Ocean Engineering 146 (2017) 21–28

Contents lists available at ScienceDirect

Ocean Engineering
journal homepage: www.elsevier.com/locate/oceaneng

Investigation of impact forces on pipeline by submarine landslide using

material point method
Youkou Dong a, b, Dong Wang b, c, Mark F. Randolph d, *
a
Institut für Geotechnik und Baubetrieb, Technische Universit€
at Hamburg (TUHH), Harburger, D - 21079 Hamburg, Germany
b
Formerly Centre for Offshore Foundation Systems, University of Western Australia, Perth, WA, Australia
c
Shandong Provincial Key Laboratory of Marine Environment and Geological Engineering, Ocean University of China, 238 Songling Street, Qingdao, 266100, China
d
Centre for Offshore Foundation Systems, The University of Western Australia, 35 Stirling Highway, Perth, WA, 6009, Australia

A R T I C L E I N F O A B S T R A C T

Keywords: Quantitative assessment of impact forces by submarine landslide is significant for the safe operation of pipelines
Dynamic analysis that must cross potential runout paths. In this paper, the transient process of a submarine landslide impacting a
Impact pipeline is modelled using the material point method (MPM) with an enhanced contact algorithm. For simplicity,
Material point method the partially-embedded pipeline is assumed to be fixed in space. The Herschel-Bulkley rheological model is
Pipelines incorporated to reflect the dependence of the undrained shear strength of the sliding mass on the shear strain rate.
Submarine landslide
The behaviour of the mass flowing over the pipe was reproduced by allowing separation between the pipe and the
sliding mass. The horizontal impact forces predicted by the MPM are verified by comparison with those estimated
using a computational fluid dynamics approach. The impact forces are interpreted with a hybrid model consid-
ering the combined effects due to the soil's inertia, its shear strength, and also the asymmetric static pressure of
the sliding material. The coefficients for the three terms are retrieved by a best-fit to the results of an extensive
parametric study. The effect of the projected height of the pipe above the seabed is also investigated.

1. Introduction
Transportation of offshore oil and gas through pipelines requires
consideration of the risks from submarine landslides emanating from the
continental margins and slopes in the vicinity of pipeline routes. Sub-
marine landslides, which may be triggered by phenomena such as seismic
activity, dissociation of hydrate methane, diapirism etc., can lead to
runout of debris comprising a mixture of soft sediments and water at
speeds of up to 20 m/s (Jakob et al., 2012). The impact force from the
sliding material may be substantial with respect to the integrity and
functionality of the pipeline.
The magnitude of the impact force will be affected by the ‘consis-
tency’ or strength of the debris, the velocity and height of the flowing
material. Typically, the runout of submarine slides has been simulated
through depth-averaged approaches, with the debris flow material
modelled as some form of non-Newtonian fluid (Imran et al., 2001;
Iverson, 2003). This practice has led naturally to the use of computa-
tional fluid dynamics (CFD) approaches to assess potential impact forces
on seabed infrastructure, and in particular pipelines (Zakeri et al., 2009;
Liu et al., 2015). However, such approaches fail to capture either the
geometry of the debris flow (such as the height, relative to the diameter
of a pipeline), or to distinguish between different contributions to the
impact force, such as arising from inertial drag and what may be referred
to as ‘geotechnical’ resistance. Published studies have also tended to
focus on conditions where the pipe has been engulfed fully by the slide
material, rather than initial conditions where the pipe is partially
embedded in the seabed. The present study will help to determine
whether pipelines will remain stable (partially buried) during
slide impact.
In this paper, the impact forces of submarine slides on partially-buried
pipes are investigated using the material point method (MPM) in
geotechnical engineering. The pipelines are simplified as planar, since
the effort of three-dimensional simulations would prove unacceptable
given that a fine mesh and large sliding domain are needed. The slide
material is characterised using a form of Herschel-Bulkley (H-B) model
(Deglo de Besses et al., 2003), hence exhibiting strain rate dependency of
strength. Attention is focused on the pipe-slide interaction during the
early stage of impact under low or medium sliding velocities. A wide
range of conditions have been explored in respect of the slide height and
velocity, pipeline exposure above the seabed and different rheological

* Corresponding author.
E-mail addresses: 21193935@student.uwa.edu.au (Y. Dong), dongwang@ouc.edu.cn (D. Wang), mark.randolph@uwa.edu.au (M.F. Randolph).

