Coastal Engineering 116 (2016) 103–117

Contents lists available at ScienceDirect

Coastal Engineering

journal homepage: www.elsevier.com/locate/coastaleng

Effect of limited sediment supply on sedimentation and the onset of

tunnel scour below subsea pipelines
Qin Zhang a,b,⁎, Scott Draper b,c, Liang Cheng b, Hongwei An b
a
School of Engineering, Ocean University of China, no.238 Songling Road, Qingdao, Shandong, China
b
School of Civil, Environmental and Mining Engineering, The University of Western Australia, 35 Stirling Highway, Crawley, WA 6009, Australia
c
Centre for Offshore Foundation Systems, The University of Western Australia, 35 Stirling Highway, Crawley, WA 6009, Australia

a r t i c l e i n f o a b s t r a c t

Article history: This paper summarises the results of a series of experiments performed to investigate the onset of tunnel scour
Received 16 February 2015 below subsea pipelines in steady currents. The experiments were performed on a model seabed that extended
Received in revised form 10 May 2016 different lengths upstream of the pipeline to assess the effects of sediment supply on sedimentation around
Accepted 22 May 2016
the pipeline and the potential for onset of tunnel scour. In each experiment, the flow velocity and pipeline em-
Available online 2 July 2016
bedment were recorded continuously from inside and outside of the model pipe. In general, the results show
Keywords:
that following evolution of the seabed profile around the pipeline due to sedimentation and changes in sediment
Onset of tunnel scour supply, the onset of tunnel scour may still occur even when the initial embedment is larger than the critical value
Sediment supply obtained by an existing empirical formula according to Sumer et al. (2001). This result suggests the potential for
Pipelines tunnel scour beneath deeply embedded pipelines in the field where the sediment supply may be interrupted by,
for example, rock outcrops or upward-sloping seabed on the upstream side of the pipe. It also demonstrates that
onset of tunnel scour is possible for a pipeline on a seabed that is not flat on either side of the pipe, provided that
the flow conditions are sufficient to promote piping. To complement the experiments, a series of numerical sim-
ulations have been conducted to investigate the seepage flow and dynamic pressure difference upstream and
downstream of the pipeline prior to the onset of tunnel scour. The numerical results show that despite significant
alteration of the surrounding seabed topography due to local scour for a deeply embedded pipe with limited sed-
iment supply, the pressure gradient across the pipe is still sufficient to cause piping compared with the flat seabed
case, and the maximum pressure gradient at the downstream side of the pipeline is consistent with the break-
through point observed in the physical experiments.
© 2016 Elsevier B.V. All rights reserved.

1. Introduction and motivation
Interaction between a submarine pipeline and an erodible seabed
has attracted much attention because of its importance in offshore engi-
neering. Of particular interest is the ‘onset of tunnel scour’, which de-
fines the point at which sediment is washed away beneath a pipeline.
Onset of tunnel scour results in the development of a scour hole
below a pipeline, followed by spanning of the pipeline and possibly sub-
sequent self-burial (Sumer and Fredsøe, 2002) or pipeline breakout.
Prediction of the onset of tunnel scour is therefore important in under-
standing and predicting the stability of a subsea pipeline.
For the case of a pipeline partially embedded in a flat seabed, a num-
ber of investigations have been reported regarding the onset of tunnel
scour. Chiew (1990), for example, conducted a series of physical exper-
iments and found that piping (backwards erosion of sediment) leads to
the onset of tunnel scour. Sumer et al. (2001) investigated this
⁎ Corresponding author at: School of Engineering, Ocean University of China, no.238
Songling Road, Qingdao, Shandong, China.
E-mail address: zhangqin200000@gmail.com (Q. Zhang).
mechanism in more detail and presented an empirical expression to
predict the onset of tunnel scour in steady currents for a pipeline par-
tially buried in a flat seabed. This expression is given as
  0:5 
U 2cr e
¼ 0:025 exp 9 : ð1Þ
gDð1−nÞðs−1Þ D
where Ucr is a ‘critical’ undisturbed steady current velocity (measured at
the level of the top of the pipeline), above which the onset of tunnel
scour due to piping will occur; g is the acceleration due to gravity; D is
the pipe diameter; n is the porosity of the sediment; s is the specific
gravity of sediment grains; and e is the burial depth of the pipeline.
Numerical studies on the onset of tunnel scour below subsea pipe-
lines have also been reported in the literature. Liang and Cheng
(2005), for instance, were the first to establish a numerical model of
the onset of tunnel scour below subsea pipelines subject to steady cur-
rents. In that work, the pressure gradient that governs the seepage flow
below the pipeline was determined by solving the two-dimensional
Reynolds-averaged continuity and Navier–Strokes equations with the
standard k − ε turbulence closure. The seepage flow was calculated

http://dx.doi.org/10.1016/j.coastaleng.2016.05.010
0378-3839/© 2016 Elsevier B.V. All rights reserved.
104 Q. Zhang et al. / Coastal Engineering 116 (2016) 103–117

