Pipe-Clamping Mattress to Stop Flowline Walking
Sebastiaan Frankenmolen and Sze-Yu Ang, Shell Global Solutions; Ralf Peek, Peek Solutions; Malcolm Carr and
Ian MacRae, Crondall Energy; David White, University of Western Australia; Jeffrey Rimmer, Shell Philippines
Exploration and Production
Copyright 2017, Offshore Technology Conference
This paper was prepared for presentation at the Offshore Technology Conference held in Houston, Texas, USA, 1–4 May 2017.
This paper was selected for presentation by an OTC program committee following review of information contained in an abstract submitted by the author(s). Contents of
the paper have not been reviewed by the Offshore Technology Conference and are subject to correction by the author(s). The material does not necessarily reflect any
position of the Offshore Technology Conference, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Offshore Technology Conference is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of OTC copyright.
Thermal gradients from a heating front travelling down a flowline at start-up can cause a flowline to walk
much like a worm creeps by repeated contractions and expansions of its body. To stop this for the Malampaya
flowline, pipe-clamping mattresses (PCMs) were invented, developed, and deployed within a period of 12
months. The objective of this paper is to share the knowledge and experience from this novel but effective
solution to mitigate pipeline walking.
PCMs provide a cost-effective alternative to rockdump or conventional mattresses to axially restrain
a pipeline at a location chosen so that the required restraint capacity is minimized. They are inspired by
conventional mattresses and bear some similarity to them, but they are designed so that the weight of the
mattress acts to clamp the pipeline with a high leverage. Thus 100% of the weight of the mattress is effective
in generating axial friction with the seabed. This solution can be applied at any point along the line (chosen
to minimize the required resistance) without requiring flanges or collars on the pipeline.
From the most recent survey results 15 PCMs with a dry weight of around 9 tons per PCM, plus 7
tons for the logmat installed over every PCM appear to be effective to stop the walking of the Malampaya
flowline. This performance is as expected from extensive analysis (FE and otherwise) to reproduce the
observed walking behavior prior to restraining, to estimate the required restraint capacity, and to estimate
the resistance provided by the PCMs.
This paper describes the PCM, the clamping forces they generate by leveraging the weight of the PCM and
logmats installed over them, and how the friction generated with the soil is estimated from interface shear
tests on samples collected from the site, considering cyclic pore pressure generation and dissipation effects.
It also briefly covers FE analyses to reproduce the observed walking behavior, and determine the required
restraint capacity, the PCM fabrication, installation, and monitoring of the post-installation performance.
Pipelines and flowlines can walk or creep, a bit like a worm does by axial contractions and expansions
of its body. This phenomenon has become well-known [Konuk (1998), Tornes et al. (2000), Peek (2002),
Carr et al. (2006), Bruton et al. (2010), Carneiro et al. (2014)]. Restraints to prevent walking can be large
and expensive. They are often installed because finite element (FE) simulations indicate the flowline could
walk, if not restrained. Walking was less well known around 1996-97, when the twin 16-inch diameter
Malampaya flowlines were designed. The flowlines carry multi-phase gas with some condensate from the
Malampaya and Camago fields in around 800m water depth, over 28.3km to a platform in about 40m water
depth. The condensate is then separated from the gas and stored for periodic off-loading to tankers, and the
gas is taken via the 500km-long Gas Export Pipeline to Batangas, near Manila, on the Isle of Luzón, where
it covers about 30% of the power demand on the island. No special provisions were made at the design stage
to prevent walking. Walking did occur, however, with about 28mm displacement per shut-down and restart
cycle, which had accumulated to a displacement of 1.8m, at the Pipeline End Structure (PLES). This started
raising concerns over the integrity of the jumper connection from the PLES to the manifold, where the
production from 11 wells is collected and distributed over the two flowlines. (Only one of the two Flowlines
experienced significant amount of walking. The other was not in use for much of the time.)
This paper first describes the walking phenomena as observed for the flowline, including the interactions
with lateral buckles, and the finite element simulations of them. Then it describes a novel way of restraining
such walking, by use of pipe-clamping mattresses (PCMs). It describes circumstances and considerations
leading to inventing the PCMs, their design, fabrication, trials, and deployment, all completed within a
period of less than 12 months. Finally, performance monitoring results since the time of installation in
December of 2015 are also included. A patent application has been filed for the PCMs.
