Академический Документы
Профессиональный Документы
Культура Документы
Abstract: Uplift resistance is a key parameter against upheaval buckling in the design of a buried pipeline. The mobilization of uplift
Downloaded from ascelibrary.org by Nottingham Trent University on 08/04/19. Copyright ASCE. For personal use only; all rights reserved.
resistance in dense sand is investigated in the present study based on finite-element (FE) analysis. The prepeak hardening, postpeak softening,
density-dependent, and confining pressure–dependent soil behavior are implemented in FE analysis. The uplift resistance mobilizes with
progressive formation of shear bands. The vertical inclination of the shear band is approximately equal to the maximum dilation angle at the
peak and then decreases with upward displacement. The force–displacement curves can be divided into three segments: prepeak, quick
postpeak softening, and gradual reduction of resistance at large displacements. Simplified equations are proposed for mobilization of uplift
resistance. The results of FE analysis, simplified equations, and model tests are compared. The importance of postpeak degradation of uplift
resistance to upheaval buckling is discussed. DOI: 10.1061/(ASCE)GT.1943-5606.0001957. © 2018 American Society of Civil Engineers.
Author keywords: Pipeline; Mohr-Coulomb model; Finite-element analyses; Dense sand; Upward movement.
Introduction (Dutta et al. 2015) or pipelines in ice gouging areas (Pike and
Kenny 2016).
Buried pipelines used for transporting oil usually operate at During installation of offshore pipelines in sand, ploughs de-
high temperature and pressure. Temperature-induced expansion, posit backfill soil in a loose to medium dense state (Cathie et al.
together with vertical out-of-straightness, might cause global up- 2005); however, it could be subsequently densified due to environ-
heaval buckling (UHB). Field evidence suggests that significantly mental loading. For example, Clukey et al. (1989) showed that the
large vertical upward displacement could occur in the buckled sec- sandy backfill of a test pipe section densified from relative density
tion and, in the worst cases, it might protrude above the ground sur- (Dr ) less than around 57% to around 85–90% in 5 months, which
face (Palmer et al. 2003). For example, Aynbinder and Kamershtein has been attributed to wave action at the test site in the Gulf of
(1982) showed that an approximately 70-m section of a buried Mexico. The uplift resistance offered by soil (Fv ) depends on up-
pipeline displaced vertically up to a maximum distance of around ward displacement (v) and generally consists of three components:
4.2 m above the ground surface. Sufficient restraint from the soil (1) submerged weight of soil being lifted (W s ); (2) vertical com-
above the pipeline could prevent excessive displacement and up- ponent of shearing resistance offered by the soil (Sv ); and (3) suc-
heaval buckling. Because burial is one of the main sources of tion under the pipe (Fsuc ). The component Fsuc could be neglected
pipeline installation cost, proper estimation of soil resistance is nec- for a drained loading condition at low uplift velocities (Bransby and
essary to select the burial depth—typically expressed as the embed- Ireland 2009; Wang et al. 2010).
ment ratio (H ~ ¼ H=D), where D is the diameter and H is the depth The force–displacement behavior is generally expressed in
of the center of the pipe. Pipelines embedded at 1 ≤ H ~ ≤ 4 in dense normalized form using N v ¼ Fv =γHD and v~ ¼ v=D, where γ is
sand are the focus of the present study, although it is understood the effective unit weight of soil, which is the dry unit weight in
that in some special scenarios H ~ could be outside this range, physical model tests and finite-element (FE) modeling of uplift
for example, for surface-laid offshore pipelines in deep water behavior presented in this study. Physical experiments showed that
N v increases with H~ and Dr (Trautmann 1983; Bransby et al. 2002;
Chin et al. 2006; Cheuk et al. 2008). A close examination of physi-
1
Pipeline Stress Specialist, Northern Crescent, Inc., 816 7 Ave. SW, ~ ≤ 4 shows that N v increases
cal model test results in dense sand at H
Calgary, AB, Canada T2P 1A1; formerly, Ph.D. Candidate, Faculty of quickly with v~ and reaches the peak (N vp ) at v~ of around
Engineering and Applied Science, Dept. of Civil Engineering, Memorial
0.01–0.05. A quick reduction of N v occurs after the peak, followed
Univ. of Newfoundland, St. John’s, NL, Canada A1B 3X5.
2
Professor and Research Chair in Seafloor Mechanics, Faculty of
~ The american lifelines alliance
by gradual reduction of N v at large v.