suggests that the onset of tunnel scour is unlikely in weak ambient

Notation
field conditions (smaller than the critical velocity Ucr) for pipelines on
most soils if they are embedded to a certain level. Field survey data for
ADV Acoustic Doppler velocimeter
subsea pipelines, however, have suggested that even if the initial (as-
Cp Pressure coefficient
laid) pipeline embedment is relatively large (Westgate, 2013) or the
D Diameter of pipe
ambient flow is relatively weak, the onset of tunnel scour may still
d50 Average grain size
occur. This implies that other mechanisms may contribute to the onset
e Burial depth of pipeline
of tunnel scour. Moreover, in practise, it is unlikely that the seabed
Fs Seepage force
will remain flat prior to the onset of tunnel scour. This is because cur-
g Acceleration due to gravity
rents less than the critical velocity for piping can lead to local sedimen-
ks Nikuradse equivalent sand grain roughness
tation, thereby altering the local seabed morphology.
n Porosity of the sediment
In this paper, the onset of tunnel scour below subsea pipelines was
p Local pressure
revisited by conducting a series of experiments and subsequent numer-
p∞ Pressure at far field
ical analysis in steady currents. In the physical experiments, specific at-
qflyover Volumetric transport rate over the pipe
tention is paid to the effect of upstream sediment supply on local
qlee Volumetric transport rate due to the lee wake vortex
sediment morphology and the potential for the onset of tunnel scour
qluff Volumetric transport rate due to the luff vortex
following changes to the local seabed morphology. This work is applica-
qin Volumetric transport rate from the far field flow
ble for better understanding of the potential for scour of pipelines in the
qout Volumetric transport rate leaving the pipe
field, particularity where changes in seabed erosion properties or sur-
RE Reynold's number for the near-bed wave-orbital
rounding rocky outcrops and sloping seabeds may limit the upstream
motion
sediment supply. It is also applicable to understanding the potential
Rec Pipe Reynolds number in current
for onset of tunnel scour following variations in the surrounding seabed
s Specific gravity of sediment grains
profile and thus the potential for de-burial of submarine pipelines. The
SP Distance of seepage flow path
numerical simulations where conducted to improve the understanding
Uc Velocity induced by current component at the height of
of the pressure distribution around a pipeline on a seabed that is not flat.
D
Ucr Critical undisturbed steady current velocity
2. Physical experiments
u⁎c Friction velocity based on grain roughness in current
Vup Volume sediment transport for the upstream side of the
2.1. Experiment setup
pipe
Vdown Volume sediment transport for the downstream side of
The physical experiments were conducted in a recirculating flume
the pipe
(Mini O-tube) at the University of Western Australia. The Mini O-tube
W Submerged weight of the soil element
(MOT) test facility comprises a motor–impeller system, unplasticised
X Seepage flow path
polyvinyl chloride (uPVC) tube sections, two honeycomb transitions at
z0 Bed roughness length
each end of the test section, and one straight test section; the main com-
α Soil permeability
ponents are indicated in Fig. 1a. The propeller in the MOT is driven by a
γ Specific weight of water
5.5 KW three-phase induction motor with rated speed 2885 RPM (rev-
ε Void fraction, porosity
olutions per minute). The rotational speed of the motor is controlled by
θc Current Shields parameter
the Danfoss VLT 2800 frequency converter, which is managed in the
k Von Karman's constant (=0.41)
LABVIEW software application on a local computer. The diameter of
ρ Fluid density
the uPVC tube is 0.17 m, and there is a tapered section measuring
τc Current shear stress
0.4 m in length to connect the circular uPVC tube to each end of the rect-
ν Kinematic viscosity of water
angular test section. The test cross-section of the MOT is 0.2 by 0.3 m
and has a length of 1.8 m. The bottom 0.1 m of the working section
can be filled with sediment or a false floor. To reduce the length scale
of turbulent structures introduced into the test section, the length of
using the Laplace equation, and the free water surface was tracked in the honeycomb is 0.2 m and is composed of 81 PVC tubes in the form
the model. The critical incoming flow velocity for the onset of tunnel of a 9 × 9 array.
scour was then calculated, and the results were found to compare well In the experiments, a smooth PVC model pipe 50 mm in diameter
with experimental data. Zang et al. (2009) also developed a numerical was fixed in the middle of the test section (see Fig. 1a). Whitehouse
model for the onset of tunnel scour by solving the flow field with the (1998) mentioned that artificially high blockage in a laboratory model
k − ω turbulence model. The average pressure gradient along the buried can be avoided if the rate of flume cross-sectional area to model cross-
pipe surface was employed in the evaluation of the onset condition with sectional area is no less than 6. However, Mao (1986) stated that the
a calibration coefficient. Zang et al. (2009) also studied the influence of blockage effect of the pipe on the flow is very limited, provided that
flow parameters, including water depth, embedment depth, boundary the ratio is less than 3.5. In this study, the ratio of O-tube cross-
layer thickness, Reynolds number (Re), and Keuleagan–Carpenter (KC) sectional area to model pipe cross-sectional area is approximately 4,
number, on the pressure variation across the pipeline. Gao and Luo and the results of validation tests (as will be shown later) are consistent
(2010) proposed a flow–pipe–seepage sequential coupling finite with previous investigations (Sumer et al., 2001).
element method (FEM) model to simulate the coupling between the Three types of siliceous sediment were used in the present study.
water flow field and soil seepage field. They indicated that the dimen- Soil properties, together with critical shear stress τcr for incipient mo-
sionless critical flow velocity changes approximately linearly with the tion in steady current conditions, were measured at the beginning of
soil internal friction angle for a submarine pipeline partially embedded the study. Relevant data are listed in Table 1, and the particle size distri-
in a sandy seabed. butions are shown in Fig. 2.
Collectively, this body of experimental and numerical work provides The physical experiments were conducted with different sediment,
a clear description of the onset of tunnel scour for a pipeline in a flat sea- pipeline embedment and lengths of seabed upstream of the pipeline.
bed, and Eq. (1) provides a particularly valuable predictive formula. It For convenience, the experiments have been divided into two
 Q. Zhang et al. / Coastal Engineering 116 (2016) 103–117 105

Fig. 1. Experimental setup. (a) Seabed model 1 (standard experiment); (b) seabed model 2 (limited sediment supply experiment).

categories: type (1) standard experiments and type (2) limited sedi- edge of the test section and the sand, a ramp with 1:2 was placed within
ment supply experiments. For the standard experiments, the sediment the sand for the standard experiments (see Fig. 1a). This was effective at
was placed across the whole (1.8 m long or 36 D) test section (see limiting local scour at the entrance of the test section and the size of an
Fig. 1a), the sediment bed is approximately 18 D on the upstream side associated bed form. In contrast to the standard experiments, the limit-
of the pipe. To limit local scour at the transition area between the ed sediment supply experiments were performed to investigate the
onset of tunnel scour below the pipe when the sediment supplied
from upstream was interrupted. For the limited sediment supply exper-
Table 1 iments, a box was made of 9 mm thick transparent plastic and was de-
Key characters of test samples.
signed to sit flush within the false floor of the working section, as shown
A sand B sand C sand in Fig. 1b. The box allowed only 0.47 m of the 1.8 m long working section
Source Factory Factory Marine
to be filled with sediment and extended only 5 D upstream of the pipe,
so the upstream sediment supply became limited quickly during the
Grain size d50 (mm) 0.24 0.60 0.22
tests. For the limited sediment supply experiments, the scour profile
d84.1 (mm) 0.32 0.72 0.37
d15.9 (mm) 0.17 0.47 0.12 upstream of the pipe roughly remained smooth throughout the experi-
Geometric standard deviation 1.37 1.24 1.76 ments, so no attempt was made to use a sloping false bed in this second
σg = (d84.1/d15.9)1/2 type of experiment.
Fines percentage (b75μm) b1% b1% 7.35%
To see beneath the pipe during the onset of tunnel scour, some type
Specific gravity of sediment grains s 2.67 2.65 2.72
Porosity of the sediment n 36.4%–40.4% 28.3%–49.7% 57.0% (2) experiments were repeated with a transparent 50-mm-diameter
Ucr at 0.05 m (m/s) 0.36 0.48 0.34 model pipe with three micro cameras inside. Because the distance
Critical shear stress, τcr (N/m2) 0.32 0.79 0.29 between the camera lens and the bottom of the pipe was limited, a slen-
Note: the notation dx indicates the grain diameter for which n% of the grains by mass is der mirror was attached to the inner wall of the transparent model pipe
finer. (see Fig. 3). Each camera had a viewing area of 70 mm along the bottom
106 Q. Zhang et al. / Coastal Engineering 116 (2016) 103–117

