Williams, E. S. et al. (2013). Geotechnique 63, No. 11, 946956 [http://dx.doi.org/10.1680/geot.12.P.

067]

Pipe uplift in saturated sand: rate and density effects

E . S . W I L L I A M S  , B. W. B Y R N E a n d A . B L A K E B O RO U G H

A key parameter for assessing the unburial behaviour of offshore pipelines, such as during upheaval
buckling, is the soil uplift resistance. Despite much research on uplift response, two variables remain
relatively unexplored: the effect of low relative density backfill and partially drained soil response. To
address this shortcoming, an investigation into the plane-strain uplift resistance of pipe segments in
saturated sand was carried out. The model tests, targeting drained and partially drained soil behaviour,
demonstrate that soil state strongly influences the pipesoil failure mechanism, and hence the
mobilised uplift resistance. In very loose sand the normalised peak uplift resistance remained constant
beyond a critical embedment depth, indicating a possible transition between different failure mechanisms. Faster rate tests result in positive excess pore pressures, which serve to further reduce the uplift
resistance. Tests at varying uplift rates and densities identified a transition relative density between net
contraction and dilation for a given embedment. These findings suggest that a minimum backfill
density should be ensured prior to pipeline operation, to avoid the negative effects associated with
contractive behaviour of loose soil.
KEYWORDS: liquefaction; model tests; offshore engineering; pipelines; sands; soil/structure interaction

INTRODUCTION
The exploitation of offshore oil and gas reserves requires a
vast network of subsea pipelines, such as those used to
connect newly discovered oil fields to existing production
facilities. These pipes are often buried to some depth of soil
cover for thermal insulation and pipeline stability, and to
protect them from mechanical damage due to exposure
above the seabed (Schupp, 2009). The required embedment
depth must be sufficient to prevent pipeline unburial behaviour, but should be minimised for economy of design
(deeper embedment depths are generally associated with
greater installation costs). One of the main pipeline unburial
mechanisms is upheaval buckling; a structural effect caused
by high operating temperatures and pressures. A key parameter in assessing pipeline stability against upheaval buckling
is the vertical resistance provided by the soil to upward
movement of the pipe. Previous research on this topic has
focused on understanding the deformation mechanisms and
the forcedisplacement response during pipe uplift, generally assuming drained conditions. This has led to prediction
models that form the basis of recommendations for pipeline
designers, such as in DNV-RP-F110 (DNV, 2007).
Although numerous studies on uplift resistance have been
carried out over the last three decades, there are some issues
that are not yet fully resolved, and therefore are not properly
accounted for in the industry guidelines: the impact of the
backfill relative density, particularly in very loose sand, on
the induced failure mechanisms, and the corresponding uplift
resistance. Furthermore, partially drained conditions stemming from faster uplift rates or pipe vibration may also be
relevant when installation methods such as jet-trenching
result in very loose, liquefiable backfill soils. In this instance
it is important to account for both density and rate effects,
and appropriate provisions for these conditions should be
included in design guidelines. This paper reports a comprehensive investigation into these two effects.

BACKGROUND
Prediction methods for drained resistance
The basis for a prediction model for peak uplift resistance
is a realistic understanding of the failure mechanism for a
pipe moving upwards in the soil. Early work, such as that by
Trautmann et al. (1985) or Schaminee et al. (1990), concluded that the primary failure mechanism consists of the
upward heave of a sliding block of soil bounded by shear
planes/zones that, at shallow burial depths, extend to the
ground surface. More sophisticated imaging techniques used
in recent years have confirmed this assumption, and provide
a better understanding of its range of applicability (e.g.
White et al., 2001; Byrne et al., 2008).
Based on these observations, the simplest model is the
vertical slip model (VSM), which assumes that the failure
planes of the soil extend vertically upwards from the edges
of the pipe to the ground surface, as shown in Fig. 1(a). The
uplift resistance is composed of the weight of the soil prism
above the centre of the pipe less the soil displaced by top
half of the pipe and the shear resistance of the slip
surfaces caused by the lateral earth pressure. A typical
calculation is given by:


D
H
9HDL
(1)
K tan 9
F ult 1 
8H
D
where D is the pipe diameter, H is the embedment depth
to the pipe centre (note that this is different from the
depth of cover defined in the industry, which is usually
measured to the pipe crown), and 9 and g9 are the
effective friction angle and unit weight respectively. Typically, K is taken as the at-rest lateral earth pressure coefficient (K0 1  sin 9); alternatively, K tan 9 can be
replaced by an empirical uplift factor f, based on broad
soil and density classifications. An adapted version of this
model (the aVSM) defined by Schupp (2009), which does
not subtract the displaced soil, is used in the DNV guidelines (see equations (5) and (6)).
The VSM is appropriate for loose sands at shallow depths,
as observed by Schupp (2009) and Wang et al. (2010).
However, as the density of the soil or the depth of cover
increases, the shear planes appear to deviate from the

Manuscript received 11 May 2012; revised manuscript accepted

22 January 2013. Published online ahead of print 18 April 2013.
Discussion on this paper closes on 1 February 2014, for further details
see p. ii.
 Department of Engineering Science, University of Oxford, UK.

