Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Modeling of Shallowly Embedded Offshore Pipelines

in Calcareous Sand
Jianguo Zhang1; Douglas P. Stewart2; and Mark F. Randolph3

Abstract: The results of centrifuge modeling of pipe–soil interaction for shallowly embedded offshore pipelines are presented. A
non-associated bounding surface model is constructed in vertical–horizontal (V – H) load space on the basis of test data and the theory of
Downloaded from ascelibrary.org by Cambridge University on 09/21/22. Copyright ASCE. For personal use only; all rights reserved.

plasticity to simulate the response of a pipeline embedded in sandy soil under combined 共vertical and horizontal兲 monotonic loading,
taking into account possible pre-loading effects. The model needs nine parameters that can be back calculated from model or field tests
and in some cases estimated theoretically. It provides a suitable basis for modeling the load-displacement response of shallowly embedded
offshore pipelines. The model reproduces the key features of the load-displacement response of pipelines observed in centrifuge model
tests. In particular, the adoption of bounding surface plasticity allows a gradual transition from elastic to plastic response to be simulated
and the introduction of a non-associated flow rule allows the model to predict the strain-softening behavior of pipes under horizontal
loading. The lateral breakout resistance predicted by the model agrees very well with experimental data.
DOI: 10.1061/共ASCE兲1090-0241共2002兲128:5共363兲
CE Database keywords: Offshore pipelines; Calcareous sand; Bounding surface; Models; Plasticity.

Introduction In an attempt to achieve good predictions of pipe–soil inter-


action for unburied or shallowly embedded 共under self-weight or
Increasing offshore oil and gas exploration and extraction activi- hydrodynamic loads兲 pipelines, a variety of physical model tests
ties have led to the construction of increasing lengths of offshore have been performed within three major projects: the PIPESTAB
pipelines. The cost of a pipeline represents a significant portion of 共Wolfram et al. 1987兲, the AGA 共Allen et al. 1989兲 and the DHI
the development budget of a new hydrocarbon field, and the en- 共Palmer et al. 1988兲 projects, and different predictive methods
vironmental consequences of pipeline failure are enormous. developed 共for example, see the work of Brennodden et al. 1989兲.
In the early days of offshore developments, trenching of pipe- However, the physics of the pipe–soil interaction process of
untrenched offshore pipelines are not well understood, and most
lines below the seabed level was invariably demanded. Trenching
pipe–soil interaction models have been constructed on an entirely
and the expected natural backfill by seabed sediment were meant
empirical basis, as noted by Hale et al. 共1991兲, by Verley and
to be a protection against trawling gear or ships’ anchors and to
Sotberg 共1994兲, and by Zhang et al. 共1999兲. No explanation has
aid in stability. However, several extensive investigations indi-
been given as to why a pipeline moves vertically downward in
cated that a properly designed concrete cover was strong enough some cases and upwards in others under the action of a horizontal
to protect steel pipes against such loads 共Moshagen and Kjeldsen load when it bears a constant vertical load. It is also not clear as
1980, Bergan and Mollestad 1981兲. As a result, the certifying to what will happen to the pipeline when cyclic loading induces
authorities approved deployment of pipelines directly on the sea- excess pore pressure in the soil. These models do not simulate the
bed without subsequent trenching or other means of protection whole process of pipeline movement, they only provide the final
共Karal 1987兲. Untrenched pipelines are exposed directly to the settlement and the break-out load. Thus it is impossible to fully
hydrodynamic environment, and it is therefore imperative to ac- simulate the pipeline response under realistic loading conditions
curately evaluate the geotechnical response of the pipeline on the with these kinds of models.
seabed in order to assess its stability during extreme loading from Shallowly embedded pipelines may be viewed as strip footings
currents. with a circular cross section, as proposed in the American Gas
Association 共AGA兲 pipeline design guidelines. While the vertical
1
Engineer, Advanced Geomechanics, 4 Leura St., Nedlands, bearing capacity of a pipeline may be evaluated using a conven-
WA 6009, Australia. E-mail: zhang@civil.uwa.edu.au tional bearing capacity factor approach, recent investigations of
2
Principal Engineer, Golder Associates, 182 Lord St., Perth, WA 6000, the response of shallow foundations to combined vertical and
Australia. horizontal loading have focused on the construction of a load
3
Professor, Director of the Special Centre for Offshore Foundation interaction diagram 共see, for example, the work of Butterfield and
Systems, Univ. of Western Australia, Nedlands, WA 6907, Australia. Gottardi 1994兲. Under inclined loading, this interaction diagram,
Note. Discussion open until October 1, 2002. Separate discussions or yield envelope, can be described by a function F 关 F
must be submitted for individual papers. To extend the closing date by
⫽ f (V,H,V max)⫽0兴 in vertical and horizontal load space, with its
one month, a written request must be filed with the ASCE Managing
Editor. The manuscript for this paper was submitted for review and pos- size defined by the maximum vertical load, V max the foundation
sible publication on March 3, 2000; approved on August 8, 2001. This has experienced 共Gottardi and Butterfield 1995, Gottardi et al.
paper is part of the Journal of Geotechnical and Geoenvironmental 1999兲, as shown in Fig. 1. This function provides memories of the
Engineering, Vol. 128, No. 5, May 1, 2002. ©ASCE, ISSN 1090-0241/ loading history of a footing and may be viewed as a bounding
2002/5-363–371/$8.00⫹$.50 per page. surface. For a conventional plasticity model, an elastic response is