Fig. 2. Particle size distribution of the model sand in the MOT. (a) A sand; (b) B sand; (c) C sand.

of the model pipe; hence, the entire bottom of the model pipe was cov-
ered collectively by all three cameras.
To capture the evolution of the seabed topography before the onset
of tunnel scour, a video camera was used in each experiment. Addition-
ally, limited sediment supply test O26 was also repeated, and the flow
was stopped several times to allow the seabed to be scanned using a
handheld three-dimensional scanner. The scanner allows us to obtain
metric information about the 3D morphological feature of the seabed,
which was calibrated by a known scale (length, width, and height) ref-
erence object, and the precision is approximately 2 mm.
The steady current velocity profile during the tests was measured by
a SonTek/YSI 16-MHz 3D MicroADV (Acoustic Doppler velocimeter)
0.4 m upstream of the model pipe; an example profile is shown in
Fig. 4. To investigate the influence of the presence of the velocimeter,
some experiments were repeated without the ADV, and the scour re-
sults were similar.
2.2. Testing conditions
For every test in this study, the pipe was partially buried on the flat
seabed, and the initial burial depth of the pipeline was measured verti-
cally from the bottom of the pipeline to the surrounding level of the sea-
bed (see Fig. 5a). A complete list of the test conditions and results is
given in Table 2, in which Uc is the undisturbed steady current flow

Fig. 3. Schematic of camera pipe. Fig. 4. Flow velocity profile in test O26.
 Q. Zhang et al. / Coastal Engineering 116 (2016) 103–117 107

Fig. 5. The definition of embedment. (a) is for the flat sand bed; (b) is for the scoured sand bed.

velocity measured at the top of the pipeline, and θc is the Shields param- curve of Eq. (1) and the critical condition tests for steady currents re-
eter in steady current flow conditions defined by ported by Sumer et al. (2001). It can be seen that the results of type
(1) test O7 for e/D = 0.2 and O10 for e/D = 0.1 are in good agreement
u2c with the empirical curve mapped out by Eq. (1). This implies that the
θc ¼ ð2Þ
gðs−1Þd50 present experimental setup is capable of producing results consistent
with earlier studies in the literature.
where u∗ c is the current friction velocity, which can be determined In addition to experiments O7 and O10, several Type 2 experiments
based on the logarithmic velocity profile also exhibited the onset of tunnel scour within 30 s on a flat seabed. Of
  these, Test O24 for e/D = 0.2 led to very similar results to Type 1 Test O7,
uc z
U c ðzÞ ¼ ln : ð3Þ whereas Test O19 for e/D = 0.1 was similar to Test O10. These results
κ z0 imply that the amount of sediment upstream (and therefore the up-
stream sediment supply) has no effect on the critical velocity required
where z is the height above the seabed, z0 is bed roughness length, and κ
to cause the onset of tunnel scour, provided that the onset of tunnel
is von Karman's constant (=0.441). The bed roughness length z0 is es-
scour occurs quickly owing to the introduction of a given velocity and
timated using the expression of Christoffersen and Jonsson (1985):
that the seabed is flat. This result is not surprising given that the onset
   of tunnel scour occurs sufficiently fast in these experiments that sedi-
ks −uc ks ν
z0 ¼ 1− exp þ ; ð4Þ ment has not started to transport from the upstream side of the pipe.
30 27ν 9uc
The remaining experiments listed in Table 2 (in addition to O7, O10,
where ν is the kinematic viscosity of water (taken to be 1 × 10−6m2/s O19, and O24) that experienced onset of tunnel scour within 30 s are
throughout) and ks is the Nikuradse roughness, which is taken as plotted in Fig. 7 together with Eq. (1). In all of these cases, the seabed
2.5d50 in this study. was also nearly flat when onset of tunnel scour occurred. Fig. 7 shows
The pipe Reynolds number is given by that for these experiments, the onset of tunnel scour was observed
when the velocity was at or above the critical velocity indicated by
Uc D Eq. (1), which is consistent with the observation of Sumer et al. (2001).
Rec ¼ : ð5Þ
ν
3.2. Onset of tunnel scour on an uneven seabed
Each of the experiments was run until at least one of the following
three conditions occurred: In addition to the observations of tunnel scour within 30 s of the flow
being initiated, the onset of tunnel scour was also observed to occur in
1) onset of tunnel scour;
some experiments after a relatively long duration. Over this longer
2) no onset of tunnel scour, but an equilibrium profile was achieved
time period, a significant amount of sedimentation occurred, leading
(i.e., tests O3, O6, and O18);
to changes in the local seabed profile around the pipeline. To provide
3) in the type (1) standard experiments, the test had run for so long
an example, Fig. 8 displays a photograph at the point of onset of tunnel
that the sediment supply from upstream started to decrease (ob-
scour for Test O19, which occurred quickly when the seabed was still
served as a lowering of the upstream seabed; i.e., test O15).
flat, and Test O13, which occurred after several minutes on an uneven
Table 2 also shows the duration of each test in minutes and (where bed. In Fig. 8b, it is clear that the local seabed in Test O13 is significantly
relevant) the ‘time to onset’ in seconds, which is the time from the start different from the initial flat seabed condition. It can also be seen that a
of a test to the time when the onset of tunnel scour was observed. mixture of sand and water breaks through the sediment slope at the
downstream side of the pipe, which is consistent with the description
3. Experimental results of the piping process given in the literature (Chiew, 1990; Sumer
et al., 2001). To follow the definition of embedment on the flat seabed,
3.1. Onset of tunnel scour on a flat seabed the definition of embedment on the uneven seabed is shown in
Fig. 5b, which is started from the joint point of the water, sediment,
In previous experiments, Sumer et al. (2001) observed that the onset and pipe on the upstream side of pipe to the joint point on the down-
of tunnel scour occurred on the order of 10 s after the flow was intro- stream side of the pipe.
duced and that the seabed around the pipeline remained flat prior to To summarise the collective set of experimental observations of tun-
the onset of tunnel scour. A similar phenomenon was observed in nel scour following sedimentation, all experiments for which the onset
some type (1) standard experiments and type (2) limited sediment sup- of tunnel scour was observed following a flow duration of more than
ply tests. In this study, the tests in which the seabed remained flat prior 30 s are plotted in Fig. 7 It can be observed in this figure that the onset
to the onset of tunnel scour are regarded as flat seabed onset of tunnel of tunnel scour in these experiments occurred when the velocity was
scour tests, all of which featured a time to the onset of tunnel scour of below the critical velocity indicated by Eq. (1). Moreover, most of the
less than 30 s (see Table 2). Among these, those with the lowest flow ve- experiments are of the type (2) limited sediment supply experiments.
locity for e/D = 0.1 and e/D = 0.2 are listed in Fig. 6 with the critical For the uneven seabed onset of tunnel scour test O1, which is a type
108 Q. Zhang et al. / Coastal Engineering 116 (2016) 103–117

Table 2
Test conditions for onset of scour tests.

