Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/262726677

Numerical Simulations of Dynamic Embedment


During Pipe Laying on Soft Clay

Conference Paper June 2009


DOI: 10.1115/OMAE2009-79199

CITATIONS READS

4 142

3 authors:

Dong Wang David J White


Ocean University of China University of Western Australia
67 PUBLICATIONS 589 CITATIONS 229 PUBLICATIONS 3,206 CITATIONS

SEE PROFILE SEE PROFILE

Mark Randolph
University of Western Australia
328 PUBLICATIONS 6,909 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Instrumented Free Fall Sphere View project

Effects of consolidation under a penetrating footing in carbonate silty clay View project

All content following this page was uploaded by Dong Wang on 31 May 2014.

The user has requested enhancement of the downloaded file.


Proceedings of the ASME 2009 28th International Conference on Ocean, Offshore and Arctic Engineering
OMAE2009
May 31 - June 5, 2009, Honolulu, Hawaii, USA

OMAE2009-79199

NUMERICAL SIMULATIONS OF DYNAMIC EMBEDMENT


DURING PIPE LAYING ON SOFT CLAY

D. Wang D.J. White


Centre for Offshore Foundation Systems Centre for Offshore Foundation Systems
University of Western Australia University of Western Australia
Perth, Western Australia Perth, Western Australia

M.F. Randolph
Centre for Offshore Foundation Systems
University of Western Australia
Perth, Western Australia

ABSTRACT embedment combines static estimates based on the pipe


Prediction of the as-laid embedment of a pipeline, which weight and local stress concentration factor in the touch down
affects many aspects of pipeline design, is complicated by the zone, with a multiplicative dynamic embedment factor to
dynamic motions that occur during the lay process. These account for the effects of cyclic pipeline motion during laying
motions cause pipelines to embed deeper than predicted based (Bruton et al. 2006). Back-analysis of the observed
on static penetration models, as the seabed soils are both embedment of pipelines indicates a wide scatter in the
softened and physically displaced by the pipeline motion. This apparent dynamic embedment factor (Lund 2000, Bruton et al.
paper describes the results of 2D numerical analyses using a 2007). Some of this scatter is inevitable, due to the variability
large displacement finite element approach aimed at associated with the dynamics of the lay process. However,
quantifying pipeline embedment due to cyclic lateral motion at other causes of this scatter can potentially be tackled by
various fixed vertical load levels. The validity of the improved calculation methods for pipeline embedment
numerical results is first assessed by comparison with perhaps bypassing the concept of a multiplicative dynamic
published data from centrifuge model tests in two different embedment factor and better techniques for assessing soil
types of clay. A parametric study varying the normalized conditions close to the mudline.
vertical load is then presented, which suggests a simple Model tests, either on the laboratory floor or in a
approach for estimating an upper limit to the dynamic geotechnical centrifuge, provide a means of assessing dynamic
embedment. pipeline embedment under well controlled conditions, but it is
not generally practical to cover all of the relevant range of
INTRODUCTION pipe weights and patterns of cyclic motion that may occur
Dynamic pipe embedment: background during the laying process. Numerical analysis has the potential
The embedment of a subsea pipeline affects the on- to augment such studies, although it is essential to first
bottom stability of the pipeline during environmental loading, validate, and calibrate, the numerical approach.
and also the lateral buckling and axial walking behavior The aim of the present paper is to present a numerical
during thermal cycles (Bruton et al. 2007). The as-laid approach, based on large deformation finite element analysis,
embedment depends on a number of factors including the to evaluate the static and dynamic embedment of pipelines.
near-mudline shear strength profile, the submerged pipe The approach allows modeling of both softening of the soil as
weight, the pipe-lay configuration and the wave-induced it is remolded and also the physical scraping of the soil aside
motion of the lay vessel and the resulting pipeline movements by the lateral pipeline motion. Results from the numerical
at the seabed. The general approach for predicting pipeline approach are compared with data from centrifuge model tests

