You are on page 1of 10

ICSE6 Paris - August 27-31, 2012 Griffiths et. al.

ICSE6-199

Development of a Novel 2D Pipe-Soil-Fluid Interaction Model for


Subsea Pipeline Stability Design

Terry GRIFFITHS, Mengmeng XU, Wenwen SHEN, Marvin RAMSAWMY,


Amar SHAH

Wood Group Kenny


Wood Group House, 432 Murray Street, Perth, Australia - e-mail :terry.griffiths@jpkenny.com.au

For over a decade it has been recognised that our existing models and tools for subsea pipeline stability
design fail to account for the fact that non-cohesive seabed soils tend to become mobile before the onset of
pipeline instability. Despite ample evidence from both laboratory and field observations that sediment
mobility has an important role to play in pipeline/soil interaction, very few models have been presented
which account for the tripartite interaction between the fluid and the pipe, the fluid and the soil, and the pipe
and the soil.
This paper presents details of a novel 2D pipe-soil-fluid (PSF) interaction algorithm, which has been
developed to offer a more realistic and accurate model of the evolution of soil profiles around the pipeline
compared to existing hysteresis friction spring approaches (such as the Verley model) which ignore sediment
transport and scour. The PSF model has been specifically developed to minimise computational cost
compared to continuum soil FEA approaches, but still enable the profile of the soil around the pipe to be
established. To achieve the simplicity of the model, a large number of parametric 2D CFD models were run
to generate seabed shear stress profiles as a function of seabed and pipe geometry under different wave and
current flow conditions. By developing algorithmic approaches based on Shield's criteria to replicate the
results of these CFD analysis results, the PSF model also incorporates sediment suspension and
transportation into a pipe-soil response model, without requiring the concurrent solution of the Navier
Stokes equations in a CFD model.
The model has significant potential to be of use to operators who struggle with conventional stabilisation
techniques for the pipelines, such as those which cross Australia's North West Shelf, where shallow water
depths, highly variable calcareous soils and extreme metocean conditions driven by frequent tropical
cyclones result in the requirement for expensive and logistically challenging secondary stabilisation
measures.

Key words
Pipe-Soil-Fluid Interaction; Scour; Erosion; Subsea Pipeline Stability.

I NOMENCLATURE
The following symbols and nomenclature are used in this paper:

a, b,w Sinusoidal Component Coefficients Fr Passive Horizontal Soil Reaction Force (N/m)
AMP Sinusoidal Component Amplitude Fv Vertical Soil Reaction Force (N/m)
AOF Sinusoidal Component Amplitude Offset grav Gravitational Acceleration (m/s2)
Ca Suspended Sediment Volumetric Concentration KC Keulegan Carpenter Number
CFD Computational Fluid Dynamics LES Large Eddy Simulation
CLM Sinusoidal Component Time Length Multiplier Lb Soil Berm Lateral Offset from Pipe / Pipe Diameter
d50 Median Grain Diameter (m) LSP Sinusoidal Component Length Start Point
D Pipe Hydrodynamic Diameter (m) Lwave Wave Length (m)
Dr Relative Soil Density NPos Soil Nodal Position
Dstar Non-dimensional Grain Diameter PAS Sinusoidal Component Wave Phase Angle Start
FE Finite Element PSF Pipe-Soil-Fluid

1141
ICSE6 Paris - August 27-31, 2012 Griffiths et. al.

T Wave Period (s) Zp Pipe Elevation / Pipe Diameter (up=+ve)


TA Sinusoidal Component Time Amplitude Zs Soil Elevation at Pipe Position / Pipe Diameter
TAO Sinusoidal Component Time Amplitude Offset Steady / Total Current Ratio (Wave=0, Current=1)
Uc Steady Current Velocity (m/s at 1m above seabed) Rouse Parameter
Ustar Shear Velocity (m/s) cr Critical Shields Number
Uw Wave Velocity (m/s at 1m above seabed) Von Karmans Constant (=0.4)
Vloc Local Velocity (m/s) Seabed - Pipe Coulomb Friction
ws Sediment Fall Velocity (m/s) Seawater Kinematic Viscosity (m/s2)
y Axis Along Seabed Perpendicular to Pipe Axis
s, w Soil grain and seawater density (kg/m3)
z Elevation Axis Above Seabed (up = +ve) w
Wave Phase Angle
za Reference elevation for Ca (m)
Zb Soil Berm Elevation / Pipe Diameter (up=+ve)