Description of the Walking Behavior and FE Simulations Thereof
Walking is caused by cyclic expansions and contractions of the pipeline, together with a mechanism that
makes the movement at each cycle in one direction a bit more than in the other. In this case, the contraction
at the inlet upon cooling each time the line is shut down is a bit more than the expansion displacement upon
restart. Most recently (before the remediation by the PCMs) the difference was about 28mm. Over almost
200 cycles, this had lead a total displacement of the inlet of about 1.8m in the downstream direction (this is
in the opposite direction to the natural expansion of the pipeline, which was anticipated to be about 2m in
the upstream direction). Evidence of walking could be seen from marks in the seabed left by pile guides on
the PLES (see Figure 1) and was further monitored by flags, as shown in Figure 1.
Figure 1—Evidence of Movement of the PLES at the inlet of the flowline, and flags used to measure displacements
The PLES includes an upper moving part with flowline and spoolpiece connection and a lower part
resting on the seabed. the upper part can slide on rails on the lower part, so that the usual thermal expansions
and contractions can be accommodated without the PLES sliding over the seabed each time. However, the
walking made the slider bear hard against its end, upon contraction. Thus, the flowline started to pull the
entire PLES with it, resulting in the displacements shown in Figure 1.
In this case the main driver of walking are thermal gradients. The worst-conceivable thermal gradients
involve a heating front travelling down the line upon start-up [Peek (2002)]. Thermal gradients can be
shown to be equivalent to applying an axial force proportional to the thermal gradient acting in the direction
of decreasing temperature. Typically, the start-up fronts create thermal gradients travelling down the line
that are much larger than those at steady-state operations. To capture this, it is essential to first perform
multiphase flow simulations of the start-up process, in order to obtain the correct inputs to the FE analysis
for temperature, pressure, and density of contents as a function of time t and location x along the flowline.
Matters are further complicated by the bathymetry and liquid accumulation: The inlet is at a low point.
This means that gravity tends to drive walking in the opposite direction to that in which it occurred, i.e. in
the upstream instead of downstream direction. Both FE simulations and observations at the PLES indicate
that walking in the upstream direction is indeed what happens in the first few cycles (see Figure 5), but
after about 50 cycles, the walking changes direction, heading downstream, apparently driven more strongly
by the thermal gradients (acting in the downstream direction), than by the gravity (acting in the upstream
direction). Gravity loads are not the only effect of the bathymetry, however. It also causes accumulation of
liquids at the inlet upon shut-down. Upon start-up, this large slug is pushed down the line. This influences
the thermal gradients. Also, the associated changes in weight influence the axial pipe-seabed interaction,
with significant impact on the walking behavior.
The walking pipeline must go somewhere. For a short flowline, the entire length can walk. However, for
the 28.3km-long Malampaya flowline, the walking feeds mostly into lateral buckles. The first one occurs
at around KP 1 (i.e. 1km from the inlet), and a total of 8 buckles are observed with the last one 8km from
the inlet. The spacing between lateral buckles ranges from 0.5 km to 1.5 km. The first buckle occurs at a
local high point that acts as a trigger. That buckle has also developed very large lateral displacements (~14
m) together with a very large wavelength of 400m. (Here the "wavelength" means the length of the pipe
over which it is laterally displaced.) FE simulation revealed that the bathymetry not only triggered the first
buckle, but also creates a lateral walking mechanism as follows: when the pipe goes into compression due
to the temperature rise, the vertical curvature makes it lift off the seabed, or, at least, it reduces the lateral
friction that can be generated with the seabed, thus reducing the resistance to lateral movement. However,
when the pipe cools, it goes from compression into tension. This tension, acting through the curvature
associated with the bathymetry, causes the pipe to dig downward into the soil at the apex of the buckle. The
result is that the apex is restrained from pulling back laterally upon cooling. Instead the pipe gains greater
straightness by additional outward lateral displacements away from the apex of the buckle. Then when the
pipe heats up again, it starts from this outward position, and develops larger lateral displacements at the
apex than at the previous cycle. Ultimately this results in the very large lateral displacement over a very
long wavelength. The mechanism is illustrated in Figure 2.
Figure 2—Growth of buckle at KP1.0 (Curves labelled "Heat" or "Cool" are from FE analyses, after
the stated number of cycles, whereas those labelled "OOS" correspond to the ROV survey performed
in the year shown. The OOS 2000 trace shows the position of the pipeline prior to operation.
The corresponding number of cycles are about 7 for the year 2002, and 116 for the year 2009.)
Indeed, the FE simulations showed that, without the formation of huge berms (such as the one shown in
Figure 3), the lateral walking simply continues to pull more and more pipe into the buckle.
