Vous êtes sur la page 1sur 10

744

ARTICLE
Uplift soilpipe interaction in granular soil

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 11/27/14


For personal use only.

Jai K. Jung, Thomas D. O'Rourke, and Nathaniel A. Olson

Abstract: Soilpipeline interaction for uplift in granular soil is evaluated by means of a two-dimensional, nite element (FE)
continuum model with a MohrCoulomb (MC) yield surface for peak strength, a strain-softening relationship tied to critical void
conditions, and an equivalent modulus that is consistent with soil deformation at maximum uplift resistance. The model
accounts for soil migration beneath the pipe through FE mesh adjustment coordinated with upward pipe displacement. A
systematic comparison of model results with multiple full-scale test measurements of pipe in dry sand show excellent agreement
both with respect to maximum force and forcedisplacement relationships, including post-peak performance. The relationship
between peak upward force and pipe depth is developed for various sand densities, all of which show maximum force at a
depth-to-diameter ratio of 30. Hyperbolic and bilinear models for vertical upward force versus displacement are presented. The
analytical approach described in this paper benets from its adaptation to MC strength selection available in many commercial
software packages.
Key words: pipeline, MohrCoulomb, nite element analyses, soilpipeline interaction, upward movement, earthquake.
Rsum : Linteraction soltuyau, responsable du soulvement dans un sol granulaire, est value a` laide dun modle continu
par lments nis (EF) en deux dimensions avec une surface de rupture a` la rsistance au pic de MohrCoulomb (MC), une
relation dadoucissement des dformations lie aux conditions critiques des vides, et un module quivalent consistant avec les
dformations du sol a` la rsistance maximale au soulvement. Le modle tient compte de la migration du sol sous le tuyau par
des ajustements de maillage des EF en lien avec les dplacements vers le haut du tuyau. Les rsultats du modle ont t compars
systmatiquement avec des mesures provenant de nombreux essais a` lchelle relle de tuyau dans du sable sec, et les modlisations et mesures dmontrent une excellente concordance en termes de la force maximale et des relations forcedplacement,
incluant la performance post-pic. La relation entre la force au pic vers le haut et la profondeur du tuyau est dveloppe pour
plusieurs densits de sable. Pour toutes les densits values, la force maximale se retrouve a` un ratio profondeur-diamtre de
30. Des modles hyperboliques et bilinaires de force de soulvement vertical versus le dplacement sont prsents. Lapproche
analytique dcrite dans cet article est facilite par son adaptation au critre de rsistance MC disponible dans plusieurs logiciels
commerciaux. [Traduit par la Rdaction]
Mots-cls : tuyau, MohrCoulomb, analyses par lments nis, interaction soltuyau, mouvement vers le haut, sisme.

Introduction
This paper focuses on the uplift behavior of underground pipelines. The characterization of uplift behavior is important for
modeling soilpipeline interaction in response to settlements
caused by earthquake-induced liquefaction, landslides, and normal faulting. Pipeline uplift in response to ground deformation
not only applies to earthquakes, but also occurs in response to
oods, landslides, tunneling, deep excavations, and subsidence
triggered by dewatering and mining (O'Rourke 2010). Uplift behavior involves material nonlinearities with signicant soil yielding
under large vertical displacement of the pipe. Such behavior also
involves nonlinear geometric behavior as soil migrates from a
position above to below the pipe in response to relative upward
movement.
Previous research investigations (e.g., Yimsiri et al. 2004; Klar
et al. 2005; Vorster et al. 2005; O'Rourke 2010; Chou et al. 2011; Xie
et al. 2012) used continuum models to address the complexities of
soilpipeline interaction under permanent ground deformation.
Pipeline uplift behavior is similarly modeled in this work with
continuum nite element (FE) models. Plane strain conditions are
simulated so that the analytical results can be compared directly
with the results of large-scale experiments on the uplift displace-

ment of pipe in granular soil as provided by Trautmann and


O'Rourke (1983).
One of the objectives of this study is to work with relatively
simple models for soil behavior that are readily accessible in or
adaptable to commercially available software. If a relatively
simple model is shown to provide good results, it will benet
from the ease of use with a number of different software packages. An elastoplastic model was adopted with MohrCoulomb
(MC) strength parameters, a nonassociated ow rule based on
the smooth triple symmetric ellipse function proposed by
Menetrey and Willam (1995), and a strain-softening model developed by Robert and Soga (2009) on the basis of direct shear
(DS) test results.
The paper describes the MC and strain-softening model in conjunction with DS test results and the conversion of DS to plane
strain strength parameters. The method of selecting a straincompatible modulus is explained, based on a hyperbolic representation of stress versus strain consistent with peak soil strength
mobilized at maximum upward pipe force. The method for FE
mesh adjustment is described that accounts for soil migration
beneath the pipe during upward pipe displacement. The analytical results are compared with multiple full-scale test measurements of pipe in dry sand. The relationship between peak upward

Received 27 September 2012. Accepted 23 April 2013.


J.K. Jung. Virginia Polytechnic Institute of Technology, Blacksburg, VA, USA.
T.D. O'Rourke. Cornell University, Ithaca, NY, USA.
N.A. Olson. Stephens Associates Consulting Engineers, Brentwood, NH, USA.
Corresponding author: Jai K. Jung (e-mail: jai.k.jung@gmail.com).