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / MAY 2002 / 363

J. Geotech. Geoenviron. Eng., 2002, 128(5): 363-371


Fig. 3. Particle size distribution
Fig. 1. Yield envelope and plastic potential surface in V – H space
Downloaded from ascelibrary.org by Cambridge University on 09/21/22. Copyright ASCE. For personal use only; all rights reserved.

assumed within the yield surface. However, in a bounding surface were conducted at 50 times the Earth’s gravity using a 20 mm
model, the elastic range shrinks to a point and there is a gradual diam smooth model pipe, thus representing a 1 m diam prototype
transition from elastic to plastic states within this envelope. The pipe. A detailed description of the test apparatus and procedures
proportion of plastic strain or displacement is controlled by the was given by Zhang et al. 共1999兲, and the centrifuge facility was
relative distance between the current load point and a reference described by Randolph et al. 共1991兲. Experimental results are pre-
point, termed the ‘‘image point’’ 共Dafalias and Herrmann 1982兲, sented throughout this paper in terms of equivalent prototype
on the bounding surface. As such, the bounding surface model units or in non-dimensional form.
avoids an abrupt jump from elastic to plastic behavior. It allows Tests were conducted using seabed calcareous sand from the
plastic strains to occur within the bounding surface and makes it North West Shelf of Australia. The particle size distribution of the
possible to simulate the accumulation of irreversible displace- material, after removing the small quantity of coarse particles
ments under cyclic loading when the loads do not reach the 共nominally larger than 1 mm in size兲, is shown in Fig. 3. Several
bounding surface. types of tests were conducted in order to clarify the fundamental
A bounding surface model for pipelines that incorporates the features of pipe–soil interaction. These tests included vertical
above features is described in this paper. The model is developed loading tests, sideswipe tests 共where the pipe is penetrated to a
on the basis of experimental data that have mainly been obtained certain depth and then the vertical position of the pipe is held
from pipe–soil interaction tests using a pipe segment shallowly constant while it is displaced laterally兲, and probe tests 共where the
embedded in calcareous sand. Additional published experimental vertical load is kept constant as the pipe is pushed laterally兲.
data 共Gazetas et al. 1985; Gazetas and Tassoulas 1987; Verley and In order to describe the problem conveniently, the terms ‘‘nor-
Sotberg 1994; Wallace 1995; Montrasio and Nova 1997; Gadre mally loaded’’ and ‘‘overloaded’’ are introduced here to describe
and Dobry et al. 1998; Byrne and Houlsby 1999; Cassidy and the loading history of a pipeline. Normally loaded means the cur-
Houlsby 1999兲 are also used. The setup of the model pipeline rent vertical load is the highest vertical load the foundation has so
experiments and the formulation and performance of the theoret- far experienced, while overloaded means that the initial vertical
ical model are described here. A major advantage of the model is load (V 0 ) of the current loading event is less than the highest
that it has a strong theoretical basis that links displacements to the vertical load (V max) the foundation has previously experienced.
loads applied through plasticity theory. The ratio of V max /V0 is called the overloading ratio and is ex-
pressed by R.
A typical result from a monotonic penetration test is shown in
Fig. 4. The data indicate a linear relationship between vertical
Experimental Observations on Model Pipelines
displacement and vertical load. Initially this may seem surprising,
A series of pipe tests have been conducted on the beam centrifuge since the area of soil contact would be expected to change as the
at The University of Western Australia. The general arrangement pipe penetrates the soil. However, the contact area increases very
of the test apparatus is shown schematically in Fig. 2. The tests

Fig. 2. Arrangement of pipe tests at the prototype scale Fig. 4. Vertical response of the model pipeline

364 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / MAY 2002

J. Geotech. Geoenviron. Eng., 2002, 128(5): 363-371


Downloaded from ascelibrary.org by Cambridge University on 09/21/22. Copyright ASCE. For personal use only; all rights reserved.

Fig. 5. Normalized loading paths obtained from sideswipe tests Fig. 7. Initial gradient of pipeline movement in probe tests