Figure 3—Trench and berms created by the sweeping action of the pipeline at the first lateral buckle
To simulate these and other features of the behavior, the FE model, constructed with Abaqus, has the
•The pipe is modeled with "pipe" elements in Abaqus, which are essentially beam-type elements,
but also account for internal and external pressure, and the effect that the associated hoop stress has
in the stress-strain relation in the axial direction. The model covers the first 15 km of the pipeline
within which all of the lateral buckles form. The length of the model is sufficient to capture the
axial feed-in for the last buckle.
•It accounts for material and geometric non-linearity, including a frictional model for pipe-soil
interaction with different friction coefficients in the axial and lateral directions. The best fit to the
"observed" buckling and walking behavior was obtained with a friction coefficient of 1.0 for both
the axial and lateral directions, together with an uncoupled model in which the axial resistance
depends only on the axial displacements and the lateral resistance depends only on the lateral
displacement, though both are frictional and thus depend also on the vertical component of force
between the seabed and the pipe. Subsequent additional geotechnical data collection and special
soil-pipe interface shear tests conducted at Fugro AG in Perth led to a median estimate of the axial
resistance after full consolidation of 0.94 times the weight of the pipe, which is excellent agreement
with the axial friction factor from the matching exercise.
•The formation of berms and trenches as shown in Figure 3 is modeled approximately by lateral
springs that are created and removed repeatedly as judged appropriate, with resistances chosen to
match the surveyed buckle amplitudes, and, if necessary increased as a function of cycle number
to whatever it takes to reduce the lateral walking to the measured levels of lateral displacement.
•Changes in temperature, pressure and contents density as function of KP location x and time t are
accounted for. These parameters are obtained by bilinear interpolation between a grid of points (x,t)
for which the results of the multiphase flow start-up and shut down simulation (calculated with
Olga) were transferred as input to the FE simulation. For this it is important to use a sufficiently
fine grid of points (x,t), and especially small increments in time (t) initially. Also, sufficient mesh
refinement is needed for the Olga model to capture the rather sharp start-up transients.
•It accounts for bathymetric changes along the line, but bathymetric changes in the direction normal
to the pipeline route are neglected, by assuming the seabed is flat in that direction.
•As-laid geometry is captured by forcing an initially straight pipe into the surveyed shape, and then
releasing it onto the seabed. Upon release, this FE simulation smooths the survey data, providing
the starting point for hydrotest and operational loads to be applied. During the initial forcing
and release, the pipe is assumed to remain elastic. (Otherwise unrealistic plastic deformations
associated with short wavelength errors in the survey data would occur during the forcing. After
release, the stresses were within the yield stress of the pipe everywhere). With this modeling of the
lateral and vertical out of straightness, most FE model buckles formed at the observed locations.
Where this was not the case, lateral loads were applied in a subsequent analysis to trigger the
buckles at the correct locations and in the correct direction. These artificial triggering loads are
reduced back to zero before the operating temperature and pressure are reached, and are applied
only at the start of the first heating cycle only, and not for subsequent cycles.
A limited number of sensitivity analyses, varying inputs to the FE and Olga models within their range
of uncertainty produced considerable changes in the results. Therefore, a matching process was applied,
to arrive at the results that best matched the survey data. The matched model was then used in subsequent
analyses to determine the restraint capacity required to stop the walking. This matching process produced
a reasonable match for both the lateral buckling displacements (Figure 4) as well as for the walking
displacements at the PLES (Figure 5).
Figure 4—Lateral buckling profiles from surveys and FE simulation for buckles at KP1.0 and KP3.6 (Curves labelled
with "OOS" followed by the year are from ROV surveys; those labelled by "Heat" followed by the cycle number are from
FE analysis. The "Heat 120" results correspond to the number of cycles which occurred prior to the 2010 OOS survey.)
Figure 5—History of axial displacements at the PLES as a function of
cycle number, from observations as in Figure 1 and from FE simulation.
This matching process was not based on formal optimization. Instead it relied largely on engineering
judgment to select suitable inputs, guided by results of earlier runs.
One of the matching challenges arose as FE simulations did not readily carry the walking displacements
to feed the more downstream buckles as appeared to be the case from the survey data. This was improved
by sharpening the temperature fronts. I.e. instead of using the Olga outputs directly in the FE simulation, the
histories were modified slightly to create sharper fronts, as might be produced by the slug of liquid being
pushed away from the inlet in a way that may not have been accurately captured in the Olga simulations.
Optimum Location of Restraint and Estimation of Force Acting on it
The optimum restraint location is the location where walking can be stopped with the minimum required
load carrying capacity of the restraint. This is not at the PLES where large cyclic displacements occur,
but further downstream, about two thirds of the way to the first buckle. To find this optimum location,
the cyclic axial displacements from the FE analysis without the restraint were considered (Figure 6). The
location where this cyclic displacement is smallest was selected as the optimum restraint location, and this
was confirmed by additional analyses including a restraint at different locations and calculating the forces
generated at the restraints. A combination of sensitivity analysis, and engineering judgment considering the
uncertainties in the various inputs to the FE simulations was used to estimate the probability distribution
for the maximum force on the restraint shown in Figure 7. It then remains to create a restraint with this
required capacity at that location.
