Engineering Structures 132 (2017) 139151

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Design of steel pipe-jacking based on buckling analysis by finite strip

method
Liang Zhen a, Pizhong Qiao a,b,, Junbin Zhong c, Qingyuan Chen a, Jin-Jian Chen a, Jian-Hua Wang a
a
State Key Laboratory of Ocean Engineering, Collaborative Innovation Center for Advanced Ship and Deep-Sea Exploration, School of Naval Architecture, Ocean and Civil
Engineering, Shanghai Jiao Tong University, Shanghai 200240, PR China
b
Department of Civil and Environmental Engineering, Washington State University, WA 99164-2910, USA
c
Shanghai Municipal Engineering Design Institute (Group) Co. Ltd., Shanghai 200092, PR China

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

Article history: In practice, the steel pipe-jacking can be regarded as a thin-walled cylindrical shell mainly subjected to
Received 28 May 2016 jacking force in the axial direction and surrounded by the soil which is usually simplified and modeled as
Revised 22 August 2016 an elastic foundation. In this paper, the elastic buckling behavior of steel jacking pipes primarily under
Accepted 5 November 2016
axial compression and with the Pasternak foundation is analyzed by the finite strip method (FSM). The
elastic foundation is considered in the stiffness matrix through the strain energy, and the deformation
in the longitudinal direction is simulated by the series functions in FSM. A parametric study is conducted
Keywords:
to analyze buckling of cylindrical shells embedded in different elastic foundations. It indicates that the
Buckling design
Steel pipe-jacking
Pasternak foundation is more conducive to prevent buckling of cylindrical shells under axial compression.
Cylindrical shells The critical length and the lower bound of buckling loads are obtained, and they offer the basis for opti-
Compression mal design of steel pipe-jacking. Finally, the case study combined with the buckling accident in the steel
Elastic foundation pipe-jacking event is presented. The present buckling analysis of soil-embedded cylindrical shells under
Finite strips axial compression provides design guidance for steel pipe-jacking construction.
2016 Elsevier Ltd. All rights reserved.

1. Introduction The construction techniques for underground pipelines can be

Because of advantageous properties, such as high strength, high
plasticity (good performance of deformation), good self-sealing
and low construction cost, steel pipes are increasingly used in
underground pipeline engineering. Pipe-jacking is the technique
for installing pipelines through the use of the hydraulic jacking
of a pipe string from a launch shaft to a receiving shaft [1], as
shown in Fig. 1. To meet the growing demand for infrastructure
construction, the sizes (e.g., both the diameter and length) of
pipelines continuously increase and the geometrical scale of steel
pipe-jacking in some of these new construction projects have even
exceeded the existing engineering standards. In general, design for
these projects is just based on experience without considering size
effect. As a consequence, buckling problems are more likely to
occur in the pipelines of large diameter and long length, which
may endanger engineering safety.
Corresponding author at: State Key Laboratory of Ocean Engineering, Collab-
orative Innovation Center for Advanced Ship and Deep-Sea Exploration, School of
Naval Architecture, Ocean and Civil Engineering, Shanghai Jiao Tong University,
Shanghai 200240, PR China.
E-mail addresses: qiao@sjtu.edu.cn, qiao@wsu.edu (P. Qiao).
divided into two types: the buried pipe method and the pipe-
jacking method. Local buckling occurs more frequently in the bur-
ied pipeline due to high water and earth pressure. Upheaval buck-
ling as another form appears in the buried pipe construction due to
the reason of temperature change of its surroundings or the effect
of buoyancy. While for the jacking pipeline, the jacking force is
usually dominant among all the external forces. It mainly leads
to global buckling (including upheaval buckling), particularly for
relatively long pipelines. If water and earth pressure is high, the
jacking force will also induce the occurrence of local buckling [1].
The buckling modes are the primary difference between the buried
pipeline and the jacking pipeline. Especially, the huge jacking force
acted on a string of pipes in axial direction easily causes unpre-
dictable buckling accidents.
Steel jacking pipes can be regarded as a kind of elastic or elasto-
plastic cylindrical shells. The mechanism of buckling of cylindrical
shells under the axial compression is still very complicated, though
many researchers have contributed to this work. The buckling of
underground steel pipelines involves the deflection of pipelines
and soil. The surrounding soil is not only as a load acted on pipeli-
nes, but also provides resistance to prevent pipeline from deform-
ing outward. The surrounding soil is generally considered as an