Engineering and Applied Science, Dept. of Civil Engineering, Memorial (ALA) guideline for design (ALA 2005) does not explicitly consider
Univ. of Newfoundland, St. John’s, NL, Canada A1B 3X5 (corresponding the postpeak reduction of N v , and the maximum N v ¼ ϕ 0 H=44 ~ is
0
author). Email: bipul@mun.ca recommended, where ϕ is a representative angle of internal friction
3
Associate Professor, Faculty of Engineering and Design, Dept. of (in degrees). However, det norske veritas (DNV 2007) recognized
Civil and Environmental Engineering, Carleton Univ., 1125 Colonel By the postpeak reduction of N v and recommended a N v –v~ relation us-
Dr., Ottawa, ON, Canada K1S 5B6. ing four linear line segments in which N v reduces linearly from the
4
Professor and Canada Research Chair in Infrastructure Engineering, peak to a residual value with v~ and then remains constant.
GeoEngineering Centre at Queen’s—Royal Military College, Queen’s
The force–displacement curves obtained from model tests
Univ., Kingston, ON, Canada K7L 4V1.
Note. This manuscript was submitted on April 17, 2017; approved on
evolve from complex deformation mechanisms and the stress–
April 26, 2018; published online on August 23, 2018. Discussion period strain behavior of soil above the pipe. To understand these mech-
open until January 23, 2019; separate discussions must be submitted for anisms, the particle image velocimetry (PIV) technique (White
individual papers. This paper is part of the Journal of Geotechnical et al. 2003) has been used in recent model tests (Cheuk et al. 2008;
and Geoenvironmental Engineering, © ASCE, ISSN 1090-0241. White et al. 2008; Thusyanthan et al. 2010; Wang et al. 2010).
0.5D
0.5D
Downloaded from ascelibrary.org by Nottingham Trent University on 08/04/19. Copyright ASCE. For personal use only; all rights reserved.
0.8D
≈ 2.2D
≈
4.2D 2.8D D
(a) (b)
Fig. 1. Finite-element modeling: (a) finite-element mesh; and (b) mobilized friction and dilation angles.
the soil around the pipe. With loading, the soil elements around the silica sand was pulled up slowly at 10 mm=h to capture soil de-
pipe experience high shear strains that cause a reduction of ϕ 0 . formation using two digital cameras. However, in FE modeling,
Therefore, assuming a looser soil condition, μ ¼ 0.32 is used. the pipe is pulled at approximately 10 mm=s by maintaining
Furthermore, μ has a little influence on the uplift resistance and quasi-static simulation conditions.
μ ¼ 0.2–0.6 gives less than 2% variation in the peak resistance. Direct shear tests show that Leighton Buzzard (LB) silica sands
The numerical analysis is conducted in two steps. In the first step, has ϕc0 of 32° (Cheuk et al. 2008). Because ϕc0 in PS condition could
geostatic stress is applied under K 0 ¼ 0.5, where K 0 is the at-rest be around 2°–4° higher than in direct shear conditions (Lings and
earth pressure coefficient. The value of K 0 does not significantly Dietz 2004), ϕc0 ¼ 35° is used, which is around 3° higher than DS
affect the uplift resistance in FE analysis (Jung et al. 2013). tests results reported by Cheuk et al. (2008). For quartz and sili-
In the second step, the pipe is displaced up by specifying a displace- ceous sands, Q is around 10 1 (Bolton 1986; Randolph et al.
ment boundary condition at the reference point (center of the pipe). 2004). Although the values are within this range, Chakraborty
The MMC model is implemented in Abaqus by developing a and Salgado (2010) showed a trend of increasing Q with initial
user subroutine VUSDFLD written in FORTRAN. The stress and confining pressure (<196 kPa). In this study, Q ¼ 10 and R ¼ 1 are
strain components are called in the subroutine in each time incre- used. Bolton (1986) suggested Aψ ¼ 5 and kψ ¼ 0.8 for PS con-
ment. The mean effective stress (p 0 ) is calculated from the three ditions based on analysis of a large number of laboratory tests
principal stresses. The strain components are transferred to the prin- results on different sands. Roy et al. (2016) calibrated the present
cipal strain components and stored as state variables. The plastic MMC model against laboratory test results on Cornell filter (CF)
strain increment (Δγ p ) in each time increment is calculated as sand and obtained the values of C1 , C2 , and m to model the varia-
Δγ p ¼ ðΔεp1 − Δεp3 Þ, where Δεp1 and Δεp3 are the major and minor tion of ϕ 0 and ψ with γ p , and then conducted a FE simulation of the
principal plastic strain components, respectively. The value of γ p physical model tests of Trautmann (1983). Cheuk et al. (2008) did
is calculated as the sum of Δγ p over the period of analysis. In not provide any stress–strain curve of LB sand used in the physical
the subroutine, γ p and p 0 are defined as two field variables. The modeling. Both of these physical model test programs used
mobilized ϕ 0 and ψ are defined in the input file as a function of
uniform/poorly graded sand, although the mean particle size (D50 )
γ p and p 0 in tabular form, using the equations in Table 1. During
of the coarse fraction of LB sand is larger (D50 of around 2.24 mm)
the analysis, the program accesses the subroutine and updates the
in Cheuk et al. (2008) than that of CF sand (D50 of around 0.5 mm)
values of ϕ 0 and ψ with field variables. Although I D is not updated
in Trautmann (1983). However, based on laboratory test results,
in each time increment, the volumetric change in soil elements
Cheuk et al. (2008) recognized a minimal influence of particle size
due to shearing and its effects on ϕ 0 and ψ are captured in the
on frictional characteristics of LB sands—the peak and critical state
MMC model.