Figure 6—Cyclic axial displacements, defined as the change in axial
displacement upon start-up from cold to steady-state operating conditions.
Figure 7—Estimated probability distribution for the force acting on- and the resistance of the restraint to
stop the walking. The resistance is based on 15 PCMs reaching fully consolidated (i.e. drained) conditions.
The PCM Concept and Load Capacity of PCMs
One way to restrain the pipe axially is to place rock dump on it. The amount of rock needed to generate the
required restraint capacity is not large, but mobilizing a suitable vessel capable of accurately placing the
rock in 800m water depth to the Philippines is expensive.
As an alternative, ordinary concrete mattresses placed over the pipe were considered. Simple and FE
analyses indicated that such mattresses were effective if some membrane tension in the mattress across the
pipe were maintained. However, if the pipe works itself deeper into the soil, the mattress could go from
membrane tension to compression, essentially arching over the pipe. In this event, the standard mattress
provides very little axial restraint, if any at all. Improved mattress concepts which could improve the
wedging action were also considered, but also suffered the drawback that they tend to push the pipe down
deeper into rather soft soil, and become ineffective.
These considerations together with an increasing urgency to come up with a solution finally lead to the
invention of the Pipe-Clamping Mattress (PCM). Instead of simply resting on the pipe, the PCM (Figure
8) clamps the pipe into it, using its weight at a high leverage. Indeed, the PCMs are designed so that the
clamping capacity between the PCM and the pipe overmatches the sliding capacity of the PCM with the
pipe over the soil. Thus 100% of the weight of the PCMs is effective in generating axial frictional resistance.
A log mattress is provided over every PCM to provide additional weight and especially additional clamping
force. The log mattress is more efficient in generating clamping force, because its weight is delivered to the
outer extremities of the PCM, whereby leverage is maximized.
Figure 8—(a) Pipe-Clamping Mattress (PCM) with logmat. (b)
Equilibrium considerations to determine the PCM clamping force Q
Simple equilibrium calculations (Figure 8) were found to give accurate estimates of the clamping force
generated and the axial capacity against sliding between the PCM and the pipe, when compared to 2-
dimensional FE analyses in which the soil is modeled as an elasto-plastic continuum. These FE models
employ a slice of elements normal to the pipe, with boundary conditions set in such a way that both plane
strain and antiplane shear deformations are captured. This was achieved by making the displacements of
corresponding nodes at each side of the slice the same.
Of course, the stresses in the pipe due to clamping by the PCM were calculated. They are well within the
allowable limits even if the clamping force were concentrated as line loads at the 3 and 9 o'clock positions.
Also, the effect of the shear and normal forces on the coating had to be considered, together with the effect
of temperature on the 3-layer-PP coating strength, and concentrations in the PCM-pipe contact stresses due
to geometric variability within specified tolerances.
An essential requirement for the development of a suitable restraint system is adequate understanding
of flowline-soil and mattress-soil interaction for representative loading conditions. Time- and loading-
dependent friction factors between the mattresses and the seabed and the pipe and the seabed were key
items required to quantify the weight and number of PCMs, and to verify the modelling of the history of
expansion and buckling. Initially, no geotechnical data were available for the proposed restraint location.
The seabed at the proposed PCM location was expected to comprise very soft soil.
Site specific geotechnical data were acquired to develop reliable site-specific geotechnical parameters for
the development of the restraint. Notwithstanding the challenging constraints of limited time and budget,
the geotechnical SI was conducted in September 2014 and high quality samples were successfully recovered
by piston corers. Based on standard laboratory tests, the seabed was classified as siliceous CARBONATE
SILT in accordance with the classification systems proposed by Clark and Walker (1977). It is reasonably
uniform and laterally homogenous within the proposed PCM location.
To develop the time-dependent soil resistance to cyclic axial pipe movements, monotonic and cyclic
interface shear tests were also performed. These tests are similar to direct shear tests on soil only, but instead
of creating a shear plane within the soil, they create a shear plane at the interface between the soil and
a surface with the same roughness as the coating of the pipe. The general arrangement of this apparatus
and a post-test view of a sample are shown in Figure 9. These tests aim to reproduce the loading history
at the pipe-soil interface, including generation and dissipation of pore pressures. They were performed in
Fugro AG's Perth laboratory using a direct shear apparatus incorporating modifications to suit testing with
the low stress levels and episodic movements representative of pipelines. These modifications include a
lightweight hangar, a low friction carriage, and careful calibration of corrections for machine friction before
each test series.