rapidly with penetration, due to heaving of the soil. The data also probe tests conducted on pipelines with different overloading ra-
show very little hysteresis and curvature in an unload–reload tios is plotted in Fig. 7. The data indicate a downward movement
cycle. Results from a group of loading–unloading tests give gra- under lateral loading when R is less than 10 and an upward move-
dients of about 350 and 7000 kPa for load and unload lines, ment when R is greater than 10. This suggests that the direction of
respectively. For the calcareous soil samples used in these tests, plastic displacement vectors is not normal to the bounding surface
the gradient of the loading line was found to be approximately determined from sideswipe tests, implying that a non-associated
equal to the gradient of the cone penetration test 共CPT兲 tip resis- flow rule is necessary in modeling the load-displacement re-
tance with depth. sponse.
Typical sideswipe test results are shown in Fig. 5 for different The response of unburied pipelines revealed from these tests,
embedment depths. The data are plotted in normalized form, with and also from previous full scale model tests 共Brenodden et al.
the vertical and horizontal loads divided by the maximum vertical 1989兲, is quite different from that reported for buried pipelines
load prior to lateral movement. The data suggest that the bound- 共Audibert and Nyman 1977, Trautmann and O’Rourke 1985兲.
ing surface of this shallowly buried pipe is approximately para- While post-peak strain-softening behavior was restricted to very
bolic in shape. The vertical load corresponding to the peak hori- dense sand for buried pipelines, for shallowly buried pipelines it
zontal load is about 0.47V max , and the peak H/V max ratio is related more to the loading history of the pipe–soil system and
increases with an increase in embedment. It appears from Fig. 5 can occur in relatively loose compressible soil. The response ob-
that the loading paths have a positive intersect with the horizontal served can be modeled only through a non-associated work hard-
load axis, indicating some passive resistance of a partially embed- ening model.
ded pipeline. On the basis of the test data illustrated above, a plasticity
Fig. 6 shows typical probe test results. Large displacements model for pipe–soil interaction was formulated, and is described
were used here in order to evaluate the ultimate capacity of the in detail in the following.
laterally loaded pipelines. For a normally loaded pipeline in this
soil, the peak horizontal load is about 0.8V 0 , which is reached
after lateral displacement of about one diameter and additional Formulation of Plasticity Model
penetration of about one diameter. The average value of the initial
displacement gradient (dz/dx) obtained from probe tests con- It is assumed that the pipe is rigid and is placed on a flat surface
ducted on normally loaded pipelines is about 0.5. When the pipe of homogeneous isotropic soil. The pipe is initially embedded
is overloaded, the lateral stiffness increases dramatically and the into the soil under a certain vertical load. The resultant load is
peak horizontal load is also larger due to the increase in embed- inclined toward the angle of inclination ranging from 0° 共purely
ment. The initial displacement gradient (dz/dx) obtained from vertical loading兲 to 90° 共purely horizontal loading兲. As such, the
interaction considered here is defined purely in vertical and hori-
zontal load space 共moment loading is not considered兲.

Hardening Law
The vertical load-displacement response shown in Fig. 4 can be
modeled by means of the following simple relationship:
V max⫽ 共 k v p z⫺Vk v p /k v e 兲 / 共 1⫺k v p /k v e 兲 (1)
where V max⫽maximum vertical load 共for zero horizontal load兲
corresponding to current displacement; k v e and k v p ⫽elastic and
plastic stiffness, respectively, for vertical loading; and V and z⫽
current vertical load and penetration depth, respectively. Taking
␦z p , the plastic vertical displacement increment, as the hardening
parameter, this expression can be used to define the hardening
Fig. 6. Lateral response from probe tests, V 0 ⫽4 kN/m and R
behavior of the pipe response, which in the context of this model
⫽V max /V0
is the evolution of the bounding surface:

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / MAY 2002 / 365

J. Geotech. Geoenviron. Eng., 2002, 128(5): 363-371


k vek vp
␦V max⫽ ␦z p . (2)
k v e ⫺k v p

Bounding Surface Derived from Pipe Tests


The sideswipe test data shown in Fig. 5 suggest a parabolic
bounding surface, which can be described as follows:
F⫽H⫺␮ 共 V⫺V min兲共 1⫺V/V max兲 ⫽0, (3)
where ␮ is the gradient of H/V at small loads 共like a friction
coefficient of the footing–soil interface兲, 共V, H兲 the image point
on the bounding surface, V max the highest vertical load the pipe
has experienced, V min the negative intersect at H⫽0, and V min⫽
Downloaded from ascelibrary.org by Cambridge University on 09/21/22. Copyright ASCE. For personal use only; all rights reserved.

⫺␤Vmax , where ␤ is a constant.


The parabolic shaped yield or bounding surface has been pro-
posed by a number of researchers 共e.g., by Nova and Montrasio
1991兲. However, it should be noted that, for an embedded pipe,
Fig. 8. Proposed bounding surfaces
Eq. 共3兲 ensures a small horizontal load capacity when the vertical
load drops to zero. This would be expected due to the passive
resistance of soil adjacent to the pipe.
A yield surface equation was proposed for embedded footings
by Montrasio and Nova 共1997兲, with the parameter ␮ increasing tests with V 0 ⫽V max 共Fig. 5兲, since the load path indicates non-
with footing embedment, alignment of the incremental displacement vector with the normal
to the yield surface 共Zhang et al. 1999兲. The plastic potential
H⫺␮V 1⫺ 冉 V
V max 冊
⫽0, (4) function takes a similar form to the bounding surface function,
and is given below.

␮⫽␮ 0 ⫹␬
z0
B
, (5) G⫽H⫺␮ t 冉 V

V min
V max V max 冊 m
共 V max⫺V兲 ⫺C⫽0, (6)

where z 0 ⫽ footing embedment. This expression has the effect of where ␮ t shape parameter and C ⫽ constant that adjusts the po-
increasing the lateral resistance with an increase in embedment, sition of the surface to allow it to pass through the loading point.
which models the contribution of passive soil resistance. How- The exponent m has the effect of distorting the parabolic shape
ever, the equations give a zero horizontal capacity for V⫽0 even such that the value of vertical load where the displacement vector
if the embedment is not zero. For very shallow embedment, the becomes perpendicular to the vertical load axis can be adjusted.
error is small and can be neglected. For larger embedment, the This enables the displacement behavior observed that is summa-
passive soil pressure on the side of the foundation can certainly rized in Fig. 7 to be modeled.
provide significant horizontal resistance, as reported by Gadre and The normalized bounding surface given by Eq. 共3兲 and the
Dobry 共1998兲. This shortcoming can be overcome by introducing normalized plastic potential surface given by Eq. 共6兲 with C⫽0
a negative intersect on the vertical load axis, as defined by Eq. are shown in Figs. 8 and 9, respectively, for ␮⫽0.4– 0.6, ␤
共3兲. ⫽0.05– 0.1, and ␮ t ⫽0.5, ␤⫽0.05– 0.1, m⫽0.1– 0.2.