Test no. Test type Sand type Initial Flow Sand size Specific Porosity Shields Pipe Reynolds Final Run Time to Note
velocity gravity parameter number Local⁎ time onset

e/D Uc d50 s n θc Rec U 2C e/D

gðs−1Þð1−nÞD
(×104)
(m/s) (mm) (min) (s)

O1 1 A 0.2 0.66 0.24 2.67 0.384 0.28 3.30 0.86 0.16 3 60

O2 1 A 0.2 0.58 0.24 2.67 0.384 0.22 2.90 0.66 0.81 30 – –
O3 1 A 0.2 0.52 0.24 2.67 0.384 0.17 2.60 0.53 0.45 40 – –
O4 1 B 0.2 0.58 0.60 2.65 0.39 0.12 2.90 0.68 0.83 20 – –
O5 1 B 0.2 0.69 0.60 2.65 0.39 0.17 3.44 0.95 0.79 40 – –
O6 1 B 0.2 0.71 0.60 2.65 0.39 0.18 3.55 1.01 0.51 50 – –
O7 1 B 0.2 0.83 0.60 2.65 0.39 0.25 4.15 1.39 0.20 5 b30 –
O8 1 B 0.1 0.32 0.60 2.65 0.39 0.03 1.60 0.21 0.35 10 – –
O9 1 B 0.1 0.39 0.60 2.65 0.39 0.05 1.97 0.31 0.41 10 – –
O10 1 B 0.1 0.46 0.60 2.65 0.39 0.07 2.31 0.43 0.10 2 b30
O11 2 B 0.2 0.69 0.60 2.65 0.39 0.17 3.44 0.95 0.16 3 234
O12 2 B 0.2 0.64 0.60 2.65 0.39 0.15 3.18 0.81 0.15 30 265/467⁎⁎⁎ Repeat⁎⁎
O13 2 B 0.2 0.56 0.60 2.65 0.39 0.11 2.82 0.64 0.13 30 508/1157⁎⁎⁎ Repeat⁎⁎
O14 2 B 0.2 0.51 0.60 2.65 0.39 0.09 2.54 0.52 0.11 30 1265
O15 2 B 0.2 0.45 0.60 2.65 0.39 0.07 2.24 0.40 30 – –
O16 2 B 0.1 0.52 0.60 2.65 0.39 0.09 2.60 0.54 0.12 1.7 90
O17 2 B 0.1 0.48 0.60 2.65 0.39 0.08 2.40 0.46 0.10 4 134
O18 2 B 0.1 0.39 0.60 2.65 0.39 0.05 1.95 0.31 4 – –
O19 2 A 0.1 0.50 0.24 2.67 0.384 0.16 2.48 0.48 0.10 1 b20
O20 2 A 0.1 0.58 0.24 2.67 0.384 0.22 2.88 0.65 0.10 1 b10
O21 2 A 0.1 0.65 0.24 2.67 0.384 0.27 3.24 0.82 0.10 1 b20
O22 2 A 0.2 1.17 0.24 2.67 0.384 0.92 5.87 2.71 0.20 1 b20
O23 2 A 0.2 0.95 0.24 2.67 0.384 0.60 4.74 1.77 0.20 1 b5
O24 2 A 0.2 0.79 0.24 2.67 0.384 0.41 3.97 1.24 0.20 1 b10
O25 2 A 0.2 0.72 0.24 2.67 0.384 0.34 3.59 1.01 0.16 5 213/90⁎⁎⁎ Repeat⁎⁎
O26 2 A 0.2 0.60 0.24 2.67 0.384 0.23 3.00 0.71 0.13 3 102
O27 2 A 0.2 0.52 0.24 2.67 0.384 0.17 2.59 0.53 0.13 15 722/840⁎⁎⁎ Repeat⁎⁎
O28 2 C 0.2 0.38 0.22 2.72 0.57 0.24 1.88 0.39 0.2 1 30
O29 2 A 0.4 1.02 0.24 2.67 0.384 0.69 5.10 2.06 1
O30 2 A 0.6 1.02 0.24 2.67 0.384 0.69 5.10 2.06 3
O31 2 A 0.6 0.97 0.24 2.67 0.384 0.62 4.84 1.86 3
O32 2 A 0.6 0.92 0.24 2.67 0.384 0.06 4.60 1.68 3
O33 2 A 0.5 0.92 0.24 2.67 0.384 0.17 4.60 1.68 2
O34 2 A 0.3 0.72 0.24 2.67 0.384 0.34 3.61 1.03 1
O35 2 A 0.3 0.92 0.24 2.67 0.384 0.06 4.60 1.68 1
O36 2 A 0.7 1.02 0.24 2.67 0.384 0.17 5.10 2.06 3

Note: the model pipe diameter is D = 50 mm.

⁎ Final local embedment for the test without the onset of scour or local embedment before the onset of scour, estimated with a video camera inside and outside. From O29, the flow
velocity is relatively high, and the test water becomes very cloudy, so no final local embedment is estimated.
⁎⁎ The same condition was repeated with a transparent pipe containing a micro camera.
⁎⁎⁎ The time to onset in the repeated test. The tests with no onset of scour are labelled with ‘–’ in the ‘Time to onset’ column.