friction angles are 52° and 32°, respectively, for a coarse and a
fine fraction of LB sand. Furthermore, in Cheuk et al. (2008),
Model Verification the force–displacement curves for the coarse and fine fractions
of LB sands were similar, including the peak and postpeak degra-
A FE simulation is first performed for a physical model test con- dation. Therefore, in the present study, the values of C1 , C2 , and m
ducted by Cheuk et al. (2005, 2008) at the University of Cambridge of LB sand are assumed to be the same as CF sand. Table 2 provides
and is called the CD (coarse dense sand) test. A 100-mm-diameter the geotechnical parameters used in FE analyses. Fig. 1(b) shows
model pipe section embedded at H ~ ¼ 3 in dry Leighton Buzzard the typical variation of ϕ 0 and ψ with plastic shear strain.
Relative density, Dr (%) 92 highly concentrated γ p zone. White et al. (2008) suggested that
Unit weight, γ (kN=m3 ) 16.87 θ ∼ ψp when the peak resistance is mobilized. Because ψp varies
Interface friction coefficient, μ 0.32
~ with p 0 (Table 1), they calculated a single representative value of
Depth of pipe, H 3 (1, 1.5, 2.0, 2.5, 3.5, 4.0)
the peak dilation angle (ψRp ) using the in situ p 0 at the springline of
Note: Numbers in parenthesis in right column show the values used for the pipe [ð1 þ 2K 0 ÞγH=3]. For the geotechnical parameters listed
parametric study.
in Table 2, ψRp ¼ 25°, which is approximately the same as θ ob-
tained from the present FE analysis [Fig. 3(a)]. The complete slip
mechanism develops at v~ > v~ p when a considerable postpeak deg-
4 radation of N v occurs [Fig. 3(b)]. Similar types of curved failure
A' planes shown in Figs. 3(b–e) were also observed in model tests
B' (Stone and Newson 2006; Cheuk et al. 2008; Huang et al. 2015).
3.5 The formation of complete slip planes after v~ p can be attributed
ALA 2005 to noticeable vertical displacement of the ground surface after
A C'
3 =45 N vp in model tests (Dickin 1994; Bransby et al. 2002; Huang
B D' =40 et al. 2015).
Normalized force, Nv
2.5 DNV 2007 E' Although it is a different type of loading, because of progressive
development of shear bands, the attainment of peak load before the
C D formation of a complete failure mechanism was also found in
2 E model tests and numerical modeling for footing in dense sand
(Tatsuoka et al. 1991; Aiban and Znidarčić 1995; Loukidis and
1.5 Salgado 2011). However, in the simplified limit equilibrium
method (LEM), a complete slip mechanism is assumed to calculate
1 the peak load irrespective of burial depth; for example, White et al.
(2008) used the LEM to fit test data for H ~ < 8.0.
Cheuk et al. 2008 The slight increases in N v in Segment CD in Fig. 2 can be ex-
0.5 Present FE analysis (MMC) plained using γ p plots in Figs. 3(a–d). In Segment ABC of the
Present FE analysis (MC)
N v –v~ curve, the shear resistance (τ f ) gradually reduces along
0 the inclined shear band that was formed during initial upward dis-
0 0.2 0.4 0.6 0.8 1
placement [e.g., Figs. 3(a–c)]. However, the location of the shear
Normalized displacement, band shifts considerably to the right at v~ of around 0.18–0.4. Be-
cause the new shear bands form through the soil where τ f has not
Fig. 2. Comparison between FE simulation and model test results. been reduced by softening, N v increases slightly in Segment CD.
After Point D, the location of the shear band does not change sig-
nificantly with v~ (θ remains at around 8°). Therefore, the gradual
Force–Displacement Behavior decreases of N v with v~ after Point D is due to strain softening in the
shear band and the reduction of soil cover depth.