946

PIPE UPLIFT IN SATURATED SAND: RATE AND DENSITY EFFECTS

947

Soil surface

F friction

W weight

F friction

W weight

W weight

(a)

(b)

(c)

Fig. 1. Pipeline geometry and simplified failure mechanisms (after Schupp, 2009; White et al., 2008): (a) vertical slip surface; (b) inclined
slip surface; (c) flow mechanism

vertical. In medium-dense to dense sand conditions they

appear oriented at approximately the angle of dilation from
the vertical (White et al., 2001; Cheuk et al., 2008). Several
researchers have proposed theories that adapt the VSM by
using these inclined shear planes. The formulation by White
et al. (2001), based on the mechanism in Fig. 1(b), appears
to agree well with data in the literature, particularly for
medium to dense soils. This inclined slip surface model
(referred to herein as the ISM) assumes that the normal
stress acting on the shear planes is equal to the in situ (atrest) value, meaning that the stress level above the pipe
increases as the uplift resistance is mobilised. Vertical
equilibrium of the modified sliding block produces nearly
the same equation as in the VSM, but with K tan 9 replaced
by f (White et al., 2001, 2008), and is given by



1 K 0 (1  K 0 )(cos 2)

f tan tan peak  tan
2
2
(2)
where the peak friction and dilation angles are determined
following Bolton (1986).
At greater embedment depths it has been observed that
the sliding-block mechanism does not develop at small
displacements. Tests conducted by Schupp (2009) and Byrne
et al. (2008) showed that localised compression, followed by
a flow-around mechanism, may occur on initial displacement
under these conditions, as shown in Fig. 1(c). As the pipe
moves closer to the surface, the sliding block again develops
in a similar manner as described for shallower embedment
depths. The initiation of this flow-around mechanism has
also been observed in the post-peak response of dense soil
(White et al., 2001), which suggests that the transition
between flow mechanism and sliding-block mechanism may
be related to the state of the soil around the pipe.
Effect of pullout rate
Previous research, as described above, has focused on the
drained behaviour of sand attained through tests in either
dry sand, or saturated sand with an uplift rate sufficiently
slow to allow adequate drainage of the pore water. Planestrain tests reported by Byrne et al. (2008, 2013), Bransby
& Ireland (2009), and Schupp (2009), investigated the
partially drained conditions that develop during faster rate
uplift tests, and the consequent effect on the mobilised soil
resistance. In very loose sand tests, described by Byrne et
al. (2013), as the rate of uplift was increased, the resistance
was found to decrease significantly almost to zero in some

of the faster rate tests. Pore pressure transducers showed

significant positive excess water pressures around the pipe
circumference, which resulted in significantly reduced effective stresses acting on the pipe. In some instances the excess
pore water pressure was found to increase to the value of
the initial vertical effective stress (ue 9H), suggesting that
static liquefaction had occurred. Conversely, in denser sand,
the peak uplift resistance was observed to be significantly
larger than predicted values for drained conditions, and this
was also observed by Bransby & Ireland (2009). This suggests that negative excess pore pressures were developed,
resulting in an increase in the effective stress operating
around the pipe.
Current design guidelines
The Det Norske Veritas recommended practice DNV-RPF110 (DNV, 2007; referred to below as DNV) is commonly
used in the offshore pipeline industry. This recommended
practice defines global buckling as a load response rather
than a failure mode, but one that may lead to an ultimate
failure mode such as local buckling. However, as little
guidance is given on post-buckled pipeline integrity, excess
upward displacement is considered to be an ultimate limit
state (ULS) failure for the purpose of buried pipeline design.
The recommended design procedure is non-linear finiteelement analyses of the buried pipe configuration to determine the operating failure temperature, with incorporated
safety factors. An important modelling input is the uplift
resistance of the soil, which is represented by a lower-bound
characteristic curve. The DNV guidelines propose a tri-linear
forcedisplacement curve for calculating uplift resistance,
and this is shown in Fig. 2. The peak uplift resistance,
denoted as Rmax , is usually determined from an aVSM, with
a recommended uplift resistance factor, f, depending on the
density of the sand. There are three categories specified:
loose, medium/dense and rock. The parameters used to
define this curve for loose and dense sand are provided in
Table 1. Important factors affecting uplift resistance, such as
rate effects and liquefaction susceptibility, are not specifically addressed in these guidelines.
EXPERIMENTAL METHOD
This paper summarises an experimental research programme investigating the plane-strain uplift resistance of
buried pipes in saturated sand. The tests were designed to
determine the influence of backfill relative density and uplift
rate across a range of embedment ratios up to H/D of 4.5.

WILLIAMS, BYRNE AND BLAKEBOROUGH

948
12

Normalised resistance, R/Rmax

Characteristic curve
10
085
08

Design curve

075

1/UR

06
/UR
04
02
0

02

04
06
08
10
Normalised displacement, d/df

12

14

Fig. 2. DNV tri-linear uplift resistance curve (after DNV (2007))