II INTRODUCTION
The present state-of-the-art in pipeline stability design consists of either applying empirically calibrated
design methods or undertaking dynamic time-domain FE simulations of pipeline response to simulated
metocean conditions, as is well summarized by Zeitoun et. al. (2008, 2009). Either approach may be
employed under the DNV subsea pipeline design code and associated family of recommended practices, of
which DNV-RP-F109 (DNV, 2011) addresses on-bottom stability.
Conventional pipeline stability design methods treat the seabed as stable with respect to all fluid-induced
interactions. As described in Zeitoun et. al. (2008), non-cohesive pipe-soil interaction is almost always
modelled using either a simplified Coulomb friction response or the Verley model, which incorporates
history-dependent non-linear hysteresis resistance. The Verley model does not account for any sediment
transport or piping (tunnel erosion) influencing the implied seabed profile and therefore the passive
resistance predicted. Seminal work by various researchers including Palmer, Teh and Damgaard brought to
the subsea pipeline industrys attention a fundamental issue with the above approaches and their omission to
account for seabed mobility, described in the seminal paper A Flaw in the Conventional Approach to
Stability Design of Seabed Pipelines (Palmer, 1996). In essence it concluded that it must follow that the
seabed must become grossly unstable long before the extreme design conditions. The traditional model is
irrelevant: it makes no sense to consider the stability of a stationary pipeline on a stationary seabed
Despite the absence of well documented methods to account for these effects in stability design, theoretical
and empirical models of sediment transport in coastal and river systems are well developed, including a
widely used model for predicting the onset of seabed sand motion by Shields from around 1936. These are
summarized well by Soulsby (1997).
In 2000, Li and Cheng (2000) reported on the development of a RANS-based CFD solver, which used LES
for its turbulence closure model and the suspended sediment concentration near the seabed are predicted
using an empirical relationship by Zyserman and Fredsoe which is based on Shields parameter and hence
related to seabed shear stress. Further information on this model has been presented by Li and Cheng (2001),
considering numerical tests of a pipe approaching an erodible seabed.
These publications demonstrate the capability for CFD to be used to model sediment transport around
subsea pipelines, however the approach is anticipated to be of limited use in engineering applications as a
design tool due to the relatively high computational demands of CFD. Dynamic stability design engineering
models typically consider a 3-hour simulation with a pipe section length of 1000m in order to capture the 3D
effects of finite wave crest lengths and inclined attack angles to the pipe. The conclusion is inescapable that
to extend the CFD approach to modelling pipeline stability in real engineering applications is impractical
without a step-change in computational capacity such as large-scale use of low cost processors.
To be successful, therefore, it is perceived that an engineering model is required which is capable of
modelling the tripartite interaction between pipe, soil and fluid. This model must include sediment transport
and evolution of the seabed profile, without this being directly linked to the concurrent solution of CFD for
fluid flow modelling or FE in the soil domain. The PSF model presented in this paper was conceived by
posing the question: What is the simplest model which is capable of following the seabed profile either side
of the pipe, without requiring CFD in the fluid domain, nor FE in the soil domain, but is capable of tracking
sediment transport and predicting seabed reaction forces?

1142
ICSE6 Paris - August 27-31, 2012 Griffiths et. al.

III PSF MODEL

III.1 PSF Model Structure


In 2D, the PSF model consists of an arbitrary series of soil nodes either side of the pipe, with the option for
mesh refinement in the vicinity of the pipe. Each node has only 2 data associated with each node to be stored
and passed from one timestep to the next: soil elevation; and suspended sediment concentration. In addition,
at each node located within the vertical envelope of the pipe, the vertical position of the pipe upper and lower
surface is tracked.
The model presently assumes non-cohesive sediment which is dominated by scour and erosion rather than
liquefaction processes. There is potential to extend the model to include liquefaction or cohesive soils in
future developments.

III.2 PSF Model Analysis Sequence


PSF model analysis sequence is presented in Figure 1. This flowchart structure is used to calculate the
hydrodynamic and soil reaction forces on the pipe and iteratively determine the resulting pipeline vertical
and horizontal dynamic response. The model can be implemented iteratively to solve for the lateral and
vertical pipe displacements, or one or both of these can be prescribed. The key elements of the PSF
interaction are described in detail in the following sections.