Fig. 2 shows the FE simulated force–displacement curves for Fig. 2 also shows that an FE simulated N v –v~ curve with the
H~ ¼ 3, on which the points of interest for further explanation are MMC model compares well with the model test results of Cheuk
labeled A–E for the MMC and A 0 –E 0 for the MC model. Adaptive et al. (2008). A slight increase in N v after a quick postpeak reduc-
meshing could not maintain a high-quality mesh at a very large pipe tion is also observed in model tests at intermediate depth of embed-
displacement. Therefore, the force–displacement curves only up ment, as in the one shown in Fig. 2 and also in other studies
to v~ ¼ 1.0 are presented in this study. For MMC, N v increases (Bransby et al. 2002; Stone and Newson 2006; Chin et al. 2006;
quickly and reaches the peak at v~ of around 0.03 and then quickly Cheuk et al. 2008; Saboya et al. 2012; Eiksund et al. 2013; Huang
decreases to Point C, primarily due to the strain-softening behavior et al. 2015). However, it does not happen at shallow burial depths.
of soil. After a slight increase between Points C and D, N v de- A similar trend is also observed in model tests for the bearing
creases again at a slower rate than in Segment AC. In the present capacity of footing in sand, which has been attributed to
study, Segment AC of the N v –v~ curve is termed the softening progressive formation of slip planes (Aiban and Znidarčić 1995).
segment, and the segment after Point C is called the large defor- The vertical inclination of the shear band gradually reduces with
mation segment. The values of N v at the peak and after softening ~ and at v~ ¼ 0.32, θ is around 8° [Fig. 3(c)]. However, θ does
v,
(i.e., Points A and C) are defined as N vp (¼ Fvp =γHD) and N vs not reduce further at v~ > 0.32 [Figs. 3(c–e)]. As discussed sub-
(¼ Fvs =γHD), respectively, where Fvp and Fvs are the peak and sequently, in the limit analysis θ ¼ 0 is assumed at large v; ~
Fig. 3. Shear band formation: (a–e) modified Mohr-Coulomb model corresponding to the points A–E marked at Fig. 2; and (f–j) Mohr-Coulomb
model corresponding to the points A'–E' marked at Fig. 2.
however, the present FE analysis shows that the shear band does not simulation with the MMC model and physical model test results.
become completely vertical even at large v~ (e.g., v~ ¼ 0.5). Because In order to explain this force–displacement behavior, γ p at five v~
of change in mobilized ϕ 0 and ψ with loading, the failure mecha- is plotted in Figs. 3(f–j). The inclination of the shear band (θ)
nism changes from an inclined slip plane [Fig. 3(b)] to a flow remains almost constant (around 25°) during the whole process
around mechanism [Fig. 3(e)], which is also indicated by the veloc- of upward displacement of the pipe. The linear postpeak reduction
ity vectors in the inset of Fig. 2. Based on PIV results, similar fail- of N v with the MC model is due to the reduction of cover depth
ure mechanisms have been reported from physical experiments with v. ~
(Bransby et al. 2002; Cheuk et al. 2008). In summary, the postpeak reduction of N v with the MMC model
for this burial depth occurs due to the combined effects of three
factors: (1) decreases in size of the failure wedge, (2) reduction
Limitations of Mohr-Coulomb Model of shear resistance with γ p , and (3) reduction of cover depth.
To show the advantages of the MMC model, FE simulation is also The MC model cannot capture the effects of the former two. How-
performed with the MC model. Based on Cheuk et al. (2005, 2008) ever, the proposed MMC model can simulate the effects of all three
laboratory test results ϕ 0 ¼ 52° and ψ ¼ 25° are used for the MC factors. Moreover, the simulations with the MMC model are similar
model. Although not explicitly mentioned in the design guidelines, to physical model test results.
equivalent values for these two parameters should be carefully se- DNV (2007) suggested the following equations to develop the
lected because they vary with γ p . In general, the equivalent values force–displacement curve for dense sand for 2.5 ≤ H ~ ≤ 8.5 where
of ϕ 0 and ψ should be smaller than the peak and higher than the N vp ¼ 1 þ fH; ~ N vs ¼ 1 þ αf fH; ~ v~ p ¼ ð0.5%–0.8%ÞH; ~ and
critical state values. A number of previous studies simulated pipe– v~ s ¼ 3v~ p . The prepeak behavior is defined by a bilinear relation,
soil interaction using constant equivalent values of ϕ 0 and ψ for the where the slope changes at ðβ v~ p ; αN vp Þ. Based on the DNV (2007)
MC model (e.g., Yimsiri et al. 2004). An equivalent ϕ 0 has also recommendations for dense sand, f ¼ 0.6, αf ¼ 0.75, v~ p ¼
been recommended for other geotechnical problems in dense sand, 0.008H, ~ α ¼ 0.75, and β ¼ 0.2; the force–displacement curve is
for example, the bearing capacity of shallow foundations (Loukidis plotted in Fig. 2. Although only one test is simulated, DNV
and Salgado 2011) and the lateral capacity of pile foundations (2007) gave considerably lower N vp, higher N vs, and lower v~ s than
(API 1987). the physical model test and present FE results with the MMC
Fig. 2 shows that the MC model calculates slightly higher N vp model.
than the MMC model. This difference will be reduced if lower The maximum N v based on ALA (2005) (¼ ϕ 0 H=44) ~ is shown
equivalent values of ϕ 0 and ψ are considered. However, the by two horizontal arrows on the right vertical axis for two ϕ 0 . ALA
key observation is that N v decreases almost linearly with v~ after (2005) requires a constant equivalent ϕ 0 and does not consider any
the peak for the MC model, which is very different from the postpeak reduction of resistance.