This involved two series of scale model tests exploring

drained and partially drained behaviour. The overall aims of
the investigation were
(a) to improve understanding of the mechanisms involved in
mobilising the soil resistance for buried pipes installed in
loose sand by trenching
(b) to evaluate prediction models presented in the literature
(c) to assess the validity of the recommendations in the
current design guidelines for these conditions.
The experimental rig used for the plane-strain pipe uplift
tests is described in detail by Schupp (2009); the main
features are outlined as follows. The rig is shown in Fig. 3.
It consists of a testing tank measuring 1 m 3 0.3 m 3 1 m,
equipped with a soil fluidisation system by means of an
upward hydraulic gradient. This is achieved by cycling water
upwards through the soil sample using a pump, which draws
from a reservoir mounted to the side of the tank and
distributes water through the base of the tank via a system
of drainage channels. Even pressure distribution and fluidisation of the sample are ensured by the use of a geotextile
filter mounted at the base of the tank.
The tank is fitted with a fully instrumented model pipe,
which can be moved vertically at different rates using a
stepper motor driven actuator. The 100 mm diameter pipe
consists of a 200 mm central active test section, and two
100 mm dummy sections at either end, which sit flush with
the tank sides. The active test section of the pipe is fitted
with an internal, moment-compensated load cell, which
allows measurement of the vertical forces exerted on the
central section of the pipe. Excess pore pressure measure-

Fig. 3. Experimental set-up (from Byrne et al., 2008; courtesy of

OTC)

Table 1. DNV suggested soil parameters for tri-linear uplift curve (DNV, 2007)
Category
Loose sand (3.5 < H/D < 7.5)

Medium/dense (2.0 < H/D < 8.0)

Post-peak resistance at displacement

Parameter

Range

f
df


f
df


fr
f
dfr

0.10.3
0.5%Hc 0.8%Hc
0.750.85
0.2
0.40.6
0.5%Hc 0.8%Hc
0.650.75
0.2
f . f
0.650.75
3 . df

PIPE UPLIFT IN SATURATED SAND: RATE AND DENSITY EFFECTS

ments are taken using five pore pressure transducers evenly
spaced around the half-circumference of the pipe. A longrange displacement transducer is used to track the pipe
displacement through the height of the tank, and a shortrange, high-resolution transducer is used to measure the
initial small displacement response.
The tests described in this paper all follow a procedure
that is predominantly automated, allowing systematic
changes to input variables such as backfill density, embedment ratio and uplift rate. All tests were conducted using
Redhill 110, a poorly graded silica sand, in fully saturated
conditions. The sand was selected for use as it resembles a
typical North Sea backfill sand; relevant soil parameters are
provided in Table 2, and the grading curve is shown in Fig.
4. To simulate jet-trenching installation, the sand is first
fluidised to a target specific gravity and the pipe is lowered
to the required depth. The soil then settles over the pipe,
while a load-control routine (mimicking the self-weight of
the pipe) is implemented. The soil typically settles at its
loosest state (i.e. zero relative density), corresponding to an
effective unit weight of 7.82 kN/m3 : If required, the backfill
is densified through vibration of the tank, after which the
pipe is pulled upwards at a constant rate. This densification
process mimics the possibility that the backfill material can
be densified either by pipe vibrations or by wave and current
effects. The relative density is determined by assuming that
the settlement of the soil occurs uniformly throughout the

tank, which is not unreasonable for the settlement of very

loose sands.
DRAINED RESPONSE
The first test programme investigated the influence of
backfill relative density, at various embedment depths, on
the peak uplift resistance and the forcedisplacement response under drained conditions. The rate of pipe uplift was
set at 0.002 mm/s, a speed judged to be sufficiently slow to
ensure adequate drainage of the soil during the test. This
was confirmed by the negligible response of the pore
pressure transducers on the pipe circumference during uplift.
A summary of the tests conducted is provided in Table 3.
Peak uplift resistance/breakout factors
The results for peak uplift force are presented in Figs 5
and 6 in the form of a soil breakout factor, Nult , which
normalises the force with respect to the effective unit weight
of the soil, the embedment depth and the pipe dimensions.
N ult

F ult
9HDL

N ult,VSM

Value

Source

Maximum void ratio, emax

Minimum void ratio, emin
Specific gravity, Gs
Critical friction angle, crit

1.04
0.64
2.63
368

Schupp (2009)
Schupp (2009)
Schupp (2009)
Villalobos et al. (2005)

(3)

Figure 5 presents the peak breakout factors for backfill at

various initial relative densities across a range of embedment
ratios. The prediction models chosen for comparison are the
VSM and aVSM (using the critical-state friction angle), as
well as the DNV-recommended ranges (Fig. 5(c)). These
models are normalised in a similar manner as follows.

Table 2. Soil properties for Redhill 110

Property

949

1  D=8H K 0 tan 9H=D9HDL

9HDL

D
H
1
K 0 tan 9
8H
D
H
N ult,aVSM 1 K 0 tan 9
D
H
N ult,DNV 1 f
D

(4)

(5)
(6)

100

where f 2 [0.1, 0.3] for loose sand and [0.4, 0.6] for
medium/dense sand.
From Fig. 5(a), it can be seen that the VSM is an
excellent predictor of the uplift resistance for soil at its
loosest state, up to an embedment ratio of approximately 2.
Beyond this depth, the breakout factors remain constant,
indicating a possible transition to a flow-around failure
mechanism. As the relative density increases, the applicability of both the VSMs extends to greater depths, after which
similar plateaus in the normalised resistance are observed
(Fig. 5(b)). The results indicate that, as the relative density
increases, the transition from sliding-block mechanism to
another mechanism (probably the flow-around mechanism)
occurs at greater values of H/D. It would be possible to fit a
mathematical expression empirically to this plateau behaviour, but, given only a small number of data points obtained,
it is not felt appropriate to do so. For practical purposes the