Can. Geotech. J. 50: 744753 (2013) dx.doi.org/10.1139/cgj-2012-0357

Published at www.nrcresearchpress.com/cgj on 29 April 2013.

Jung et al.

745

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 11/27/14


For personal use only.

Fig. 1. Mohr circle for stress at peak direct shear and plane strain
shear in dry sand. =, effective stress; =, effective shear stress;
p , effective peak shear stress.

force and pipe depth is developed for various sand densities, and
hyperbolic and bilinear models for vertical upward force versus
displacement are presented for use in the analytical models most
frequently used in current design methods.

To make use of eq. (1), one needs to determine ds-p and p.


Olson (2009) performed numerous DS tests to determine strength
and dilatancy properties of Cornell University (CU) Filter sand
used by Trautmann and O'Rourke (1983) in large-scale tests of
soilpipe interaction under vertical upward movement. The test
soil was an angular to subangular glaciouvial sand, characterized by a mean grain size, D50 = 0.59 mm, and the coefcient of
curvature, Cc = 1.26. The grain-size distribution curve for the sand
is shown in Fig. 2, with an inset photo of the sand particles. The
sand is classied as poorly graded sand (SP) according to the Unied Soil Classication System (ASTM 2011). For various dry unit
weights, d, Olson (2009) plotted ds-p and p for CU Filter sand as
shown in Fig. 3. The linear regression equations for ds-p and p
versus d are provided in the gure.
As dilatancy is diminished by increasing conning stress, it is
necessary to characterize the relationship between p and N.
Olson (2009) provided data for p versus N from DS tests that
were normalized with respect to a reference stress N = 2.1 kPa
and plotted in Fig. 4. The nonlinear regression equation for
p/p@ N Ref. (where p@ N Ref. is the dilation angle at the reference
stress) is presented in the gure. The p values from Figs. 3 and 4
was used to obtain ds-p, using the relationship provided by Lings
and Dietz (2004) as

Finite element model and soil strength properties


Many constitutive models for soil behavior (e.g., Mohr
Coulomb, DruckerPrager, Modied DruckerPrager/Cap, CamClay, and Nor-Sand) are available. As previously discussed, the
authors selected an elastoplastic model with an MC yield surface
due to its simplicity and extensive use in soilstructure interaction analyses (Ellis and Springman, 2001; Hu and Pu 2003; Yimsiri
et al. 2004; Sun et al. 2006; Robert and Soga 2009). In addition, an
MC strength model is easy to apply in many commercial software
packages (e.g., ABAQUS, FLAC, and PLAXIS).
The elasto-plastic MC model used in this work requires the
following principal input parameters: (i) soil strength as expressed by the plane strain peak friction angle ( ps-p), (ii) peak
dilation angle (p), and (iii) the secant soil modulus (E) at a strain
compatible with the maximum uplift pipe load. Each input parameter is discussed under a separate subheading that follows.

(2)

Plane strain soil strength parameters


To account for soil behavior under plane strain conditions it is
necessary to dene the MC yield surface in terms of ps-p. This
friction angle can be obtained from DS tests by recognizing the
coaxiality between incremental strain and stress, and using the
horizontal plane of the DS apparatus as the direction of zero
linear strain. Davis (1968) rst recognized and described these
conditions to derive a relationship that has been conrmed subsequently by numerous researchers (e.g., Bolton 1986; Jewell and
Wroth 1987; Lings and Dietz 2004; O'Rourke 2010), as follows:

(3)

(1)

sin ps-p

tan ds-p
cosp (sinp) (tan ds-p)

in which ds-p is the DS peak friction angle.


Figure 1 shows a Mohr circle of stress at peak DS and plane
strain shear for dry sand. In the gure, N, ds-p, and ps-p are
normal stress, direct shear peak shear stress, and plane strain
peak shear stress, respectively. Given that the DS test failure plane
coincides with zero extension and that incremental stress and
strain are coaxial, p is oriented at p with respect to the vertical
radius of the Mohr circle at ( 1 + 3)/2, where 1 and 3 are the
major and minor principal stress, respectively, and ps-p is dened
at the point of maximum obliquity with respect to the Mohr
circle.

tan ds-p

sin crit sinp


cosp

in which crit is critical friction angle. Olson (2009) reported crit


for CU Filter sand as 38.6.
Secant soil modulus
Implicit in using an elastoplastic model to predict upward soil
pipe behavior at maximum pipe load is the selection of a Young's
modulus that is strain-compatible with the level of stress in the
soil at peak pipe load. The secant modulus (Esec) of sand is related
to the fraction of the peak soil strength mobilized at maximum
lateral pipe force. Duncan and Chang (1970) have shown that the
stressstrain curve of sand can be approximated by hyperbola as

( 1 3)

a b

where is strain and a and b are dened in Fig. 5. In the gure, Ei


is the initial tangent modulus and ( 1 3)ult is the asymptotic
value of the principal stress difference at innite strain. The stress
difference at the peak lateral pipe force, ( 1 3)f, is expressed as
(4)

( 1 3)f Rf( 1 3)ult

where Rf is a reduction factor. The value of Rf = 0.9 reported by


Trautmann and O'Rourke (1983) was used in this study. Assuming
that is the fraction of the peak soil strength mobilized at maximum upward pipe force, Jung (2010) related the secant modulus to
(E) as
(5)

E (1 Rf)Ei

From eq. (5), Esec associated with any percentage of the maximum stress level can be calculated from the initial tangent modulus. Finite element simulations of lateral pipe movement in sand
were used to estimate the relationship between Ei and Esec at peak
pipe load. As the maximum lateral pipe force is independent of
the Youngs modulus, E, FE results can be used without a priori
knowledge of E to determine . The appropriate value of Esec can
Published by NRC Research Press

746

Can. Geotech. J. Vol. 50, 2013

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 11/27/14


For personal use only.