https://doi.org/10.1016/j.oceaneng.2017.09.008
Received 14 October 2016; Received in revised form 15 April 2017; Accepted 7 September 2017
Available online 28 September 2017
0029-8018/© 2017 Elsevier Ltd. All rights reserved.
Y. Dong et al.
properties for the slide material. Eventually the steady force, following an
initial peak impact force, is then quantified with a hybrid model
considering the combined effects due to the inertia, shear strength and
static pressure of the sliding material.
2. Design frameworks
Estimation of impact forces on pipelines has generally focused on
pipelines that are fully engulfed by (i.e. suspended within) slide material.
Computational fluid dynamics (CFD) approaches, in which the sliding
mass is regarded as an incompressible viscous fluid, have been used to
quantify steady state forces, or average pressure, on elements of a pipe
suspended within a moving fluid (Zakeri, 2009; Zakeri et al., 2009; Liu
et al., 2015). The average pressure p, the horizontal force divided by the
projected area of the pipe, is then expressed in terms of a drag coefficient
CD as
p ¼ 0:5CD;Re ρv2
where ρ and v are the density and horizontal (free-field) velocity of the
sliding mass, respectively. The drag coefficient, CD,Re, is then expressed
as a function of the non-Newtonian Reynolds number, Re ¼ ρv2 =su
(Zakeri, 2009; Liu et al., 2015), where su is the mobilised shear stress
(or strength).
For non-Newtonian fluids, the rate-dependent shear strength may be
characterised by the Herschel-Bulkley (H-B) rheological model,
expressed in its original form as (Deglo de Besses et al., 2003)
su ¼ su0 þ K_γn
where su0 is the yield strength at negligible strain rate, K a ‘consistency’
parameter, n the ‘shear-thinning’ index and γ_ the shear strain rate. In
geotechnical applications, a normalised form of the H-B model has ten-
ded to be adopted, expressed as (Boukpeti et al., 2012a)


n 
γ_
su ¼ su0 1 þ μ
γ_ ref
where γ_ ref is the reference shear strain rate and μ the viscosity coefficient.
H-B fitting of rate-dependent penetrometer data presented by Boukpeti
et al. (2012b) gave ranges of μ and n of 0.3–0.7 and 0.1 to 0.4 respec-
tively, with γ_ ref as 0.06 s1.
For the slide-pipeline impact problem, a convenient shear strain rate
may be expressed as γ_ ¼ v=D. The non-Newtonian Reynolds number
then becomes
ρv2 ρv2
Re ¼ ¼   n 
su su0 1 þ μ v=D
γ_ ref
Relationships between drag coefficient and non-Newtonian Reynolds
number have been proposed on the basis of laboratory flume experiments
(Zakeri et al., 2008) and CFD analyses (Zakeri et al., 2009; Liu et al.,
2015). Although these relationships work reasonably at moderate to high
velocities of the slide, Equation (1) is not appropriate to estimate inter-
action forces accurately at low velocities. In general the force exerted on
a pipe engulfed within slide material is influenced by two components: a
drag force resulting from inertial effects, and a geotechnical resistance
related to the shear strength (Randolph and White, 2012). The latter
contribution becomes more significant at low slide velocities (Georgiadis,
1991; Zakeri et al., 2011; Sahdi et al., 2014).
Randolph and White (2012) suggested that for planar pipes fully
engulfed by slides, the impact pressure may be expressed as
p ¼ 0:5CD ρv2 þ Nc su
where Nc is a conventional geotechnical resistance factor and su is the
Ocean Engineering 146 (2017) 21–28
undrained shear strength at the relevant shear strain rate (or v/D). Since
the effect of viscosity is captured within the value of su, both CD and Nc
may be taken as constants, independent of the slide velocity. Equation (5)
was proposed on the basis of pipes fully engulfed within the sliding mass,
i.e. with no gap between the pipe and the slide. A modified form of this
relationship is proposed later for the conditions considered here, with
slide material breaking over a partially embedded pipeline, allowing a
gap to be sustained at the rear side of the pipe.
3. Methodology
3.1. Material point method
The material point method (MPM), originated from the particle-in-
cell method in CFD (Harlow, 1964). It can be regarded as a combina-
tion of finite element and meshfree methods, providing an acceptable
balance between computational cost and accuracy for large deformation
(1) analysis. The MPM has an inherent advantage for large deformation
problems such as run-out of landslides (Andersen and Andersen, 2010)
and large-amplitude displacement of structural elements through soil
(Phuong et al., 2016), since it discretises the soil as Lagrangian particles.
The material mechanical and kinematic properties (mass, volume, ve-
locities, deformation gradients and stresses) are recorded and updated at
the particles, while a fixed rectilinear background mesh is used just for
the calculation of each incremental step. Since the mesh is fixed in space,
mesh entanglement that can occur in conventional finite element
methods is avoided. The MPM analyses presented here, for slides impact
on a fixed pipeline, were undertaken using an in-house program that
(2)
stems from the open-source package Uintah (Guilkey et al., 2012). The
Uintah package was enhanced with a contact algorithm ‘Geo-contact’
(Ma et al., 2014) and a GPU parallel computing strategy (Dong et al.,
2015). The GPU parallelisation strategy allows for two-dimensional
simulations with up to 20 million particles. The explicit updated
Lagrangian calculation is based on the generalised interpolation material
point method presented by Bardenhagen and Kober (2004).
(3)
3.2. Contact algorithm
The contact between the pipe and the sliding mass was implemented
with an algorithm termed ‘Geo-contact’ (Ma et al., 2014). Compared with
the contact algorithms presented in Bardenhagen et al. (2000, 2001), the
Geo-contact reduces numerical oscillation in the quantitative contact
forces effectively. The pipe was simplified as a rigid body due to its much
higher stiffness than the sliding mass. According to the Geo-contact al-
gorithm, the pipe and the sliding mass may be in contact at element
nodes if non-zero particle masses from the two bodies are projected onto
(4)
a given node. For a specific node i of the sliding mass in contact, the
relative normal velocity to the pipe is Δvi ¼ ðvi  v0 Þni , where vi is the
velocity at node i of the sliding mass, v0 is the velocity of the pipe and ni is
the unit vector from node i to the centre of the pipe. Node i can be
distinguished as approaching (Δvi > 0) or departing from (Δvi < 0) the
pipe according to the sign of the relative normal velocity.
In Geo-contact, the relative normal velocities for the nodes may be
reduced to close to zero in order to eliminate (or minimise) inter-
penetration or, if required, separation. Alternatively, if separation is
allowed, no adjustment of negative relative normal velocities is neces-
sary, leaving the sliding mass free to flow away from the pipe. When the
sliding mass comes in contact with the pipe, the value of Δvi is
reducing using