Cumulative percentage passing: %

90
80
70
60
50
40
30
20
10
0

001

01
Grain size: mm

Fig. 4. Grading curve for Redhill 110

Table 3. Summary of drained uplift tests

Set no.
Drained 1
Drained 2
Total drained

Embedment ratio, H/D

Relative density: %

Uplift rate: mm/s

Number of tests

0.54.5
1.54.5

020
2160

0.002
0.002

62
13
75

35

Breakout factor, Nult

Sliding-block mechanism
15
Flow-around mechanism

10
VSM
aVSM
Measured data RD 0%
05

05

10

15
20
25
30
Embedment ratio, H/D
(a)

35

40

Breakout factor, Nult

15

20
15
10
0

RD 20%
RD 15%

RD 10%

RD 0%

15

VSM
aVSM
Measured data RD 0%
Measured data RD 10%
Measured data RD 15%
Measured data RD 20%
Flow mechanism approximation

10

05

10

15
20
25
30
Embedment ratio, H/D
(b)

35

40

45

30
DNV medium/dense
25

Breakout factor, Nult

10

20

30
40
Relative density: %

50

60

Fig. 6. Breakout factors against relative density, compared with

inclined slip surface model by White et al. (2001)

20

20
aVSM
15

DNV loose

10

VSM
Measured data RD 2535%
Measured data RD 4055%
DNV ranges
Extended DNV ranges
Schupp (2009): Dry, RD 30%, D 100 mm
Schupp (2009): Dry, RD 17%, D 220 mm

05

D
H/

25

45

25

05

30

10

20

30
40
50
Embedment ratio, H/D
(c)

60

the difference in stress levels is minimal, and in addition the

potential effects of dilation are minimised at the range of
relative densities targeted in these tests.
Figure 6 compares the peak breakout factors with prediction curves based on the ISM proposed by White et al.
(2001). At an embedment ratio of 1.5 the breakout factors
are very consistent with this model, across the entire range
of densities. However, similar to the VSMs, as the embedment depth increases, the ISM increasingly overpredicts the
peak force response. The results demonstrate that for very
loose soil conditions an increase in embedment depth,
beyond a critical value, does not necessarily result in any
significant improvement in the breakout factor.
Forcedisplacement response
Examples of forcedisplacement curves across the range
of embedment depths tested are shown in Figs 7 and 8,
together with VSM and aVSM curves for comparison. In
Fig. 7, the measured and predicted forces are normalised by
the soil effective unit weight and pipe diameter, as this
preserves the shape of the forcedisplacement curve,
through the uplift test. Once again, for the loose sand at 0%
relative density, it can be seen that the VSM closely matches
the forcedisplacement response of the pipe at low embedment ratios, whereas at greater depths it overpredicts the
response. Tests at an embedment ratio of 4 demonstrate how
an increase in relative density to 20% is required for the
force to reach the VSM curves, while the response at
RD 30% is higher than the predicted values, but appears

70

Fig. 5. Summary of breakout factors compared with vertical slip

models: (a) very loose sand, relative density 0%; (b) loose sand;
relative density 020%; (c) loose to dense sand, relative density
055%

plateau value of Nult could be related to either relative

density or some estimate of the soils friction angle accounting for stress-level effects.
Figure 5(c) also includes dry sand data obtained by
Schupp (2009) using the same rig, which demonstrates that
the above trends can be observed at higher embedment ratios
and higher stress levels (,23), as well as for greaterdiameter pipes. Given that full-scale pipelines are generally
two to three times the diameter of the model pipe in these
tests, scaling effects are felt to be negligible. For example,

1200

VSM
aVSM
Flow-around RD 0%
Measured data

1000

RD 30%

800

Force/D3

Breakout factor, Nult

20

2
5

H/D 15
H/D 25
H/D 35
H/D 40
H/D 45

40

H
/D

25

H
/D

H
/D 45

4
H
0
/D

3
5

WILLIAMS, BYRNE AND BLAKEBOROUGH

950

RD 20%

RD 0%

600
400
RD 20%
200
0
05

RD 0%
10

15

20
25
30
35
Embedment ratio, H/D

40

45

50

Fig. 7. Representative curves of force plotted against embedment

ratio

PIPE UPLIFT IN SATURATED SAND: RATE AND DENSITY EFFECTS

220

Measured data
H/D 35
Increasing relative
density (055%)

160
140
120
100
80

H/D 15
Increasing relative
density (055%)

60
40
20
0

3
4
5
Displacement, H z: mm

Fig. 8. Initial forcedisplacement response at selected embedment ratios

to approach these curves after approximately 100 mm of

displacement.
In Fig. 8, the initial displacement response is shown for
varying initial densities at embedment ratios of 1.5 and 3.5
(displacement is equal to the pipe embedment at any given
time, z, subtracted from the original embedment, H). As the
relative density increases from 0%, an increasingly distinct
peak response occurs, followed by rapid strain-softening.
Estimates of the initial stiffness (taken as a linear leastsquares fit of the curve up to 50% Fult ) indicate that stiffness
increases with relative density, but is virtually unaffected by
embedment ratio (see Fig. 9). Regardless of the initial
relative density and embedment depth, the tests show that,
after the initial displacement, the soil resistance converges to
a consistent load response (matched by the VSM at H/
D , 2). The displacement required for this convergence is
dependent on the embedment depth, where a greater uplift
distance is required for convergence at deeper embedment
depths.
Mobilisation displacement
The measured mobilisation displacements, defined as the
pipe displacement corresponding to peak resistance, are
normalised by both H (Fig. 10(a)) and D (Fig. 10(b)). Both
are presented for comparison with data in the literature;
however, owing to the potential for localised failure mechanisms, it is felt that pipe diameter (a local length scale) is the