Previous time Wave /


step / Initial current
Condition velocity
timeseries

Estimate new pipe Interpolate Update new Update Calculate new pipe based on:
acceleration, new soil soil profile to new soil Hydrodynamic loads
velocity and profile from account for profile to
position based on old profile pipe motion account for Soil reaction loads
loads (sweep, scour Accounting for embedment, trench
suck, slump) profile, and pipe velocity

No Yes
Converged? Next time step

Figure 1: PSF Model Flowchart

III.3 Pipe-Soil Interaction Soil Deformation


The PSF model structure incorporates the option for mesh refinement of the soil nodes in the vicinity of the
pipe and depending on the case being considered, potentially large lateral pipe displacements could be
required. Therefore the model operates on the basis of the mesh acting as a moving reference frame centred
on the pipe. The first analysis step is therefore to interpolate from the seabed profile at the preceding time-
step to the same profile at the present time-step. The algorithm applies a weighted interpolation to conserve
soil volume.
The model calculates the deformation of the soil profile using a 3-step process which obeys 3 governing
rules:
1. Soil cannot be created or destroyed (conservation of volume);
2. Soil cannot exist inside the pipe;
3. Soil cannot maintain a profile exceeding its natural angle of repose (unless it is supported by the
pipe).
The first step uses an explicit method to calculate the swept area of sediment in front of the pipe. The
displaced area of sand is placed to the side of the pipe in the direction of motion as shown in Figure 2. The
second step checks to see whether there is a void created behind the pipe into which sand would be sucked.
Any suck area required to fill the void is drawn from either side of the pipe, again apportioned between
each side depending on the direction of pipe movement. By this means, the model is able to handle a wide

1143
ICSE6 Paris - August 27-31, 2012 Griffiths et. al.

variety of conditions, including transitions from/to full burial or partial embedment, spanning above the
seabed, and motion in any direction.
1 1

Pipe Pipe
0.8
A 0.8
B
Centre Centre

0.6 0.6

0.4 0.4

0.2 0.2

0 0
-1.5 -1.25 -1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1 1.25 1.5 -1.5 -1.25 -1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1 1.25 1.5

-0.2 -0.2
1 1

Pipe Pipe
0.8 C 0.8 D
Centre Centre

0.6 0.6

0.4 0.4

0.2 0.2

0 0
-1.5 -1.25 -1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1 1.25 1.5 -1.5 -1.25 -1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1 1.25 1.5

-0.2 -0.2
1 1

Pipe Pipe
0.8 E 0.8
F
Centre Centre

0.6 0.6

0.4 0.4

0.2 0.2

0 0
-1.5 -1.25 -1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1 1.25 1.5 -1.5 -1.25 -1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1 1.25 1.5

-0.2 -0.2
1 1

Pipe
G Pipe
H
0.8 0.8
Centre Centre

0.6 0.6

0.4 0.4

0.2 0.2

0 0
-1.5 -1.25 -1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1 1.25 1.5 -1.5 -1.25 -1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1 1.25 1.5

-0.2 -0.2

Figure 2: Soil Sweep, Suck and Slump Volume Conservation

The final step in calculating the soil profile displaced by the motion of the pipe is to allow the soil to slump
should it exceed the internal repose angle of the soil, through a mechanism which distributes a proportion of
that nodes sand to any nodes either side which are lower than it. This is the only part of the algorithm which
requires an iterative approach, with experience suggesting around 10 iterations is sufficient when around
50% of the sand is slumped each iteration. This soil deformation model is described in more detail in
Griffiths (2012).