http://dx.doi.org/10.1016/j.engstruct.2016.11.016
0141-0296/ 2016 Elsevier Ltd. All rights reserved.
140 L. Zhen et al. / Engineering Structures 132 (2017) 139151
Ground Surface
Reception Shaft
Tunnelling Machine
Fig. 1. Diagram of pipe-jacking [1].
elastic foundation in modeling and design analysis, so that the
effect of pipe-soil interaction is not neglected.
Many researchers have studied the buckling problem of a
cylindrical shell embedded in an elastic medium. Forrestal and
Herrmann [2] conducted early research about the stability of a
cylindrical shell embedded in an elastic medium and loaded by a
far-field hydrostatic pressure [3]. Luscher [4] investigated the
failure of the flexible soil-surrounded tubes using a combined
experimental and theoretical approach. Soil was assumed to be a
thick-walled cylinder surrounding tubes. The study indicated that
buckling rather than compressive yielding might be the controlling
mode of failure for thin-walled, smooth metallic tubes supported
evenly, though the medium-dense sand and the surrounding soil
increased the buckling resistance of a flexible tube tremendously
over that of an unsupported tube. Duns and Butterfield [5] devel-
oped a simple theoretical prediction of buckling load for cylinders
buried in an ideal elastic medium. The solution was derived from
Donnells stability equation for cylindrical shells extended to
include the effect of the surrounding medium. Yun and Kyriakides
[6] analyzed buckling of buried pipelines under compressive loads
induced by seismic action through beam and shell modes. Che-
ney [7] built a 2-D model and assumed that the Winkler spring
constant was taken as a function of the mode number in buckling.
The solution represented an upper bound on local buckling of bur-
ied flexible tubes that might also be affected by imperfections in
geometry and residual internal stresses. Muc [8] studied the influ-
ence of unilateral friction boundary on shell prebuckling deforma-
tion which was illustrated by the example of a cylindrical shell
loaded locally by external pressure and restrained by a rigid or
elastic outer wall. Moore et al. [9] presented the solutions capable
of assessing the elastic stability of circular structures in square, cir-
cular and rectangular zones of elastic solid. Fok [3] used the energy
method together with a RayleighRitz trial function to analyze
buckling of a long cylindrical shell embedded in an elastic material
and loaded by a far-field hydrostatic pressure. If the surrounding
medium is incompressible, the solution is very similar to that given
by Forrestal and Herrmann [2]. Otherwise, the solution gives lower
predictions for the buckling load and provides better agreement
with experiments. Kang et al. [10] used a pipe-spring model for
buckling analyses of buried corrugated steel pipes. The spring coef-
ficients in the pipe-spring model were calculated using the static
analyses of soil-structure models. The ultimate or critical strengths
determined were compared well with those from the American
Iron and Steel Institute (AISI). The majority of the above studies
focused on the buckling problem of pipes subjected to internal
Slurry Mixing
and Pumping
Plant
Slurry Pipe
Reaction Walls
Steel Pipe Launch Shaft
pressure or external pressure, and they are more suitable for the
study of buried pipes. However, the axial load is dominated for
steel pipe-jacking in construction stage, and buckling of steel jack-
ing pipes is more prone to occur.
Mandal and Calladine [11] conducted self-weight buckling
experiments and non-linear finite element analysis of thin, open-
top, fixed-base, small-scale silicone rubber cylindrical shells, and
the material, structure and loads considered were still fundamen-
tally different from those of steel pipe-jacking. Sheng et al. [12],
Bagherizadeh et al. [13] and Shen [14] analyzed functionally-
graded cylindrical shells embedded in Pasternak elastic foundation
under axial load. Sheng et al. [12] placed emphasis on the eigen-
value solution for buckling of cylindrical shells. Bagherizadeh
et al. [13] achieved the closed-form solutions for the critical
mechanical buckling loads of the FGM cylindrical shells sur-
rounded by elastic medium based on a higher-order shear defor-
mation shell theory (HSDT). While Shen [14] put forward the
boundary layer theory and applied it to analyze postbuckling of
composite cylindrical shells surrounded by tensionless Pasternak
elastic foundation. Similar to the work by Shen [14], Li and Qiao
[15] recently studied the buckling and postbuckling of an anisotro-
pic laminated thin cylindrical shell of finite length subjected to
combined loading of external pressure and axial compression using
the boundary layer theory. Although these studies are similar to
the condition of the steel pipe-jacking, the more practical
method is still needed for design and construction of pipe-
jacking projects.
Numerical methods have been widely used in modeling and
analysis of buckling of cylindrical shells. In addition to the tradi-
tional methods, like finite differences, finite elements method
(FEM), boundary element method (BEM), etc., differential quadra-
ture (DQ) [16], discrete singular convolution (DSC) [17,18] and
meshless method [1921] have also gradually risen. However,
the complicated modeling, tedious mathematical formulas, and/
or programming applicability existed in the above methods, cause
inconvenience when used in practice. The finite strip method
(FSM) is used extensively for reducing partial differential equations
to ordinary or partial differential equations of a lower order. Con-
sequently, much shorter computing time is achieved for solution
with comparable accuracy [22,23]. Especially suitable for the struc-
tures which can be divided into strip elements, the series functions
are defined along the longitudinal direction instead of longitudinal
element division in FEM or other numerical methods. Thus, the
model in FSM is much simplified as well, particularly useful for
preliminary design and analysis.
 L. Zhen et al. / Engineering Structures 132 (2017) 139151 141

In this study, according to the general characteristics of steel X

r
e
u Ym
u fC u gfdgm
pipe-jacking projects, pipes are assumed as elastic cylindrical
m1
shells under axial compression, while the surrounding soil is pos-
Xr
tulated as the two-parameter elastic foundation in which the v Ym e
v fC v gfdgm
transverse shear effect is considered. The stability of steel jacking m1
pipes embedded in different elastic foundations is analyzed using Xr
e
the finite strip method (FSM) based on the first-order shear defor- w Ym
w fC w gfdgm 1
m1
mation plate theory (FSDPT). The critical length and lower bound
X r
of the buckling load are obtained through a parametric study for /x Ym e
/x fC /x gfdgm
the sake of optimal design analysis of steel jacking pipes. Finally, m1
the buckling accident in the practical steel pipe-jacking project is Xr
e
illustrated, and the reasonable design and remedy are accordingly /y Ym
/y fC /y gfdgm
proposed. m1

where Y m m m m m
u , Y v , Y w , Y /x and Y /y are the longitudinal series functions,
2. Theoretical background closely associated with the end boundary conditions. Reddy [24]
defined the general form of the series function according to the
The elastic foundation model is frequently used for simulating buckling eigen-function. For the boundary condition of two simply
pipe-soil interaction. The Winkler foundation model is one of the supported ends, the function can be expressed as:
simplest forms, and it only considers compressive deformation of 8 m l y
m
the surrounding soil as a kind of discontinuum. While the Paster- < Y w Y /x sin ma
nak foundation model considers the shear effect by setting the  0   lm mp; m 1; 2; . . . 2
: Y m Y m lm cos lm y
shearing layer between the foundation and the structure, and it /y /x a a

more closely emulates practical situation. However, it is not easy

where fdgem is a vector representing the mth term nodal displace-
to be implemented in the general simulation software. FSM has
ment parameters at the node lines of the finite strip element. For
been proven to be efficient and accurate for buckling analysis of
the low order finite strip with three node lines (LO3) [22,23] as
plates. The steel jacking pipe can be equivalently approximated
shown in Fig. 3(b), the following expression is held:
by a series of long plates with an equilateral polygon section as
 T
long as the number of these plates tends to infinity or be suffi- fdgem uim v im wim /xim /yim ukm v km wkm /xkm /ykm ujm v jm wjm /xjm /yjm
ciently large (see Fig. 2). Therefore, FSM based on FSDPT is used 3
herein for analyzing buckling of steel jacking pipes (cylindrical
shells) simply supported at two ends and embedded in the Paster- where fC u g, fC v g, fC w g, fC /x g and fC /y g are the transverse interpo-
nak foundation. The connection joint is neglected because of good lation shape functions, and they are given as:
integration of adjacent pipes.
fC u g f C 1 0 0 0 0 C2 0 0 0 0 C3 0 0 0 0g
Buckling analysis of plates under uniaxial compression is a clas-
sical problem, and it is performed using the FSDPT-based FSM. It fC v g f 0 C 1 0 0 0 0 C2 0 0 0 0 C3 0 0 0g
assumes that the length (a) of a plate is in y-direction and the fC w g f 0 0 C 1 0 0 0 0 C2 0 0 0 0 C3 0 0g
width (b) is in x-direction as shown in Fig. 3(a). fC /x g f 0 0 0 C 1 0 0 0 0 C2 0 0 0 0 C3 0g
The displacements of the middle surface of the plate ux; y,
fC /y g f 0 0 0 0 C 1 0 0 0 0 C2 0 0 0 0 C3 g
v x; y and wx; y, and rotations of the normal to the middle sur-
face /x x; y and /y x; y are expressed by the interpolation polyno- 4
mial function in x-direction and smooth series functions in y- where C 1 1  3x 2x ; C 2 4x  4x ; C 3 x 2x ; x x=be .
2 2 2
direction: The strain vector feg, curvature vector fjg, and transverse shear
strain vector fcg are given in the vector form as:
8 9 2 3
> u;x > /x;x " #
< = w;x /x
6 7
feg
>
v ;y >
; fjg 4 /y;y 5; fcg 5
: ; w;y /y
u;y v ;x /x;y /y;x