Non-Associated Flow Rule Elastic Deformation


It is commonly accepted that a non-associated flow rule is essen- For the plane strain problem described here, the elastic deforma-
tial in modeling the behavior of shallow foundations on dense tion may be expressed as
sand. However, less progress has been made on establishing the
form of the plastic potential function. Nova and Montrasio 共1991兲
proposed a plastic potential equation of similar shape to the
bounding surface, but using ␮ t ⫽2 – 4⫻␮. The model predictions
with the flow rule defined by this plastic potential function were
reasonably good. However, only normally loaded footings were
investigated in the tests performed by Nova and Montrasio 共1991兲
and by Montrasio and Nova 共1997兲. Since the overloaded condi-
tion is commonly encountered in offshore engineering through
preloading, further investigation is required. It has been shown
earlier that, for pipelines with a high overloading ratio, there was
an upward movement that led to strain-softening behavior. The
transition from downward to upward movement occurred at about
R⫽10. A similar response was reported for spudcan footings on
dense sand by Allersma et al. 共1997兲, although the transition was
at about R⫽50.
It was found necessary to adopt a non-associated flow rule in
Fig. 9. Proposed plastic potential surfaces
order to model the post-peak strain softening during sideswipe

366 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / MAY 2002

J. Geotech. Geoenviron. Eng., 2002, 128(5): 363-371


Table 1. Model Parameters Table 2. Key Features of Pipe–Soil Interaction under the Normally
Parameter Description Loaded Condition
Sideswipe event Probe event
kvp Plastic stiffness of vertical loading
k ve Elastic stiffness of vertical loading Ratio Expression Ratio Expression
k he Elastic stiffness of horizontal loading ␮ 1 m␮共1⫹␤兲2
Hmax Hmax
␭ ratio of k he /k v e 共1⫹␤兲2
Vmax 4 1⫺0.5共1⫹␤兲kvp /kve Vmax 共1⫹m兲共m⫺␤兲
␤ Intersect of the normalized bounding surface on the


vertical load axis
Shape parameter for the bounding surface
V兩H⫽Hmax
Vmax
1⫺␤
2⫺共1⫹␤兲kvp /kve 冉 冊
dH
dx initial
␹k v p /k v e
k
1⫹␹k v p /k v e he
a

冉冊
␮t Shape parameter in the plastic potential equation
Hres 4m dz ␮ t 共 1⫹␤ 兲 m
m Exponent in the plastic potential equation
H max 共1⫹m兲2 dx 1⫹␹k v p /k v e
␬ Gradient of the parameter ␮ increase with embedment initial


冉 冊 ␹
Downloaded from ascelibrary.org by Cambridge University on 09/21/22. Copyright ASCE. For personal use only; all rights reserved.

Exponent in interpolation of the plastic modulus to dH


k a
control the decay of stiffness dx initial
1⫹␹ he
a
␹⫽␮␮ t (1⫹␤) 1⫹m k v e /k he (k v e /k v e ⫺k v p ).

再 冎
äV e
再 冎冋
äz k ve
äH ⫽ 关 D 兴 äx ⫽ 0 k hc
0
册再 冎
äz
äx (7) 再 冎 再冉
äV
äH ⫽
k ve
0
0
k he

冉 冊冎再 冎
where ␦V and ␦H and ␦z and ␦x⫽increments of the vertical and
1 abk 2v e bk v e k he äz
horizontal load and increments of vertical and horizontal dis- ⫺
placements, respectively; 关 D e 兴 ⫽elastic stiffness matrix; and k v e K f g⫹abk v e ⫹k he ak v e k he k 2he äx
and k he ⫽elastic stiffness coefficients for vertical and horizontal
directions, respectively 共with no cross coupling between the two兲. (10)
In this model, the horizontal stiffness k he is related to k v e by Eq. 共10兲 is solved using an explicit method with very small
k he ⫽␭k v e , where ␭ is a constant. displacement increments, typically 0.1% of the pipe diameter, in
order to avoid any influence of the size of the loading step. All
parameters for the model are described in Table 1. Equations for
Constitutive Relationship
key performance ratios derived from the plasticity model are
Using a radial mapping rule 共Dafalias and Herrmann 1982兲, the listed in Table 2.
consistency condition on the bounding surface (␦F⫽0), the
hardening equation, non-associated flow rule, and the plastic
modulus K b at the image point can be obtained. Evaluation of the Model Parameters
k v e k v p bc The model described above contains nine independent param-
K b⫽ , (8)
k v e ⫺k v p f g eters, as shown in Table 1. These parameters can be evaluated
where f ⫽ 冑1⫹a 2 and g⫽ 冑1⫹b 2 since technically ⳵F/⳵H from back analysis of a series of loading tests at model or full
⫽⳵G/⳵H⫽1, and a⫽⳵F/⳵V, b⫽⳵G/⳵V, and c⫽⳵F/⳵V max . scale. The background as to how this may be done and how these
The plastic modulus, K, at the load point 共V, H兲 is expressed as parameters can be estimated in the absence of model or field test
follows to ensure dimensional consistency and to guarantee that K data are given below.
has an infinite value at the beginning of reloading when ␩ is zero
共and V m ⫽V 0 兲. This simulates the elastic response initially and the Vertical and Horizontal Stiffness k vp , k ve , and k he
plastic response when K⫽K b at V m ⫽V max .