III.4 Pipe-Soil Interaction Soil Reaction Loads


A number of conventional established methods are incorporated into the PSF model to estimate the soil
vertical and horizontal reaction forces. In fact, a wide variety of soil reaction models can easily be included.
The PSF model considers the profile of the soil around the pipe and calculates the contact angle on the left
and right sides of the pipe to provide the embedment depth and bearing width. It should be noted that in most
of the publications referenced, the soil reaction forces have been published with respect to pipe penetration

1144
ICSE6 Paris - August 27-31, 2012 Griffiths et. al.

although careful review shows that this refers to the vertical downward movement of the pipe from initial
point contact on a flat seabed (defined as Zp later in this paper, see Figure 3). Since the PSF model finds the
contact points between the pipe relative to the locally observed soil level (which might result from sediment
transport or pipe movement), the model takes this value and converts it to an equivalent penetration for
calculation of soil forces.
For the soil vertical reaction force, one of the theoretical methods included is to invert equations 3.28 and
3.24 from F109 (DNV, 2011) for passive resistance and initial penetration in sand. The model also provides
for including an elastic component of the vertical reaction force to aid in the numerical stability of the
solution. The elastic mobilization distance is user-defined, but generally a value of 5% of D appears to work
well.
For the horizontal reaction force, the penetration depth is calculated based on the soil contact angle in the
direction of pipe motion. One of the theoretical models included within the PSF model is taken from Verley
et. al. (1990), where the soil reaction force comprises frictional and passive components, where the passive
resistance Fr is given in Eq. (1) and k1 given by Eq. (2), which differs slightly to the coefficients given in
AGA (PRCI, 1993). The parameters and values for k2 and K are also taken from Verley et. al. (1990).

Fr 1.5 2
k1 + k2 K
z k1 0.6 Dr 1.4 Dr + 0.79
(2)
Fv D
(1)

III.5 Pipe-Fluid Interaction Hydrodynamic Loads


The PSF model is able to make use of the same pre-processing tool as used for J P Kennys SIMULATOR
software as described by Zeitoun et. al. (2009), to generate a near-bed random wave velocity time-history
and force time-history. Both the velocities and forces are based on the pipeline sat proud on a flat and rigid
seabed.
Load reduction factors to account for trench shielding, partial embedment, spanning and pipe movement
are calculated based on DHI (Jacobsen, 1988b, Jacobsen et. al. 1989) and by Wilkinson et. al. (1988). The
seabed profile is analysed to determine the degree of embedment if the pipeline is contacting the seabed, or
gap if not touching, and the height and angle of any trench to be estimated.

III.6 Fluid-Soil Interaction Sediment Transport Model


Sediment transport theoretical basis
The present fluid-soil interaction model is focused on sediment transport and it does not yet account for
cyclic-loading induced liquefaction of the soil. The model is based on a similar approach described by Li and
Cheng (2000, 2001) and draws extensively on the work of Soulsby (1997). The approach is to use Shields
criteria to predict the onset of motion, which for a flat seabed is given by Eq. (3), where Dstar is defined in Li
and Cheng (2001) including their method for accounting for the effect of seabed slope angle. The behaviour
of Calcareous soils can be accounted for by project-specific modifications to the coefficients of this formula.
0.30
cr
1 + 1.2 Dstar
( (
+ 0.055 1 exp 0.020 Dstar ))
(3)
The critical Shields criteria for onset of sediment motion can be related to the friction or shear velocity
using Eq. (4) which can be related to the current velocity at a given reference height above the bed using a
logarithmic profile.
2
Ustar
cr
s
grav 1 d 50
w (4)
Once the critical Shields number is exceeded, the sediment is transported by 2 different methods
bedload transport where the grains remain essentially in contact with the bed (taken to be less than 2.d50

1145
ICSE6 Paris - August 27-31, 2012 Griffiths et. al.

above the bed); and suspended sediment transport. The PSF model includes bedload transport in accordance
with Li and Cheng (2001) and Soulsby (1997).
The equilibrium volumetric concentration of suspended sediment Ca above the seabed is given by Eq. (5) at
an elevation za taken to be equal to 2.d50. The suspended sediment concentration profile is then found using
Eq. (6) using the Rouse coefficient () defined by Soulsby (1997). The total volume of suspended sediment
above each node is then found by integration.