Fig. 2. CU Filter sand: (a) photo and (b) particle-size distribution (after Olson 2009).


Fig. 3. p and ds-p
versus d for dry CU Filter sand at N = 2.1 kPa
(after Olson 2009). n, number of data points; r2, coefcient of
determination.

Fig. 4. Plot of p/p@ N Ref. versus N at d = 16.5 kN/m3.

Fig. 5. Hyperbolic stressstrain curve (after Duncan and Chang


1970).

marized K and n values obtained from triaxial tests on sand, and


reported mid-range values of K and n equal to 800 and 0.55, respectively, for dense sand. Jung (2010) performed a series of FE
simulations for vertical pipe movement to show 0.91 0.98,
from which = 0.94 was selected for analysis. Values of secant soil
modulus at a strain compatible with the maximum lateral pipe
load are summarized in Table 1 for each depth-to-diameter ratio,
Hc/D, where Hc is the depth from the soil surface to the center of
the pipe and D is the external diameter of the pipe. The modulus
is a function of conning stress and changes with depth as E
increases from 4002800 kPa and 7005200 kPa for medium and
very dense dry CU lter sand, respectively, for Hc/D = 1.513.
Strain-softening behavior
To represent strain-softening, the model proposed by
Anastasopoulos et al. (2007) was used to diminish linearly both
ps-p and p to residual values of crit and 0, respectively, from the
plastic strain at ps-p to the plastic strain at crit, using the results
of DS testing as

then be chosen from the hyperbolic model, using from initial FE


simulations.
Janbu (1963) proposed the relationship between Ei, and 3 as Ei =
Kpa( 3/pa)n, in which K is a constant, pa is atmospheric pressure,
and n is an exponent determining the rate of variation of Ei with
3. Duncan and Chang (1970) and Wong and Duncan (1974) sum-

(6)

pf

dxp dxy
dxf dxp

H
dFE

where pf is the the plastic strain at crit; dxy, dxp, and dxf are the DS
test horizontal displacements at yield, peak strength, and crit at
Published by NRC Research Press

Jung et al.

747

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 11/27/14


For personal use only.

Table 1. Summary of secant soil modulus for upward pipe movement.


Soil

Hc/D

E (kPa)

Medium dry CU
Filter sand
Very dense dry
CU Filter sand

1.513

4002800

1.513

7005200

Fig. 6. Geometry of the numerical analysis for upward pipe


movement test.

which full softening occurs, respectively; H is the thickness of the


DS specimen, and dFE is the FE element size. The FE element size,
dFE, in the rened part of the nite element mesh was taken as
20d50, where d50 is the median grain size. This thickness compares
favorably with the shear band thickness as reported by several
investigators as 820d50 (e.g., Roscoe 1970; Vardoulakis and Graf
1985; Muhlhaus and Vardoulakis 1987; Huang et al. 2002). A FORTRAN subroutine developed by Robert and Soga (2009) to apply
the Anastasopoulos et al. (2007) model in the MC model in
ABAQUS was used in the two-dimensional (2D) FE simulations.

Finite element modeling


FE analyses were run with a semi-innite soil medium using
innite elements as developed by Zienkiewicz et al. (1983).
Figure 6 shows the typical geometry of the model used for simulating vertical upward soilpipe interaction. In the model, eightnode biquadratic, plane strain, quadrilateral, reduced integration
elements (element type CPE8R) are used to represent the soil
around the pipe from A to B, and from the surface to C. Five-node
quadratic, plane strain, one-way innite quadrilateral elements
(element type CINPE5R) are used to represent the semi-innite
soil. The innite elements are attached to CPE8R elements and are
1.3 m in width. The innite elements at the base of the model are
1.3 m in height. The pipe was modeled as a rigid cylinder with
outsider diameter of 102 mm. The interface between the soil and
pipe has a friction angle of 0.6 ds-p. To promote numerical stability, all simulations were performed with a nominal cohesion, c ps =
0.10.3 kPa. Sensitivity analyses were performed with different
c ps values to show that the range of c ps = 0.10.3 kPa had negligible
effect on the results. The interface interaction is modeled through
surface-based contact, in which separation and slip between soil
and pipe are allowed. A rened mesh was used within a distance
of approximately two pipe diameters from the center of the pipe.
The thickness of the elements within the rened mesh was taken
as the shear band thickness of 12 mm, which is consistent with the
rupture zone observed during DS tests (Roscoe 1970; Vardoulakis
and Graf 1985; Muhlhaus and Vardoulakis 1987; Huang et al. 2002).
Approximately 1500 to 5000 elements were used in the meshes to
simulate different Hc/D conditions. Geostatic loading was applied
to the soil and pipe at the beginning of the analysis under Ko =
1 condition, where Ko is the initial horizontal to vertical stress
ratio. Sensitivity studies for Ko showed that the maximum force
changed approximately 3% and the computational time increased
approximately 10 times when Ko changed from 1 to 0.3. Upward
pipe movement was imposed as a vertical upward displacement of
all pipe nodes.
In the FE model, vertical effective stress at the horizontal pipe
centerline ( vc), shown in Fig. 6, is taken as N. Characterizing p
in terms of a single vc is a simplication that does not account for
variable p and ps-p with depth or link p with the appropriate N
at any given depth. Jung et al. (2012) performed a series of FE
simulations for layered soil in which p and ps-p were varied with
depth, and the analytical results were compared with those using
p and ps-p linked to a single vertical stress at the pipe centerline.
The comparisons showed only 0.2%3.3% difference in the peak
lateral forces for a wide range of depths. Thus, characterizing p