k
minðsi ; hÞ
Δv0i ¼ f i Δvi where f i ¼ 1  (6)
h
(5)
where h represents the square element size, si is the distance from the
node i in the contact region to the surface of the pipe and k is a penalty
power. Introduction of the penalty function fi permits slight
22
Y. Dong et al. Ocean Engineering 146 (2017) 21–28

for objectivity, the stresses being measured with the Cauchy stress and
updated with the Jaumann rate σ_ J according to

σnew ¼ σ þ Δtσ_ J (8)

where Δt is the time step. The Jaumann rate is calculated by

σ_ J ¼ ðσW  WσÞ þ CD (9)

where C is the fourth-order stiffness tensor, D is the deformation rate

" !T #
1 X X
Fig. 1. Schematic for submarine landslide across pipe with MPM (non-scaled).
D¼ ∇SiP vnew
i þ ∇SiP vnew
i (10)
2 i i
interpenetration between the pipe and the sliding mass, but moderates
unwanted oscillations in computed contact forces if Δvi were to be set
and W is the vorticity
directly to zero. An appropriate value for the power k was suggested as 6
for quasi-static analysis (Ma et al., 2014), but was taken as 1 for the " !T #
current dynamic analyses. 1 X X
W¼ ∇SiP vnew
i  ∇SiP vnew
i (11)
2 i i
3.3. Initial conditions
The superscript T in Equations (10) and (11) denotes the transpose of
the tensor. The time step Δt was determined through the Courant-
A schematic of the initial conditions is provided in Fig. 1. A rectan-
Friedrichs-Lewy stability condition
gular block of soil, height H and length L (taken here as 28 m) was given
an initial horizontal velocity, v, along a smooth rigid base. A pipe of αl
diameter D (taken as 0.8 m) was partially buried immediately in front of Δt ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (12)
ðλ þ 2GÞ=ρ
the moving soil mass. The pipe was regarded as smooth and (zero ten-
sion) separation permitted between soil and pipe. Thus, in contrast to
where G and λ are the Lame parameters, and α is the Courant number.
previously quoted studies that focused on pipeline fully engulfed within a
sliding soil mass, the focus here was on initial breakout conditions,
4. Results and discussions
specially to evaluate the lateral force exerted on an in service partially
buried pipe. The fixed condition should provide an upper bound on the
4.1. Verification of MPM analysis
estimated impact force.