The strain energy of the finite strip element can be expressed as

ZZ 
y 1 T 1 T 1 T
Ue feg Afeg jT Bfeg fjg Dfjg fcg Hfcg dxdy
2 2 2
z
6

where
2 3 2 3 2 3
r A11 A12 A16 B11 B12 B16 D11 D12 D16
45 6 7 6 7 6 7
A 4 A12 A22 A26 5; B 4 B12 B22 B26 5; D 4 D12 D22 D26 5
o x A16 A26 A66 B16 B26 B66 D16 D26 D66


H11 H12
and H , in which Aij are the extensional stiffness
H12 H22
coefficients, Bij are the bending-extension coupling stiffness coeffi-
cients, Dij are the bending stiffness coefficients, and Hij are the
transverse shear stiffness coefficients.
Fig. 2. The model of cylindrical shells by the finite strip method based on the first- Substituting the related variables into Eq. (6), it is can be
order shear deformable plate theory. obtained that:
142 L. Zhen et al. / Engineering Structures 132 (2017) 139151

b1 b2 be
i k j
b be/2 be/2

(a) (b)
Fig. 3. The low order finite strip with three node lines (LO3).

U U e1 U e2 U e3 U e4 7 @ Pe @U e V e W e
e K ip e K io e K op e
@fdg @fdge
Considering the finite strip element, the strain energy due to the
Pasternak foundation is given by Xiang et al. [25] as K ts e K ef e  kK g e fdge 0 10
RR h 2
i
where k is a unknown load factor; 0 is the zero vector. It is neces-
V e 12 kw Gw2;x w2;y dxdy
sary to introduce the drilling rotation freedom /z (the rotation
X
r X
r 8
about the z axis) and a fictitious stiffness relating to the drilling
fdgem K ef mn fdgen
m1 n1
rotation [22,23] in the present thin-walled structural analysis.
From the governing equations of the plate strip element, the
 e  e  e T
where K ef mn
Ref mn Ref mn and governing equations of the whole structure in the linear stability
analysis can be obtained as:
ZZ h
 e 1
Ref mn
kfC w gT Y m n T m n
w Y w fC w gGfC w g;x Y w Y w fC w g;x K  kK g  0 11
2
i
fC w gT Y m n where K is the stiffness matrix, and K K ip  K io  K op 
w;y Y w;y fC w g dxdy
K ts  K ef , respectively, representing in the plane stiffness matrix,
where k is the coefficient of subgrade reaction, G is the shear mod- in and out of plane coupling stiffness matrix, out of plane stiffness
ulus of the subgrade. matrix, transverse shear stiffness matrix, and elastic foundation
The Winkler foundation can be regarded as a special form of the stiffness matrix; K g  is the geometric matrix. Interested readers
Pasternak foundation if the shear modulus (G) is set to 0. may refer to the works of Cheung [22,23] for more details of FSM.
In the buckling process, the potential energy W e due to the
applied initial stresses (r0x , r0y and r0xy ) can be expressed as [23,26]: 3. Validation
ZZ
t 3.1. Numerical model
We  r0x w2;x r0y w2;y 2r0xy w;x w;y dxdy
2
ZZ
t3 Cylindrical shells embedded in the Winkler foundation and
 r0x /2x;x /2y;x r0y /2x;y /2y;y
24 Pasternak foundation under axial compression with two simply
X r X r supported ends are analyzed using the aforementioned FSDPT-
2r0xy /x;x /x;y /y;x /y;y dxdy  fdgem K g mn fdgen 9 based FSM. The cylindrical shell is approximatively modeled as a
m1 n1
series of plates, and a 3-D pipe-soil model is simplified as a 2-D
where K g emn K g1 emn K g2 emn , and profile of the pipe based on FSM. The effect of the elastic founda-
tion is considered in the stiffness matrix through the strain energy,
 e  e T and the longitudinal deflection is simulated by series functions.
K g1 emn Rg1 mn Rg1 mn
ZZ h Hence, the cylindrical shell can be modeled as a regular polygon
 e t
Rg1 mn r0x fC w gT;x Y mw Y nw fC w g;x r0y fC w gT Y mw;y Y nw;y fC w g as shown in Fig. 4. Each side of the regular polygon denotes a strip
2
i element. The low order finite strip with three node lines (LO3)
2r0xy fC w gT;x Y m n
w Y w;y fC w g dxdy (Fig. 3) is chosen, of which the deflection and the rotation are taken
into account.
 e  e T
K g2 emn Rg2 mn Rg2 mn 3.2. Verification and comparison
ZZ h
t3
Rg2 emn r0x fC /x gT;x Y m/x Y n/x fC /x g;x fC /y gT;x Y m/y Y n/y fC /y g;x In theory, the increasing number of plate strips emulating the
24
T m n T cylindrical shells will lead to higher accuracy and relatively low
r y fC /x g Y /x ;y Y /x ;y fC /x g
0
fC /y g Ym n
/y ;y Y /y ;y fC /y g
i computational efficiency. The different cases with cylindrical shells
T m n
2r xy fC /x g;x Y /x Y /x ;y fC /x g
0
fC /y gT;x Y m n
/y Y /y ;y fC /y g dxdy
embedded in the different elastic foundation are compared with
the specific example given in the paper of Shen [14], and the com-
The governing equation of the plate strip element can be formu- parisons with the FSM predictions are shown in Fig. 5. The results
lated by applying the principle of minimum potential energy, demonstrate that the buckling load converges quickly when the
which is in the partial differentiation as: number of strips increases. It is sufficient to meet the accuracy
 L. Zhen et al. / Engineering Structures 132 (2017) 139151 143

z z diameter and d is the inner diameter. Thus, Eq. (13) is related to

n1 the inner diameter and the wall thickness of cylindrical shells. Fur-
n8 n1 n2 thermore, the slenderness ratio is represented as L/q. At the same
n3 time, the effects of the length-diameter ratio (L/d) and the
n7 r 45 n2 r 45 diameter-thickness ratio (d/t) are considered as well.
x x
n6 n3 4.2. Buckling of cylindrical shells

n5 n4 In practice, the diameter-thickness ratio (d/t) of the steel jack-

(a) (b) ing pipes is usually about 100 according to the design. Therefore,
if d/t is given, the slenderness ratio (L/q) varies with respect to
Fig. 4. The cylindrical shell model in FSM: (a) octagon, and (b) polygon.
the length-diameter ratio (L/d). The cylindrical shell models with
the inner diameter (d) of 2 m are analyzed using both of FSM (ele-
ment type: LO3, see Fig. 3) and FEM (element type: S4 in the com-
3000 mercial finite element software ABAQUS) as shown in Fig. 6. The
(k1,k2) */(kN) 36 strips 60 strips 72 strips 80 strips 100 strips Shen (2013)
(0,0) 2441.65 2367.5 2358.39 2354.94 2349.6 2422.52 simply supported steel jacking pipe is assumed as an ideal elastic
2900 (100,0) 2480.92 2406.77 2397.82 2394.21 2389.02 2462.66 material with the Youngs modulus (E) of 210 GPa and Poissons
ratio (v) of 0.3, subjected to axial compression.
(100,10) 2782.35 2709.47 2700.98 2697.84 2693.13 2769.3
2800 Shen[14] (100,10)
B u c k li n g l o a d ( k N )