(
0.331 s cr ) 1.75 z h za
Ca C( z) Ca
1 + 0.720 ( s cr)
1.75
za h z (6)
(5)

All of the above enables us to predict the mass of suspended sediment above a flat seabed, which when
combined with the fluid velocity and bedload transport gives us the rate of advection of sediment. In order
for this to be applied to a non-flat seabed in the presence of a pipeline, the PSF model takes the approach of
using CFD models to map the distribution of seabed shear stresses on both sides of the pipe. Making use of
the relationship between seabed shear and friction velocity, the contour of seabed shear stress can be
converted back into a local velocity contour around the pipe which is then used to model sediment
transport the local velocity yields both the seabed shear stress and hence the mass of suspended sediment,
and also the local advection rate of that mobile sediment mass. The accretion or erosion of the seabed at each
soil node can therefore be found at each time-step by comparing the equilibrium suspended sediment at that
time at that node with the net suspended sediment which has been advected in and out of that node.
One enhancement to this method is applied by curtailing the rate at which the suspended sediment can be
redeposited onto the seabed. It was found that on occasion there could be very rapid changes in seabed shear
stress, leading to unrealistically high redeposition. By calculating the time taken for sediment to advect from
one node to the next, the settling velocity can be used to find a maximum distance by which the suspended
sediment concentration profile may translate down.
CFD modelling of seabed shear stresses and local velocities
In order to predict sediment transport, the requirement is therefore to have a map of the seabed shear
stresses around an arbitrary seabed profile accounting for pipeline embedment, spanning or trenching (Figure
3). This was accomplished by drawing on the results of a number of student projects (Xu, 2010 and Shen,
2011) which undertook a parametric range of CFD models including both geometric parameters and a variety
of wave and current conditions. The student projects were undertaken through the UWA School of Civil
Engineering as Cooperative Education for Enterprise Development (CEED) projects. The models were
developed using the parameters defined in Figure 3. The ranges of each parameters considered to date are
presented in Table 1.
An example of the CFD model results is presented in Figure 4 for a spanning pipeline (Case 24). The
colour contours above the seabed are y-velocity component, while the purple areas below the seabed indicate
the magnitude of seabed shear stress. The high velocities under the small span gap are clearly associated with
very high seabed shear around the seabed under the pipe. While not the direct focus of the CFD analysis,
information on the variation of hydrodynamic forces on subsea pipelines was also able to be extracted
(Griffiths et. al. 2012).

All parameters non- Parameter Lower Bound Upper Bound


dimensionalised by D D (m) 0.2 1.2
Lb
z-axis

Zp -1.5 1.5
Zb Zs -1.5 1
Zs Zp
D Lb 1 5

y-axis Zb 0 0.5
Uc (m/s) 0 1.5
Uw (m/s) 0 2.5
T (s) 8 16

Figure 3: 2D Parametric Pipe / Seabed Geometry Table 1: CFD Case Parameter Range

1146
ICSE6 Paris - August 27-31, 2012 Griffiths et. al.

Y-velocity

Figure 4: Snapshot of Velocity Contours and Seabed Shear Stress Around Pipe (Xu, 2010)

Scaling investigation of CFD local velocity results


Having converted the seabed shear stresses to equivalent local velocities along the seabed, the results of
different cases were compared as shown in Figure 5 (left). This figure shows the results for cases with the
pipe resting proud on a flat seabed over a range of steady current velocities and pipe diameters. By scaling
the amplitude of local velocities with the far-field reference velocity and the lateral distribution with the pipe
diameter D, the results are replotted in Figure 5 (right). It can be seen that the results now show a high degree
of similarity. By digitizing these curves, the local velocities over a wide range of pipe diameters and steady
current magnitudes could be reproduced. Similar scaling comparisons have been undertaken for waves,
mixed wave and currents and for the geometric parameters shown in Figure 3.
1.5
Local Velocity / Reference Velocity

2
Local Velocity (m/s)

1.5 1

1 1 1
0.5 2
2
0.5 3 3
0
79 79
0 -10 -5 0 5 10 15 20
81 81
-40 -30 -20 -10 0 10 20 30 40 -0.5 Lateral Position / Pipe Diam
83 83
-0.5
Lateral Position (m) 85 85
-1
-1

-1.5 -1.5

Figure 5: Parametric Steady Current and Diameter Results Unscaled (Left) and Scaled (Right)

Local velocity algorithmic model


The key objective of the PSF model is to capture the sediment transport and soil deformation around the
pipe. About 100 2D parametric CFD cases have been run, representing many weeks of modelling and run-
time. This would be totally impractical as an engineering tool to run concurrently while trying to model the
stability of a subsea pipeline. An approach is therefore required to distil these local velocity profiles into a
parametric algorithm which is defined ideally in terms of only the 5 geometric and 3 flow parameters.
It was recognized that the non-dimensionalised local velocity profile presented in Figure 5(right) could be
generated by superimposing a number of sinusoidal components onto the ambient far-field reference
velocity. Figure 6(left) shows how the components can be superimposed to build up a close approximate to
the CFD results, with the comparison presented in Figure 6(right). The solid blue line shows the CFD results
for this geometry (CaseCFD), the blue dashed curve the algorithmic model (CaseMOD) and red dashed
curve shows the error associated with the algorithm for this case (CaseERR).