冉 冏 冏冊 冉 冊
The vertical stiffness parameters may be obtained directly by

k vek vp M V max⫺V m means of a vertical loading–unloading test on a pipeline as de-
K⫽K b ⫹ 1⫹ , (9)
共 k v e ⫺k v p 兲 f g ␩ V max⫺V 0 scribed in the ‘‘Constitutive Relationship’’ section, where k v p and
k v e equal gradients of the loading and unloading lines. As an
where V⫽intersect of the loading surface on the V axis;
alternative, k v p and k v e could be calculated from theory using an
V 0 ⫽initial vertical load of the current loading event;
estimate of the ultimate bearing capacity of a footing. Since k v p
(V max⫺Vm)⫽current reference distance between the loading sur-
Ⰶk v e , Eq. 共1兲 can be approximated by V max⬇kvpz. For a footing
face and the bounding surface; (V max⫺V0)⫽maximum reference
on calcareous soil, the ultimate bearing capacity is V max /B
distance between the initial loading point and the bounding sur-
⫽␥⬘Nqz 共Finnie 1993兲, where N q is a conventional bearing capac-
face; ␳⫽plastic stiffness interpolation factor; ␩⫽ratio of H/V;
ity factor. Hence k v p is equivalent to ␥ ⬘ N q . Estimated k v p
M⫽ratio of H/V at the point where horizontal load peaks; and
M ⫽0.5 ␮ 关 (1⫹␤) 2 /(1⫺␤) 兴 . The exponent ␳ is a parameter that
adjusts the rate of decay of stiffness with displacement and will
generally lie in the range 3–5. Table 3. Estimated k v p for Footings on Calcareous Sand
The incremental form of the load-displacement relationship
␾ ⬘ 共°兲 32 34 36 38 40 42 44
can be derived by adopting the same framework as that in a
␥ ⬘ 共kN/m3兲 7.0 7.5 7.5 8.0 8.0 8.5 8.5
conventional model 共Zhang et al. 1999兲, noting the difference in
k v p 共kPa兲 130 220 280 390 510 730 980
plastic modulus:

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / MAY 2002 / 367

J. Geotech. Geoenviron. Eng., 2002, 128(5): 363-371


Table 4. Reported Ratios of Elastic to Plastic Stiffness
Reference Footing type Soil type k v e /k v p
Byrne and Houlsby 共1999兲 Flat, circular Dense 5
quartz sand
Gottardi et al. 共1999兲 Flat, circular Dense 10
quartz sand
Zhang et al. 共1999兲 Pipeline Calcareous 20
sand

(⫽␥⬘Nq) values for footings on calcareous sand are listed in


Downloaded from ascelibrary.org by Cambridge University on 09/21/22. Copyright ASCE. For personal use only; all rights reserved.

Table 3. Some reported ratios of k v e to k v p are listed in Table 4. Fig. 10. Variation of ␮ with embedment
The horizontal elastic stiffness, k he , could be obtained directly
from a horizontal loading test on a pipeline or footing. For the
pipeline–soil system tested here k he ⫽k v e was adopted, whereas
Cassidy and Houlsby 共1999兲 adopted k he ⫽0.87k v e for modeling
foundations for jack-up units on quartz sand. Gazetas and his The variation of ␮ with embedment can be examined using
co-workers 共1985, 1987兲 investigated both vertical and horizontal conventional earth pressure theory. Adding friction on the base of
stiffness of arbitrarily shaped embedded footings in an elastic the pipe to the net soil resistance on the side of the pipe,
half-space, and proposed simple expressions for the elastic stiff-
H⫽␮ 0 V⫹ 21 ␥ ⬘ z 20 共 K p ⫺K a 兲 (12)
ness of surface footings. The ratios of k he to k v e calculated from
these expressions are listed in Table 5 for various values of the where K p and K a ⫽passive and active earth pressure coefficients,
Poisson ratio. These ratios compare well with those reported respectively. From Eq. 共3兲, H⬇␮␤k v p z 0 for V⫽0, which may be
above. combined with Eq. 共12兲 to give
␥ ⬘ 共 K p ⫺K a 兲
Parameters ␤ and ␮ ␬⫽ (13)
2␤k v p
Parameters ␤ and ␮ define the shape of the bounding surface, and For the pipeline experiments described here, ␥ ⬘ ⫽8 kN/m3 , ␤
may be derived directly through sideswipe tests on pipelines or ⫽0.06, and k v p ⫽350 kPa. Using the Rankine earth pressure co-
footings described earlier. efficient for simplicity 共the frictionless vertical wall兲, ␬⫽0.75 is
From the gradient ⳵V/⳵H 兩 V/V max⫽0.47⫽0, a value for ␤ of 0.06 derived for ␾ ⬘ ⫽38° 共typical of dense sand兲. This compares very
is achieved. Test data also suggested that the ratio of peak hori- well with the value of 0.72 suggested by Montrasio and Nova
zontal load to maximum vertical load (H max /Vmax) increased with 共1997兲 for flat footings, and also with the value of 0.65 given in
pipe embedment, shown in Fig. 5. By substituting these ratios Eq. 共11兲 for pipelines taking into consideration the effect of the
together with ␤⫽0.06 into the bounding surface equation, values curved side wall in reducing passive soil resistance.
of ␮ corresponding to different pipe embedment are obtained. The
parameter ␮ may be related to pipe embedment by
Parameters m, ␮ t , and ␳
z0
␮⫽0.4⫹0.65 (11) The plastic potential shape parameters m and ␮ t can be derived
B
from back analysis of pipeline or footing tests, although there is
to fit test data, as shown in Fig. 10. In Eq. 共10兲, z 0 is less than 0.5 no means by which to estimate them on the basis of simple con-
B, indicating that it is applicable for shallow embedments. ventional soil mechanics theory. As described earlier, probe test
For comparison, individual values of ␮ 0 between 0.44 and indicated that the gradient of the plastic potential function
0.52 have been reported by Butterfield and Gottardi 共1994兲, by changes sign when the overloading ratio R was about 10. This
Byrne and Houlsby 共1999兲, and by Gottardi et al. 共1999兲 from property may be expressed mathematically as