1900

H/D 10
H/D 15
H/D 20
H/D 25
H/D 30
H/D 35
H/D 45

1800

Stiffness: N/mm

1700
1600
1500
1400
1300

951

RD 0%
RD 10%
RD 15%
RD 20%
RD 3060%
Thusyanthan et al. (2010)

35
30
25
20
15

DNV 08%Hc
DNV 05%Hc

10
05
0
05

10

15

20
25
30
Embedment ratio, H/D
(a)

35

40

45

Normalised mobilisation distance, df /D: %

180

Normalised mobilisation distance df /H: %

VSM

200

Force: N

40

RD 0%
RD 10%
RD 15%
RD 20%
RD 3060%
Thusyanthan et al. (2010)
DNV 08%Hc
DNV 05%Hc

6
5
4
3
2
1
0
05

10

15

20

25

30

35

40

45

Embedment ratio, H/D

(b)

Fig. 10. Mobilisation displacements for various relative densities,

normalised by: (a) H; (b) D

more appropriate normalisation. The plotted displacements,

df /H or df /D, appear to be dependent on both relative
density and embedment ratio, although the overall variation
is small, with values ranging from 0.1%H to 1.7%H. This is
consistent with values found by several researchers (Matyas
& Davis, 1983; Trautmann et al., 1985; Bransby et al.,
2002; Cheuk et al., 2008), which range from 0.1%H to
1.5%H, but is at the lower bound of values reported by
Thusyanthan et al. (2010) and Wang et al. (2012). The
mobilisation displacements tend to be highest in loose tests
at low embedment ratios; at deeper burial depths relative
density has a lesser influence.
The DNV guidelines suggest mobilisation displacements
for modelling purposes in the range 0.5%Hc to 0.8%Hc ,
where Hc is measured to the pipe crown. This was challenged in a recent study by Thusyanthan et al. (2010), in
which it is suggested that instead this value could be related
to H/D as follows.
 
df
H
(7)
0:02 exp
2D
D

1200
1100
1000

10

20

30
40
Relative density: %

50

60

Fig. 9. Initial soil stiffness at various embedment ratios and

relative densities

Both models are compared with the experimental data in

Fig. 10, and it can be seen that while the DNV limits appear
to match the data at greater embedment ratios, they are not
conservative for loose sand at shallow embedment ratios.
The model proposed by Thusyanthan et al. (2010) predicts
higher mobilisation displacements than observed in these
experiments, particularly at higher values of H/D.

WILLIAMS, BYRNE AND BLAKEBOROUGH

952

PARTIALLY DRAINED RESPONSE

The second series of tests explored the effect of varying
rates of uplift on soil at different relative densities (see Table
4) to determine the effect of soil initial state on pipe
response.
Rate effects
Figure 11 presents the forcedisplacement curves and
corresponding pore pressure response for tests conducted at
an embedment ratio of 3.5, with a soil relative density of
0%. It was observed that faster rates of uplift resulted in
higher excess pore pressures around the pipe. At its loosest
state, the soil is clearly contractive, creating positive excess
pore pressures and a decrease in resistance on shearing. It
can be seen in Fig. 11(b) that the excess pore pressure rises
rapidly on initial displacement of the pipe, followed by
gradual drainage. At faster rates, a larger displacement is
required for the excess pore pressures to dissipate and the
uplift resistance to increase to its drained residual value. This
is consistent with the conclusion by Schupp (2009) that the
time required for dissipation of excess pore pressures in
the loose sand is independent of the rate of uplift. At 5 mm/s
the uplift rate is sufficiently rapid for the initial excess pore
pressure to approach the vertical effective stress of the soil,
120

Density effects
To explore the effect of density on the uplift resistance a
series of tests with an uplift rate of 1 mm/s were carried out.
This was chosen as a representative rate at which a partially
drained response occurred. The forcedisplacement results
are shown for a range of relative densities in Fig. 12. In
these tests, a gradual transition from a contractive to dilative
180

VSM
aVSM
Schupp model
Measured data

100

140

60
40
20

50

100
150
200
250
Embedment depth, z: mm
(a)

80
60
Increasing relative
density (035%)
0

50

100
150
200
250
Embedment depth, z: mm
(a)

300

25
Increasing uplift rate
(00025 mm/s)

20
15
Vertical effective
stress

10
05
0
0

50

100

150
200
250
Embedment depth, z: mm
(b)

300

350

Fig. 11. Rate effects at low relative density: (a) force response;
(b) pore pressure response

25
20
Increasing relative
density (035%)

15
Vertical effective
stress

10
05
0

05

50

100
150
200
250
Embedment depth, z: mm
(b)

300

Embedment ratio, H/D

Relative density: %

Uplift rate: mm/s

Number of tests

3.5
3.5
1.54.5

0
035
055

0.015
0.1
1, 5

8
2
44

PD 1
PD 2
PD 3
Total partially drained

350

Fig. 12. Density effects on fast-rate tests: (a) force response;

(b) pore pressure response

Table 4. Summary of partially drained uplift tests

Set no.