and ps-p in terms of a single vc is used, due to its simplicity and
straightforward characterization of soil strength.
The soil displacements measured for vertical upward pipe
movements in the medium CU Filter sand at shallow depth
(Trautmann and O'Rourke 1983) show that the soil above the pipe
was pushed upward and outward, as the pipe was displaced vertically. The soil at the sides moved downward and owed into the
void created under the pipe, in much the same manner as described by Kananyan (1966) for uplift tests on disks. Soil displacement during the large-scale tests was obtained by measuring
movement of 150 mm long by 6 mm diameter wooden dowels that
were placed normal to the side of the test box. The large-scale test
box was constructed with a glass window through which measurements were made (Trautmann and O'Rourke 1983). The measurements

= 43.8 and
were performed for tests in medium to very dense sand ( ps-p
48.6, respectively) for 1.5 Hc/D 13.
The average maximum upward ground surface displacement
for eight tests was 25% of the pipe displacement with a range of
10% to 45% of the pipe placement. In contrast, the surface displacement from nite element analysis (FEA) was approximately equal
to the vertical pipe movement. As a consequence, the simulation
was modied to be consistent with the observed soil movement
during the tests. Figure 7 shows the deformed shape of the FE
mesh for medium sand at Hc/D = 1.5. The surface elements that
heaved more than 25% of the pipe displacement were removed to
be consistent with the observed surface displacement. This reduction was accomplished by exercising the Remove option in
ABAQUS during three to four steps of the simulation. As illustrated later in the paper by comparisons of model results with
particle image velocimetry (PIV) measurements, the analytical
simulations are able to replicate soil displacement elds and
shear zone formation at both peak and post-peak conditions of
upward pipe movement. Although the modied numerical model
is not able to account fully for soil that moves into the void beneath the pipe, it is able to simulate zones of distributed shear
above the pipe that generate forces restricting upward pipe movement. The reduction of surface elements in the analyses reduces
soil weight while replicating shear transfer mechanisms consistent with those observed experimentally and thus captures the
post-peak loss of resistance to vertical movement in the full-scale
tests.
In contrast to the conditions for relatively low Hc/D, FEA without surface modication for Hc/D = 8 and 13 provided results consistent with the experimental data, including agreement with
Published by NRC Research Press

748

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 11/27/14


For personal use only.

Fig. 7. Example of deformed shape of FE analysis for vertical


upward pipe movement test after modication process.

Can. Geotech. J. Vol. 50, 2013

Table 2. Summary of dry CU Filter sand strength parameters.


Density

d
(kN/m3)


ds-p
()


ps-p
()

Medium

16.4

36.0

43.8

4.0

Very dense

17.7

40.6

48.6

10.9

p
()

Full-scale
test results
Trautmann and
O'Rourke (1983)
Trautmann and
O'Rourke (1983)

Fig. 9. Comparisons of NqVU-fea and NqVU-m for vertical upward


movement.

Fig. 8. Comparisons of vertical upward NqVU-m and NqVU-fea in medium


and very dense CU Filter sand using the unmodied (regular) FE
analysis.

Fig. 10. Summary plot of vertical upward pipe movement NqVU


versus Hc/D. F, force.

observed surface displacements. To identify the smallest Hc/D


where FEA without modication compares favorably with experimental data, a series of simulations was performed to compare
the maximum dimensionless vertical upward movement force,
NqVU-fea = Fpeak/(dHcDL), with the measured maximum dimensionless vertical upward movement force, NqVU-m. In this equation,
Fpeak is peak force, L is length of the pipe, and d, Hc, and D have
been dened previously.
Figure 8 shows the percentage difference between NqVU-m and
NqVU-fea with respect to NqVU-m versus Hc/D. The FEAs without modication overpredict NqVU-m at relatively shallow depths (Hc/D = 1.5
and 4) by 17%42%, whereas the FEAs without modication predict
NqVU-m reasonably well for greater depths. The percentage difference between NqVU-m and NqVU-fea for both medium and very dense
sand is less than 10%NqVU-m for Hc/D = 5.5. Therefore, the modication process was not required for Hc/D 5.5.