冋 册
sideswipe tests on rough footings on the surface of dense quartz
sand. Eq. 共5兲 has been proposed by Montrasio and Nova 共1997兲 to ⳵V
⫽0. (14)
define the increase of ␮ with embedment. They used ␮ 0 ⫽0.35 ⳵H V/V max⫽0.1
and 0.48 for smooth and rough surface footings, respectively,
with ␬⫽0.72. This gives m⫽0.18 for ␤⫽0.06. For the normally loaded pipe
probe tests, an equation for the initial gradient of pipeline move-
ments is listed in Table 2; this expression gives ␮ t ⬇0.68 for ␮
⫽0.5, ␤⫽0.06, (dz/dx) initial⫽0.5, m⫽0.18, k v e ⫽20k v p , and
k he ⫽k v e .
Table 5. Ratios of k he /k v e 共Gazetas et al. 1985; Gazetas and Tassou- The plastic stiffness interpolation parameter ␳ changes the rate
las 1987兲 of decay of stiffness with a change in load. Appropriate values for
Footing type Square Circle Strip this parameter must be determined by suitable curve fitting of the
sideswipe or probe test results under overloaded conditions. It is
v ⫽0.3 0.82 0.83 1.13
found through curve fitting that a value of 5 is appropriate for
v ⫽0.4 0.74 0.75 1.03
simulating the response of the pipeline–soil system in this study.
v ⫽0.5 0.66 0.67 0.91
Lower values of ␳ lead to a softer response in sideswipe tests.

368 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / MAY 2002

J. Geotech. Geoenviron. Eng., 2002, 128(5): 363-371


Downloaded from ascelibrary.org by Cambridge University on 09/21/22. Copyright ASCE. For personal use only; all rights reserved.

Fig. 11. Comparison of model prediction and test results

Model Performance Good agreement is achieved for the horizontal load-displacement


response and the values of peak soil resistance. The computed
displacement gradients are very similar, although the predicted
Ultimate Lateral Resistance
initial penetration is somewhat different from that measured. The
From the plasticity model it can be shown that, at a given depth, model predicts upwards pipe movement and strain softening for
the residual lateral resistance ratio H res /V max in a sideswipe test is overloaded pipes when the overloading ratio is larger than about
equal to the peak resistance ratio H max /Vmax in a probe test. The 10. The transition from elastic to plastic response was simulated
normalized value of this resistance is given by successfully.
H
␥ ⬘z 2⫽
共 1⫹m 兲 2 冉 冊
m␮ 共 1⫹␤ 兲 2 k v p
␥ ⬘z
(15) Comparison with the Conventional Model
where H⫽H res for sideswipe tests and H⫽H max for probe tests. A conventional plasticity model had been proposed by the authors
This expression is shown for comparison with the experimental 共Zhang et al. 1999兲, in which fully elastic response is assumed
data in Fig. 11 using m⫽0.18, ␮⫽0.4⫹0.65z 0 /B, ␤⫽0.06, k v p
⫽350 kPa, and ␥ ⬘ ⫽8 kN/m3 . The expression provides an excel-
lent fit to test data from both current and previous centrifuge
modeling conducted in calcareous sand 共Wallace 1995兲, and also
from full-scale model tests performed in silica sand 共Verley and
Sotberg 1994兲. While Wallace’s data were from sideswipe tests,
Verley and Sotberg’s data were from simple break-out tests where
the pipe was embedded in sand under self-weight followed by
lateral loading to its ultimate state.