1 Copyright 2009 by ASME


in two different soil types. The approach is then used in a The accuracy of LDFE analyses depends largely on the
parametric study to bracket the dynamic embedment strategy used to map field variables. The mapping consists of
anticipated for a range of normalized pipe weights and a extrapolation of field variables from the old Gauss points to
particular soil strength profile. The resulting embedments are the old element nodes and then interpolation from the old
quantified through a simple relationship to the static nodes to the new Gauss points. For the extrapolation, a
embedment, but based on the remolded soil strength profile technique termed Superconvergent Patch Recovery (SPR,
rather than the intact strength profile. Zienkiewicz & Zhu 1993) is employed, the robustness of
which has been verified in our previous LDFE studies
Objectives and scope of this study (Randolph et al. 2008; Wang et al. 2008c). After the stresses
Dynamic pipeline embedment is typically a large and strains at the old element nodes are recovered, the old
deformation process which features a combination of cyclic deformed element that contained the coordinates of each new
horizontal and vertical loading. The purpose of this paper is to Gauss point is identified and the corresponding variable S at
simulate, in an idealized manner, the response of an element of new Gauss points is then interpolated as
pipeline during undrained dynamic laying using a two-
dimensional large deformation finite element (LDFE) method, S = NS* (1)
termed remeshing and interpolation technique with small
strain (RITSS, Hu & Randolph 1998). The parameters that where N is the shape function and S* represents the recovered
affect the embedment behavior are assessed, including the variables at nodes of the old element. Application of the SPR
pipe weight relative to the soil strength, the number of technique requires optimal sampling points which own at least
horizontal cycles of movement and softening of the soil one-order higher convergence than other positions. The
around the pipe. existence of optimal sampling points in quadratic triangular
The dynamic embedment of a pipeline is a three- elements has not been proved in theory, however, the Gauss
dimensional soil-structure interaction process, which is driven points were found as optimal sampling points empirically
by the prevailing metocean conditions, the geometry of the (Zienkiewicz & Zhu 1993).
pipeline catenary, the characteristics of the lay vessel, and the To mimic pipe-soil interaction during the dynamic lay
response of the seabed. This study focuses on the geotechnical process, the sequence of small strain increment, mesh
aspect of this behavior the interaction between the pipeline regeneration and mapping is repeated until the predefined
and the seabed. This interaction has been idealized as displacement has been reached. The overall scheme of the
monotonic vertical load applied on the pipe and subsequent LDFE analysis is controlled by a master program written in
cyclic horizontal motion whilst the vertical load is held Fortran. The mapping of field variables is carried out
constant. This approach is an idealization of the load externally by subroutines in Fortran. ABAQUS is called by
conditions applied as a section of pipe moves through the the master program to build the small strain finite element
touchdown zone when laid on the seabed. This process is model, which is automatically implemented by a Python file
described in more detail by Randolph & White (2008a). prepared beforehand. Python is the built-in script language of
ABAQUS (HKS 2004). Another Python file is coded to
NUMERICAL TECHNIQUES extract the necessary data from the ABAQUS result files.
Large deformation finite element analysis More detailed procedures for the RITSS method coupled with
The RITSS method is composed of Lagrangian steps ABAQUS can be found in our previous papers (Wang et al.
incorporating convection of field variables between meshes. 2006; 2008a; 2008b).
The method divides the initial vertical penetration and
subsequent horizontal oscillations into a series of incremental Strain-softening and rate-dependent soil model
steps. The pipe displacement in each step has to be small The profile of intact undrained strength, su0, for most
enough to avoid element distortion in the vicinity of the pipe. seabed clays can be described as a linear function:
The small strain calculation then can be performed reasonably,
followed by mesh regeneration according to the deformed s u0 = s um + kz (2)
geometry. The field variables including the stresses, strains
and material properties are mapped from the old mesh to the where sum is the shear strength at the mudline, k is the strength
new mesh. The mapping is similar to the flow of solid in an gradient and z is the soil depth. The soil strength reduces
Eulerian mesh, so the RITSS method can be viewed as a during episodes of remolding. This strength reduction can be
special Arbitrary Lagrangian-Eulerian method. The advantage estimated by means of cyclic penetrometer tests and is
of the RITSS approach is that it can be coupled with any quantified by a cyclic sensitivity factor, ST-bar, which is defined
standard finite element package. In this paper, ABAQUS has as the ratio of the initial downward penetration resistance to
been selected to conduct the small strain calculation and the the steady value of penetration resistance reached after many
mesh generation process, with the soil being composed of cycles of movement. Numerical studies (Zhou & Randolph
quadratic triangular elements. 2009) and field data (Yafrate et al. 2009) have shown that ST-