350

30

Average excess PWP, ue: kPa

Average excess PWP, ue: kPa

100

350

30

05

RD 21%

20

300

RD 35%

120

40

Increasing uplift
rate
(00025 mm/s)

VSM
aVSM
Measured data

160

Uplift force: N

80

Uplift force: N

reducing the resistance temporarily to zero. The rate is also

such that the excess pressure does not have sufficient time to
dissipate over the course of the test, and consequently the
drained resistance is never achieved.
Tests with uplift rates greater than 0.1 mm/s were also
compared with a model proposed by Schupp (2009) and
Byrne et al. (2013) to predict the partially drained uplift
response. This model is based on the VSM, but where the
9 term in the uplift equation is replaced by 9  (ue /H).
The excess water pressure is given an initial value equal to
the effective stress, and is assumed to decay based on
Terzaghis one-dimensional consolidation equation. The results of the analysis plotted in Fig. 11 demonstrate that the
model can successfully capture the pore pressure response in
fast-rate tests, when the soil is at its loosest state, and for
embedment depths for which the VSM is appropriate.

54

PIPE UPLIFT IN SATURATED SAND: RATE AND DENSITY EFFECTS

soil response was identified, as evidenced by the excess pore
pressure response. For backfill soil with a relative density
ranging from 0% to 15%, positive excess pore pressures
develop, although the magnitude and effect of these decreased as the density increased. This is also evident in the
recorded force curves. At 21% relative density the pipe
experiences positive excess pore pressures, indicating contraction, but this is immediately followed by negative excess
pore pressures, indicating dilation. In this instance the force
reaches a higher resistance than the drained peak value at
the same density. Finally, at a relative density of 35%, there
is an immediate large negative pore pressure, indicating a
dilating soil response, and a single peak resistance much
larger than the drained value for this density.
COMPARISON BETWEEN DRAINED AND PARTIALLY
DRAINED; TRANSITION DENSITIES
Figure 13 shows a comparison between drained and
partially drained tests at an uplift rate of 1 mm/s for relative
densities of 0% and 30% respectively. In these two figures
the average excess pore pressure is converted to an equivalent force, based on a half-circumference of the pipe. The
sum of the measured soil force on the pipe and the excess

200

pore pressure force for partially drained tests is then compared with the drained forceresponse curves for each
density. At a relative density of 0%, it can be seen that the
difference between the drained and partially drained response is almost exactly equal to the average excess pore
water pressure around the pipe (i.e. the sum curve and the
drained curve are effectively equivalent). This suggests that
contraction of soil occurring predominantly around the pipe
is the main cause of decreasing uplift resistance. In contrast,
at a relative density of 30%, the difference between the
drained response and partially drained response is greater
than the excess pore pressure force. This indicates that, in
addition to dilation around the pipe, dilation elsewhere in
the tank (particularly on the shear planes) is acting to
increase the soil resistance. It is expected that this additional
dilation effect would increase as the relative density and/or
embedment ratio further increased.
Trends for peak uplift resistance due to variations in uplift
rate and relative density for a given embedment ratio
(H/D 3.5) are summarised in Fig. 14. From both plots,
estimates can be made regarding the transition density between net contractive and dilative behaviour, based on
differences observed between the drained and partially
drained responses during uplift. Note that the drained resis-

VSM
Measured force: drained
Measured force: v 1 mm/s

150

Excess pore pressure (force): v 1 mm/s

Force: N

Pressure measured force: v 1 mm/s

Difference between drained and

partially drained resistance

100

50

250

270

290
310
Embedment depth, z: mm
(a)

330

350

200
Difference between
drained and partially
drained peak resistance

Force: N

150

Increase due to dilation

near pipe
Additional increase

100

50

250

270

953

290
310
Embedment depth, z: mm
(b)

330

350

Fig. 13. Comparison of drained and partially drained response for a fast-rate test: (a) relative
density 0%; (b) relative density ,30%

WILLIAMS, BYRNE AND BLAKEBOROUGH

954
40

25
35
20

30

Relative density: %

Transition range

25
Nult 20
15
10

v 0002 mm/s
v 01 mm/s
v 1 mm/s
v 5 mm/s

05
0

10

20
Relative density: %
(a)

30

40

10

0
10

15

20

25
30
35
Embedment ratio, H/D

40

45

50

Fig. 15. Comparison of transition densities

RD 0%
RD 15%
RD 30%

30
25
Nult 20
15
10
05
0

15

40
35

RD to reach aVSM
RD to reach VSM
Transition RD contraction/dilation
Plateau RD drained tests