1147
ICSE6 Paris - August 27-31, 2012 Griffiths et. al.

Vloc (m/s)
Pipe on Flat Seabed Uc = 1m/s CaseCFD
1.5 2 CaseERR

Vloc (m/s)
1 CaseMOD
1.5
0.5
0 1
-20 -15 -10 -5 -0.5 0 5 10 15 20 25 30
y/D 0.5
-1
-1.5 0
-20 -15 -10 -5 0 5 10 15 20 25 30
-2 y/D
-0.5
1 2 3 4 5 6 -2.5
7 8 9 10 11 12 13 14 15 CaseMOD
-1

Figure 6: Superposition of Sinusoids to Form Local Velocity and Resulting Comparison with CFD Results

By extending this approach to wave cases, arbitrary sinusoidal components can be defined in time (or wave
phase angle) and space based on defining 5 coefficients: the amplitude, the length and start point laterally
around the pipe and the wave phase angle start point and phase length in time. The definition of these
generalized sinusoidal components are in accordance with Eq. (7). The wave phase coefficients PAS (Phase
Angle Start) and CLM (Component Length Multiplier), and the spatial coefficients LSP (Length Start Point)
and LSW (Length Sine Wave) are variables, as is AMP (amplitude). Other parameters are fixed. It should be
noted that only whole sinusoids are incorporated, although any part of the sinusoids extending below a wave
phase angle of 0 or above 360 is discarded and not wrapped. The sinusoids are arranged to avoid any
step-changes in amplitude in the modelled profile.
AMP sin 2 NPos LSP + + AOF TA sin w
PAS
Vloc + + TAO
LSW 2 180 CLM 2 (7)

For each sinusoidal component we can conclude that a set of 5 coefficients is needed to define the 5
geometric and 3 flow parameters from Table 1. To achieve this, the generalized equation (8) has been written
based on the scaling analysis of the CFD results which is used to calculate each of the Active coefficients
presented in Eq. (7). The equation contains the key flow parameters and contains the key geometric
parameters, where is the steady/(wave+steady) current ratio so that = 1 for pure current and 0 for pure
wave conditions.
grav T2
0.5

Param a Uc + b + ( c Uw + d ) e + f
Lwave
( 2
( g KC + h) i + j + )
k l Zp + m Zp + n o Zs + p Zs + q r Lb + s Lb + t u Zb + v Zb + w

2

2

2

2

(8)
The task is therefore to define the sinusoidal components using coefficients a to w so that they match
the CFD results. Engineering experience was applied to set up initial sinusoidal components and then an
iterative perturbation process was used to optimize the coefficients to minimize the RMS error across the
whole matrix of test case results. The optimisation process is ongoing and the RMS errors are trending
towards minimised values for both the cumulative total and individual peak error.
Note that by holding a = c = 0 for all active parameters other than AMP, and b = d = 0 for AMP, the
general equation (8) produces values for Eq. (7) which are dimensionally correct in degrees for phase
angles, normalized by D for LSW and LSP and in m/s for AMP.
Piping and spanning algorithm
It was found that the CFD cases needed to be segregated into those cases with gaps under the pipe, and
those with pipe contact on the seabed. As can be seen in Figure 4, cases with gaps under the pipe generate
very high local shear and thence scour through the gap. The coefficients which generate the sinusoidal
components have therefore been developed in 2 sets, for gap and no-gap cases. The algorithm is easily
able to test the soil profile to determine whether there is soil contact at any point along the pipe, and if not
then select the gap set of coefficients.
The algorithm also requires a method to describe when the onset of piping occurs. The model uses the
method described in F109 (DNV, 2011) based on estimating the differential pressure across the soil below
the pipe from the horizontal hydrodynamic force (drag + inertia) over the exposed area of pipe.
After extensive optimization using 24 sinusoidal components for wave and current conditions, the
algorithm is able to replicate the CFD local velocity results for a wide range of geometric and flow
conditions. An example result for steady current is presented in Figure 6, where the sinusoidal components
are actually generated from the algorithm rather than specifically for that case. An example for a pure wave
with Uw = 1m/s and T = 15s is presented in Figure 7.