2 Copyright 2009 by ASME


bar is typically 20 to 40 % lower than the true sensitivity of the created by heave, but can also incorporate a strain softening
soil, due to partial softening during the initial penetration. A model for the soil strength.
ratio of remolded to intact strength in the range 0.2 0.5 is The pipe-soil interface interaction is simulated by setting
typical (i.e. a true sensitivity, St, of 2 5). An empirical the pipe and soil surfaces as a master-slave pair. The length of
expression that is widely used to model this progressive loss contact around the pipe perimeter is permitted to vary as the
of strength during remolding is (Zhou & Randolph 2007; pipe moves. The maximum shear stress on the pipe-soil
Wang et al. 2008c) interface is defined as max = sum, where denotes the
roughness factor. A value of = 0 corresponds to a fully
s u = rem + (1 rem ) e-3 95 s u0 (3) smooth interface that cannot sustain shear stress.
In order to confirm the effectiveness of the RITSS
implementation, the monotonic penetration of a pipeline i.e.
where rem = 1/St; denotes the accumulated absolute plastic a rigid cylinder with a horizontal axis is first simulated
shear strain and 95 represents the value of at which the soil under various different conditions. An intact soil strength
has undergone 95% of the reduction in strength due to profile of su0 = (0.75 + 1.6z) kPa (where z is in meters) and a
remolding. Typical values of 95 are in the range of 10 30 pipe diameter of 0.8 m have been adopted, to match the
(Zhou & Randolph 2007). centrifuge model tests back-analyzed later in this paper. This
The shear strength at the mudline is a key parameter for diameter is relatively large compared to field conditions, but
assessing the behaviour of on-bottom pipelines. However, the the resulting normalized loads and embedments are typical.
strength of the surficial soil is not easily measured in many Strain softening of the soil is not considered in this benchmark
actual applications, and the exact strength profile in the depth problem. The numerical responses during monotonic
range relevant for pipelines is often difficult to determine. penetration are compared with the upper bound solution
In the following LDFE studies, the soil is regarded as an (Randolph & White 2008b) in Fig. 1, where w represents the
elastic-perfectly plastic material obeying the Tresca yield embedment of the pipe invert and the value of su0 at the level
criterion. The Youngs modulus of the two clays is taken as of the pipe invert has been used for normalization. In
E = 500su, and the Poissons ratio is taken as 0.499 to weightless soil (' = 0), the soil heaves that forms over the
approximate constant volume undrained conditions. The shoulders of the pipe does not produce a surcharge load, thus
submerged unit weight of the soils is taken as ' = 6 kN/m3. the normalized load-displacement curve from the LDFE
method is very close to the theoretical solution by Randolph &
MONOTONIC PIPELINE EMBEDMENT White (2008b).
Pipelines laid untrenched on the seabed embed more
deeply than can be calculated from the self-weight alone. This
V/s u0D
is due to two factors. Firstly, the contact stress between the
0 1 2 3 4 5 6 7
pipe and the soil is enhanced through the touchdown zone due
to the catenary shape. Secondly, dynamic movement of the 0
pipe during the lay process leads to disturbance of the seabed.
This motion remolds the seabed soil, causing a reduction in 0.1
the strength. Also, if the horizontal movement is of sufficient LDFE
amplitude, the soil is pushed aside and a trench forms around 0.2
=0.2, 'D/s um=6.4
w/D

the pipe.
The additional penetration from remolding and trenching 0.3 smooth, 'D/s um=6.4
is principally attributed to the horizontal oscillations and also
smooth, w eightless
depends strongly on the pipe weight (Lund 2000; Cheuk & 0.4
White 2008). Therefore, accurate assessment of the smooth, w eightless
relationship between the monotonic embedment and the pipe 0.5
(Randolph & White)
weight forms the initial basis for any simulation of the
dynamic laying response and the overall dynamic embedment Figure 1. Monotonic penetration resistance
process. Limit plasticity analyses have been proposed by (non-softening soil strength)
Murff et al. (1989) and Randolph & White (2008b) to predict
the embedment induced by monotonic vertical load (i.e. pipe However, when the soil self-weight is included, the
weight). In their studies the laying process is neglected and the penetration resistance is significantly increased (Fig. 1). For
pipe is assumed to be wished-in-place, and the soil is modeled the adopted value of submerged unit weight ('D/sum = 6.4),
as a rigid plastic solid. The full penetration process and the the monotonic penetration resistance of the smooth pipe is
resulting soil heave and remolding are neglected in this type enhanced by 49% in the LDFE results compared to weightless
of analysis. By contrast with limit plasticity analyses, the soil at w/D = 0.5. This contrast shows that the penetration
LDFE method can not only track the deformed seabed surface, resistance may be underestimated if the surcharge loading

3 Copyright 2009 by ASME


created from the self-weight of the soil heave is neglected, stable value of H/su0D 0.9, where su0 refers to the value of
especially in softer soils for which the non-dimensional the intact strength profile at the current embedment of the pipe
parameter 'D/sum is higher. In contrast, the penetration invert. This steady value indicates that the theoretical increase
resistance increases by less than 6% when the roughness in normalized lateral resistance created by the increased
factor is adjusted from = 0 to 0.2 (Fig. 1). embedment is countered by progressive remolding of the soil.