Comparison of analytical and experimental results


The analytical results are compared with large-scale test results
reported by Trautmann and O'Rourke (1983). Dry glaciouvial
sand was used in the tests with unit weights of 16.4 and
17.7 kN/m3, and depth-to-diameter ratios of 1.5, 4, 8, and 13. All
tests were performed with 102 mm diameter pipe. The soil
strength properties associated with each series of tests are summarized in Table 2. Average values of d as well as ps-p and p are
given. The preponderance of the measurements obtained by

Trautmann and O'Rourke (1983) show a clear peak load. For tests
with no clear peak, the procedure used by Trautmann and
O'Rourke (1983) by tting the data to a hyperbolic curve and sePublished by NRC Research Press

Jung et al.

749

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 11/27/14


For personal use only.

Fig. 11. Comparison of analytical and measured incremental displacement elds for (a) FE simulation at peak load, (b) experiment at peak
load (Cheuk et al. 2008), (c) FE simulation at 0.4D pipe movement, and (d) experiment at 0.5D pipe movement (Cheuk et al. 2008).

lecting peak force as the product of the hyperbolic asymptote


multiplied by a reduction factor, Rf = 0.9, was used.
Figure 9 shows NqVU-fea plotted with respect to NqVU-m. There are
excellent agreements between NqVU-fea and NqVU-m, with 88% of
NqVU-fea within 10% of NqVU-m. Figure 10 compares the FE results
and 2-D test measurements of Trautmann and O'Rourke (1983), in
which the vertical upward dimensionless force (NqVU) is plotted as
a function of the dimensionless depth. Nonlinear regression lines
are plotted for the experimental data and numerical results. In
general, the FEA results overpredict measured force by 5%13%,
with an average difference of 7.7%. Also plotted in Fig. 10 are data
expressed in dimensionless terms from centrifuge tests reported
by Dickin (1994) for sand with grain size characteristics and unit
weight, d = 16 kN/m3, similar to those of the sand used by
Trautmann and O'Rourke (1983) for the medium sand (d =
16.4 kN/m3) results plotted in Fig. 10. The maximum dimensionless upward forces predicted by FE simulation and those obtained
by Dickin (1994) compare very well.
As an additional check, the velocity elds determined by PIV
and photogrammetry for pipe uplift tests reported by Cheuk et al.
(2008) and White et al. (2008) for pipe in dense sand with Hc/D = 3
are compared with the analytical results for very dense Cornell

Filter sand (d = 17.7 kN/m3) for Hc/D = 4. Figures 11a and 11b show
the incremental displacement elds for pipe movement at peak
uplift force from FE simulations and the pipe uplift tests, respectively. Figures 11c and 11d provide a similar comparison for postpeak conditions when the pipe movement, , is 0.4D to 0.5D.
Incremental vector displacements are normalized by the upward
pipe movement between the movement at peak resistance, peak,
and 0.5D, which are scaled by a factor of 5 for better visualization.
The FEA incremental displacements from peak resistance to 0.4D
pipe movement are also scaled by a factor of 5. The FEA displacements are conned to 0.4D to eliminate the effects of local mesh
instability at higher values. The displacement elds at peak uplift
resistance show similarly sized zones of distributed shear (A) that
curve outward towards the surface (B), and have similar average
angles of inclination of 18 with respect to vertical, as dened by
Cheuk et al. (2008). The post-peak displacement elds show similar shear bands that have narrowed above the pipe (C), accompanied by downward rotational displacement zones (D) of similar
dimensions at the pipe level. From an examination of the displacement elds, Cheuk et al. (2008) concluded that normality is violated, thus requiring a nonassociated ow rule when modeling
the soilpipe interaction. The FEA model in this study uses the
Published by NRC Research Press

750

Can. Geotech. J. Vol. 50, 2013

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 11/27/14


For personal use only.

Fig. 12. Vertical upward dimensionless force versus displacement curves for medium CU Filter sand.

nonassociated ow rule proposed by Menetrey and Willam (1995).


The comparisons indicate that the FEA model is able to replicate
the experimental displacement elds and shear zone dimensions
in a favorable manner. The model results in post-peak shear band
formation that matches the experimental measurements even as
soil is removed from the surface during modeling to account for
the effects of sand migration beneath the pipe.
Comparisons of FEA with measured force versus displacement
plots are provided in Figs. 12 and 13 for medium and very dense CU
Filter sand, respectively. The vertical upward pipe forces are
shown as dimensionless force F= = F/(dHcDL), and upward pipe
displacements are shown as dimensionless displacement Z= = Z/D,
in which F is force and Z is the relative vertical upward displacement between pipe and soil. The results from modied numerical
simulations for shallow depth show sudden changes in the postpeak behavior as elements on the top surface were removed from
the FE mesh. The irregular pattern of the upward forces matches
that of the full-scale test measurements. Overall, the predicted
forcedisplacement behavior from the FEA compares very well
with the measured data in terms of peak force and post-peak
behavior, with analytical peak forces within 10% of the measured
ones.