Robustness of the MPM was verified by comparison with CFD simu-

3.4. Slide material
lations using a commercial finite volume package ANSYS FLUENT
(ANSYS, 2011). Since separation between the smooth pipe and the
Impact from a submarine slide would occupy only a short time, so that
sliding mass passing over the pipe was not allowed in the CFD simulation,
the response of the slide material would be essentially undrained. As
no-separation contact was also employed in the MPM for verification
such, the slide material was modelled using a simple total stress von
purpose. By contrast, separation was allowed in the subsequent para-
Mises failure envelope, but with a rate-dependent shear strength deter-
metric study to quantify impact forces.
mined from the H-B rheological model according to Equation (3).
An initially rectangular planar slide, 28 m in width and 6 m in height
Compared with power law and Bingham models of rate dependency, the
H, was placed close to a fixed pipe with a diameter D of 0.8 m (Fig. 1).
H-B model is considered superior in being able to capture the variation of
The pipe was half-buried, i.e. the projected height e ¼ 0.5D. The distance
shear strength over a significant range of strain rates (Locat and Lee,
between the slide front and the pipe was specified as small (0.01 m) to
2002). Also, no softening was incorporated in the soil model, on the
allow for essentially immediate impact. A roller boundary condition was
assumption that the material would be fully remoulded by the stage at
imposed along the bottom of the domain. The whole slide was launched
which impact might occur.
with an initial horizontal velocity of v ¼ 6 m/s, and then ran forward due
In the MPM analyses, the H-B model was implemented by updating
to its momentum.
the undrained strength at each particle at the end of each incremental
In the MPM analysis, the element size was selected as D/80. A 4  4
step, while the strength was maintained constant during the subsequent
particle configuration was allocated for each element fully occupied by
incremental step. The shear rate γ_ at the end of each step was calculated
the soil or the pipe prior to the calculation. This particle density is finer
as (Guilkey et al., 2012)
than the 2  2 particle configuration used in Ma et al. (2014) and Dong
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
!2
u et al. (2015), to improve the numerical accuracy. In total, there were 19.2
u X X X X
γ_ ¼ t ∇SiP;x vi;x  new
∇SiP;y vi;y new
þ4 ∇SiP;y vnew ∇SiP;x vnew
million slide particles and 3936 pipe particles. The density of the slide
i;x i;y
i i i i
was ρ ¼ 1500 kg/m3. The acceleration due to gravity was g ¼ 9.81 m/s2.
Considering buoyancy, the body force of the submarine slide was
(7)
ðρ  ρw Þg, where ρw is the density of water (1000 kg/m3).
The viscosity coefficient μ and the shear-thinning index n in the H-B
where ∇SiP is the gradient of the shape function at node i evaluated at
model were taken as n ¼ 0.23, μ ¼ 0.45, with a reference shear strain rate
particle P; vnew represents the velocity at node i after adjustment with the
P of γ_ ref ¼ 0.06 s1 and threshold shear strength of su0 ¼ 0.5 kPa. Poisson's
i
contact algorithm; represents a summation over all nodes related to
i ratio of the soil was taken as 0.49 to approximate constant volume hence
the particle; and the subscripts x and y represent the horizontal and undrained conditions. Young's modulus was taken as a nominal value of
vertical components, respectively. The detailed expression of the shape 300su0, but also the effect of an arbitrarily large value of 8000su0
function can be found in Guilkey et al. (2012). explored. The time step Δt was determined with a Courant number of 0.3.
The definition of the stresses σ follows finite strain theory taking In the CFD analysis, the whole domain was discretized with 117,990
account of incremental rotation of the configurations between time steps quadrilateral elements, with minimum and maximum sizes of the