3
0.01H from model tests on dense sands. For the range of soil prop-
2.5 erties and burial depths considered in the present FE analysis, v~ p
does not vary significantly with H ~ between 1 and 4. However, FE
2 simulations show a significant increase in v~ p with H ~ for deep burial
conditions (Roy et al. 2018). Fig. 5 shows that N vp for the MMC
1.5
model increases almost linearly with H. ~ Moreover, N vp obtained
1 from the present FE analysis is comparable with available physical
model tests and FE results.
Downloaded from ascelibrary.org by Nottingham Trent University on 08/04/19. Copyright ASCE. For personal use only; all rights reserved.
Fig. 4. Comparison between simplified equations and FE results for Proposed Simplified Equations for Uplift
~
different H. Force–Displacement Curve
The solid lines in Fig. 4 show the proposed N v –v~ relation for sim-
plified analysis, which consists of a bilinear curve up to N vs fol-
lowed by a slightly nonlinear curve at large displacements. DNV
9 (2007) recommended that N v remains constant after N vs (Fig. 2).
Present FE analysis (MMC, D=100mm) [90%] The parameters required to define the proposed N v –v~ relation are
Wang et al., 2010 (Centrifuge, D=258mm) [Saturated, 85%]
Trautmann, 1983 (Laboratory test, D=102mm)[80%] Fvp , vp , Fvs , and vs .
8
Bransby et al.,2002 (Laboratory test, D=48mm)
Dickin, 1994 (Centrifuge, D=1000mm) [77%]
Chin et al., 2006 (Centrifuge, D=190mm) [85%] Peak Resistance
7 Cheuk et al., 2008 (Centrifuge, D=100mm) [92%]
Saboya et al., 2012 (Centrifuge, D=500 mm) [70%] Depending on slip plane formation, inclined and vertical slip
Yimsiri et al., 2004 (FE, MC, D=102mm)[80%]
plane models are commonly used to calculate uplift resistance
Peak dimensionless force, Nvp
cated on the right side of this line and curve outward near the Moreover, R is almost independent of pipe diameter. The overesti-
ground surface. Therefore, the weight of the lifted soil wedge is mation of uplift resistance in LEM is significant at large embed-
less in FE simulations than the LEM, especially for a large H ~ ment ratios—for example, the LEM calculates an approximately
(e.g., H~ ¼ 4). Moreover, although ϕ 0 ¼ ϕp0 is used in the LEM, 22% higher peak resistance than FE calculated value for H ~ ¼ 4.
this is valid only for a small segment of the slip plane [e.g., near
Point A in Fig. 6(a) for H ~ ¼ 3]. Below this point, ϕ 0 < ϕp0 because
Uplift Resistance after Initial Softening
the large plastic shear strain (γ p ) causes strain softening. Above this
point, γ p is not sufficiently large (i.e., γ p < γ pp ) to mobilize ϕp0 ; Similar to Eq. (3), a simplified equation is proposed for the uplift
therefore ϕ 0 is less than ϕp0 also in this segment of the slip plane. force after initial softening, Fvs [Eq. (4)]. At a large displacement,
The ratio between the prepeak and postpeak segments of the slip the failure planes reach the ground surface [Fig. 3(c)] and therefore
plane increases with embedment ratio. R ¼ 1 is used. Because significant strain softening occurs, ϕ 0 along
Overestimation of W s and ϕ 0 gives a higher Fvp in the LEM the slip planes reduces almost to ϕc0 . Considerable ground surface
(Fvp LEM ) than FE simulation (Fvp FE ). In order to investigate this heave occurs at this stage [Fig. 3(c)], which increases with pipe
(a)
1
Reduction factor for peak resistance, R
0.95
0.9
0.85
0.8
0.75
0.7
D=100 mm
0.65 D=300 mm
D=500 mm
0.6
1 1.5 2 2.5 3 3.5 4
(b)
Fig. 6. Effect of burial depth on peak resistance: (a) soil failure; and (b) reduction factor R.
6 6
MMC result (N
(Nvp) Nvp
Nvp (R=1)
vp)
Nvpp
Nvs
MMC result (N
(Nvpp)
vs) DNV (2007) [f=0.6, 0.45] Weight
W component of Nvpp
s of Nvs
5 5
Shear
Sv of Ncomponent
vs of Nvpp
Eq. (3) R=1.0 Weight
W component of Nvp
s of Nvp
Shear
Sv of Ncomponent
vp of Nvp
Normalized force, Nv
4 4
Normalized force, Nv
Eq. (3)
3 3
2 2
Eq. (4)
1 1
Fig. 7. Performance of simplified equations: (a) comparison with FE analysis; and (b) contribution of weight and shear components.