Performance of Modeling Load-Displacement


Response
Comparisons between experimental data and model simulations
for sideswipe and probe tests are made using k v p ⫽350 kPa, k v e
⫽7,000 kPa, k he ⫽7,000 kPa, ␮⫽0.4⫹0.65z 0 , ␮ t ⫽0.6, ␤
⫽0.06, m⫽0.18, and ␳⫽5. Comparisons are shown in Figs. 12共a
and b兲, respectively, for normally loaded and overloaded pipe
sideswipe tests. Good agreement between the simulations and ex-
perimental data is achieved for normally loaded pipes. At small
embedments the agreement is very good, although the model pre-
diction is about 10% lower than the experimental data at higher
loads. For overloaded cases, the agreement between model simu-
lation and test results is reasonably good. The initial response is
well predicted although the peak horizontal resistance is progres-
Fig. 12. Comparison of model simulation 共thin lines兲 and test results
sively more underestimated as overloading ratio R increases.
共thick lines兲: 共a兲 normally loaded sideswipe events; 共b兲 overloaded
Model behavior of drained probe tests for pipelines of over-
sideswipe events
loading ratios R⫽1, 10, and 20 at V⫽4 kN/m is shown in Fig. 13.

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / MAY 2002 / 369

J. Geotech. Geoenviron. Eng., 2002, 128(5): 363-371


Downloaded from ascelibrary.org by Cambridge University on 09/21/22. Copyright ASCE. For personal use only; all rights reserved.

Fig. 13. Comparison of model simulation 共thin lines兲 and test results
共thick lines兲 for probe events, V⫽4 kN/m: 共a兲 horizontal response;
共b兲 pipeline movement

Fig. 14. Comparison of model simulations: Conventional model


共thin lines兲 and bounding surface model 共thick lines兲: 共a兲 sideswipe
within the yield surface. Model simulations given by the conven- events; 共b兲 probe events
tional and bounding surface models are compared in Fig. 14 for
sideswipe and probe tests. It can be seen from this comparison
that the bounding surface model gives a more gradual transition duces the general behavior observed in centrifuge model tests on
from elastic to plastic states, resulting in significant reductions in a pipeline segment in calcareous sands. The introduction of a
maximal horizontal resistance for sideswipe tests, and a slight non-associated flow rule allows the model to predict the displace-
reduction for probe tests on heavily overloaded pipelines. By ment character for a pipe under horizontal loading. The lateral
comparing Figs. 12共b兲 and 14共a兲, it appears that the conventional break-out resistance predicted by the model agrees very well with
YS model does a better job than the BS model in simulating the experimental data.
overloaded sideswipe tests. However, the probe event is probably At this stage, only drained behavior has been addressed in this
more similar to the loading situation in practice and the BS model paper. Hence the loads used in all equations are total values. The
does a much better job than the YS model in this case. Further- model is being developed further to simulate the effective load
more, the bounding surface model predicted accumulated pipe and displacement response under partially drained conditions by
settlement for probe events that take place inside the bounding relating k v p to a normalized consolidation parameter. Although
surface, which is a significant advantage of the model for future attention was restricted mainly to pipelines, or strip foundations,
development in simulating pipeline response under small ampli- the model also has potential for application to other shapes of
tude of cyclic horizontal loading—an issue of great importance in foundation, at least in the V – H loading plane.
offshore engineering. Caution should be used when using this model so that funda-
mental parameters are determined from experimental observa-
tions as they are affected by the geometry and boundary condi-
Conclusions tions such as pipe diameter, pipeline embedment, roughness of
the pipe surface, and soil conditions.
A bounding surface model was proposed for pipe–soil interaction
of offshore pipelines. The hardening behavior is related to the
plastic vertical displacement in a linear relation. The bounding Acknowledgments
surface takes the shape of a parabola and gives an intercept on the
horizontal load axis at zero vertical load. A non-associated flow This paper was prepared when the first writer was a research
rule was adopted through a bullet-shaped plastic potential surface. student and the second writer was a lecturer at The University of
A radial mapping rule is employed to relate the loading point Western Australia 共UWA兲. The work described here is a part of a
within the bounding surface and the image point on the bounding continuing research program on offshore foundation systems at
surface, allowing the model to simulate a gradual elastic–plastic UWA. Research funding from the Australian Research Council is
transition for loading paths that start inside the bounding enve- gratefully acknowledged. The first writer was supported by a Uni-
lope. versity postgraduate award and an Ernest and Evelyn Havill
The model provides a suitable basis for modeling the load- Shacklock Scholarship in Civil Engineering at The University of
displacement response of unburied offshore pipelines. It repro- Western Australia.