23
Y. Dong et al.
Fig. 2. Normalised impact pressures predicted by CFD and MPM simulations.
Fig. 3. Touchdown of sliding mass with no-separation contact by MPM.
Fig. 4. History of impact pressure.
elements of 0.01 m and 0.05 m, respectively. The mesh was testified as
sufficiently fine through trial calculations. The sliding mass was consid-
ered to give rise to laminar flow using a ‘no-turbulence’ model (Raie and
Tassoulas, 2009). The pressure and velocity fields were computed with a
scheme termed ‘pressure implicit with splitting of operator’. The time
step in the explicit calculations was estimated by (ANSYS, 2011)
24
Ocean Engineering 146 (2017) 21–28
βhs
Δt ¼ (13)
v
where the coefficient β was taken as 0.2 and hs represents the minimum
mesh size.
The normalised impact pressures predicted by the MPM and CFD are
shown in Fig. 2. The pressure increases rapidly at the very early stage of
impact. The peak pressure appears earlier in the CFD analysis than for the
MPM with E/su0 ¼ 300. This is consistent with the essentially infinite
elastic modulus of the slide material in the CFD analysis. When E/su0 in
the MPM analysis was increased to 8000, the MPM curve shows excellent
agreement with the CFD before t ¼ 0.024 s.
A nearly steady pressure is reached after t ¼ 0.045 s. The steady
pressures predicted by the MPM are close to each other, ~10% lower
than the CFD prediction. The sudden fall of the pressure around 0.055 s is
due to touchdown of the sliding mass at the base (Fig. 3).
4.2. Flow process
By contrast with the assumption in the above comparison, if the
interface between pipe and soil is changed to allow separation, a gap is
sustained at the rear of the pipe. Eventually the gap becomes filled with
back-flowing mass. Therefore, for the main parametric analyses with the
MPM, separation was allowed between the pipe and the slide since that
was considered a more realistic condition.
Keeping other input parameters the same, the rigidity ratio for the soil
was varied between E/su0 ¼ 300, 1000 and 8000. The element size was
D/80. The history of the normalised pressure in terms of E/su0 ¼ 300 is
shown in Fig. 4, which took ~4 h to calculate. The peak pressure is
reached rapidly in <0.01 s. After that, the pressure reduces to a nearly
steady state. During the steady stage, the soil flows over the pipe with a
gap sustained behind the pipe. This is illustrated in Fig. 5 which shows
velocity contours at times A, B and C in Fig. 4. The soil in front of the pipe
is stagnated while that passing over the pipe crown is accelerated (Fig. 6).
The flow head falls onto the ground at ~0.5 s (just after time C), followed
by backflow towards the pipe. The presence of a gap behind the pipe is
consistent with previous numerical and experimental studies for slide
impact of suspended pipes (Zakeri et al., 2008, 2009).
In Fig. 7, the peak impact pressure is shown to increase with the
Young's modulus, although the steady pressure after the peak value is
nearly independent of the Young's modulus. The steady pressure lasts a
relatively long period, and so is considered as the controlling factor in
assessing the stability of the pipeline. The peak pressure may also need to
be considered in design, although its short-lived nature will limit the
contribution (relative to the steady state value) to cause significant mo-
tion of the pipe. The peak impact pressures are not considered further in
this study. E/su0 ¼ 300 was used in the subsequent analyses to obtain the
steady pressure. Regarding the mesh density, as shown in Fig. 8, an
element size of D/80 was established as sufficiently fine to achieve
convergence of the steady impact pressure. An element size of D/80 was
thus used in the following.
5. Interpretation of impact force
A wide range of scenarios were explored with respect to the geometry
and properties of the slide material, keeping the pipe diameter constant
at D ¼ 0.8 m. The ranges of geometry and soil properties are summarised
in Table 1. The total number of scenarios was 650.
The steady impact pressures obtained from the MPM simulations are
normalised as p/ρv2 in the following discussion. If Equation (1) is used to
predict the impact force, the drag coefficients obtained are scattered in a
wide range of 0.44–5.82, and with no consistent trend with Reynolds
number. As such, the approach would be unreliable in practical appli-
cations to quantify the pipe-slide interaction forces at low and medium
flow velocities.
Given that a gap has been shown to be sustained at the rear side of the
Y. Dong et al. Ocean Engineering 146 (2017) 21–28

Fig. 5. Contour of velocities.

pipe for a significant period, an additional term related to the static

pressure of the sliding mass was added to Equation (5), resulting in

p ¼ 0:5CD ρv2 þ Nc su þ Cγ ðρ  ρw ÞgH (14)

where Cγ is referred to as a ‘static coefficient’. The rate-enhanced shear

strength su is estimated by Equation (3) with a nominal shear rate
γ_ ¼ v=2e. Equation (14) may thus be re-written as

p 1 ðρ  ρw Þ 1
¼ 0:5CD þ Nc þ Cγ (15)
ρv2 Re ρ F2r
pffiffiffiffiffiffiffi
where Fr is the Froude number, Fr ¼ v= gH (Masson et al., 2006) and
Reynolds number is Re ¼ ρv2 =su . In the MPM analyses, Re varies in the
range 10–100, Fr in the range 0.18–3.32. The normalised impact pres-
sures p/ρv2 from the analyses fell in the range 0.22–2.91.
Fig. 6. Velocity vectors of soils around pipe at time 0.12 s. The drag coefficient CD, the bearing capacity factor Nc and the static