Figure 9—(a) Interface shear box apparatus (b) View of sample after
completion of test and removal from apparatus (White et al. 2012)
Thus the following effects could be captured using the approach described in White et al. (2015), and
in Atkins (2015): (i) the effect of stress level on drained interface friction angle; (ii) the effect of loading
rate on interface strength for normally consolidated soil and overconsolidated soil; (iii) the effect of over-
consolidation on undrained interface properties and (iv) the cyclic hardening (or softening) behavior of the
interface through cycles of undrained shearing and consolidation. The cyclic hardening, through episodes of
undrained movement followed by consolidation (pore pressure dissipation) prior to a subsequent movement,
is the process that occurs during the operating life of the flowline, and leads to a change in the interface
Figure 10 shows an example result from the interface shear box test program. The sample was initially
consolidated to an effective stress of 25 kPa. Then two cycles of +/- 9 mm shear displacements were applied
at a rate of ~1 mm/s to ensure essentially undrained shearing. This is followed by a 30-minute dissipation
period. The sequence of undrained shearing followed by dissipation is then repeated. The resulting shear
stress vs. shear displacement relations from such a test are shown in Figure 10. The steady or residual
interface shear strength rose from 0.45 times the total vertical stress in the initial cycle pair to about 0.8
times the vertical stress in the final cycle. The initial value represents the normally-consolidated undrained
interface strength ratio, Rint (which can be denoted (su-int/σ'no)nc), and the final value represents the drained
strength, tan(δres), corresponding to this stress level (and the particular interface material and roughness
used). The plot also shows that this final value is already reached for shearing cycles 11 & 12, which take
place after 5 cycles of dissipation. The shearing in this case represents a shut-down/restart sequence, which
could be of short duration (of around 1 day), whereas the dissipation cycle represents the longer period (of
around 1 month) of operating at approximately steady-state conditions.
Figure 10—Typical results of episodic interface direct shear test, with 30min of dissipation at an
effective stress of σ'no= 25 kPa initially and after every pair of essentially undrained dissipation cycles.
This pattern of cyclic hardening is consistent with previous reported studies but the specific values of
the soil parameters – including the high values of Rint and δres – are different from typical values for non-
carbonate clays and reflect the carbonate mineralogy of the seabed sediment in the Malampaya region. The
site-specific sampling and testing was therefore valuable in identifying the interface strength parameters
relevant for this site.
A further important parameter for the geotechnical analysis is the coefficient of consolidation, cv, at
near-surface depths of a few tens of centimeters. This parameter is important since it controls the time
required for dissipation of any generated excess pore pressures and the associated changes in effective stress
and undrained strength beneath the pipeline and mattress after installation and following any undrained
movements. Near-surface measurements of cv are difficult to obtain using conventional in situ tests such as
the cone penetrometer dissipations. The design range of cv was estimated based on the consolidation stage
of the interface shear box tests supported by extrapolation of the trend from deeper measurements.
The same interface shear tests also served to estimate the sliding resistance between the PCMs and the
soil. This was supplemented by a few extra tests in view of the higher roughness of the PCM surface, as
opposed to the PP-coated pipe. The assessment indicated that the PCMs are initially supported mostly by
the pore water trapped in the soil, which provides very little resistance to sliding. A dissipation time of about
one month was estimated for these pore pressures, based on the consolidation times for the interface shear
tests, and scaling this to account for the longer drainage paths for pore water trapped under the PCMs, as
opposed to that in the much smaller interface shear tests. (The PCMs have holes covered with geotextile
to enable quicker dissipation of pore pressures.)
Upon first consolidation, under the weight of the PCMs alone, the PCM sliding resistance already
increases significantly. It increases even more after several cycles of shearing and re-dissipation of the excess
pore pressures generated by each shearing event, until finally the drained sliding resistance applies (even
for undrained movements), after about 5 cycles of dissipation or more. Since 5 cycles of walking could be
tolerated, the design was based on this drained sliding resistance. The estimated probability distribution for
the drained sliding resistance is shown in Figure 7, together with the probability distribution of the load
acting on the restraint. Other results from the geotechnical assessment include: flowline force-embedment
relationship applicable during installation (undrained conditions), short- and long-term embedment of the
PCMs, and mobilization displacement for the load-displacement relation associated with sliding of the PCM
over the seabed. In all cases, low, median, and high estimates were made.