SIMULATION OF DYNAMIC LAYING


0
Benchmark results: centrifuge model tests
The following large deformation simulations of the
dynamic lay process will focus on reproducing the phenomena 0.1

observed in four centrifuge model tests reported by Cheuk &


White (2008). The pipe diameter was D = 0.8 m at prototype 0.2

scale. Two tests were carried out in Kaolin Clay (denoted KC)

w/D
and the other two in a natural high plasticity clay (denoted 0.3

HP). The linear strength profiles derived from 70 deep T-bar


tests conducted in these samples are shown in Table 1. 0.4

Table 1. Soil strength profiles in centrifuge model tests 0.5

Soil sample Mudline Strength Cyclic T-bar


strength, gradient sensitivity, 0.6
sum ST-bar -1 -0.5 0 0.5 1
Kaolin Clay (KC) 0.75 kPa 1.6 kPa/m 2.3 H/su0D
High plasticity clay (HP) 0.4 kPa 2.5 kPa/m 1.9 Figure 2. Measured lateral resistance (test KC05)

In the centrifuge, the pipe was rigidly attached to the Numerical modeling: analysis parameters
actuator by means of a vertical load cell and a loading arm that During horizontal oscillations, the soil may be bonded to
was instrumented to measure the horizontal load applied to the the rear face of the pipeline due to the tensile resistance
pipe. The pipe was firstly penetrated into the soil under a sustained through negative excess pore pressure. Alternatively,
monotonic vertical load, V, which represents a steady vertical immediate breakaway may occur on the pipe-soil interface
pipe-soil contact pressure. This value was maintained constant when net tension is created. The bonded contact leads to a
during the sequence of horizontal displacement cycles which double-sided failure mechanism. In a similar centrifuge test
simulated the dynamic embedment process. The values of V that simulated large-amplitude lateral displacement of a
and the corresponding monotonic embedment obtained for the pipeline, Dingle et al. (2008) used particle image velocimetry
four tests are listed in Table 2. In all tests the pipe was and close range photogrammetry techniques (White et al.
subjected to 40 horizontal displacement cycles of 0.05D. 2003) to capture the instantaneous velocity fields in the soil.
This lateral amplitude was chosen to represent typical pipeline They observed that the double-sided mechanism remained
movements during laying in a modest seastate. In practice, this until the pipe lateral displacement exceeded 0.06D. The
amplitude depends on the lay conditions and will reduce oscillation amplitude in this study is only 0.05D so it is
through the touchdown zone until the point at which the reasonable to assume that there was no separation between the
pipeline is at rest on the seabed. soil and the rear face of the pipe. This assumption appears to
be confirmed by the monotonically increasing lateral
Table 2. Simulated pipe weight and monotonic embedment resistance mobilized during each cycle, with no indication of
Test identifier KC04 KC05 HP06 HP07 the brittleness that would accompany breakaway at the rear of
Pipe weight, V/sumD 1.93 3.68 4.53 8.16 the pipe (Cheuk & White 2008).
Monotonic embedment, 0.08 0.12 0.10 0.22 The loading mode adopted during the monotonic
w/D penetration stage was displacement-rate controlled, rather than
load-controlled, so the maximum vertical load was not exactly
To illustrate the typical observed behavior, the lateral equal to the vertical load given in Table 2, which was
resistance measured during test KC05 is compared with the sustained during the cyclic phase, but the difference was less
accumulating pipe embedment in Fig. 2. The monotonic than 3.5%.
embedment under the imposed vertical load was only 0.12 The strain softening of the soil strength, as described by
diameters, but the pipe embedment increased rapidly during Equation 3, has been incorporated into the large deformation
the first few cycles of horizontal motion (whilst the vertical simulations to capture the role of remolding in the laying
load was held constant). The rate of embedment reduced with behavior. The brittleness parameter, 95 is taken as 10. Four
continued cycling, and the horizontal resistance approached a simulations of test KC05 have been conducted, adopting four

4 Copyright 2009 by ASME


values of the soil sensitivity, St. For test KC05 with St =1 and
2.3, the relationship between the horizontal resistance and the 0
pipe embedment are shown in Fig. 3(a) and (b), respectively. 0.1
(a) St=1