25
Y. Dong et al. Ocean Engineering 146 (2017) 21–28

Fig. 7. Influence of Young's modulus on impact pressure.

Fig. 9. Fitted drag coefficient CD.

Fig. 8. Influence of mesh size on impact pressure.

Fig. 10. Fitted bearing capacity factor Nc.

Table 1
Ranges of geometry and soil parameters considered.
exposure e/D. The following main conclusions may be drawn.
e/D su0 μ n H/D Inflow
(kPa) velocity
(1) In Fig. 9, for a given pipe projected height, the fitted value of CD
(m/s)
increases with H/D until a nearly stable value is approached at a
0.25 0.15 0.35, 0.45 0.15, 1.25, 1.75, 2.5, 5, 7.5, 2–12 critical H/D; the value is 5 for e/D ¼ 0.25, 7.5 for e/D ¼ 0.5 and
and and 0.65 0.23 and 10, 12 and 15
0.5 0.5 0.4 1.25, 1.75, 2.5, 3.5, 5,
10 for e/D ¼ 0.75. The stable value of CD increases from 0.63 to
7.5, 10, 12 and 15 ~1.04 as e/D increases from 0.25 to 0.75. This range may be
0.75 2.5, 3.5, 5, 7.5, 10, 12 compared with data from CFD analyses for a fully engulfed pipe
and 15 (Zakeri, 2009), reinterpreted by Randolph and White (2012), who
quote deduced CD values of 0.4–0.6. By contrast, experiments
reported by Sahdi et al. (2014) for a fully engulfed pipe suspended
coefficient Cγ in Equation (15) are the key concerns that quantify the
within a moving soil mass quote a best fit value of CD ¼ 1.06.
contributions of drag force, ‘geotechnical’ resistance and static pressure,
(2) From Fig. 10, a minimum H/D is also required to achieve a stable
respectively. These three coefficients will be affected by the projected
capacity factor Nc; the critical value of H/D is 2.5 for e/D ¼ 0.25,
height of the pipe, and the initial height, velocity and strength properties
3.5 for e/D ¼ 0.5 and 5 for e/D ¼ 0.75. The stable value of Nc
of the sliding mass. The coefficients were fitted by regression analysis of
increases from ~4 for e/D ¼ 0.25 to ~4.5 for e/D ¼ 0.75. Two
the impact forces retrieved from the MPM analyses. For each combina-
additional analyses were conducted, penetrating a pipe with e/
tion of the normalised pipe projected height and the slide height, CD, Nc
D ¼ 0.25 and 0.75 through a sliding mass with H/D ¼ 10 at very
and Cγ were fitted with R2 values not less than 0.97, over the full range of
slow velocity. These gave static Nc values of 4.2 and 4.8 for e/
soil strength properties and slide velocities.
D ¼ 0.25 and 0.75, respectively. Since the pipe is exposed at the
The fitted values of the coefficients are considered as functions of the
rear side (see Fig. 5), the predicted value of Nc corresponds to
initial height ratio H/D of the sliding mass and the normalised pipe

26
Y. Dong et al. Ocean Engineering 146 (2017) 21–28

Fig. 11. Static coefficient with ratio (H-e)/D.

shallow embedment of the pipe into soil, and is much lower than
the value of 9.18 for a deeply embedded cylinder with a static full-
flow mechanism (Randolph and Houlsby, 1984). Wang et al.
(2010) reported Nc ¼ 4.25 for a pipe penetrating to a depth of
0.5D, which is close to the value of 4.2 at e/D ¼ 0.5 shown in
Fig. 10.
(3) Contrasting with the approach for CD and Nc, the static coefficient
Cγ was analysed as a function of the ratio (H-e)/D as shown in
Fig. 11. This resulted in the relationship