The offshore trials provided a rare (if unplanned) opportunity to confirm low initial resistance to sliding.
For this, a friction coefficient of less than 0.05 was estimated when the PCM or pipe is initially placed,
and mostly supported by pore pressures from water trapped in the soil below. A joint of pipe was lowered
to the seabed at a location which had a slope of approximately 1 in 12. The pipe joint was observed to
slide away axially as the supporting sling went slack, indicating an interface friction coefficient of < 0.08.
Fortunately, no axial resistance was needed from the PCMs so early on. Nevertheless, this illustrates an
important point about pipeline on soft, normally-consolidated seabeds. Before sufficient time has elapsed
for the pore pressures from the pipe weight to dissipate, the available axial friction coefficient is very low.
This may have implications for pipeline stability when laying around curves or on slopes.
Structural Reliability Assessment (SRA)
In lieu of an established design procedure for restraints intended to stop walking, structural reliability
assessment (SRA) was used to define the safety margins required to cover the uncertainties involved.
For this purpose, all relevant uncertainties were estimated explicitly and combined to calculate the failure
probability. The number of PCMs was then chosen to keep this failure probability below a tolerated level.
Here a simplified, approximate approach was used, so that a few sensitivity analyses with the FE model were
sufficient, instead of having to incorporate the FE model within one of the algorithms used for calculation
of failure probabilities, such as first or second order reliability calculations, Monte-Carlo simulation, or
All identified and relevant uncertainties were included, even if, in lieu of statistical data, the estimates
of the probability distributions had to be mainly based on engineering judgment. The SRA involves a load
side and a resistance side for the PLES displacement.
On the resistance side, the probability distribution for the PLES displacement at which the connecting
spoolpiece fails is estimated using an FE model of the spoolpiece. This FE model leads to the bending and
torsional moments at the spool-piece-to-PLES connector, which is the potential weak link. Uncertainties
considered for this include: the moment capacity of the connector, the initial position of the sliding part of
the PLES on its base, the initial stresses in the spoolpiece at the time it is connected (arising for instance
when the connector pulls the connection together), the friction factor between the spoolpiece and the seabed,
the settlement of the PLES and the associated increase in the frictional resistance between the spoolpiece
and the pipe, and eccentricity of soil resistance to sliding of the PLES, which would in part be resisted by
bending moments in the spoolpiece that prevents the PLES from rotating about a vertical axis. (The yield
strength of the spool piece was not judged relevant, since the most likely failure scenario involves loss of
integrity of the connector before any yielding of spoolpiece. Thus, the one uncertainty for which statistical
data are available was not relevant in this case.)
On the load side, the additional displacement that might develop before the soil is sufficiently
consolidated for the PCMs to stop the walking was estimated considering uncertainties in: the consolidation
times under the PCMs, in the load and resistance of the PCM restraint, and how much movement might
occur at each cycle if the load (assuming no movement) exceeds the resistance.
Every one of the above ingredient uncertainties could in-itself require a more detailed investigation.
For instance, since information available on the McPac connector was limited, OneSubsea, the supplier of
the connector, was engaged, and performed detailed local FE analyses of the connector, including factors
that could affect the metal-to-metal contact seals under bending and pressure loads. This was then used to
estimate the probability distribution of the bending moments on the restraint that might cause a leak due
to loss of a metal-to-metal contact seal.
Combining all of the above in a probabilistic model, led to the conclusion that the most likely failure
scenario (if failure occurs) would be that walking continues even after the soil under the PCMs is fully
consolidated. The probability of this scenario, can be estimated much more simply, by comparing the
maximum load that develops on an immobile constraint with the actual drained resistance provided by the
PCMs, as is done in Figure 7.
Even if this "failure" scenario did develop, additional PCMs could be added before it is too late, or
more weight could be added on top of the existing PCMs (to the extent that the design margins from the
structural design of the PCMs themselves allow it). For this reason, a probability of "failure" of 2% was
judged tolerable, since "failure" only means the need to take further action.
The matter of safety margins required to cover uncertainties for restraints to stop walking is a more
general issue for the industry, for which no standardized industry approach is yet available. The uncertainties
are mostly geotechnical (i.e. large compared to typical structural uncertainties), and these geotechnical
uncertainties are compounded, as they occur on both the load- and resistance sides. If, despite these large
uncertainties, a low failure probability is demanded, the required design margin will be large, resulting in
large and expensive restraints. However, with suitable monitoring, and a contingency plan the consequences
"failure" can be minimized, so that a higher probability of "failure" can be tolerated. Indeed, for such cases
"restraint deficiency" is a better term than "failure". The safety margins for restraints to stop walking is one
of the issues currently being addressed in the APT JIP led by Crondall Energy, Crondall Energy (2016).