The normalized maximum horizontal resistance, Hmax/su0D


0.2
reaches 2.73 and 1.55 in these two cases. These values

w/D
significantly exceed the experimental result Hmax/su0D = 0.9 0.3

(Fig. 2, and Cheuk & White 2008). This over prediction can 0.4
be attributed to the T-bar test sensitivity of St = 2.3 being an 0.5
underestimate of the softening of the surficial soil. This
underestimation may be for two reasons: (i) under-estimation 0.6
-3 -2 -1 0 1 2 3
of the true soil sensitivity due to partial remolding during the H/s u0D

initial penetration phase (Zhou & Randolph 2009, Yafrate et (a) St = 1


al. 2009), or (ii) additional loss of soil strength due to water 0
entrainment into the near-surface soil. (b) St=2.3
When cyclic T-bar tests are used to determine the 0.1

sensitivity factor, the T-bar is usually cycled up and down at 0.2

depth within the soil since the T-bar capacity factor is based

w/D
0.3
on the fully local failure mechanism. This local mechanism
0.4
cannot be mobilized if the T-bar is close to the soil surface.
Although the soil at depth is fully remolded by cyclic 0.5

movement of the T-bar, the event takes place in undrained 0.6


conditions. Different from T-bar tests, on-bottom pipelines lay -2 -1.5 -1 -0.5 0 0.5 1 1.5 2
H/s u0D
on the soil surface, where the soil has access to the free water
which can be mixed with the soil during horizontal (b) St = 2.3
oscillations of the pipe. The surficial soil can be weakened by 0
the increasing water content, leading to a higher apparent 0.1
(c) St=4

sensitivity than at depth.


Two further analyses are shown in Fig. 3(c) and 3(d), with 0.2

higher values of soil sensitivity: St = 4 and 5. In Fig. 3(a) to


w/D

0.3

3(d), all of the final values of pipe embedment after 40 cycles 0.4
are close to the experimental result of w/D = 0.52, but the
0.5
maximum horizontal resistance decreases with increasing
sensitivity factor. In order to replicate the observed horizontal 0.6

resistance accurately, the sensitivity factor found by means of -2 -1.5 -1 -0.5 0


H/s u0D
0.5 1 1.5 2

a deep T-bar test needs to be modified to consider the effect of


water entrainment. There is currently no established routine (c) St = 4
for capturing the changing strength of clay through this type 0

of remolding, with a change in water content in addition to the 0.1


(d) St=5

accumulation of shear strain.


0.2
In Fig. 3(c) and 3(d) the lay response is broadly consistent
w/D

with the centrifuge test results (Fig. 2), but the maximum 0.3

horizontal resistance Hmax/su0D = 1.22, remains approximately 0.4


35% higher than the experimental value. This discrepancy
0.5
may partly be attributed to the uncertainty surrounding the
sensitivity factor but it should also be noted that the 0.6

interpretation of the near-surface strength profile (within the -2 -1.5 -1 -0.5 0


H/s u0D
0.5 1 1.5 2

upper 10 mm of the sample) is also subject to uncertainty in


the centrifuge test. The final embedment for the case of St = 4 (d) St = 5
is 0.49D, which is very slightly shallower than the value of Figure 3. Numerical simulations of test KC-05, using
0.5D found for St = 2.3. The reason for greater embedment different values of soil sensitivity
being recorded for the less sensitive soil is that the actual
value of monotonic vertical load achieved numerically in the
former case is 4% higher. As will be discussed later, the
dynamic lay embedment is sensitive to monotonic vertical
load.

5 Copyright 2009 by ASME


SOFTENING
SOFTENING
0.95
0.95 0.90
0.90
0.86 0.84
0.81 0.79
1 0.77 1 0.73
0.72 0.68
0.68 0.63
0.63 0.57
0.59 0.52
0.54 0.47
0 0.50 0.41
z/D

0.45 0 0.36

z/D
0.30
0.25

-1
-1

-2
-2 -1 0 1 2
x/D -2
-2 -1 0 1 2
Prior to horizontal cycles x/D
th
SOFTENING At 30 cycle
0.95
0.90 SOFTENING
0.84 0.95
0.79 0.90
1 0.73 0.84
0.68 0.79
0.63 1 0.73
0.57 0.68
0.52 0.63
0.47 0.57
0.41 0.52
0 0.36
z/D

0.47
0.30 0.41
0.25 0

z/D
0.36
0.30
0.25

-1

-1

-2
-2 -1 0 1 2
x/D -2
th -2 -1 0 1 2
At 5 cycle x/D
SOFTENING th
0.95 At 40 cycle
0.90
0.84
0.79
Figure 4. Distribution of remolding during lay
1 0.73
0.68
process (KC05)
0.63
0.57
0.52
0.47
0.41
0
The distribution of the soil strength relative to the intact
z/D