102

100
Uplift speed: mm/s
(b)

Fig. 14. Summary of breakout factors with respect to (a) relative

density and (b) uplift speed (H/D 3.5)

tance is identified by the curve relating tests at a rate of

0.002 mm/s. At this depth the transition range appears to
occur at a relative density between approximately 17% and
21%, above and below which net dilation and contraction
occur respectively. It should also be noted that in drained
tests at an embedment ratio of 3.5, RD 20% is sufficient
for the data to match the aVSM a similar value to the
transition range identified above.
Figure 15 shows a plot of the transition densities determined for a 100 mm diameter pipe using fast-rate tests for
several embedment ratios (as described above), along with
the required relative densities to reach the VSM and aVSM
in drained tests. It can be seen that, at embedment ratios
greater than 2, the transition density between contraction
and dilation falls within the range of densities required for
the peak resistance to reach the VSM predictions. Furthermore, H/D values at which a plateau in the drained breakout
factors was observed for several relative densities are plotted
in the figure, and appear to be consistent with this trend.
DISCUSSION
Soil behaviour
The test results demonstrate that the volumetric behaviour
of the backfill soil, which is dependent on its initial state,
strongly influences both its drained and its partially drained
uplift resistance. In the drained tests, a plateau in the peak

breakout factors observed in very loose soil indicates a

possible transition from a sliding-block failure mechanism
(which is well captured by the VSM) to a flow-around
failure mechanism, where shear planes do not extend to the
surface. The mobilised mechanism appears to be dependent
on both relative density and embedment ratio for example,
looser backfills experience a transition at shallower embedment ratios which suggests that this behaviour may be
determined by a combination of the soils critical-state
parameters and the burial depth.
Tests conducted at fast rates, inducing partially drained
conditions, allowed for relative quantification of dilation
(and contraction) through measurement of the pore pressure
response around the pipe and its contribution to the overall
uplift force. Soil that contracts during shear sees the
development of positive excess pore pressure and a subsequent reduction in peak force, whereas dilation results in the
opposite effect. The existence of a localised flow-around
mechanism is supported by evidence that, at very low relative densities, contraction appears to occur primarily around
the pipe, whereas at higher densities dilation away from pipe
contributes to the resistance increase. Accordingly, the density (for a given embedment ratio) at which the transition
from net contraction to dilation occurs was determined based
on total resistance, rather than on pore pressure measurements. This relationship between the volumetric transition
density and embedment depth was found to be consistent
with the assumed transition density between failure mechanisms occurring in the drained tests (as shown in Fig. 15).
Evaluation of prediction models
For drained conditions the adapted VSM is an excellent
predictor of the peak resistance for loose sand when the
sliding-block failure mechanism is mobilised. This mechanism appears to occur for all relative densities up to an
embedment ratio of approximately 2, after which an increasing density is required to reach the aVSM for increasing
embedment depths. The prediction model proposed by White
et al. (2001) has applicability for a greater range of densities, as it accounts for dilation along its failure planes.
However, as with the aVSM, this model overpredicts the
resistance at greater embedment ratios once the flow-around
mechanism is mobilised and the failure planes no longer
extend to the surface.
For partially drained conditions, currently the only available prediction model is that proposed by Schupp (2009)

PIPE UPLIFT IN SATURATED SAND: RATE AND DENSITY EFFECTS

and Byrne et al. (2013), which adds a term to the VSM to
account for the pore pressure response. This is applicable
only to soil at its loosest state, probably because in this state
the excess pore pressure development appears to occur
primarily around the pipe. At higher relative densities the
phenomenon becomes more complex, as dilation effects
further away from the pipe contribute to the overall uplift
resistance.

Implications for design against upheaval buckling

The DNV guidelines currently specify two broad ranges
to predict the peak uplift resistance for loose and medium/
dense sand in drained conditions, using empirical frictional
factors (shown in Fig. 5(c)). For the sand tested here, the
upper bound of the loose sand region is equivalent to the
calculated aVSM, and virtually all the data for 020% RD
are contained within this band when it is extended to lower
embedment ratios. The data for relative densities of 25
55% also fit very well within the medium/dense region.
Thus, as a broad estimate, the DNV provides very good
guidance for the range of breakout factors, although within
these bands more accurate predictions may be required. The
data from this study also agree well with the DNV-recommended range for mobilisation distances to be used in
finite-element analyses for upheaval buckling; however, engineering judgement should be used when extrapolating the
displacement values from these model tests to larger-scale
applications.
Generally, current design practice treats the drained case
as conservative, assuming that if fast rates of uplift were to
occur, the soil would be sufficiently dense to dilate and
cause an increase in resistance. The findings of this study,
however, show that in very loose sand fast-rate tests may
cause liquefaction, and drastically reduce the uplift resistance. Given that it is an extremely complex task to quantify
the pore pressure effects across varying densities, and estimate the rate of uplift that might occur during buckling, it
makes little sense to attempt to predict this partially drained
response. Instead, it follows that the design should stipulate
a minimum backfill density for a given embedment ratio,
which will ensure that dilation occurs upon shearing of the
soil. This, in turn, means that the flow-around failure
mechanism at peak resistance will probably be avoided, and
the conservative drained resistance can then be predicted
using the aVSM and the DNV guidelines.

CONCLUSIONS
The results from this scale study of pipe uplift resistance
at various embedment ratios, uplift rates and relative densities illustrate the importance of the backfill density in
controlling the failure mechanism and corresponding peak
resistance. Given these findings, it is practical to ensure that
a minimum relative density of the backfill is reached to
avoid the negative effects of contraction of very loose soil.
This study identifies approximate transition densities for one
soil (and pipe diameter); further work is required to relate
this transition between mechanisms to measurable parameters, and thus extend the framework to a variety of soils.
Additional research is needed to determine field installation
backfill densities for pipes installed by jet-trenching and
ploughing, as well as more details of the time-dependent
densification processes after installation. Finally, it is acknowledged that the application of this work in design
would require consideration of scaling issues, as the experiments were conducted on smaller-diameter pipes than might
typically be used in the field.