(1) standard test, the upstream sediment supply was not limited artifi-
cially. However, the time to onset for test O1 was only 60 s, and it was
observed that during this time period, local scour around the pipe devel-
oped fairly rapidly. In contrast, the rate of sediment transport to the
front of the pipe was relatively slow. Both effects led to a temporary sed-
iment supply shortage and the immediate onset of tunnel scour. These
results indicate that if the upstream sediment supply is limited, the
onset of tunnel scour can still occur when the combination of velocity
and initial pipeline embedment is insufficient to cause scour according

Fig. 6. Experiments on onset of tunnel scour. Test results from Sumer et al. (2001) and
results from this investigation (O15 and O20) are shown with the curve of Eq. (1). Fig. 7. Flat and uneven onset of tunnel scour tests.
 Q. Zhang et al. / Coastal Engineering 116 (2016) 103–117
Fig. 8. Snapshot of onset of tunnel scour during tests. (a) Flat seabed onset of tunnel scour
tests (Test O19); (b) uneven onset of tunnel scour tests (Test O13).
to Eq. (1). In the remainder of this section, we focus on understanding
why this is possible.
3.3. Sediment transport and sedimentation around pipeline
In the longer-duration experiments (i.e., the experiments that ex-
hibited the onset of tunnel scour on an uneven seabed), the seabed
formed a distinctive profile prior to the onset of tunnel scour as a result
of local sedimentation. This profile is illustrated in Fig. 9b and was
characterised by a scour hole on the upstream side of the pipe, and
Fig. 9. Sediment transport around pipeline. (a) Flat seabed; (b) uneven seabed.
109
sediment build up on the downstream side of the pipe. To explain
how this characteristic profile formed, we describe the sediment trans-
port processes that were observed during the experiments in this
section. To facilitate this discussion, Fig. 9 sketches the main flow struc-
tures that were observed around the pipeline in the experiments both
prior to any sedimentation and following significant changes to the
local scour profile. For the case of a flat seabed, this figure is consistent
with that presented by Mao (1987), who documented three main vorti-
ces: a luff vortex A in front of the pipe, and vortices B and C formed in the
lee of the pipe. As the seabed profile evolved, vortices A and C persisted,
and vortex B disappeared.
Fig. 9a also indicates several sediment fluxes, labelled qin, qout, qluff,
qflyover, and qlee. These fluxes encapsulate the volumetric transport of
sediment observed in the experiments. The first of the fluxes, qin, de-
fines the sediment transport (in the bed load and suspended load)
that approached the pipeline from upstream. In addition to this influx
of sediment, it was also observed that vortex A led to entrainment of
sediment upstream of the pipeline. In some of the experiments, vortex
A was sufficiently strong that this entrained sediment was advected
over the pipe, contributing to the transport of sediment qluff. This
entrained sediment then combined with all (or some fraction of) qin,
leading to qflyover. On the downstream side of the pipe, vortex C
transported sediment from the bed load back towards the pipe, leading
to qlee. In addition, some sediment in the far wake combined with
sediment passing over the pipeline was seen to be advected away
downstream. This total downstream transport is denoted qout.
In each of the experiments, it was seen that relative changes in these
different sediment fluxes led to changes in the seabed profile. More spe-
cifically, defining the volume of sediment upstream and downstream of
the pipe as Vup and Vdown, respectively, these changes in the sediment
profile were related to the fluxes according to


d V up qin −qflyover
¼ ; ð6Þ
dt 1−n
and
dðV down Þ qflyover þ qlee −qout
¼ : ð7Þ
dt 1−n
Consequently, in the present experiments, if the upstream supply qin
was zero or small (due to limited sediment supply) relative to qflyover,
110 Q. Zhang et al. / Coastal Engineering 116 (2016) 103–117
then dVup/dtb 0, and the formation of a scour depression in front of the
pipe could be observed, leading to the upstream scour hole in Fig. 9b. Al-
ternatively, downstream of the pipe, it was generally found that the nu-
merator on the right-hand side of Eq. (7) was positive and dominated by
qlee. Consequently dVdown/dt N 0, which led to local sediment build-up
downstream of the pipe evident in Fig. 9b.
Throughout the experiments, the rates implied by Eqs. (6) and (7)
were not constant but varied in time. For instance, for the Type 2 exper-
iments, the sediment supply decreased over time, as sediment up-
stream of the pipeline eroded. As such, the local profile evolved over
time towards that shown in Fig. 9b. This evolution is explained in
more detail in the next subsection for the pipeline experiments, which
eventually led to onset of tunnel scour after a long duration of testing.
3.3.1. Description of experiments with limited sediment supply
Owing to sedimentation, the local pipeline embedment changed
prior to the onset of tunnel scour. In this section, we review these
changes for four experiments (Test O12, O13, O25, O26 and O27),
each of which experienced the onset of tunnel scour after a long dura-
tion. To capture the changes in embedment due to sedimentation in
Fig. 10. The scour profile evolution for Test O26. (a) initial embedment (e/D≈0.2); (b) embedment build-up (e/D≈0.37); (c) embedment reduction (e/D≈0.2); (d) just before onset (e/
D≈0.13).
these experiments, Fig. 5b defines a measure of embedment depth
based on the chord length associated with the perimeter of the pipe in
contact with the seabed. For convenience we split the sedimentation
process into four stages and describe changes to embedment observed
during the tests with reference to Fig. 10.
• Stage 1: In the first stage of the experiments, the sediment in the front
of the pipe (qin) was transported by the steady current to the corner
between the pipe and bed on the upstream side of the pipe, increasing
the local embedment. Meanwhile, vortex A eroded the sand bed at the
corner between the pipe and bed on the upstream side of the pipe,
leading to a local scour pit in the increased sand bed (see the scour
profile on the upstream side of the pipe in Fig. 10b). On the down-
stream side of the pipe, vortex B transported sediment towards the
pipe and led to a local build-up of sediment (see Fig. 10b for an exam-
ple). The net result of the upstream erosion and downstream sedi-
ment build-up was to cause an increase in the embedment of the pipe.
• Stage 2: In the second stage, the sand bed level increased owing to qin
and decreased owing to qluff on the upstream side of the pipe tend to
balance each other. On the downstream side of the pipe, the sediment
continued to build up the sand bed level. At some point, the
 Q. Zhang et al. / Coastal Engineering 116 (2016) 103–117 111

downstream build-up of sediment reached a maximum value, which

coincided with an equilibrium downstream scour profile equivalent
to that shown in Fig. 10c. The maximum embedment of the pipe
was observed.
• Stage 3: In the third stage, owing to the limited sediment supply, there
was a noticeable reduction in sediment supplied from upstream
(i.e., qin decreased), which was associated with renewed growth of
the scour depression in front of the pipe (suggesting that qluff N 0).
No change in the downstream profile was observed, so the embed-
ment of the pipe (driven by upstream scouring) began to decrease;
see the 3D scour profile in Fig. 10c.
• Stage 4: Finally, in the last stage, a mixture of sand and water broke
through from one point on the downstream side of the pipe, leading
to the onset of tunnel scour. The mixture of sand and water was sim-
ilar in resemblance to that for the flat seabed onset of tunnel scour
tests (see the screenshot from the camera inside the model pipe in
Fig. 12. The embedment change with time for various experiments reported in Table 2.
Fig. 11).