370 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / MAY 2002

J. Geotech. Geoenviron. Eng., 2002, 128(5): 363-371


References rigid surface footing on dense sand under general planar loading.’’
Soils Found., 35共3兲, 71– 82.
Allen, D. W., Lammert, W. F., Hale, J. R., and Jacobsen, V. 共1989兲. Gottardi, G., Houlsby, G. T., and Butterfield, R. 共1999兲. ‘‘Plastic response
‘‘Submarine pipeline on-bottom stability: Recent AGA research.’’ of circular footings on sand under general planar loading.’’ Geotech-
OTC 6055, Proc., 21st OTC, 121–132. nique, 49共4兲, 453– 469.
Allersma, H. G. B., Hospers, B., and den Braber, J. G. 共1997兲. ‘‘Centri- Hale, J. R., Lammert, W. F., and Alen, D. W. 共1991兲. ‘‘Pipeline on-bottom
fuge tests on the sliding behaviour of spudcans.’’ Can. Geotech. J., 34, stability calculations: Comparison of two state-of-the-art methods and
658 – 663. pipe-soil model verification.’’ OTC 6761, Proc., 23rd Annual OTC,
Audibert, J. M. E., and Nyman, K. J. 共1977兲. ‘‘Soil restraint against 567–582.
horizontal motion of pipes.’’ J. Geotech. Eng., 103共10兲, 1119–1142. Karal, K. 共1987兲. ‘‘Trends in pipeline technology development.’’ Offshore
Bergan, P. G., and Mollestad, E. 共1981兲. ‘‘Impact response behaviour of platforms and pipelines, Series on Rock and Soil Mechanics, Mazurk-
offshore pipelines,’’ OTC 4065, Annual OTC, Houston. iewicz, ed., Vol. 13, Trans Tech, Amersdorf, Switzerland, 319–345.
Brennodden, H., Lieng, J. T., Sotberg, T., and Verley, R. L. P. 共1989兲. ‘‘An Montrasio, L., and Nova, R. 共1997兲. ‘‘Settlements of shallow foundations
energy based pipe-soil interaction model.’’ OTC 6057, Proc., 21st on sand: Geometrical effects.’’ Geotechnique, 47共1兲, 49– 60.
OTC, 147–158.
Downloaded from ascelibrary.org by Cambridge University on 09/21/22. Copyright ASCE. For personal use only; all rights reserved.

Moshagen, H., and Kjeldsen, S. P. 共1980兲. ‘‘Fishing gear loads and effects
Butterfield, R., and Gottardi, G. 共1994兲. ‘‘A complete three-dimensional on submarine pipelines.’’ OTC 3782, Annual OTC, Houston.
failure envelope for shallow footings on sand.’’ Geotechnique, 44共1兲, Nova, R., and Montrasio, L. 共1991兲. ‘‘Settlements of shallow foundations
181–184. on sand.’’ Geotechnique, 41共2兲, 243–256.
Byrne, B. W., and Houlsby, G. T. 共1999兲. ‘‘Drained behaviour of suction
Palmer, A. C., Steenfelt, J. S., Steensen-Bach, J. O., and Jacobsen, V.
caisson foundations on very dense sand.’’ Proc., OTC 1999, OTC
共1988兲. ‘‘Lateral resistance of marine pipelines on sand.’’ OTC 5853,
10994.
Proc., 20th OTC, 399– 408.
Cassidy, M. J. and Houlsby, G. T. 共1999兲. ‘‘On the modelling of founda-
Randolph, M. F., Jewell, R. J. L., and Brown, T. A. 共1991兲. ‘‘Establishing
tions for jack-up units on sand.’’ Proc., OTC 1999, OTC 10995.
a new centrifuge facility.’’ Centrifuge ’91, Boulder, Colo. 2–9.
Dafalias, Y. F., and Herrmann, L. R. 共1982兲. ‘‘Bounding surface formu-
lation of soil plasticity.’’ Soil mechanics—Transient and cyclic loads, Trautmann, C. H., and O’Rourke, T. D. 共1985兲. ‘‘Lateral force-
Pande and Zienkiewicz, eds., 253–282. displacement response of buried pipe.’’ J. Geotech. Eng., 111共9兲,
Finnie, I. M. S. 共1993兲. ‘‘Performance of shallow foundations in calcar- 1077–1092.
eous soil,’’ PhD thesis, Univ. of Western Australia, Nedlands, WA, Verley, R. L. P., and Sotberg, T. 共1994兲. ‘‘A soil resistance model for
Australia. pipelines placed on sandy soils.’’ J. Offshore Mech. Arct. Eng., 116,
Gadre, A. D., and Dobry, R. 共1998兲. ‘‘Lateral response of square embed- 145–153.
ded foundation in dry sand.’’ Centrifuge’98, Kimura, Kusakabe, and Wallace, L. T. I. 共1995兲. ‘‘Pipeline performance in calcareous soil.’’ Hon-
Takemura, eds., 465– 470. ors thesis, Dept. of Civil and Resource Engineering, Univ. of Western
Gazetas, G., and Tassoulas, J. L. 共1987兲. ‘‘Horizontal stiffness of arbi- Australia, Nedlands, WA, Australia.
trarily shaped embedded foundations.’’ J. Geotech. Eng., 113共5兲, 440– Wolfram, Jr., W. R., Getz, J. R., and Verley, R. L. P. 共1987兲. ‘‘PIPESTAB
457. project: Improved design basis for submarine pipeline stability.’’ OTC
Gazetas, G., Dobry, R., and Tassoulas, J. L. 共1985兲. ‘‘Vertical response of 5501, Proc., 19th OTC, 153–158.
arbitrarily shaped embedded foundations.’’ J. Geotech. Eng., 111共6兲, Zhang, J., Randolph, M. F., and Stewart, D. P. 共1999兲. ‘‘An elasto-plastic
750–771. model for pipe–soil interaction of unburied offshore pipelines in cal-
Gottardi, G., and Butterfield, R. 共1995兲. ‘‘The displacement of a model careous sand.’’ ISOPE-99, Brest, Vol. 2, 185–192.

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / MAY 2002 / 371

J. Geotech. Geoenviron. Eng., 2002, 128(5): 363-371

You might also like