2.5
Uplift force, Fv
1.5
1
Downloaded from ascelibrary.org by Nottingham Trent University on 08/04/19. Copyright ASCE. For personal use only; all rights reserved.
0.5
0
-4 -3.5 -3 -2.5 -2 -1.5 -1 -0.5
(a) (b)
Fv A
B′
C
O vp vs ~H v
(c)
displacements, which is shown schematically by the dashed line coating thickness (tc ) of 50 mm, steel yield strength (σy ) of
(BC) in Fig. 8(c). These processes could not be simulated using 448 MPa, and steel thermal expansion coefficient (α) of 11 ×
the present numerical technique. Therefore, for structural response 10−6 °C−1 . The pipe is buried in dense sand (Dr ¼ 90% and
of the pipeline presented in the following sections, the post- γ 0 ¼ 10 kN=m3 ) at an embedment ratio (H) ~ of 3. The density
peak segment of the force–displacement curve is defined by AB 0 C of steel, concrete, seawater, and oil in the pipe are 7,850, 2,800,
[Fig. 8(c)], where Fv at B 0 is calculated using Eq. (4) without heave 1,025, and 800 kg=m3 , respectively, which gives submerged pipe
and it mobilizes at v ¼ vs . weight (oil-filled) of 1.6 kN=m.
Wang et al. (2012) showed that the postpeak segments of the To initiate upheaval buckling response associated with increas-
uplift curves for loose sand for varying burial depths tend to follow ing oil temperature (T), two initial imperfection ratios (v0 =L0 ) of
a backbone curve similar to Eq. (4). There is only one postpeak 0.005 (v0 ¼ 0.16 m and L0 ¼ 31.56 m) and 0.011 (v0 ¼ 0.45 m
segment in loose sand. However, an Fv –v~ curve for dense sand and L0 ¼ 41.05 m) are considered, where v0 is the maximum ini-
has two postpeak segments—a quick reduction of Fv just after tial vertical imperfection and L0 is the initial imperfection length.
the peak, followed by the gradual reduction after v~ s. Fig. 8(a) shows The initial shape of the pipe is defined using Taylor and Tran
that, for dense sand, the postpeak segments even after Fvs do not lie (1996)’s empathetic model. A 3,500-m-long pipe is simulated to
on a unique line. avoid boundary effects in the buckled section. The modified Riks
method is used to capture any snap-through buckling response that
Effect of Postpeak Degradation of Uplift Resistance may occur (Abaqus; Liu et al. 2014).
on Upheaval Buckling The force–displacement behavior of soil is defined using three
sets of nonlinear independent spring formulations that do not con-
Finite-element analysis is performed to investigate the structural sider load coupling or interaction (e.g., Kenny and Jukes 2015). For
response of a steel pipeline having the following properties: outside the modeling of upward resistance, two types of force–displacement
diameter (D) of 298.5 mm, wall thickness (t) of 12.7 mm, concrete relations are used. In Model 1, the Fv –v relation is defined as
Ty v0/L0 ~0.011
• The angle of inclination of the slip planes to the vertical (θ) is
approximately equal to the peak dilation angle when the peak
0 resistance mobilizes. However, it decreases with upward displa-
0 1 2 3 4
cement due to decreases in the dilation angle. The angle θ
Buckle amplitude, vm (m)
significantly influences the weight of the soil wedge and thereby
Fig. 9. Effect of postpeak reduction of uplift resistance on permissible uplift resistance.
temperatures (solid lines for Model 1 and dashed lines for Model 2). • Uplift resistance at large displacement does not remain constant
but decreases with upward displacement.
• Unlike peak displacement, displacement required to complete
initial softening increases significantly with the H=D ratio.
• Postpeak reduction of uplift resistance could significantly
OAB 0 C, as shown in Fig. 8(c). Using Eqs. (3) and (4), respectively, reduce the permissible temperature during operation.
the uplift resistances at Points A (9.14 kN=m) and B 0 (5.16 kN=m)
are calculated with vp ¼ 9.3 × mm and vs ¼ 61.5 mm, as dis-
cussed previously. The Model 2 is same as the Model 1 but Acknowledgments
without postpeak degradation where Fv remains constant after
Point A (i.e., elastic and perfectly plastic behavior). Based on ALA The works presented in this paper have been supported by the
(2005), the axial and vertical downward soil resistances of 4.62 Research and Development Corporation of Newfoundland and
and 607.5 kN=m, respectively, are calculated, which mobilize at Labrador, Chevron Canada Limited, and the Natural Sciences
3 and 30 mm displacements, respectively. and Engineering Research Council of Canada (NSERC).