0.36
0.30
0.25
value is indicated in Fig. 4 for different stages in the analysis,
with St = 4. The weakening of the soil due to the horizontal
-1
oscillations of the pipe is clearly evident. The strain softening
appears around the pipe-soil interface during monotonic
-2
penetration and none of the soil is fully remolded. The
-2 -1 0
x/D
1 2
softening zone extends around the shoulders of the pipe
th
At 10 cycle during the horizontal cycles, and the soil in the heaved zones
SOFTENING
0.95
becomes heavily remolded. These zones are pushed outwards
0.90
0.84
gradually, becoming flatter.
1
0.79
0.73 The area of the softened zone remains approximately
0.68
0.63 constant after 30 cycles, which is the same stage at which the
0.57
0.52
0.47
rate of pipe embedment reduced (Fig. 3(c)). The small-
0
0.41
amplitude cycles generate a triangular pool of softened soil,
z/D

0.36
0.30
0.25 within which the soil flow mechanism becomes localized. The
maximum width and depth (below the pipe invert) of the
-1 heavily remolded zone are 1.6D and 0.2D, respectively.

-2
-2 -1 0 1 2
x/D
th
At 20 cycle

6 Copyright 2009 by ASME


Number of cycles
for the simulated load level, which corresponds to a
0 10 20 30 40 monotonic embedment in the range of 0.03D to 0.19D. Any
0
additional cycles would not significantly deepen the pipe. The
steady embedment is an important concept for the assessment
0.1 of dynamic embedment in practice, since it narrows down the
HP06
0.2 input parameters necessary for a simplified analysis of the lay
0.3
KC04 process. It may not be necessary to specify precisely the
w/D

number of cycles that will be imposed during the lay process


0.4
KC05 if that number is sufficient for the embedment to approach the
0.5 steady value.
0.6 HP07 The final embedment from the centrifuge tests and the
numerical simulations are compared in Table 3. The
0.7
embedment depths predicted by the LDFE approach are
(a) LDFE analyses smaller than the measured values, but the average discrepancy
Number of cycles
is less than 10%. The numerical error for the HP tests is higher
0 10 20 30 40 than for the KC tests. This may be due to two reasons: (1) the
surficial strength of HP profile may not have been assessed
0
accurately, (2) an additional reduction in strength due to water
HP06
0.1 entrainment during the remolding process is not incorporated
0.2 accurately within the sensitivity factor adopted in the LDFE
0.3
analyses.
w/D

KC04
0.4 Table 3. Comparison of final embedment ratio, w/D
0.5 Test identifier KC04 KC05 HP06 HP07
KC05
0.6 Centrifuge model 0.34 0.52 0.44 0.70
HP07
0.7 LDFE analyses 0.33 0.49 0.37 0.63
(b) Centrifuge test results (Cheuk & White 2008) Difference (%) 2.9 5.8 15.9 10.0

Figure 5. Cyclic evolution of pipe embedment


Parametric study of the effect of pipe weight
Comparison with experimental data A final set of analyses were conducted to explore in more
The evolution of pipe embedment found by the LDFE detail the effect of pipe weight on the dynamic embedment
method is compared with the experimental data in Fig. 5. A behavior. Fig. 6 shows the accumulation of embedment with
sensitivity factor of St = 4 has been adopted for both clays in displacement cycles of 0.05D under different monotonic
these LDFE analyses, which used values of pipe weight and vertical loads, ranging from V/sumD = 0.98 up to
intact strength profiles that replicated all four centrifuge tests V/sumD = 3.56. The markers show the pipe embedment at the
reported by Cheuk & White (2008). end of each cycle. In each case, the pipe is monotonically
The pipes penetrate rapidly during the first 10 horizontal penetrated to the specified vertical load then is cycled with a
cycles in the LDFE analyses. The ratios of the dynamic horizontal displacement of 0.05D on soil with the KC
embedment after 10 cycles to the final dynamic embedment strength profile (Table 1), using a sensitivity of St = 4. The
are 74% (HP06), 66% (HP07), 65% (KC04) and 77% (KC05), dynamic embedment is significantly affected by the
respectively. In particular, the first five cycles can triple monotonic vertical load. In normalized vertical load
(KC04, KC05) or double (HP06, HP07) the monotonic embedment space, the responses move downwards from the
embedment. monotonic embedment curve. The shape of these curves arises
In contrast, the final 10 of the 40 cycles give rise to only a from the changing intact soil strength with depth. The vertical
slight increase in embedment. For example, the embedment load, V, was held constant during the dynamic movement.
increment of KC05 is 0.006D, about 1.5% of the final The embedment achieved after 40 cycles of lateral
dynamic embedment. movement exceeds the embedment under the same monotonic
Similar results are apparent from the centrifuge model vertical load by a factor of 28 for the lowest vertical load case
tests. Compared with the numerical results, the initial rate of (V/sumD = 0.98) but by a factor of only 4 for the highest load
embedment is slightly less rapid, and the rate of embedment case (V/sumD = 3.56). This ratio between the static and
during the final cycles is slightly higher. dynamic embedment is commonly termed the dynamic
For the horizontal cyclic displacement of 0.05D, all embedment factor, fdyn (Bruton et al. 2006), but does not
pipes approached a steady embedment after 30 cycles, at least