Fabrication and Installation on a Live Flowline
Whereas Shell with support from Crondall Energy took the responsibility to ensure that the PCMs would
indeed stop the walking without damaging the pipe or pushing it too deep into the soil before the PCMs
close on it, Subcon took the responsibility for the detailed design and fabrication of the PCMs and logmats
themselves. The detailed design of the PCMs was complicated by the need to ensure a suitable weight
distribution within the PCM (to maximize clamping force) and the requirement for holes in the mattresses to
reduce the time required for drainage to occur. Further, the PCM comprises two basic halves, with internal
structural wires to tie the two halves together in the same way that the blocks of ordinary concrete mattresses
are tied together. To ensure reliable clamping performance tight tolerances are required to control the gap
between the two halves, and for the parts of the mold that form the clamping surface. Other complications
included the design of suitable lifting points for fabrication, transportation and installation. As a result,
fabrication required a very complex mold. The concrete was poured with the PCM upside down, and
subsequently flipped over once curing had achieved a suitable strength. The final design also included
markings to aid installation and allow measurement of embedment and any future movement, Figure 11.
The mattresses fabrication took place at a temporary facility at Batangas in the Philippines, developed
specifically for the project.
Figure 11—Final detailed PCM design, showing key installation and monitoring details
Installation of the PCMs was undertaken by DOF subsea from the Skandi Hawk. Some preparatory work
was required prior to installation of the PCMs. Initially the surface of the pipe was cleaned to ensure that the
axial friction generated between the clamp and the pipe was not compromised by any marine growth. At the
target location, the pipe was typically embedded to between 30% and 40% of its diameter. To ensure that
soil entrainment would not compromise the clamping operation, a plough was used to remove excessive
material, and reduce the local embedment to 25% OD or less. The PCMs were transferred from the deck
of the vessel to the seabed using a lifting frame, Figure 12
Figure 12—PCM deployment Offshore. (a) In-air. (b) Subsea
The rigging was designed to keep the PCM open at the optimum angle. This angle was a compromise
between two competing effects; it was as large as practical to simplify contact with and grabbing of the
pipe, but as small as practical to minimize the vertical load imposed on the pipe as the clamp closed (which
may further embed the pipe prior to full clamping). Also, moving the lifting points on the PCM towards the
hinge, instead of lifting it by the loops in the cables (as an ordinary mattress would be lifted) helps reduce
the vertical load the PCM exerts on the pipe, before it closes on the pipe.
The PCMs were lowered to just above the seabed away from the pipe, and then maneuvered over the
top of the pipe at a clearance of no more than 1 m to minimize any denting risk to the pipeline. (Simple
energy balance calculations indicated that the dent from dropping the PCM from a height up to 1m would
be tolerable.) Once over the pipeline, the PCM was slowly lowered down until contact was made with
the pipe. As contact became established, the PCM naturally closed around the pipe and self-aligned as the
installation proceeded. It was found that, once partial clamping had been established, the PCM could be
easily maneuvered along the axis of the pipe, to allow very accurate positioning of the PCM. Following
installation of the PCMs, the log mats were deployed to the seabed in a similar manner and installed over
the PCMs, Figure 13. Installation of 15 mattresses was completed in 10 days.
Figure 13—As-Installed PCM. (a) prior to installation of log-mat. (b) with log-mat installed
To minimize the axial displacements that must occur at the extremity of the PCM bank, it is desirable to
install the PCMs over as short a distance as practical. However, it was not possible to install the PCMS as a
continuous bank due to the presence of anodes (on every second pipe joint at this location) and the need to
avoid the coating field joints. A maximum of 3 PCMs could be installed on a single 12.2m pipe joint, so the
15 PCMs were installed as 5 banks of 3, Figure 14. This resulted in a total length of coverage of about 60 m.
Figure 14—MBES overview image of the as-installed PCM banks.
Checking to make sure that the PCMs do indeed stop the walking in this case is not just "nice to have"
but essential, since the estimated probability that they fail to do so even after full consolidation is around
2% based on the estimated probability distributions of Figure 7. Such a high failure probability is tolerated,
since it is always possible to add more PCMs if the ones deployed were not sufficient.
The first inspection in January 2016 did not reveal significant movement, from the installed flags, Figure
15. During the period from installation to the first inspection, 3 operating cycles occurred. Without the
PCMs, these cycles would have produced about 8cm of walking displacements, which could have been
detected by the flags. This suggests that even before full consolidation the PCMs already have sufficient
capacity to stop the walking. Nevertheless, further monitoring is planned, in case small movements are
occurring, which over the longer term might be detected by the flags.
