world where on-bottom lateral buckling solutions are cur-
• Sensitivity, preferably based on first-in/first-out and cy- rently being employed.
clic T-bar response
Large-scale model tests typically use 3–5m3 of soil col-
• Submerged soil unit weight
lected from the field and re-consolidated in a large tank.
• Laboratory test data using disturbed soil samples and The model pipe is placed into the soil bed and swept axi-
specialist testing methods, particularly the assessments ally and laterally to assess the response. A number of such
of pipe-soil interface friction at appropriate low stress project-specific tests have been carried out at Cambridge
levels. University (UK) and the Norwegian Geotechnical Institute
Full-flow penetrometers, such as the T-bar, currently offer (NGI)18 over recent years. The practical limitations of such
the best method of assessing the strength of soft near-sur- tests include the time taken to reconsolidate the soil in the
face sediments for pipeline design. A T-bar penetrometer tank and the maximum pipe diameter of typically 12in.
comprises a cylindrical bar, typically 40mm in diameter Nevertheless, such tests provide detailed measurements of
and 250mm long, and may be viewed as a small scale seg- pipe-soil interaction in near full-scale conditions.
ment of pipe. Measurements of T-bar penetration resistance
Large-scale lateral tests have been supplemented by
can be converted to undrained shear strength in a more
small-scale centrifuge tests, which were initiated by the
straightforward manner than CPT tip resistance, and the
SAFEBUCK JIP at Cambridge University19 and have since
higher projected area of the T-bar leads to reduced noise on
been carried out by a number of international research fa-
the force measurements16.
cilities. A suite of centrifuge tests for the JIP was recently
To assess axial resistance, it is useful to conduct laboratory completed at UWA, including a set of tests that used im-
tests to assess friction angles and remoulded strengths at the age analysis9 to better understand the failure mechanisms.
very low effective stress levels relevant to pipelines. Specialist The advantage of conducting lateral pipe-soil model test-
equipment has been designed to measure pipe-soil interface ing in a centrifuge and the reason that this type of testing
resistance at these low stress levels, including the tilt table was initiated by the JIP are that only a small quantity of
device at the University of Texas at Austin17 and the Cam soil is required and the testing programme is much shorter.
shear device used at the University of Cambridge, discussed Centrifuge model tests, using a suitably advanced actuation
further in the next section. system, can also replicate the necessary sequences of pipe
The scopes for specialist laboratory and in situ pipe-soil test- laying and lateral motion.
ing are being developed and refined on a project-by-project The most advanced centrifuge model testing facilities can
basis. A specialist pipe-soil testing rig – mounted on a sea- replicate the complete load and displacement patterns im-
145
146
147
13. Zhang J, Stewart DP, Randolph MF. (2002). Modeling of shal- 18. Langford TE, Dyvik R and Cleave R. (2007). Offshore pipeline
lowly embedded offshore pipelines in calcareous sand. Proc. J. of and riser geotechnical model testing: practice and interpretation
Geotechnical and Geoenvironmental Engng. 128(5):363–371. OMAE2007-29458.
14. Martin CM and Houlsby GT. (2001). Combined loading of 19. Cheuk CY and Bolton MD. (2006). A technique for modelling
spudcan foundations on clay: numerical modelling. Géotechnique the lateral stability of on-bottom pipelines in a small drum cen-
51(8):687–699. trifuge. Proc. Int. Conf. on Physical Modelling in Geotechnics,
15. Nova R and Montrasio L. (1991). Settlements of shallow founda- Hong Kong.
tions on sand. Géotechnique 41(2):243–256. 20. Baker JHA, Bruton DAS, Matheson IC. (2006). Monitoring
16. Randolph MF. (2004). Characterisation of soft sediments for off- and effective integrity management of laterally buckled flowlines
shore applications. Proc. 2nd Int. Conf. on Site Characterisation, in deep water. OTC 17932, Proc. Offshore Technology Conf.,
Porto, 1, 209–231. Houston, USA.
17. Najjar SN, Gilbert RB, Liedtke EA and McCarron W. (2003).
Tilt Table Test for Interface Shear Resistance Between Flowlines
and Soils. Proc. Conf. on Offshore Mechanics and Arctic
Engineering.
148