Fig. 9 shows the variation of temperature increase with the
maximum buckle amplitude (vm ). For both v0 =L0 ratios, the T–vm
curve with postpeak reduction is below that without any reduction. Notation
Previous studies suggested a number of permissible temperature
The following symbols are used in this paper:
increase criteria including (1) the critical (T c ) and safe (T s ) temper-
ature for snap-through buckling response (represented by the circle Aψ = slope of (ϕp0 − ϕc0 ) versus I R curve;
and square symbols in Fig. 9); and (2) temperature required for the C1 , C2 = material constants;
onset of first yield (T y ) for stable buckling (i.e., maximum D = pipe diameter;
stress ¼ σy ) (Hobbs 1984; Taylor and Gan 1986). In this study, Dr = relative density;
the maximum stress is calculated from axial stress and bending E = Young’s modulus;
moment obtained from the numerical simulations. For the snap- Fsuc = suction force under the pipe;
through buckling response case (v0 =L0 ¼ 0.005), Fig. 9 shows Fv = uplift force;
the reduction of T c and T s of 10°C and 23°C, respectively, when Fvp = peak uplift force;
the postpeak reduction in uplift resistance is considered. For the Fvp FE = Fvp calculated by FE;
stable buckling case (v0 =L0 ¼ 0.011), the postpeak reduction
Fvp LEM = Fvp calculated by LEM;
could decrease T y by 17°C. Previous studies also recognized the
Fvs = after softening uplift force;
importance of postpeak reduction of uplift resistance and suggested
using full force–displacement curve considering large vertical H = distance from ground surface to the center of pipe;
H~ = embedment ratio (¼ H=D);
displacements (Klever et al. 1990; Goplen et al. 2005; Wang
et al. 2009). I D = relative density/100;
I R = relative density index;
K = material constant;
Conclusions K 0 = at-rest earth pressure coefficient;
kψ = slope of (ϕp0 − ϕc0 ) versus ψp curve;
The uplift behavior of buried pipelines in dense sand have been
L0 = initial imperfection length;
investigated using finite-element modeling. The stress–strain
m = material constant;
behavior of soil was modeled using a MMC model that considers
the variation of angles of internal friction (ϕ 0 ) and dilation (ψ) with N v = normalized uplift force;
plastic shear strain, density, and confining pressure as observed in N vp = normalized peak uplift force;
laboratory tests on dense sand. Comparison with a model test result N vs = value of N v after softening;
showed that force–displacement, soil deformation, and failure n = exponent;
mechanisms can be explained from the variation of ϕ 0 and ψ with p 0 = mean effective stress;
γ p = engineering plastic shear strain; Chin, E. L., W. H. Craig, and M. Cruickshank. 2006. “Uplift resistance of
γ pc = strain-softening parameter; pipelines buried in cohesionless soil.” In Vol. 1 of Proc., 6th Int. Conf.
on Physical Modelling in Geotechnics, edited by C. W. W. Ng, L. M.
γ pp = γ p required to mobilize ϕp0 ;
Zhang, and Y. H. Wang, 723–728. London: Taylor & Francis.
Δγ p = plastic strain increment; Clukey, E. C., C. R. Jackson, J. A. Vermersch, S. P. Koch, and W. C. Lamb.
Δεp1 = major principal plastic strain increment; 1989. “Natural densification by wave action of sand surrounding a
Δεp3 = minor principal plastic strain increment; buried offshore pipeline.” In Proc., Offshore Technology Conf.
ϵ̇pij = plastic deviatoric strain rate; Richardson, TX: Society of Petroleum Engineers.
μ = friction coefficient between pipe and soil; Dickin, E. A. 1994. “Uplift resistance of buried pipelines in sand.” Soils
Found. 34 (2): 41–48. https://doi.org/10.3208/sandf1972.34.2_41.
τ f = shear resistance along the shear band;
DNV (Det Norske Veritas). 2007. Global buckling of submarine
v~ = normalized upward displacement of pipe (¼ v=D); pipelines—Structural design due to high temperature/high pressure.
v~ p = v~ required to mobilize N vp ; DNV-RP-F110. Baerum, Norway: Det Norske Veritas.
v~ s = v~ required to mobilize N vs ; Dutta, S., B. Hawlader, and R. Phillips. 2015. “Finite element modeling of
θ = inclination of slip plane to the vertical; partially embedded pipelines in clay seabed using Coupled Eulerian-
ϕ 0 = mobilized angle of internal friction; Lagrangian method.” Can. Geotech. J. 52 (1): 58–72. https://doi.org/10
.1139/cgj-2014-0045.
ϕc0 = critical state friction angle;
Eiksund, G., H. Langø, and E. Øiseth. 2013. “Full-scale test of uplift re-
ϕin0 = ϕ 0 at the start of plastic deformation; sistance of trenched pipes.” Int. J. Offshore Polar Eng. 23 (4): 298–306.