7 Copyright 2009 by ASME


appear to give a consistent multiplicative factor to capture the vicinity of the pipe becomes grossly distorted, and the loss of
effect of the same dynamic motion on pipes of differing soil strength through heavy remolding cannot be captured. A
weight. large deformation method, RITSS, based on mesh
A better strategy for describing the additional embedment regeneration, was developed to investigate pipe-soil
created by the dynamic motion is to compare the dynamic interaction under complex pipe movements. A strain softening
embedment (which arises from remolding of the soil) with the soil model was incorporated into RITSS to capture the soil
monotonic embedment that would result from the soil initially remolding during dynamic pipe laying. The robustness of the
being in the full remolded state. The monotonic penetration large deformation method is verified by comparison with an
resistance from an LDFE analysis based on a non-softening upper bound solution for monotonic penetration of cylindrical
soil model using the fully-remolded strength profile (i.e. the objects such as pipelines.
intact strength divided by St = 4) is shown dotted in Fig. 6. LDFE studies of dynamic pipeline laying, were conducted
The dynamic embedment reached after the cyclic phases is to replicate the behavior observed in centrifuge model tests of
comparable to the monotonic embedment in the fully softened this process. The basic form of the tests and simulations is for
soil under a vertical load which is approximately 0.5 2.0 the pipe to be initially embedded monotonically to a specified
times higher (length OA vs. OA). vertical load (or simulated weight), before many cycles of
small amplitude lateral displacement (0.05D) are imposed, to
V/s u0D represent the dynamic effect.
0 1 2 3 4 5 6 From comparison of the two sets of simulations, it was
0
found that the sensitivity factor of the soil may be
Monotonic embedment underestimated by cyclic T-bar tests in deep soil, since
0.1 additional water becomes entrained in the soil with
0.2 oscillations of the pipe. The soil is then softened due to both
0.3 the remolding effect of accumulated shear strain and also
w/D

0.4
increasing water content. The lower lateral resistance
A'
encountered in the centrifuge test compared to the LDFE is
0.5 O attributed to this mechanism. To achieve closer agreement, the
A
0.6 Monotonic embedment impact of the water content on the soil strength is captured in
0.7 in fully-remoulded soil a simple manner by increasing the sensitivity factor. It is
shown that the steady dynamic embedment is less affected by
the sensitivity factor, compared to the horizontal resistance,
Figure 6. Cyclic accumulation of embedment for which decreases as the sensitivity factor is raised. For pipes in
different values of pipe weight kaolin clay and the offshore high plasticity clay simulated
here, the numerical evolution of pipe embedment agrees well
This observation is consistent with an interpretation that with the test data.
links the steady embedment during dynamic motion to the When small amplitude lateral cycles are imposed on the
parallel point of the corresponding theoretical failure envelope pipe, the embedment increases rapidly within the first 5
in vertical-horizontal load space. The parallel point is the cycles, with the embedment exceeding the monotonic value by
vertical load level at which the trajectory of pipe movement at a factor of 2 to 3. For a horizontal amplitude of 0.05D, an
failure under horizontal load is purely horizontal (i.e. the pipe approximately steady embedment was reached in all cases at
has no tendency to rise or fall), according to a flow rule based 30 cycles. At this stage, the LDFE analyses show that a
on normality, combined with failure envelopes derived from softened zone had formed around the pipe and remained
plasticity limit analysis. This approach is explored in more largely unchanged with subsequent cycles. Overall, the LDFE
detail by Cheuk & White (2009). Based on these observations, method has shown good potential for simulating the dynamic
it may be possible to improve the prediction of dynamic embedment of seabed pipelines.
pipeline embedment by adopting the remolded rather than
intact soil strength in embedment calculations, and by ACKNOWLEDGEMENTS
assuming that the soil reaches a lateral failure condition during This work forms part of the activities of the Centre for
laying. Offshore Foundation System at UWA, which was established
under the Australian Research Councils Special Research
CONCLUSIONS Centre scheme and is now supported by the State Government
Pipelines laid on the seabed usually embed more deeply than of Western Australia through the Centre of Excellence in
indicated by static penetration calculations due to the dynamic Science and Innovation program. This financial support is
motion as the pipe touches down particularly horizontal gratefully acknowledged.
oscillations. Conventional small strain finite element methods
cannot simulate the pipe lay process since the mesh in the