The variation of pipe embedment during sediment B (tests O12 and
O13) and sediment A (tests O25, O26, and O27) experiments are plotted
in Fig. 12 In the figure, the total embedment of the pipe was measured
from the video recordings outside the O-tube and micro cameras inside
the model pipe to obtain the best estimation of the embedment at the
centre of the pipe. It should be noted that the embedment of pipe
shown in Fig. 12 is the local embedment defined by Fig. 5. With respect
to Fig. 12, it is evident that at first, the local embedment of the pipe in-
creased (i.e., Stage 1). This increase then continued, but at a faster rate,
until the embedment of the pipe reached the maximum value (i.e., Stage
2). The local embedment then proceeded to decrease when the
upstream supply decreased (Stage 3), until the onset of tunnel scour
occurred (Stage 4).
A second observation from Fig. 12 is that the local embedment just
before the onset of tunnel scour appears to be similar to that defined

Fig. 11. Screenshot from the camera inside the pipe. (a) Before piping occurs; (b) piping occurs.
112 Q. Zhang et al. / Coastal Engineering 116 (2016) 103–117

Fig. 13. Model validation computational mesh around the pipeline. (a) e/D=0.0; (b) e/D=0.1.

by Eq. (1) despite the fact that Eq. (1) is derived from experiments on a 4.1. Governing equations and numerical method
flat seabed. To investigate why this may be the case, numerical simula-
tions are presented in the next section to investigate the pressure differ- 4.1.1. Flow field
ence driven by current flow between the upstream and downstream In this work, the incompressible flow above the seabed has been
sides of the pipe on the uneven seabed and the seepage flow around modelled using the two-dimensional Reynolds-averaged Navier–
the pipe in the uneven seabed. The simulations focus on experiment Stokes (RANS) equations, which can be written in Cartesian coordinate
O26 in particular to provide a detailed back analysis of the experimental form together with the continuity equation as
results.
∂  
2
∂ui ∂u 1 ∂p ∂ u
þ uj i ¼ − þ ν 2i þ −u;i u;j ð8Þ
4. Numerical simulations ∂t ∂x j ρ ∂xi ∂xi ∂x j

The onset of tunnel scour underneath a partially buried pipeline in ∂ui

¼0 ð9Þ
currents is dependent on two-flow processes: one is the flow around ∂xi
the pipeline, and the other is the seepage field within the porous under-
lying sediment. To investigate these two flow processes, a series of nu- where ui is the velocity component, u,i and u,j are the fluctuating veloci-
merical simulations has been undertaken using FLUENT. ties, t is the time, ρ is the density of the fluid, p is the pressure, ν is the

Fig. 14. Pressure coefficient distribution (Re=4.8×104), (a) e/D=0.0; (b) e/D=0.1.
 Q. Zhang et al. / Coastal Engineering 116 (2016) 103–117 113

Fig. 15. The computational mesh in the vicinity of the pipe. (a) N1 (e/D= 0.2(0.2)); (b) N2 (e/D= 0.37(0.2)); (c) N3 (e/D= 0.2(0.2)); (d) N4 (e/D= 0.13(0.2)).

kinematic viscosity of the fluid, and xi and xj are the coordinates in the
horizontal and vertical directions, respectively.
In the present work, the SST (shear stress transport) k − ω turbu-
lence equation developed by Menter (1994) has been used to calculate
the Reynolds stresses. In practise, the k − ε models are generally well
behaved in the far field, whereas the k − ω models are more accurate
and much more numerically stable in the near-wall region. Moreover,
the freestream independence of the k − ε model in the far field makes
the SST k − ω model more accurate and reliable for a wider class of
flows than the standard k − ω model. The specific details of SST k − ω
equations and corresponding coefficients can be found in Menter et al.
(2003) and Fluent (2012).

Fig. 16. Streamlines around the pipeline. (a) e/D=0.2(0.2)flat seabed; (b) e/D=0.37(0.2) uneven seabed; (c) e/D=0.2(0.2) uneven seabed; (d) e/D=0.13(0.2) uneven seabed.
114 Q. Zhang et al. / Coastal Engineering 116 (2016) 103–117

Fig. 17. Pressure coefficient along the seabed for four typical scour profiles prior to onset of Fig. 18. Pressure coefficient along the flat seabed with different pipe embedment.
tunnel scour. (Re=3×104).

4.1.2. Seepage flow field
In the seabed, Darcy's law is used to calculate the velocity of seepage
flow, which is given by
K P dp
U Bi ¼ −
ρν dxi
where UBi is the discharge of water per unit area in the xi direction, dp/
dxi is the pressure gradient and Kp is the specific permeability (m2),
which is governed by the properties of the porous medium, including
the grain size d50 and porosityn. In this work, the specific permeability
has been computed using the expression proposed by Ergun (1952):
boundary is defined as the symmetry plane, and the seabed and cylinder
are defined as non-slip smooth wall boundaries. The initial condition is
set to meet the Reynolds number and turbulence intensity (0.2%) in the
physical experiment. The cylinder surface is discretized by 210 nodal
points, and the minimum element size in the radial direction next to
the pipeline surface is 0.0002 D. This is equivalent to a y+ of less than
ð10Þ
1 (where y+ = u∗c Δ/ν is the non-dimensional mesh size, and u∗c is the
friction velocity). The total number of nodes employed to discretise
the computational domain is 30,600, with convergence tests indicating
that the numerical results changed little with increasing mesh density.
Finally, a non-dimensional time step of UmΔt/D = 0.0024 was used in
the calculations.