As shown in Fig. 6, even though the results of FSM overall are

2700
2600
2500 Shen[14] (100,0)
2400 Shen[14] (0,0)
2300 *k1 and k2 are the dimensionless Winkler foundation stiffness and the
dimensionless shearing layer stiffness, respectively [14].
2200
0 10 20
Fig. 5. Comparisons of buckling loads of cylindrical shells by the different number
of plate strips and foundation stiffness.
requirement if the number of strips is more than 60. As shown in
Fig. 5, the critical buckling loads obtained by FSM are smaller than
those in the paper of Shen [14] which was based on the analytical
solution. It suggests that the results from the FSM program are rel-
atively reliable and conservative for the design.
4. Parametric analysis
4.1. Dimensionless key parameters
The buckling behavior of cylindrical shells embedded in the
Winkler and Pasternak foundations subjected to the axial compres-
sive load with two simply supported ends by FSM are, respectively,
studied. To facilitate the comparison of different parameter effect
of the model, the key dimensionless parameters are introduced.
slightly smaller than those of FEM, their respective predictions
are closely matched. As expected, an apparent turning point
appears between the intersection of local and global bucking. How-
ever, there exist a few critical local buckling loads with respect to
the critical slenderness ratios (L/q) in the region of local buckling.
While as expected, the buckling loads fall fast when the global
buckling occurs. The global buckling loads are in excellent agree-
ments with the results of Euler buckling formula. The computing
time of some FSM and FEM cases on the same computer is pre-
30 40 50 60 70 80 90 100 110
sented in Fig. 7. It indicates that the computational efficiency of
The number of strips
FSM is much higher than that of FEM with the increasing size of
models when achieving the same accuracy. Especially, FSM has a
remarkable advantage for analyzing buckling of large-scale struc-
tures like steel jacking pipes.
The detailed analysis is conducted for local buckling, as demon-
strated in Fig. 8. For the comparatively small slenderness ratios (L/
q), because the buckling load is much affected by the slenderness
ratio (L/q) and the number of half-waves (n) generated along the
axial (length) direction, the buckling loads fluctuate considerably
in this region. In the local buckling region, although the curve is
composed of several downward convex waves, the general trend
of buckling load decreases when the slenderness ratio (L/q)
increases. The number of the half-wave (n) in the axial direction
changes with several downward convex waves and each of them
contains a minimum value of the local buckling load (see Fig. 8).
Both the critical buckling load and the critical buckling length or
1.1
1.0
Dimensionless buckling load x /cr

The dimensionless buckling load is represented as 0.9 The standard formula based
on the Donnell's theory
a rx =rcr 12 0.8

where rx is the actual buckling load and rcr p

E t is the clas- 0.7
31m2 r
0.6
sical elastic buckling formula, in term of the Youngs modulus E,
Local buckling
Poissons ratio v, the wall thickness t, and the inner radius r. 0.5 FSM
The gyration radius of cylindrical shells is expressed as FEM
0.4
s Euler's buckling
r q  2 0.3
I 1 2 d d 2t
q D2 d 1 0.2 Global buckling
A 4 4 d
s
 2 0.1
d 2t
1 1 13 0.0
4 d 0 20 40 60 80 100 120 140 160

4 4 p D2 2 The slenderness ratio L/

where I 64 D  d is the moment of inertia and A 4
p  d is
the cross-sectional area of cylindrical shell. D denotes the outer Fig. 6. Buckling behavior of cylindrical shells with respect to the slenderness ratio.
144 L. Zhen et al. / Engineering Structures 132 (2017) 139151

500 1.2
L/ 14.14 28.28 56.57 70.71 84.85 113.13 141.41
The standard formula based
FSM (s) 80.23 81.23 79.92 77.50 79.94 78.72 77.39

Dimensionless buckling load x/cr

450 FEM (s) 101.40 229.70 207.70 235.90 250.50 333.40 421.00
on the Donnell's theory
1.0
400
Computing time (s)

350 0.8

300
0.6
250
Local buckling Global buckling
200 0.4

150 d=1m, d/t=100

0.2 d=2m, d/t=100
100 d=3m, d/t=100
d=4m, d/t=100
50 0.0
0 20 40 60 80 100 120 140 160 0 10 20 30 40 50 60
The slenderness ratio L/ (a) The slenderness ratio L/
Fig. 7. The comparison of computational efficiency of FSM and FEM. 1.6
d/t=50

Dimensionless buckling load x/cr

1.4 d/t=100
1.1
The standard formula based d/t=150
FSM 1.2
1.0 on the Donnell's theory d/t=200
Dimensionless buckling load x /cr

n=3 FEM
0.9 n=1 The standard formula based
1.0
on the Donnell's theory
0.8
0.8 The lower bound
0.7 n=1 n=2 of local buckling
0.6 0.6

0.5
0.4
0.4 Local buckling
0.2
0.3
0.2 0.0
0 20 40 60 80
0.1 (b) The slenderness ratio L/
0.0
0 5 10 15 20 25 30 35 40 45 Fig. 9. Effect of diameter-to-thickness ratio (d/t) on buckling behavior of cylindrical
The slenderness ratio L/ shells: (a) a constant d/t, (b) changing d/t.

Fig. 8. Local buckling behavior of cylindrical shells.

Table 1
slenderness ratio exist for cylindrical shells of specific dimensions The critical length for different diameter-to-thickness ratios (d/t).