Maximum vertical upward force versus pipe depth


To explore further NqVU versus Hc/D for vertical upward pipe
movement, FEAs were performed for Hc/D varying from 5.5 to
100 for medium, dense, and very dense dry sand as characterized
in Table 2. In the case of dense sand, d is taken as 17.1 kN/m3 and

other input parameters are obtained with the procedures described earlier.
Comparisons of vertical upward NqVU versus Hc/D for medium,
dense, and very dense sand are shown in Fig. 14. The values of NqVU
were determined as described previously. The FE simulations of
vertical upward pipe movement show that NqVU reaches its maximum of 14, 17, and 20 for medium, dense, and very dense sand,
respectively, at Hc/D = 30. A review of the FE results for D = 900 mm
also show that, at depths of Hc/D = 30, the upward reaction force
attains its maximum value, after which there is a very small force
reduction at greater depth due to reduction in dilation angle with
increased conning stress. Similar trends are reported by Yimsiri
et al. (2004) for lateral pipe movement.

Force versus displacement relationships for


one-dimensional modeling of soilpipe interaction
Pipeline analysis for soilstructure interaction under permanent
ground deformation is often performed with one-dimensional
(1-D) FE models to represent the pipe and soil force versus
displacement relationships that are mobilized by various types
of ground movement. The force versus displacement relationships for upward ground movement (qz) can be represented
by a rectangular hyperbola for the nonlinear characterization
of qz or by a simple bilinear relationship between force and
displacement (Trautmann and O'Rourke 1983; ASCE 1984;
Trautmann et al. 1985).
Published by NRC Research Press

Jung et al.

751

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 11/27/14


For personal use only.

Fig. 13. Vertical upward dimensionless force versus displacement curves for very dense CU Filter sand.

Fig. 14. Summary plot of NqVU versus Hc/D for vertical upward pipe
movement.

To represent the rectangular hyperbola curve for the qz relationship, FE simulations for different Hc/D and d were performed.
The results were plotted on the dimensionless transformed axes
formed by q=/z= and z= in Fig. 15, as proposed by Kondner (1963) and
Trautmann and O'Rourke (1983). In the gure, q= = (F/dHcDL)/NqVU,
z= = (z/D)/(zmf/D), where z is the vertical upward displacement and
zmf is the upward displacement at which NqVU is mobilized. The
linear regression, provided in the gure, was tted to the transformed data to determine the constants a and b for the general
form of the hyperbola, F = z/(a + bz), from which the representative

Fig. 15. q=/z= versus z= plot.

hyperbolic model for vertical upward pipe movement was found


to be
(7)

q

z
0.1 0.88z 
Published by NRC Research Press

752

Can. Geotech. J. Vol. 50, 2013

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 11/27/14


For personal use only.

Fig. 16. qz relationships for 1-D modeling of soilpipeline


interaction for vertical upward pipe movement.

Fig. 17. zmf/Hc versus Hc/D plots for vertical upward pipe movement.

Fig. 18. z70/Hc versus Hc/D plots for vertical upward pipe movement.

The results of FE simulations of upward pipe displacement,


including those presented in Figs. 12 and 13, are plotted in dimensionless form in Fig. 16. Equation (7) is also plotted in the gure.
Based on the experimental data, Trautmann and O'Rourke (1983)
reported a representative hyperbola relationships for vertical upward pipe movement as q= = z=/(0.07 + 0.93z=), which is plotted as a
dashdot line in Fig. 16 for comparison. Equation (7) compares
favorably with the expression reported by Trautmann and
O'Rourke (1983). The displacements at the maximum force from
the simulations were normalized with respect to Hc and expressed
as zmf/Hc in Fig. 17 for various Hc/D. As shown in the gure, representative values of zmf/Hc are 1.3%, 1.1%, and 1.0% for medium,
dense, and very dense sand, respectively. The displacement zmf
from Fig. 17 can be used with eq. (7) to obtain the hyperbolic
relationships for vertical upward pipe movement.
Some computer programs use a bilinear force versus displacement relationship to model soil behavior as shown in Fig. 16. The
bilinear force versus displacement model consists of an initial
linear portion with slope K70 that intersects the nonlinear force
versus displacement relationship at 70% of the maximum force.
This method of estimating K70 was proposed by Thomas (1978) and
used by Trautmann and O'Rourke (1983).
Figure 18 shows values of z70/Hc for simulated vertical upward
pipe movement, where z70 is the displacement corresponding to
70% of maximum force for the bilinear relationship. As shown in
the gure, representative values of z70/Hc are 0.4%, 0.3%, and 0.2%
for medium, dense, and very dense sand, respectively. The representative value for z70/Hc can be used to compute K70 as
(8)

K70 0.7NqVU(Hc /z70)

Trautmann and O'Rourke (1983) reported a representative value


of z70/Hc = 0.003, or 0.3%, for all sand densities, which is similar to
the simulation values. Using eq. (8) and the values of z70/Hc from
Fig. 18, K70 for various pipe diameters, depths, and soil unit
weights can be obtained and used for modeling soilpipeline interaction.