2
4.2.2. Model results
d n3 Fig. 14 shows the calculated pressure coefficient along the cylinder
K p ¼ 50 ð11Þ
150 ð1−nÞ2 surface and on the plane boundary on either side of the pipeline, togeth-
er with the measured data of Bearman and Zdravkovich (1978) for e/
D = 0.0 and e/D = 0.1. The pressure coefficient Cp in Fig. 14 is defined as
4.2. Validation of the numerical model
p−p∞
4.2.1. Model setup Cp ¼ ð12Þ
0:5ρU 2c
The numerical model for the flow field has been validated against
experimental data presented by (Bearman and Zdravkovich, 1978) for
steady current flow past a pipeline resting on an impermeable plane where p is the pressure, p∞ is the pressure at the far field, ρ is the fluid
boundary (for Re = 4.8 × 104). In the experiments by Bearman and density, and Uc is the amplitude of the upstream steady current at the
Zdravkovich (1978), the boundary layer thickness is 0.8 D at the posi- top of the pipe. It can be seen that the calculated pressure coefficient
tion of the cylinder. To obtain a similar thickness of the boundary agrees well with the experimental data on the pipe surface and the sea-
layer in the numerical model, a long computational domain of 200 D bed surface. The pressure distribution along the seabed calculated from
in length and 10 D in height was generated without the pipe. Steady the flow field simulation is used to calculate the seepage flow in the soil.
flow was then simulated in this pipe, with a uniform upstream velocity
boundary condition. It was found that a boundary layer 0.8 D thick oc- 4.3. Flow structure and pressure distribution
curred at approximately 60 D from the inlet boundary. Therefore, a rect-
angular computational domain of 120 D in length and 10 D in height To investigate the flow structure and pressure distribution around
was discretised, and the pipeline was placed in the centre of the domain. the pipeline for an uneven seabed, four typical seabed profiles were con-
In both the long domain and the domain with the pipe, structured, sidered based on the scour profile observed in Test O26. The computa-
quadrilateral, four-node elements were used. Fig. 13 illustrates the com- tional mesh and seabed profile in the vicinity of the pipeline for these
putational mesh in the vicinity of the cylinder. The left boundary is de- profiles are shown in Fig. 15 It should be noted that the embedment
fined as the inlet, the right boundary is defined as the outlet, the top written in brackets for each subfigure is the initial embedment on a
flat seabed. In this figure, Case N1 represents the initial embedment
Table 3 condition on a flat sand bed at the start of Stage 1 (when e/D = 0.2).
Pressure difference for all numerical simulation cases. Case N2 represents a point during Stage 1 when the embedment of
Local Far field the pipe is increasing. Case N3 represents a point during Stage 3 when
Case no. e/D e/D Morphology ΔCp the embedment has decreased to an embedment equivalent to the ini-
N1 0.2 0.2 Flat 1.239
tial embedment (i.e., e/D = 0.2) but on an uneven seabed. Finally, Case
N2 0.37 0.2 Uneven 1.210 N4 represents the scour profile just before the onset of tunnel scour oc-
N3 0.2 0.2 Uneven 1.230 curred in experiment O26.
N4 0.13 0.2 Uneven 1.222 A rectangular computational domain 28 D in length and 4 D in height
N5 0.0 0.0 Flat 1.616
has been adopted for the simulations. These dimensions were chosen
N6 0.13 0.13 Flat 1.358
N7 0.3 0.3 Flat 1.080 because the measured velocity profile in Test O26 (used for the up-
N8 0.37 0.37 Flat 0.965 stream boundary condition in the numerical model) was measured
N9 0.13 0.13 Uneven 1.323 4 D upstream of the pipe, and the experimental flume extended 28 D
N10 0.3 0.3 Uneven 1.045 downstream of the pipe. The boundary condition configurations are
N11 0.37 0.37 Uneven 0.986
consistent with the validation tests, but the initial condition is
 Q. Zhang et al. / Coastal Engineering 116 (2016) 103–117
Fig. 19. Pressure distribution in the water and in the soil, seepage flow streamlines in the soil. (a) N1 (e/D = 0.2(0.2)); (b) N2 (e/D = 0.37(0.2)); (c) N3 (e/D = 0.2(0.2)); (d) N4
(e/D = 0.13(0.2)).
calculated based on the Reynolds number (Re= 3 × 104) and turbulence
intensity (5%) of the present physical experiment.
The calculated flow fields for the different seabed profiles are shown
in Fig. 16 It can be seen in Fig. 16a that the three major vortices observed
by Mao (1986) are reproduced numerically. However, as the seabed
profile changes, various changes to the smaller vortices A and B occur.
For instance, the small lee wake vortex B disappears as the berm is
built up downstream of the pipeline, whereas the luff vortex A grows
in size as the hole in front of the pipeline grows. In contrast to vortices
A and B, the large lee wake vortex C is relatively unperturbed by the
changes in profile local to the pipe.
The pressure along the seabed for each of the four scour profiles
modelled is illustrated in Fig. 17 (with the centre of the pipe located at
x = 0). This figure confirms the existence of a pressure difference up-
stream and downstream of the pipe on an uneven seabed. Fig. 17 also
indicates that the pressure remains almost constants for a length of ap-
proximately 1 D upstream of the pipe and 3 D downstream of the pipe.
The pressure difference upstream and downstream of the pipe is listed
in Table 3 for all geometries. For each of the cases, the pressure differ-
ence is very similar, with a slightly larger difference for case N1, follow-
ed by case N3 and case N4. The uneven seabed case with e/D = 0.37 has
the smallest pressure difference.
To further investigate the relationship between the pressure differ-
ence and the embedment, several supplementary flat seabed cases
were simulated with e/D=0.13, 0.3, and 0.37. The corresponding calcu-
lated results are also listed in Table 3. The pressure along the seabed in
115
the vicinity of the pipeline is shown in Fig. 18 These results indicate that
the pressure difference gradually decreases with increasing embed-
ment, in agreement with the results presented by Zang et al. (2009).
Comparing back to Fig. 17, it is therefore evident that the pressure dif-
ference is more dependent on the initial pipeline embedment than
changes to the local embedment due to sedimentation.
4.4. Stability analysis of sand grains
For the seepage field simulation, a computational domain of D in
length and 2 D in height was discretised by four-node quadrilateral
elements. The top boundary was formed by the seabed and the buried
perimeter of the pipe. The length and mesh node distribution on the
top boundary were exactly the same as those of the bottom boundary
of the flow field mesh, and the buried part of the pipe has been defined
as smooth, impermeable wall. The results for the pressure along the
seabed calculated from the flow field simulation were directly trans-
ferred to the seepage field simulation to provide the driving boundary
condition. The sides and bottom of the seepage field domain were
defined as impermeable walls. The seepage flow in the seabed was