1148
ICSE6 Paris - August 27-31, 2012 Griffiths et. al.

CaseCFD (m/s)

Wave Phase Angle


Series92
Series79
Series66
Series53
Series40
Series27
2.62
2.4
2.2 Series14
1.8
1.61
1.4
1.2
0.8
0.6
0.4
0.2
-0.2
-0.4
-0.60 Series1
-0.8
-1
-1.2

14
27
40
53
66
79
92
105
118
131
144
157
170
183
196
209
222
235
248
261
274
287
300
313
326
339
352
365
378
391
404
417
430
443
456
469
482
495
1
-1.4
-1.6
-1.8
-2
-2.2
-2.4
-2.6 Left Seabed Right Seabed
CaseMOD (m/s)

Wave Phase Angle


Series97
Series85
Series73
Series61
Series49
Series37
Series25
2.8
2.6
2.4
2.2 Series13
2
1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
-0.2 Series1
-0.4
-0.6
-0.8
-1
-1.2
12
23
34
45
56
67
78
89

-1.4
1

100
111
122
133
144
155
166
177
188
199
210
221
232
243
254
265
276
287
298
309
320
331
342
353
364
375
386
397
408
419
430
441
452
463
474
485
496
-1.6
-1.8
-2
-2.2
-2.4
-2.6
-2.8

Figure 7: Vloc Results for CFD and Algorithm for Wave Case 6

A number of features in this figure are worth noting, including:


The lateral axis is plotted by node number rather than distance, exaggerating the size and
significance of the pipe;
The CFD models applied orbital wave conditions to the inlet and top domain boundaries (Xu 2010,
Shen 2011), hence the wave phase can be seen to propagate through the domain. This also explains
the asymmetry between the forward and reverse half wave cycles. The wave phase is calculated at
the pipe position;
Despite a relatively complex local velocity surface, the algorithm is able to reproduce it well, with
only a limited number of components.

IV VERIFICATION AND VALIDATION


The PSF model is presently undergoing verification, validation and refinement. The sequence of pipe-soil
displacements illustrated in Figure 2 can be used to evaluate the efficacy of the algorithm for soil
conservation, with the final step in this sequence having a soil deficit equal to 1.08% of the pipe cross-
sectional area.
A number of benchmark tests have been performed, including comparing the PSF model to the results
obtained by Brennodden, Verley et. al. (Brennodden 1989) as presented in Table 2. As shown in Figure 2 (in
still fluid), the PSF model replicates the general behaviour predicted by the Verley model when pipeline
lateral deflections are prescribed. Good agreement is observed for the peak lateral resistances, although some
differences are apparent in pipe penetration for hard soil.

D Wsub Pipe Penetration (cm) Peak Lateral Resistance (kN/m)


(m) (kN/m) Brennodden PSF Diff % Brennodden PSF Diff %
3
Soft Soil (9 Repose Angle, Dr = 0.05, = 18.4kN/m )

0.5 0.25 0.4 0.395 -1% 0.34 0.35 3%

0.5 0.5 0.6 0.56 -7% 0.66 0.71 7%

1 1 1.6 1.24 -23% 1.4 1.43 2%

1 2 1.8 1.79 -1% 3 2.84 -6%

Hard Soil (35 Repose Angle, Dr = 0.48, = 19.4kN/m3)

1 1 0.3 1.01 237% 0.94 0.90 4%

1 2 0.4 1.22 205% 1.71 1.80 -5%

Table 2: Static Load Penetration and Breakout

1149
ICSE6 Paris - August 27-31, 2012 Griffiths et. al.

V CONCLUSIONS
A PSF model has been developed to offer a computationally much more efficient approach than CFD
modelling to capture the effects of scour, piping and erosion as very important factors affecting the stability
of subsea pipelines. While still undergoing further development and refinement, it is clear that the model has
potential to offer the subsea pipeline industry a design methodology which achieves a significant step change
improvement in predicting the stability of pipelines on erodible seabeds.