Conclusions
An elastoplastic constitutive model using MohrCoulomb (MC)
strength parameters, a nonassociated ow rule, and a strainsoftening subroutine is used to represent soil behavior for upward
pipe movement. Plane strain strength parameters converted from
direct shear (DS) test results are used in the model and the conversion process is explained. The model accounts for reduced dilatancy and peak angle of shear strength with increased depth.
The method for the selection of an appropriate strain-compatible
modulus is described, based on a hyperbolic representation of
stress versus strain consistent with peak soil strength mobilized
at maximum upward pipe force. To account for soil migration
beneath the pipe during upward movement, the nite element
(FE) mesh is adjusted. Large-scale test results are compared with
the simulation results to validate and qualify the soilstructure
interaction modeling. The predicted maximum dimensionless
forces from FE analyses match very well with those from largescale tests for various dry sand densities and depths. Moreover,
the displacement elds and shear zone dimensions for both peak
and post-peak uplift load conditions compare favorably with
those measured by Cheuk et al. (2008) and White et al. (2008) with
PIV methods. The favorable comparisons between the analytical
results and the experimental evidence from multiple sources indicate that the modeling approach is sufciently robust to analyze
many different granular soils, pipe diameters, and pipe-depth-todiameter ratios, including those at large depths. The modeling
Published by NRC Research Press

Jung et al.

process is used to characterize the peak upward force as a function of pipe depth. The results show that the dimensionless vertical upward force reaches its maximum value of approximately 14,
17, and 20 for medium, dense, and very dense sand, respectively,
at a depth-to-diameter ratio of 30. The results of FE simulations
are used to construct hyperbolic and bilinear relationships for the
force versus displacement response during upward pipe movement. The relationships from the simulation compare favorably
with the expression from the large-scale test data for both cases.

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 11/27/14


For personal use only.

Acknowledgements
The work on which this paper was based was supported by the
George E. Brown, Jr. Network for Earthquake Engineering Simulation (NEES) Program of the National Science Foundation (NSF)
under Grant No. CMMI-1041498. Any opinions, ndings, and conclusions or recommendations expressed in this material are those
of the authors and do not necessarily reect the views of the NSF.
The authors acknowledge Drs. Soga, Anastasopoulos, and Robert
for their assistance in programming the FORTRAN subroutine.

References
Anastasopoulos, I., Gazetas, G., Bransby, M.F., Davies, A., and El Nahas, M.C.R.
2007. Fault rupture propagation through sand: nite-element analysis and
validation through centrifuge experiments. Journal of Geotechnical
and Geoevironmental Engineering, 133(8): 943958. doi:10.1061/(ASCE)10900241(2007)133:8(943).
ASCE. 1984. Guidelines for the seismic design of oil and gas pipeline systems.
Committee on Gas and Liquid Fuel Lifelines, American Society of Civil Engineers, Reston, Va.
ASTM. 2011. Standard practice for classication of soils for engineering purposes
(Unied Soil Classication System). ASTM standard D2487. American Society
for Testing and Materials, West Conshohocken, Pa.
Bolton, M.D. 1986. The strength and dilatancy of sands. Gotechnique, 36(1):
6578. doi:10.1680/geot.1986.36.1.65.
Cheuk, C.Y., White, D.J., and Bolton, M.D. 2008. Uplift Mechanisms of pipes
buried in sand. Journal of Geotechnical and Geoenvironmental Engineering,
134(2): 154163. doi:10.1061/(ASCE)1090-0241(2008)134:2(154).
Chou, J.C., Kutter, B.L., Travasarou, T., and Chacko, J.M. 2011. Centrifuge modeling of seismically induced uplift for the BART Transbay Tube. Journal of
Geotechnical and Geoenvironmental Engineering, 137(8): 754765. doi:10.
1061/(ASCE)GT.1943-5606.0000489.
Davis, E.H. 1968. Theories of plasticity and the failure of soil masses. In Soil
mechanics, selected topics. Edited by I.K. Lee. Butterworth, London.
Dickin, E.A. 1994. Uplift resistance of buried pipelines in sand. Soils and Foundations, 34(2): 4148. doi:10.3208/sandf1972.34.2_41.
Duncan, J.M., and Chang, C.-Y. 1970. Nonlinear analysis of stress and strain in
soils. Journal of the Soil Mechanics and Foundations Division, ASCE, 96(5):
16291653.
Ellis, E.A., and Springman, S.M. 2001. Modelling of soil-structure interaction for
a piled bridge abutment in plane strain FEM analyses. Computers and
Geotechnics, 28(2): 7998. doi:10.1016/S0266-352X(00)00025-2.
Hu, L., and Pu, J.L. 2003. Application of damage model for soilstructure interface. Computers and Geotechnics, 30(2): 165183. doi:10.1016/S0266-352X
(02)00059-9.
Huang, W., Nbel, K., and Bauer, E. 2002. Polar extension of a hypoplastic model
for granular materials with shear localization. Mechanics of Materials, 34(9):
563576. doi:10.1016/S0167-6636(02)00163-1.
Janbu, N. 1963. Soil compressibility as determined by oedometer and triaxial
tests. In Proceedings, European Conference on Soil Mechanics & Foundations
Engineering, Wiesbaden, Germany. Vol. 1, pp. 1925.
Jewell, R.A., and Wroth, C.P. 1987. Direct shear tests on reinforced sand. Gotechnique, 37(1): 5368. doi:10.1680/geot.1987.37.1.53.