regarded as laminar flow (based on Darcy's law, as mentioned above),
so the boundary layer and wall function were not applied in this simu-
lation. A porosity of n = 0.4 was used to define the specific permeability
Kp via Eq. (10).
The onset of tunnel scour due to piping occurs on the surface of a co-
hesionless granular material when the seepage force Fs = − (∂p/∂s)V
116 Q. Zhang et al. / Coastal Engineering 116 (2016) 103–117
Fig. 20. Pressure gradient along the seabed on the downstream side of the pipeline. (a) N1 (e/D= 0.2(0.2)); (b) N3 (e/D= 0.2(0.2)).
acting on a soil element exceeds the submerged weight of the soil
element W = ρg(s − 1)(1− n)V (Sumer et al., 2001; Zang et al., 2009);
i.e., where
−ð∂p=∂sÞV ≥ρgðs−1Þð1−nÞV
where ∂p/∂s is the pressure gradient along a streamline and V is the vol-
ume of the soil element. Therefore, the point with maximum pressure
gradient along the seabed on the downstream side of the pipe is the
most likely point for the onset of tunnel scour to occur.
Fig. 19 illustrates the calculated pressure distribution in the water
and in the soil for the four scour profiles associated with Test O26. The
numerical labels in Fig. 19 indicate the pressure field and the seepage
flow velocity streamlines, i.e., the flow net.
For the flat seabed case (Fig. 19a), the flow net indicates that the
maximum pressure gradient occurs adjacent to the pipe on the down-
stream soil surface. This is in good agreement with the location of the
breakout of the sand particles observed in the experiments (see
Fig. 8a) and previous results (Gao and Luo, 2010; Sumer et al., 2001).
As noted by Liang and Cheng (2005), however, the location of this max-
imum coincides with a singular point at the intersection between the
pipeline and the seabed in the seepage flow calculation. Consequently,
the evaluation of the exit seepage gradient at the rear corner point of
the pipeline is impossible and must be obtained approximately (by
evaluating the gradient at the grid point next to the singular point, for
example, see Zang et al., 2009). Apart from the singular points, the seep-
age exit pressure gradient distribution along the seabed on the down-
stream side of the pipe for the flat seabed case is calculated based on
Eq. (13) and illustrated in Fig. 20a.
For the uneven seabed situations, such as cases N2–N4 shown in
Fig. 19b,c,d, a singular point still exists at the upstream corner between
the pipeline and seabed; however, on the downstream side of the pipe,
it can be seen in Fig. 20b that the maximum seepage exit pressure gra-
dient is not adjacent to the seabed and pipe but is located some distance
from the pipeline. This means that the exit pressure gradient is well
defined. Furthermore, this result is consistent with observations of the
location where the mixture of sand and water broke through at the
point of piping.
5. Conclusions
The conclusions derived from the physical experiments are
summarised as follows:
1) The onset of tunnel scour may occur at a velocity lower than the crit-
ical velocity defined by Sumer et al. (2001) if the upstream sediment
supply is restricted.
2) If the onset of tunnel scour occurs because of limited sediment sup-
ply, it will take a relatively long time (compared with velocities
above the critical velocity for the onset of tunnel scour on a flat
seabed).
3) When the onset of tunnel scour occurs because of limited upstream
sediment supply, the local seabed profile immediately before onset
ð13Þ is quite different from the initial profile. This is due to luff and lee
wake erosion. It should be noted that the conclusion that both luff
and lee wake erosion affect the potential for piping and tunnel
scour is consistent with comments by Sumer et al. (2001), in
which they note that the presence of vortices in front of the pipe
and in the lee wake may contribute to the onset of tunnel scour.
4) Observations from micro cameras within a transparent pipe con-
firmed that the onset of tunnel scour occurs because of piping even
when the local seabed is uneven.
From the numerical simulations, the following conclusions can be
summarised:
1) The luff vortex still exists in the scour hole on the upstream side of
the pipe prior to the onset of tunnel scour on an uneven seabed,
and the lee wake vortex maintained the sediment slope on the
downstream side of the pipe.
2) The pressure difference upstream and downstream of the pipe is
mainly dependent on the initial embedment and is not very sensi-
tive to changes in the local seabed geometry.
3) The maximum seepage exit pressure gradient for the case of uneven
seabed onset of tunnel scour is not at the joint point of the pipe and
seabed as in the case of flat seabed onset of tunnel scour but is on the
downstream slope.
Acknowledgments
The first author would like to acknowledge the support of the
Australian Government and The University of Western Australia by pro-
viding SIRF and UIS scholarships for a doctoral degree.
References
Bearman, P., Zdravkovich, M., 1978. Flow around a circular cylinder near a plane bound-
ary. J. Fluid Mech. 89 (1), 33–47.
Chiew, Y.M., 1990. Mechanics of local scour around submarine pipelines. J. Hydraul. Eng.
ASCE 116 (4), 515–529.
Christoffersen, J.B., Jonsson, I.G., 1985. Bed friction and dissipation in a combined current
and wave motion. Ocean Eng. 12 (5), 387–423.
Ergun, S., 1952. Fluid flow through packed columns. Chem. Eng. Prog. 48, 89–94.
Fluent, A., 2012. 14.5, Theory Guide. Ansys, Inc., Canonsburg, PA.
Gao, F.P., Luo, C.C., 2010. Flow-pipe-seepage coupling analysis of spanning initiation of a
partially-embedded pipeline. J. Hydrodyn. 22 (4), 478–487.
Liang, D., Cheng, L., 2005. A Numerical Model of Onset of Scour below Offshore Pipelines
Subject to Steady Currents. Taylor & Francis, Frontiers in offshore geotechnics.
Mao, Y., 1986. The Interaction between a Pipeline and an Erodible Bed. Technical Univer-
sity of Denmark, Denmark.
Mao, Y., 1987. The Interaction between a Pipeline and an Erodible Bed. SERIES PAPER
TECHNICAL UNIVERSITY of DENMARK 39.
 Q. Zhang et al. / Coastal Engineering 116 (2016) 103–117 117

Menter, F.R., 1994. Two-equation eddy-viscosity turbulence models for engineering appli-
cations. AIAA J. 32 (8), 1598–1605.
Menter, F.R., Kuntz, M., Langtry, R., 2003. Ten years of industrial experience with the SST
turbulence model. Turbul. Heat Mass Transf. 4, 625–632.
Sumer, B.M., Fredsøe, J., 2002. The Mechanics of Scour in the Marine Environment. World
Scientific, Singapore.
Sumer, B.M., Truelsen, C., Sichmann, T., Fredsøe, J., 2001. Onset of scour below pipelines
and self-burial. Coast. Eng. 42 (4), 313–335.
Westgate, Z.J., 2013. The As-Laid Embedment of Subsea Pipelines. The University of
Western Australia, Perth, Australia.
Whitehouse, R., 1998. Scour at Marine Structures: A Manual for Practical Applications.
Thomas Telford, UK.
Zang, Z., Cheng, L., Zhao, M., Liang, D., Teng, B., 2009. A numerical model for onset of scour
below offshore pipelines. Coast. Eng. 56 (4), 458–466.