0:56
He
Cγ ¼ 0:56 (16)
D
For each projected height ratio of the pipe, the normalised impact
forces from the MPM simulation are compared with those estimated
through Equation (15) (with CD, Nc, and Cγ fitted as in Figs. 9–11) in
Fig. 12. Equation (15) provides estimations for almost all scenarios with
divergence less than 15%. To verify this relationship further, another
group of scenarios, not used in the above fitting exercise, were consid-
ered. These covered input parameters of: D ¼ 0.5 m, e/D ¼ 0.5,
su0 ¼ 0.3 kPa, n ¼ 0.3, μ ¼ 0.55, H/D ¼ 9 and v ¼ 2–12 m/s. As shown in
Fig. 13, the impact forces predicted by the MPM are close to those from
Equation (15) with the suggested values of CD ¼ 0.88 (see Fig. 9),
Nc ¼ 4.2 (see Fig. 10) and Cγ ¼ 0.169 (see Equation (16)).
The contributions of the three components in the impact force are
demonstrated in Fig. 14. When the slide velocity is 2 m/s, the compo-
nents due to the shear strength (~42%) and the static pressure (~43%)
dominate. When the slide velocity reaches 12 m/s, the maximum velocity
considered in the study, the contribution of the inertial resistance in-
creases to ~85%.

6. Conclusions

The horizontal impact forces on a pipeline by submarine landslides

were investigated by means of numerical simulation with the material
point method. The models of a soil mass impacting and then flowing
across the fixed pipe were established by giving the soil an initial hori-
zontal velocity. The pipe was simplified as a rigid (fixed) cylinder
partially buried in the seabed and the flowing soil mass was characterised
using the Herschel-Bulkley rheological model. The interaction between
the pipe and the soil mass was implemented using ‘Geo-contact’
assuming a smooth interface. Inflow velocities of the landslides were
Fig. 12. Comparison of normalised pressures by MPM and Equation (15).
2–12 m/s, resulting in ranges of the non-Newtonian Reynolds number of
10–100, and Froude number of 0.18–3.32. The predicted horizontal