753

Jung, J.K. 2010. Soil-pipe interaction under plane strain conditions. Ph.D. thesis,
Cornell University, Ithaca, N.Y.
Jung, J.K., Koo, D., and Zhang, K. 2012. Verication of the pipe depth dependent
model using a nite element analysis. Tunneling and Underground Space
Technology. dx.doi:10.1016/j.tust.2012.02.024. [Posted online ahead of print
24 April 2012.]
Kananyan, A.S. 1966. Experimental investigation of the stability of bases of
anchor foundations. Soil Mechanics and Foundation Engineering, 3(6): 387
392. doi:10.1007/BF01702954.
Klar, A., Vorster, T.E.B., Soga, K., and Mair, R.J. 2005. Soilpipe interaction due to
tunnelling: comparison between Winkler and elastic continuum solutions.
Gotechnique, 55(3): 461466. doi:10.1680/geot.2005.55.6.461.
Kondner, R.L. 1963. Hyperbolic stress-strain response: cohesive soils. Journal of
the Soil Mechanics and Foundations Division, ASCE, 89(SM1): 115143.
Lings, M.L., and Dietz, M.S. 2004. An improved direct shear apparatus for sand.
Gotechnique, 54(4): 245256. doi:10.1680/geot.2004.54.4.245.
Menetrey, P., and Willam, K.J. 1995. Triaxial failure criterion for concrete and its
generalization. ACI Structural Journal, 92(3): 311318.
Muhlhaus, H.B., and Vardoulakis, I. 1987. The thickness of shear bands in granular materials. Gotechnique, 37(3): 271283. doi:10.1680/geot.1987.37.3.271.
Olson, N.A. 2009. Soil performance for large scale soil-pipeline tests. Ph.D. thesis, Cornell University, Ithaca, N.Y.
O'Rourke, T.D. 2010. Geohazards and large, geographically distributed systems.
Rankine Lecture. Gotechnique, 60(7): 505543. doi:10.1680/geot.2010.60.7.
505.
Robert, D.J., and Soga, K. 2009. Simulations of soil-pipeline interactions in unsaturated soils. Report to Tokyo Gas Co. Ltd., Cambridge University.
Roscoe, K.H. 1970. The inuence of strains in soil mechanics. 10th Rankine
Lecture. Gotechnique, 20(2): 129170. doi:10.1680/geot.1970.20.2.129.
Sun, D., Yao, Y.-P., and Matsuoka, H. 2006. Modication of critical state models
by Mohr-Coulomb criterion. Mechanics Research Communications, 33(2):
217232. doi:10.1016/j.mechrescom.2005.05.006.
Thomas, H.G. 1978. Discussion of Soil restraint against horizontal motion of
pipes by J.M.E. Audibert and K.J. Nyman. Journal of the Geotechnical Engineering Division, ASCE, 104(GT9): 12141216.
Trautmann, C.H., and O'Rourke, T.D. 1983. Behavior of pipe in dry sand under
lateral and uplift loading. Geotechnical Engineering Report 83-7, Cornell
University, Ithaca, N.Y.
Trautmann, C.H., O'Rourke, T.D., and Kulhawy, F.H. 1985. Uplift forcedisplacement response of buried pipe. Journal of Geotechnical Engineering,
111(9): 10611076. doi:10.1061/(ASCE)0733-9410(1985)111:9(1061).
Vardoulakis, I., and Graf, B. 1985. Calibration of constitutive models for granular
materials using data from biaxial experiments. Gotechnique, 35(3): 299
317. doi:10.1680/geot.1985.35.3.299.
Vorster, T.E.B., Klar, A., Soga, K., and Mair, R.J. 2005. Estimating the effects of
tunneling on existing pipelines. Journal of Geotechnical and Geoenvironmental Engineering, 131(11): 13991410. doi:10.1061/(ASCE)1090-0241(2005)
131:11(1399).
White, D.J., Cheuk, C.Y, and Bolton, M.D. 2008. The Uplift resistance of pipes and
plate anchors buried in sand. Gotechnique, 58(10): 771779. doi:10.1680/geot.
2008.3692.
Wong, K.S., and Duncan, J.M. 1974. Hyperbolic stress-strain parameters for nonlinear nite element analyses of stresses of stresses and movements in soil
masses. Report No. TE-74-3, University of California, Department of Civil
Engineering, Berkeley, Calif.
Xie, X., Symans, M.D., O'Rourke, M.J., Abdoun, T.H., O'Rourke, T.D., Palmer, M.C.,
and Stewart, H.E. 2012. Numerical modeling of buried HDPE pipelines subjected to normal faulting: a case study. Earthquake Spectra. [In press.]
Yimsiri, S., Soga, K., Yoshizaki, K., Dasari, G.R., and O'Rourke, T.D. 2004. Lateral
and upward soil-pipeline interactions in sand for deep embedment conditions. Journal of Geotechnical and Geoenvironmental Engineering, 130(8):
830842. doi:10.1061/(ASCE)1090-0241(2004)130:8(830).
Zienkiewicz, O.C., Emson, C., and Bettess, P. 1983. A novel boundary innite
element. International Journal for Numerical Methods in Engineering, 19(3):
393404. doi:10.1002/nme.1620190307.

Published by NRC Research Press

Vous aimerez peut-être aussi