Figure 15—Flags with ruler markings installed adjacent to PCMs to facilitate displacement measurement
The conditions that lead to inventing, fabricating and deploying the PCMs within a period of 12 months
•An axial restraint is needed to stop flowline walking.
•The optimal location for the restraint is away from the inlet (PLES). To restrain it at the PLES
would require a much larger restraint capacity.
•No provision had been made for collar joints on which a clamp might be applied.
•It is preferred not to remove the coating. Therefore, the coating must withstand the clamping forces
and shearing forces at the operating temperature of the pipeline.
•Rock dumping would require an expensive mobilization to a remote location.
For this the PCMs proved to be a simple, inexpensive, and effective solution. The possibility to use PCMs
also can justify "wait and see" approach for new projects, where there is uncertainty about whether the
flowline will walk, to avoid installation of unnecessary restraints. It is also always possible to add more
PCMs, if necessary. This makes the large design margins that might otherwise be needed to cover uncertainty
in load and resistance for the restraint unnecessary.
The development and successful installation of the PCMs involved vital contributions from a wide range
of companies and individuals. Within Shell Projects and Technology, Giuseppe Pagliuca performed the
transient flow simulations with Olga, using a model originally developed by Pek-San Ong, Jan van Bokhorst
provided support in regard to the capacity of the 3-layer-PP coating to resist the PCM loads at the operating
temperature of the pipeline, Leo de Mul reviewed thermoplastic materials for the clamping surface, Chris
Kettle reviewed Cathodic Protection design and impact of electrolyte access restrictions due to the PCMs,
Mike Coyne and Don Nelson performed the overall design and installability review. DOF Subsea, under
the project leadership of Heng Yeu Tan, not only performed the installation, but also held the contract with
Subcon for the detailed design and fabrication of the PCMs. Within Subcon John Holmes and Jonathon
Curry were very effective and resourceful in overcoming the many technical and non-technical challenges
involved. Finally, Fraser Bransby and Hongjie Zhou of Fugro AG contributed to the geotechnical analysis.
Atkins (2015) SAFEBUCK JIP: Safe design of pipelines with lateral buckling. Design Guideline, SAFEBUCK III.
5087471/01/C dated 15 January 2015.
Bruton, D., et al., (2010), "Lessons Learnt from Observing Walking of Pipelines with Lateral Buckles, Including New
Driving Mechanisms and Updated Analysis Models," 2010 Offshore Technology Conference, Houston, Texas, 3-6
May 2010, Paper No. OTC 20750.
Carneiro, D., Rathbone, A., Soon, K.S., Viecelli, G. (2014) "Rate Dependent Soil Resistance in FE Analysis of Pipeline
Walking," Proceedings of the ASME 2014 33rd International Conference on Ocean, Offshore and Arctic Engineering,
June 8-13, San Francisco, California, USA, Paper No. OMAE2014-23521.
Carr, M., Sinclair, F., and Bruton, D., 2006, "Pipeline Walking - Understanding Field Layout Challenges, and Analytical
Solutions Developed for the Safebuck JIP," Offshore Technology Conference, Houston Texas, USA, 1-4 May 2006,
Paper No. OTC 17945.
Clark A.R. and Walker B.F. (1977). A proposed scheme for the classification and nomenclature for use in the engineering
description of Middle Eastern sedimentary rocks. Géotechnique Vol. 27, No.1, 93–99
Crondall Energy (2016). Anchoring Pipeline Technology – a new Joint Industry Project. The APT JIP.
Konuk, I. (1998) "Expansion of Pipelines under Cyclic Operational Conditions" OMAE 1998.
Peek, R. (2002) "Crawling of Pipelines under Cyclic Thermal Loading," Shell Global Solutions International, Report
No. OP 02-30011, Reviewed by P.R. Paslay and W.T. Jones, Revised 4th January 2002. Released as unrestricted,
Tornes, K., Jury, J., Ose, B., Thompson. (2000) "Axial Creeping of High Temperature Flowlines Caused by Soil
Ratcheting", OMAE 2000.
White D.J., Westgate Z.J., Ballard J-C, de Brier C. & Bransby M.F. (2015). "Best practice geotechnical characterization
and pipe-soil interaction analysis for HPHT flowline design". Proc. Offshore Technology Conference, Houston Paper
White D.J., Campbell M.E., Boylan N.P. & Bransby M.F. (2012). A new framework for axial pipe-soil interaction
illustrated by shear box tests on carbonate soils. Proc. Int. Conf. on Offshore Site Investigation and Geotechnics. SUT,