under the certain buckling mode. Diameter to thickness ratio Lcr1 (m) Lcr2 (m) Lcr3 (m)
For the cylindrical shell with the inner diameter (d) of 2 m and d/t = 50 1.4 4 8
the diameter-thickness ratio (d/t) of 100, there are three obvious d/t = 100 4.4 11 22
local minimum (critical) values among all local buckling loads, d/t = 150 8.1 21 42
excluding the region of the comparatively small slenderness ratios d/t = 200 13 32 64
(L/q). The three local minimum values correspond to the critical
lengths of the cylindrical shells of 4.4 m, 11.0 m and 22.0 m,
respectively. If the diameter-to-thickness ratio (d/t) of the steel
with those of FEM. Moreover, the former is relatively smaller and
jacking pipe keeps a constant as 100, the buckling behavior is
thus provides a lower bound for design without considering com-
almost not affected by the dimension of the steel jacking pipe in
plex spring models in FEM.
the cylindrical shell case as shown in Fig. 9(a). As demonstrated
The coefficient of subgrade reaction (k) in the Winkler founda-
in Fig. 9(b), if the wall thickness (t) maintains the same, the transi-
tion is closely related to the geological condition and the buried
tion slenderness ratio (L/q) distinguishing the local buckling and
depth of steel jacking pipes. Two sets of the coefficient of subgrade
global bucking increases as the diameter-thickness ratio (d/t)
reaction (k), a small one and a practical one in the specific project,
increases. It suggests that the size effect indeed exists in buckling
are selected for comparison as shown in Fig. 11. The results imply
of cylindrical shells subjected to axial compression. The approxi-
that the buckling loads are enhanced when the cylindrical shells
mate critical lengths (Lcr) for different diameter-thickness ratio
are embedded in the Winkler foundation. When the coefficient of
(d/t) are presented in Table 1. For local buckling, the lower bound
subgrade reaction (k) is small, e.g., k = 10,000 N/m3, the turning
is about 60% of the classical solution based on the Donnells theory.
point is still evident between the regions of local buckling and glo-
bal bucking. The buckling loads are almost the same as the results
4.3. Buckling of cylindrical shells embedded in the Winkler foundation of cylindrical shells without the Winkler foundation in the region
of local buckling; but they increase in the region of global buckling
As shown in Fig. 10, the results of FSM for buckling of cylindrical as expected. When the coefficient of subgrade reaction (k)
shells embedded in the Winkler foundation is highly consistent approaches the value considered in the practical project, e.g.,
 L. Zhen et al. / Engineering Structures 132 (2017) 139151 145

1.1 normal and shear deformation, were developed, and the Pasternak
FSM foundation model is one of the widely recognized two-parameter
1.0
Dimensionless buckling load x/cr

FEM models. The general commercial finite element software, like ABA-
0.9 The standard formula based
QUS, cannot provide such a complex foundation model directly.
on the Donnell's theory
0.8 But it is convenient to implement this two-parameter foundation
0.7 model in FSM just through modifying the stiffness matrix.
The two parameters in the Pasternak foundation model, i.e., the
0.6
coefficient of subgrade reaction (k) and the shear modulus of the
0.5 Local buckling subgrade (G), are analyzed in Fig. 12. The shear modulus (G) of
k = 10000 N/m
0.4 the foundation is related to the elastic modulus (E) of the founda-
tion. So the shear modulus (G) of 961,500 N/m according to the
0.3
practical projects is calculated, and as a comparison, the shear
0.2 Global buckling modulus (G) of 1,923,000 N/m is also chosen. The buckling loads
0.1 are greatly increased when the Pasternak foundation is considered
0.0 in the model as shown in Fig. 12. Buckling behavior of cylindrical
0 20 40 60 80 100 120 140 160 shells is changed as well. The transition point again appears in
The slenderness ratio L/ the intersection region between local buckling and global bucking
for cylindrical shells embedded in the Pasternak foundation. For
Fig. 10. Buckling behavior of cylindrical shells embedded in the Winkler the greater shear modulus (G), the local buckling load increases
foundation.
before the global buckling occurs. It suggests that the shear effect
enhances the resistance of the foundation within a certain range
1.1
of the length and leads to increased local buckling load. Finally,
the global buckling loads trend to accord. It seems that there is
1.0
the lower bound of the local buckling load in the region of global
Dimensionless buckling load x/cr

The standard formula based

0.9
on the Donnell's theory
buckling when the Pasternak foundation is present.
0.8 By comparisons, the buckling loads of steel jacking pipes in the
pure cylindrical shell model (without any foundation support, i.e.,
0.7 Local buckling
k = 0 N/m3 and G = 0 N/m) and the Winkler foundation model (i.e.,
0.6 k = 312,500 N/m3 and G = 0 N/m) are conservative. Moreover, the
0.5 Local buckling k=0 buckling modes vary among different foundation models. The
k = 10000N/m Pasternak foundation model improves the accuracy of the estimate
0.4
k = 312500N/m since it better represents the practical situation.
0.3 With the development of the pipe-jacking technique and the
0.2 Global buckling growing demand for the construction, the maximum diameter of
0.1
the steel jacking pipe has already reached about 4.0 m. However,
the traditional design is still determined mainly according to expe-
0.0 rience of the steel jacking pipe with the diameter under 2.0 m. The
0 20 40 60 80 100 120 140 160
The slenderness ratio L/ size effect of buckling behavior of cylindrical shells embedded in
the Pasternak foundation (Fig. 13) is much different from the pure
Fig. 11. Effect of the coefficient of subgrade reaction (k) in the Winkler foundation. cylindrical shells without any foundation support (Fig. 9(b)). As
depicted in Fig. 13, the local buckling load decreases as the diam-
k = 312,500 N/m3, the global buckling is difficult to take place. It eter increases in the region of the relatively small slenderness ratio
seems that the buckling loads are independent of the lengths for (L/q). For the larger slenderness ratio (L/q), the critical local buck-
relatively long cylindrical shells. For design, the critical lengths ling still takes place with a constant critical buckling load. As the
(Lcr) corresponding to the critical local buckling loads are displayed diameter of the cylindrical shells decreases, the critical local buck-
in Fig. 11. The differences only appear for the greater coefficient of ling load decreases as well, such as from the diameter of 1 m to
subgrade reaction (k) presented in Table 2. 2 m. Whereas for the diameter of 3 m or large, the critical local
buckling load reaches a constant critical value. As observed in
4.4. Buckling of cylindrical shells embedded in the Pasternak Fig. 13, the global buckling hardly takes place when the Pasternak
foundation foundation is considered for the studied steel jacking pipes in this
study. The results in Fig. 13 imply that the local buckling is more
The Winkler foundation is modeled as a series of closely spaced, likely to occur in the steel jacking pipes with the increased diam-
mutually independent, and linear elastic spring elements. It eters currently being considered in practical projects and the
implies that the interaction between springs is neglected, namely Pasternak foundation model is more suitable for buckling analysis
that the continuity of soil is ignored. However, the shear effect of of the steel jacking pipes in practice. The critical length for differ-
the foundation plays an important role in practice. Thus, the Win- ent inner diameters changes a lot. Although no certain rule could
kler foundation cannot meet the practical need for more compli- be followed within the scope of the current analysis, the critical
cated projects. The foundation models, considering both the length can be easily determined by FSM.

Table 2
The critical length for different coefficients of subgrade reaction (k) in the Winkle foundation.