955

ACKNOWLEDGEMENTS
The first author gratefully acknowledges the generous support from the Clarendon Fund at Oxford University. Fig. 3 is
reproduced with the permission of the copyright owner, OTC,
and further reproduction is prohibited without permission.

NOTATION
D diameter of pipe
d pipe displacement
df mobilisation displacement
emax maximum void ratio
emin minimum void ratio
Fult peak uplift force
f empirical uplift factor
fr empirical post-peak uplift factor
Gs specific gravity
H initial embedment depth to pipe centre
Hc initial embedment depth to pipe crown
K lateral at-rest earth pressure coefficient
L segmental length of pipe
Nult soil breakout factor (normalised peak uplift force)
R uplift force
Rmax DNV peak uplift force
RD relative density
ue excess pore water pressure
z embedment depth
,  shape parameters for DNV tri-linear uplift curve
f , dfr post-peak shape parameters for DNV uplift curve
9 effective unit weight of soil
UR DNV uplift resistance safety factor
9 effective friction angle
peak peak friction angle
crit critical-state friction angle
dilation angle

REFERENCES
Bolton, M. D. (1986). The strength and dilatancy of sands. Geotechnique 36, No. 1, 6578, http://dx.doi.org/10.1680/geot.
1986.36.1.65.
Bransby, M. F. & Ireland, J. (2009). Rate effects during pipeline
upheaval buckling in sand. Proc. Instn Civ. Engrs Geotech.
Engng 162, No. 5, 247256.
Bransby, M. F., Newson, T. A., Davies, M. C. R. & Brunning, P.
(2002). Physical modelling of the upheaval resistance of buried
offshore pipelines. Proc. 1st Int. Comf. on Physical Modelling in
Geotechnics, St Johns, 899904.
Byrne, B. W., Schupp, J., Martin, C. M., Oliphant, J., Maconochie,
A. & Cathie, D. (2008). Experimental modeling of the unburial
behaviour of pipelines. Proc. Offshore Technol. Conf., Houston,
TX, paper OTC-2008-19573.
Byrne, B. W, Schupp, J., Martin, C. M., Oliphant, J., Maconochie,
A. & Cathie, D. (2013). Uplift behaviour of shallowly buried
pipe sections in saturated very loose sand. Geotechnique 63, No.
5, 382390, http://dx.doi.org/10.1680/geot.11.P.016A.
Cheuk, C. Y., White, D. J. & Bolton, M. D. (2008). Uplift
mechanisms of pipes buried in sand. J. Geotech. Geoenviron.
Engng 134, No. 2, 154163.
DNV (2007). DNV-RP-F110, Global buckling of submarine pipelines: Structural design due to high temperature/high pressure.
Baerum, Norway: Det Norske Veritas.
Matyas, E. L. & Davis, J. B. (1983). Experimental study of earth
loads on rigid pipes. J. Geotech. Engng 109, No. 2, 202209.
Schaminee, P. E. L., Zorn, N. F. & Schotman, G. J. M. (1990). Soil
response for pipeline upheaval buckling analyses: full-scale
laboratory tests and modelling. Proc. Offshore Technol. Conf.,
Houston, TX, paper OTC-1990-6486.
Schupp, J. (2009). Upheaval buckling and flotation of buried offshore pipelines. DPhil thesis, Univeristy of Oxford, UK.
Thusyanthan, N. I., Mesmar, S., Wang, J. & Haigh, S. K. (2010).
Uplift resistance of buried pipelines and DNV-RP-F110 guidelines. Proc. 33rd Ann. Offshore Pipeline Technol. Conf., Amsterdam, 120.

956

WILLIAMS, BYRNE AND BLAKEBOROUGH

Trautmann, A. M., ORourke, T. D. & Kulhawy, F. H. (1985). Uplift

forcedisplacement response of buried pipe. J. Geotech. Engng
111, No. 9, 10611076.
Villalobos, F. A., Byrne, B. W. & Houlsby, G. T. (2005). Moment
loading of caissons installed in saturated sand. Proc. Int. Symp.
on Frontiers in Offshore Geotechnics (ISFOG 2005), Perth,
411416.
Wang, J., Ahmed, R., Haigh, S. K., Thusyanthan, N. I. & Mesmar,
S. (2010). Uplift resistance of buried pipelines at low coverdiameter ratios. Proc. Offshore Technol. Conf., Houston, TX,
paper OTC-2010-20912.

Wang, J., Haigh, S. K., Forrest, G. & Thusyanthan, N. I. (2012).

Mobilization distance for upheaval buckling of shallowly buried pipelines. J. Pipeline Syst. Engng Practice 3, No. 4, 106
114.
White, D. J., Barefoot, A. J. & Bolton, M. D. (2001). Centrifuge
modelling of upheaval buckling in sand. Int. J. Phys. Modelling
Geotech. 1, No. 2, 1928.
White, D. J., Cheuk, C. Y. & Bolton, M. D. (2008). The uplift
resistance of pipes and plate anchors buried in sand. Geotechnique 58, No. 10, 771779, http://dx.doi.org/10.1680/geot.2008.
3692.