VI ACKNOWLEDGMENTS
The authors note and acknowledge the extensive published work of W/Prof. Liang Cheng and colleagues at
UWA who through their research over many years have provided insights into sediment transport around
subsea pipelines, a foundation on which the model presented in this paper is created. The authors also
acknowledge and are grateful to A/Prof Jeremy Leggoe for his efforts in the co-supervision of Mengmeng
Xu and Wenwen Shen during their final year projects. Finally the authors gratefully acknowledge the
financial support of the J P Kenny Technology for Business (T4B) programme which enabled the PSF
model to be developed from my concept stage into a working model.

VII REFERENCES
Brennodden, H, Lieng, J, Sotberg, T and Verley, R (1989), OTC6057 An Energy-Based Pipe-Soil Interaction Model,
Proceedings of the 21st Offshore Technology Conference.
Det Norske Veritas (2011), On-Bottom Stability Design of Submarine Pipelines, DNV-RP-F109.
Griffiths T, Shen W, Xu M and Leggoe J (2012), OMAE2012-83282, Comparison of Recent Parametric Trenched and
Partially Embedded / Spanning Pipelines With DNV-RP-F109 Load Reduction Design Curves, Proceedings of the 31st
International Conference on Offshore Mechanics and Arctic Engineering.
Griffiths T (2012), ISOPE2012-TPC-0828 Development of a Novel 2D Pipe-Soil-Fluid Interaction Model for Subsea
Pipeline Stability Design, Proceedings of the 22nd International Ocean and Polar Engineering Conference.
Jacobsen V, Bryndum M and Tsahalis D, (1988a), Prediction of Irregular Wave Forces on Submarine Pipelines,
Proceedings of the 7th International Conference on Offshore Mechanics and Arctic Engineering.
Jacobsen V (1988b), OTC5851 Forces on Sheltered Pipelines, Proceedings of the 20th Offshore Technology
Conference.
Jacobsen V, Bryndum M, Blonde (1989), OTC6056 Fluid Loads on Pipelines: Sheltered or Sliding, Proceedings of
the 21st Offshore Technology Conference.
Li F and Cheng L (2000), Numerical Simulation of Pipeline Local Scour With Lee-Wake Effects, International
Journal of Offshore and Polar Engineering, IJOPE,10(3), 195-199.
Li F and Cheng (2001) Effect of Pipeline Sagging on Local Scour Below Pipelines OMAE2001-PIPE4027,
Proceedings of the 20th International Conference on Offshore Mechanics and Arctic Engineering.
Palmer, A (1996), A Flaw in the Conventional Approach to Stability Design of Seabed Pipelines, Proceedings of the
19th Offshore Pipeline Technology Conference.
PRCI (1993), Submarine Pipeline On-Bottom Stability Volume 1 Analysis and Design Guidelines, AGA Project PR-178-
9333.
Shen W (2011), 3D CFD Investigations of Seabed Shear Stresses around Subsea Pipelines, Final Year Thesis, UWA
School of Civil Engineering.
Soulsby, R Dynamics of Marine Sands A Manual for Practical Applications, Thomas Telford 1997.
Verley R, Sotberg T and Brennodden H (1990), Break-out Soil Resistance for a Pipeline Partially Buried in Sand,
Proceedings of the 9th International Conference on Offshore Mechanics and Arctic Engineering, pp121-126.
Wilkinson R, Palmer A, Ells J, Seymour E and Sanderson N (1988), Stability of Pipelines in Trenches, Proceedings of the
Offshore Oil and Gas Pipeline Technology Seminar, Stavanger.
Xu, M. (2010), CFD Modelling of Seabed Shear Stresses Around Subsea Pipelines, Final Year Engineering Thesis (Civil),
UWA Faculty of Engineering, Computing and Mathematics.
Zeitoun H, Trnes K, Cumming G and Brankovi M (2008), Pipeline Stability State of the Art, OMAE200857284,
Proceedings of the 27th International Conference on Offshore Mechanics and Arctic Engineering.
Zeitoun H, Trnes K, Li J, Wong S, Brevet R and Willcocks J (2009), Advanced Dynamic Stability Analysis, OMAE2009
79778, Proceedings of the 28th International Conference on Offshore Mechanics and Arctic Engineering.

1150

You might also like