The length of cylindrical shells Lcr1 (m) Lcr2 (m) Lcr3 (m) Lcr4 (m)
k=0 4.4 11 22
k = 10,000 N/m3 4.4 11 22
k = 312,500 N/m3 4.4 10.8 22 43.5
146 L. Zhen et al. / Engineering Structures 132 (2017) 139151

3
1.2 where D 121
Et
m2 is the flexural stiffness. The curvature parameter
The standard formula based
on the Donnell's theory (or so-called the Batdorf parameter) for cylindrical shells under
Dimensionless buckling load x/cr

1.0 axial compression is defined as

 2
Lower bound L2  1=2 L r 1=2
0.8 Z 1  m2 4 1  m2 16
rt d t
The relationship between the dimensionless axial compressive-
0.6 Local buckling
(k,G)=(0,0) stress coefficient (k) and the curvature parameter (Z) was obtained
(k,G)=(312500N/m,0) by a large number of test results. Then, the approximate fitting
0.4 (k,G)=(312500N/m,961500N/m) curve was given. According to the characteristics of buckling and
Global buckling
Local buckling

(k,G)=(312500N/m,1923000N/m) the k-Z curve, the formula of the dimensionless axial

0.2 compressive-stress coefficient varies.
For Z < 2.85

0.0 12Z 2
0 20 40 60 80 100 120 140 160 k1 17
p2
The slenderness ratio L/
For Z P 2.85
Fig. 12. Buckling behavior of cylindrical shells embedded in the Pasternak p
foundation. 4 3
k Z 0:702Z 18
p2
1.2
and it corresponds to the case before the global buckling occurs.
The standard formula based The diameter-to-thickness ratio (d/t) in the practical pipe-
jacking engineering is usually 100, and the Poissons ratio (v) is
1.1
Dimensionless buckling stress x/cr

on the Donnell's theory

1.0 0.3. So the curvature parameter (Z) in Eq. (16) can be simplified as
Local buckling
0.9  2  2
L2  1=2 L r 1=2 L
0.8 Z 1  m2 4 1  m2 190:79 19
rt d t d
0.7
0.6 Then, for L/d < 0.123,
!
0.5
Dp2 12Z 2
0.4 rcr 1 20
tL2 p2
0.3 d=1m, d/t=100
0.2 d=2m, d/t=100 For L/d P 0.123,

0.1
d=3m, d/t=100 Dp2 p2 E
d=4m, d/t=100 rcr 0:702 2
Z 0:014 1=2
21
0.0 tL 121  m2
0 20 40 60 80 100 120 140 160
The slenderness ratio L/ However, the effect of the length-to-diameter ratio (L/d) is not
considered when L/d P 0.123 because it is not accurate any more
Fig. 13. Size effect on buckling behavior of cylindrical shells embedded in the for relatively long cylindrical shells.
Pasternak foundation.
Rotter [29] compared the different European standards used in
different countries for buckling of cylindrical shells under axial
5. Design practice compression. For the sake of convenience needed in practice, the
knock-down factor ax is defined to take imperfection of cylindrical
5.1. Current design standards shells caused by different situations into account in the ECCS
(European Convention for Constructional Steelwork) code. Hence,
The classical thin shell theory, i.e., the Donnells theory is com- the design buckling load is defined as
monly used as the standard formula in various standards for buck-
ling analysis of cylindrical shells under axial compression because
rRk ax rRc rRc rcr 22
of its simplicity. The critical buckling stress based on the Donnells wk pr
where ax is formulized as ax 0:62
, where
wk 1:44t
. wk
t
1
Q
theory is simply expressed as 11:91 t

denotes the amplitude of the largest notional imperfection, while

E t
rcr p
31  m2 r
Eq. (14) is not suitable for relatively long cylindrical shells
because the length effect is not considered [27].
Although the theory of buckling of cylindrical shells under axial
compression is widely studied and it provides a basis for design,
the theoretical results still derivate far away from the test results
because the practical cylindrical shells are not ideal structures
and imperfections exist. Many researchers tried to explain the dif-
ferences and improved the analytical results. Batdorf [28] defined
the dimensionless axial compressive stress coefficient as
rcr tL2
k 15
Dp2
14 Q is the fabrication quality parameter. For the practical pipe-
jacking engineering, rt 50 and Q = 25; thus, ax 0:473, and it
means that the design buckling load is just 0.473 of the theoretical
buckling load. The effect of the length-to-diameter ratio (L/d) is not
considered either.
Buckling analysis based on different practical standards for
cylindrical shells with two simply supported ends under axial com-
pression is briefly reviewed in the above. For the steel jacking
pipes, it can be assumed as a thin cylindrical shell mainly subjected
to the axial compressive load, and the surrounding soil can be trea-
ted as an elastic foundation in practice. Buckling is largely affected
by the pipe-soil interaction, whereas there are few appropriate
methods in the standards to solve this problem. It thus leads to
the conservative stability design for the steel jacking pipes in