ϕp0 = peak friction angle; Farhadi, B., and R. C. K. Wong. 2014. “Numerical modeling of pipe-soil
ϕμ = pipe–soil interface friction angle; interaction under transverse direction.” In Proc., Int. Pipeline. Conf.,
ψ = mobilized dilation angle; 2014–33364. New York: American Society of Mechanical Engineers.
ψp = peak dilation angle; and Goplen, S., P Strom, E. Levold, and J. Mork. 2005. “Hotpipe jip: HP/HT
buried pipelines.” In Proc., 24th Int. Conf. on Ocean, Offshore and
ψRp = representative value of the maximum dilation angle.
Arctic Engineering. New York: American Society of Mechanical
Engineers.
Hobbs, R. E. 1984. “In-service buckling of heated pipelines.” J. Transp.
References Eng. 110 (2): 175–189. https://doi.org/10.1061/(ASCE)0733-947X
(1984)110:2(175).
Aiban, S. A., and D. Znidarčić. 1995. “Centrifuge modeling of bear- Huang, B., J. Liu, D. Ling, and Y. Zhou. 2015. “Application of particle
ing capacity of shallow foundations on sands.” J. Geotech. Eng. image velocimetry (PIV) in the study of uplift mechanisms of pipe
121 (10): 704–712. https://doi.org/10.1061/(ASCE)0733-9410(1995) buried in medium dense sand.” J. Civ. Struct. Health Monit. 5 (5):
121:10(704). 599–614. https://doi.org/10.1007/s13349-015-0130-y.
ALA (American Lifelines Alliance). 2005. “Guidelines for the design of Jung, J. K., T. D. O’Rourke, and N. A. Olson. 2013. “Uplift soil-pipe in-
buried steel pipe.” Accessed March 13, 2017. https://www.american teraction in granular soil.” Can. Geotech. J. 50 (7): 744–753. https://doi
lifelinesalliance.com/pdf/Update061305.pdf. .org/10.1139/cgj-2012-0357.
API (American Petroleum Institute). 1987. Recommended practice for Kenny, S., and P. Jukes. 2015. “Pipeline/soil interaction modelling in
planning, designing and constructing fixed offshore platforms: API support of pipeline engineering design and integrity.” In Oil and gas
Recommended practice, 2A (RP 2A). 17th ed. Washington, DC: API. pipelines: Integrity and safety handbook, edited by R. W. Revie, 93.
Aynbinder, A. B., and A. G. Kamershtein. 1982. Raschet magistral’nykh New York: Wiley.
truboprovodov na prochnost’ i ustoichivost’ [Calculation of trunk pipe Klever, F. J., L. C. Van Helvoirt, and A. C. Sluyterman. 1990. “A dedicated
for strength and stability]. [In Russian.] Moscow, Russia: Nedra finite-element model for analyzing upheaval bucking response of sub-
Publishers. marine pipelines.” In Proc., 22nd Annual Offshore Technology Conf.,
Bolton, M. D. 1986. “The strength and dilatancy of sands.” Géotechnique 529–538. Texas: Society of Petroleum Engineers.
36 (1): 65–78. https://doi.org/10.1680/geot.1986.36.1.65. Lings, M. L., and M. S. Dietz. 2004. “An improved direct shear apparatus
Bransby, M. F., and J. Ireland. 2009. “Rate effects during pipeline upheaval for sand.” Géotechnique 54 (4): 245–256. https://doi.org/10.1680/geot
buckling in sand.” Proc. Inst. Civ. Eng. Geotech. Eng. 162 (5): .2004.54.4.245.
247–256. https://doi.org/10.1680/geng.2009.162.5.247. Liu, R., H. Xiong, X. L. Wu, and S. W. Yan. 2014. “Numerical studies on
Bransby, M. F., T. A. Newson, and M. C. R. Davies. 2002. “Physical mod- global buckling of subsea pipelines.” Ocean Eng. 78: 62–72. https://doi
elling of the upheaval resistance of buried offshore pipelines.” In Proc., .org/10.1016/j.oceaneng.2013.12.018.
Int. Conf. on Physical Modelling in Geotechnics. Boca Raton, Florida: Loukidis, D., and R. Salgado. 2011. “Effect of relative density and stress
CRC press, Taylor & Francis group. level on the bearing capacity of footings on sand.” Géotechnique 61 (2):
Cathie, D. N., C. Jaeck, J.-C. Ballard, and J.-F. Wintgens. 2005. “Pipeline 107–119. https://doi.org/10.1680/geot.8.P.150.3771.
geotechnics: State-of-the-art.” In Proc., Int. Symp. on Frontiers in Merifield, R. S., S. W. Sloan, A. J. Abbo, and H. S. Yu. 2001. “The ultimate
Offshore Geotechnics, 95–114. Boca Raton, FL: CRC Press. pullout capacity of anchors in frictional soils.” In Proc., 10th Int. Conf.