27
Y. Dong et al.
Fig. 13. Predictions of normalised pressures by MPM and Equation (15).
Fig. 14. Relative contributions of inertia, shear strength and static pressure components.
impact pressures on the pipe was interpreted with a hybrid expression
summing independent contributions from inertia (expressed in terms of
drag coefficient, CD), geotechnical resistance (bearing factor, Nc) and
static pressure (coefficient Cγ). Best-fit values of the coefficients CD, Nc
and Cγ were obtained from the data, taking account of the normalised
slide depth H/D and the exposed height of the pipe e/D. The resulting
design approach was validated for both the analyses used to develop it,
and also an independent set of analyses. Maximum deviations from the
design approach were less than 15%.
Acknowledgements
The research presented here was supported by the Australian
Research Council through an ARC Discovery grant (DP120102987). The
work forms part of the activities of the Centre for Offshore Foundation
Systems (COFS), currently supported as a node of the Australian Research
Council Centre of Excellence for Geotechnical Science and Engineering
(CE110001009) and through the Fugro Chair in Geotechnics, the Lloyd's
28
Ocean Engineering 146 (2017) 21–28
Register Foundation Chair and Centre of Excellence in Offshore Foun-
dations and the Shell EMI Chair in Offshore Engineering.
This work was also supported by resources provided by the Pawsey
Supercomputing Centre with funding from the Australian Government
and the Government of Western Australia and NVIDIA Corporation with
the donation of the Tesla K40 GPU for this research.
The authors would also like to acknowledge the value of preliminary
analyses of this problem carried out by Dr Jiajie Ma during his
doctoral studies.
References
ANSYS, 2011. ANSYS FLUENT 14.0 Theory Guide. v.14.0.1. ANSYS, Inc.
Andersen, S., Andersen, L., 2010. Modelling of landslides with the material-point method.
Comput. Geosci. 14 (1), 137–147.
Bardenhagen, S.G., Kober, E.M., 2004. The generalized interpolation material point
method. Comput. Model Eng. Sci. 5 (6), 477–496.
Bardenhagen, S.G., Brackbill, J.U., Sulsky, D., 2000. The material-point method for
granular materials. Comput. Methods Appl. Mech. Eng. 187 (3–4), 529–541.
Bardenhagen, S.G., Guilkey, J.E., Roessig, K.M., Brackbill, J.U., Witzel, W.M., 2001. An
improved contact algorithm for the material point method and application to stress
propagation in granular material. Comput. Model. Eng. Sci. 2 (4), 509–522.
Boukpeti, N., White, D.J., Randolph, M.F., 2012a. Analytical modelling of a submarine
slide under steady flow in relation to the impact load on a pipeline. Geotechnique 62
(2), 137–146.
Boukpeti, N., White, D.J., Randolph, M.F., Low, H.E., 2012b. Strength of fine-grained soils
at the solid-fluid transition. Geotechnique 62 (3), 213–226.
Deglo de Besses, B.D., Magnin, A., Jay, P., 2003. Viscoplastic flow around a cylinder in an
infinite medium. J. Newt. Fluid Mech. 115, 27–49.
Dong, Y., Wang, D., Randolph, M.F., 2015. A GPU parallel computing strategy for the
material point method. Comput. Geotech. 66, 31–38.
Georgiadis, M., 1991. Landslide drag forces on pipelines. Soils Found. 31 (1), 156–161.
Guilkey, J., Harman, T., Luitjens, J., et al., 2012. Uintah Code (Version 1.5.0). Computer
program; available at: http://www.uintah.utah.edu.
Harlow, F.H., 1964. The particle-in-cell computing method for fluid dynamics. Methods
Comput. Phys. 3, 319–343.
Imran, J., Harff, P., Parker, G., 2001. A numerical model of submarine debris flow with
graphical user interface. Comput. Geosci. 27 (6), 717–729.
Iverson, R.M., 2003. The debris-flow rheology myth. Debris-flow Hazards Mitigation:
Mech. Predict. Assess. 1, 303–314.
Jakob, M., Stein, D., Ulmi, M., 2012. Vulnerability of buildings to debris flow impact. Nat.
hazards 60 (2), 241–261.
Liu, J., Tian, J., Yi, P., 2015. Impact forces of submarine landslides on offshore pipelines.
Ocean. Eng. 95, 116–127.
Locat, J., Lee, H.J., 2002. Submarine landslides: advances and challenges. Can. Geotech.
J. 39 (1), 193–212.
Ma, J., Wang, D., Randolph, M.F., 2014. A new contact algorithm in the material point
method for geotechnical simulations. Int. J. Numer. Anal. Methods Geomech. 38
(11), 1197–1210.
Masson, D.G., Harbitz, C.B., Wynn, R.B., Pedersen, G., Løvholt, F., 2006. Submarine
landslides: processes, triggers and hazard prediction. Philosophical transactions of
the royal society of london a: mathematical. Phys. Eng. Sci. 364 (1845), 2009–2039.
Phuong, N.T.V., van Tol, A.F., Elkadi, A.S.K., Rohe, A., 2016. Numerical investigation of
pile installation effects in sand using material point method. Comput. Geotech. 73,
58–71.
Raie, M.S., Tassoulas, J.L., 2009. Installation of torpedo anchors: numerical modeling.
J. Geotech. Geoenviron. Eng. 135, 1805–1813.
Randolph, M.F., Houlsby, G.T., 1984. The limiting pressure on a circular pile loaded
laterally in cohesive soil. Geotechnique 34 (4), 613–623.
Randolph, M.F., White, D.J., 2012. Interaction forces between pipelines and submarine
slides - a geotechnical viewpoint. Ocean. Eng. 48, 32–37.
Sahdi, F., Gaudin, C., White, D.J., Boylan, N., Randolph, M.F., 2014. Centrifuge modelling
of active slide-pipeline loading in soft clay. Geotechnique 64 (1), 16–27.
Wang, D., White, D., Randolph, M.F., 2010. Large-deformation finite element analysis of
pipe penetration and large amplitude lateral displacement. Can. Geotech. J. 47 (8),
842–856.
Zakeri, A., Hoeg, K., Nadim, F., 2008. Submarine debris flow impact on pipelines - Part I:
experimental investigation. Coast. Eng. 55 (12), 1209–1218.
Zakeri, A., 2009. Submarine debris flow impact on suspended (free-span) pipelines:
normal and longitudinal drag forces. Ocean. Eng. 36 (6), 489–499.
Zakeri, A., Hoeg, K., Nadim, F., 2009. Submarine debris flow impact on pipelines - Part II:
numerical analysis. Coast. Eng. 56 (1), 1–10.
Zakeri, A., Chi, K., Hawlader, B., 2011. Centrifuge modeling of glide block and out-runner
block impact on submarine pipelines. In: Proceedings of the Offshore Technology
Conference. Houston, TX, USA, paper OTC 21256.