which the effect of soil is not included. Even with the conservative
stability design practice, buckling accidents still occur in practice.
5.2. Design for steel pipe-jacking
The traditional design of steel pipe-jacking is similar to the
design of buried pipes. Confining pressure which may cause the
deformation of the cross section is the main factor considered in
design. In addition, the axial compression stability analysis is based
on buckling of pure cylindrical shells, and the effect of the elastic
foundation, e.g., the pipe-soil interaction, is rarely taken into
account in stability analysis.
Cylindrical shells embedded in the Pasternak foundation with
two simply supported ends under axial compression is closer to
the practical situation of steel pipe-jacking as demonstrated in
the above parametric study. Besides, the buckling strength under
the axial compression increases sharply under the circumstances.
Although the parametric analysis in Section 4.4 is for an ideal
structure and imperfections from manufacturing errors, construc-
tion, surroundings, etc. are not considered, the buckling behaviors,
such as the buckling load, the critical length (or the slenderness
ratio), and the size effect, are still meaningful in design.
The diameter-to-thickness ratio (d/t) in the current steel jacking
pipe is usually about 100 due to the material strength and cost effi-
ciency. The representative stratum parameters (the elastic modu-
lus of the foundation of 2.5 MPa and Poissons ratio of 0.3) in
Shanghai, China are selected. Assume that the buried depth is
8 m. Thus, the coefficient of subgrade reaction (k) is computed as
312,500 N/m3, and the shear modulus of the subgrade (G) is
961,500 N/m. Different diameters with constant diameter-
thickness ratio (d/t) of 100 are analyzed in Section 4.4. Lines 14
in Fig. 14 represent the steel jacking pipes with the respective
diameters of 14 m, which are commonly used in the practical
steel pipe-jacking projects, embedded in the Pasternak foundation.
The range of the buckling load is approximately from 51.7% to
114% of the standard formula based on the Donnells theory.
Compared with line 0, which represents the pure cylindrical
shell case, if its lower bound is regarded as a benchmark of the tra-
ditional design, the envelope of pure (without any foundation
being considered) steel jacking pipes can be approximately
expressed as the following piecewise function
8
> 0:02L=q 0:95
>
>
< 0:59
rx =rcr
>
> 0:02L=q 1:21
>
: 515 m, it began to deviate from the design axis to the left. Although
0:04
1.2
1.1
Dimensionless buckling load x/cr
1.0
0.9
0.8
0.7
0.6
0.5
Lower bound
0.4
0.3
0.2
0.1 Lower bound
0.0
0 20 40
The slenderness ratio L/
Fig. 14. Buckling design of practical steel pipe-jacking.
L. Zhen et al. / Engineering Structures 132 (2017) 139151 147
The approximate envelope of steel jacking pipes embedded in
the Pasternak foundation can be expressed as:
8
> 0:009L=q 1 0 6 L=q < 13:05
>
>
< 0:88 13:05 6 L=q < 29:23
rx =rcr
> 0:024L=q 1:58
> 29:23 6 L=q < 44:37 24
>
:
0:51 44:37 6 L=q
for the lower bound
8
< 0:005L=q 1
> 0 6 L=q < 26:62
rx =rcr 0:03L=q 1:94 26:62 6 L=q < 35:24
>
: 25
0:88 35:24 6 L=q
for the upper bound
Considering the foundation effect in the steel jacking pipe anal-
ysis, the maximum increment from the critical local buckling load
is approximate 49.15% in the lower bound and 137.29% in the
upper bound, while the maximum increment from the global buck-
ling load is approximate 1175% in the lower bound and 2100% in
the upper bound. It means that the design jacking force can be
improved on the basis of the traditional design.
In addition, the critical length for different dimensions of the
steel jacking pipes can be obtained by the buckling analysis. The
minimum value corresponds to the critical length due to the
volatility of the buckling load curve. The critical length is affected
by the dimension of steel jacking pipes and the foundation param-
eters. However, it is quite difficult to obtain empirical formulas
considering these parameters within the scope of the current anal-
ysis. Therefore, re-analysis through FSM is needed for the steel
jacking pipe under the specifically given conditions.
6. Case study
6.1. Background to the engineering accident
Two sets of steel pipelines (#1 and #2) abreast with the outer
diameter (D) of 2.0 m, thickness of 20 mm, and buried depth of
79 m, were used in Shanghai water supply project for branch
pipelines. The whole branch pipeline was constructed by the
0 6 L=q < 15:59 pipe-jacking method. A total length from working well #44 to
15:59 6 L=q < 35:08 #46, shown in Fig. 15, was 1357 m. The measured value of axial
23 deflection is exhibited in Fig. 16. When pipe #2 was jacked to
35:08 6 L=q < 62:36
62:36 6 L=q the remedial treatments had been taken for returning to the right
direction, it had to stop pipe-jacking while severe local deflection
probably caused by the local buckling (shown in Fig. 17) appeared
The standard formula based
on the Donnell's theory
and some connectors in flanges were damaged. The maximum axis
deflection reached 6.452 m at the final jacking distance of 597 m.
Upper bound The remaining length was finished by jacking the pipes from the
4
3 other end (working well #46), and finally two sections were con-
nected together. The length of the affected pipeline was approxi-
2
mate 85 m. Moreover, the only intermediate jacking station
located near the jacking machine had not been used till deviation
1
of the design axial. Hence, the deflection of the pipeline is
continuous.
Line 0 represents cylindrical shells with d/t = 100.
Line 1-4 represents cylindrical shells embedded
in the Pasternak foundation with d/t = 100.
6.2. Geological condition
0 According to the geological exploration report, the area is
60 80 100 120 140 160 divided into nine geological layers as shown in Table 3. Based on
the buried depth of two pipelines, the pipelines are likely to cross
muddy silty clay sandwiched between clayey silt in 3-interlayer,
silty clay in layer 31 and muddy clay in layer 4, shown in Fig. 18.
148 L. Zhen et al. / Engineering Structures 132 (2017) 139151

River
Building
Pipe #2 River

5m
515 m 15 m Deflection area

#44 Pipe #1 Bridge #46

Jacking
Working well Lake Working well
direction

Fig. 15. Overview from working well #44 to #46.

Intermediate Jacking
jacking stations
machine
Section 9 Section 8 Section 7 Section 6 Section 5 Section 4 Section 3 Section 2 Section 1
21 m 51 m 2m 28.2 m
-660
-619.3 -645.2
-600 -587.8
-604.4
-541.7-566.2
-496.2
-500 -473.8
Axial deflection (cm)

-400 -377.2
-329.8
-300 -279.3-302.6
-239.2
-202.7-226.8
-200 -165.9
-129.1
-101.7 Design axis
-100 -79.1
-54.6
-36.2
-20 3.6 6.5 4.6 2.8 -1.6 -10.8
0 579 584
20
494 498 502 506 511 515 521 525 530 533 537 541 545 549 553 558 562 570 573 577 582 586 597
Distance from #44 working well (m)

Fig. 16. Axial deflection of pipe #2.

(a) Section 2 of pipe #2 (b) Section 6 of pipe #2

Fig. 17. Site photos of pipe #2.

6.3. Analysis of steel pipe-jacking accident
According to the practical condition, pipe #2 can be simplified
as a cylindrical shell with two simply supported ends under axial
compression embedded in elastic foundation. In practical engi-
neering, there are eight jacks providing a total of 12,000 kN of jack-
ing force although the design jacking force is 4500 kN. The steel
jacking pipe is assumed as an ideal elastoplastic material with
the Youngs modulus (E) of 206 GPa, Poissons ratio (v) of 0.3, and
yield stress (ry) of 235 MPa.
The jacking force is mainly used for overcoming the frictional
resistance. The friction factor (l) on the pipesoil contact surface
is 0.3 and the shear stress limit (smax) is 5 kPa. Because the average
buried depth is 8 m from the ground to the axis of jacking pipes
and the average unit weight of soil is 18 kN/m3, the maximum fric-
tional resistance can reach 43.2 kPa based on the friction factor (l)
of 0.3, which far exceeds the shear stress limit (smax) of 5 kPa. Thus,
the frictional resistance can be considered to be 5 kPa. When pipe
#2 was jacked to 515 m cumulatively, the frictional resistance
reached 16,171 kN in the most unfavorable conditions, which is
greater than the largest jacking force of 12,000 kN available in con-
struction stage. It is estimated that the maximum jacking force
actually acted on jacking pipes probably approached 12,000 kN
for the jacking pipes going forward smoothly. Due to the restraint
 L. Zhen et al. / Engineering Structures 132 (2017) 139151 149

Table 3
Soil properties of construction site.

Number cS e Es0.10.2 wn wp wL Cc c u N63.5

kN/m3 MPa % % % Pa
11
21 18.7 0.862 4.68 30.1 22.7 38.3 22 18.5
31 17.5 1.173 2.90 41.9 23.1 38.0 32.750 11 17.0
3-interlayer 18.1 0.983 6.25 31.6 22.9 37.6 33.358 8 24.5 5.0
31 17.5 1.173 2.90 41.9 23.1 38.0 32.750 11 17.0
4 16.9 1.374 2.11 48.8 26.0 45.0 17.300 10 12.5
511 17.4 1.204 2.90 42.5 25.7 44.6 12 12.5
512 17.9 1.047 4.12 36.6 23.3 39.1 15 16.5 9.0
6 19.4 0.719 6.41 24.7 20.7 35.4 45 18.0

Note: cs = Unit weight of soil; e = Void ratio; Es0.10.2 = Compression modulus; wn = Moisture content; wp = Plastic limit; wL = Liquid limit; c = Cohesion under direct shear
consolidation; u = Internal friction angle under direct shear consolidation; Cc = Compression index of soil; N63.5 = Blow counts of SPT.