8 Copyright 2009 by ASME


REFERENCES of plate anchor. China Ocean Engineering, 20(2): 269-
Bruton, D. A. S., Carr, M. C. and White, D. J. (2007). The 279.
influence of pipe-soil interaction on lateral buckling Wang, D., Hu, Y. and Randolph, M. F. (2008a). Three-
and walking of pipelines: the SAFEBUCK JIP. dimensional large deformation analyses of plate anchor
Proceedings of 6th International Conference on keying in clay. Proceedings of 27th International
Offshore Site Investigation and Geotechnics, London, Conference on Offshore Mechanics and Arctic
133-150. Engineering, OMAE2008-57218.
Bruton, D. A. S., White, D. J., Cheuk, C. Y., Bolton, M. D. Wang, D., Hu, Y. and Randolph, M. F. (2008b). Effect of
and Carr, M. C. (2006). Pipe-soil interaction loading rate on the uplift capacity of plate anchors.
behaviour during lateral buckling, including large Proceedings of the 18th International Offshore and
amplitude cyclic displacement tests by the Polar Engineering Conference, 2: 727-731.
SAFEBUCK JIP. Proceedings of Offshore Technology Wang, D., White, D. J. and Randolph, M. F. (2008c). Large
Conference, OTC17944. deformation finite element analysis of pipe penetration
Cheuk, C. Y. and White, D. J. (2008). Centrifuge modelling of and large-amplitude lateral displacement. Submitted to
pipe penetration due to dynamic lay effects. Canadian Geotechnical Journal, September 2008.
Proceedings of 27th International Conference on White, D. J., Take, W. A. and Bolton, M. D. (2003). Soil
Offshore Mechanics and Arctic Engineering, deformation measurement using Particle Image
OMAE2008-57923. Velocimetry (PIV) and photogrammetry. Gotechnique,
Cheuk, C. Y. and White D. J. (2009). Modelling the dynamic 53(7): 619-631.
embedment of seabed pipelines. Gotechnique. Yafrate, N., DeJong, J., DeGroot, D. and Randolph, M. F.
submitted December 2008, in review. (2009). Assessment of remolded shear strength in soft
Dingle, H. R. C., White, D. J. and Gaudin, C. (2008). clay with full flow penetrometers. Journal of
Mechanisms of pipe embedment and lateral breakout Geotechnical and Geoenvironmental Engineering,
on soft clay. Canadian Geotechnical Journal, 45: 636- ASCE (under review, revised October 2008).
652. Zhou, H. and Randolph, M. F. (2007) Computational
HKS (2004). ABAQUS Users Manual, Version 6.5, Hibbit, techniques and shear band development for cylindrical
Karlsson and Sorensen, Inc. and spherical penetrometers in strain-softening clay.
Hu, Y. and Randolph, M. F. (1998). A practical numerical International Journal of Geomechanics, 7(4): 287-295.
approach for large deformation problem in soil. Zhou, H. and Randolph, M. F. (2009). Numerical
International Journal for Numerical and Analytical investigations into cycling of full-flow penetrometers
Methods in Geomechanics, 22(5): 327-350. in soft clay. Gotechnique (accepted December 2008).
Lund, K. H. (2000). Effect of increase in pipeline soil Zienkiewicz, O. C. and Zhu, J. Z. (1993). The
penetration from installation. Proceedings of superconvergent patch recovery and a posteriori error
International Conference on Offshore Mechanics and estimates. Part 1: The recovery technique. International
Arctic Engineering, New Orleans, OMAE2000- Journal for Numerical Methods in Engineering, 33:
PIPE5047 1331-1364.
Murff, J. D., Wagner, D. A and Randolph, M. F. (1989). Pipe
penetration in cohesive soil. Gotechnique, 39(2), 213-
229.
Randolph, M. F., Wang, D., Zhou, H., Hossain, M. S., and Hu,
Y. (2008). Large deformation finite element analysis
for offshore applications. Proceedings of 12th
International Conference of International Association
for Computer Methods and Advances in
Geomechanics, 3307-3318.
Randolph M. F. and White D. J. (2008a). Pipeline embedment
in deep water: processes and quantitative assessment.
Proceedings of Offshore Technology Conference,
Houston, USA. Paper OTC19128.
Randolph, M. F. and White, D. J. (2008b). Upper-bound yield
envelopes for pipelines at shallow embedment in clay.
Gotechnique, 58(4): 297-301.
Wang, D., Hu, Y. and Jin, X. (2006). Two-dimensional large
deformation finite element analysis for the pulling-up

9 Copyright 2009 by ASME

View publication stats

You might also like