0.00 m 6.4. Practical treatment

0.00-1.20 m 1-1 The buckling section stopped the jacking due to the large deflec-
Miscellaneous fill
tion. The measurement of the residual stress shows the maximum
2-1 axial residual stress is close to 350 MPa which is far more than the
Silty clay yielding stress of 235 MPa. They are mainly distributed in the
0.60-3.00 m region of local buckling as shown in Fig. 20. It is unrecoverable
plastic deflection. In order to ensure the safety of the pipeline
3-1 and control the construction cost, the steel plates of 0.02 m in
2.30-4.80 m Muddy silty clay thickness were welded on the inner wall of the pipeline. The area
of the steel plate was more than the deformed region (see Fig. 21).
3- interlayer The remaining length was finished by the jacking pipes from the
Muddy silty clay working well #46. The accident serves to show that on one hand,
sandwiched the design jacking force is rather conservative and cannot meet
4.20-6.60 m between clayey silt the demand of the practical project; on the other hand, the local
-7.00 m buckling is more likely to occur due to the large axial jacking force.
The design of steel pipe-jacking as shown in Fig. 22 can be adopted
3-1 to improve the local buckling capacity of pipe design when the
-8.00 m Muddy silty clay Pasternak foundation model is used. The lower bound of the steel
Center of pipe #2
jacking pipes with the outer diameter (D) of 2.0 m embedded in
6.50-9.60 m
the Pasternak foundation can be expressed as:
8
-9.00 m 4 > 0 6 L=q < 29
< 1:0
Muddy clay
rx =rcr 0:03L=q 1:97 29 6 L=q < 38 26
>
:
Fig. 18. Location of pipe #2. 0:7 L=q P 38

and support of the foundation, the jacking pipes remained stable in

the beginning.
The jacking axis of pipe #2 began to deviate from the design
axis to the left when the jacking continued. According to the sup- (a) The axial local buckling mode of steel
plemental geological exploration report, the unknown concrete jacking pipe based on FSM
structure was found close to the design axis of the steel jacking
pipes at the location of 530 m far from the working well #44. It 80 70 60 50 40 30 20 10 0
2.00
causes uneven distribution of hardness in the stratum which might
Jacking direction
be the main reason to trigger the derivation from the design axis.
The original
After the derivation of pipeline from the design axis occurred, 1.98
Actual inner diameter (m)

inner diameter
the stress of the steel jacking pipes increased accordingly as the
deflection developed. The axial buckling mode of the deflection 1.96
section of 85 m steel jacking pipe embedded in the Pasternak foun-
dation is presented in Fig. 19(a) based on FSM. It suggests that local
1.94
buckling may occur, though the theoretical critical buckling load is
far greater than the actual value. Since the average confining pres-
sure reached 144 kPa, the combination of the jacking force and the 1.92
Horizontal inner diameter
surrounding soil pressure possibly led to the occurrence of local
Vertical inner diameter
buckling and inward deflection on the part of the pipeline as 1.90
shown in Fig. 17. The measured deflections of horizontal and ver- Distance from the jacking machine (m)
tical inner diameters were recorded, respectively, and they are (b) The measured diameter deflections of inner
shown in Fig. 19(b). The shape of the deformed pipeline is similar diameter of steel jacking pipe
to the half-waving of local buckling on the cylindrical shells, as
shown in Fig. 19(a). Fig. 19. Buckling analysis of steel jacking pipe.
150 L. Zhen et al. / Engineering Structures 132 (2017) 139151

Jacking direction Jacking direction

1.5 m 4.0 m

0.75 m

1.5 m

1.2 m
0.7 m

8.0 m 8.0 m
Local deflection

Local deflection
1.930 m

1.930 m
1.990 m
area

1.930 m
area

1.950 m
2.000 m
(a) Section 2 of pipe #2 (b) Section 6 of pipe #2
Fig. 20. Diagram of local buckling of pipe #2.

Local deflection Local deflection

area area

0.020 m
0.020 m

0.020 m
0.020 m

1.000 m Liner panel 1.000 m Liner panel

0.700 m 0.750 m
Wall of pipe Wall of pipe
(a) Section 2 of pipe #2 (b) Section 6 of pipe #2
Fig. 21. Reinforcement of the local buckling section.

The local buckling load increases by approximate 5.2669.49% val of 200 m. The remaining length of 780 m just needs three inter-
and the global buckling load increases by approximate 18.64 mediate jacking stations. Each of them has eight jacks providing a
1650%. Therefore, the scope of the design jacking force should be total of 12,000 kN of jacking force. According to the aforemen-
increased to 4737 kN to 7627 kN and 5339 kN to 78,750 kN accord- tioned method, the remaining length of 780 m was finally com-
ing to the lower bounds of the local buckling and the global buck- pleted in success.
ling, respectively.
A single section of steel jacking pipes is suggested to be 8 m 7. Conclusions
according to the critical length. The maximum frictional resistance
of 200 m steel jacking pipes will be 6280 kN. As shown in Fig. 22, In this study, the finite strip method (FSM) is used for buckling
the jacking pipe length of 200 m is not the most unfavorable analysis of cylindrical shells under the axial compression to simu-
length. Thus, it is suitable to modify the design jacking force to late the construction practice of steel pipe-jacking, and its effi-
7000 kN if the intermediate jacking station is set up with the inter- ciency and accuracy in design of analysis of steel jacking pipes
are demonstrated. The following concluding remarks are drawn
from this study:
1.2
The standard formula based
on the Donnell's theory  The main advantages of FSM considered in this study for buck-
Dimensionless buckling load x/cr

1.0 ling analysis of cylindrical shells under axial compression and

surrounded by the Pasternak foundation are modeling simplic-
ity and computational efficiency when compared to other
0.8
numerical methods. The longitudinal deflection is simulated
Lower bound
by series functions, and the foundation parameters are consid-
0.6 ered in the stiffness matrix through the strain energy, which
is suitable for buckling analysis of cylindrical shells embedded
0.4 (k,G)=(0,0) in the elastic foundation subjected to axial compression, like
(k,G)=(312500N/m,961500N/m) in steel pipe-jacking.
 For buckling of pure cylindrical shells, the critical local buckling
0.2
load decreases as the slenderness ratio (L/q) increases, while
Lower bound
the global buckling load is in close agreement with the Euler
0.0 buckling formula.
0 20 40 60 80 100 120 140 160
 For buckling of cylindrical shells embedded in the Winkler
The slenderness ratio L/ foundation, the local buckling load is almost consistent with
Fig. 22. Design for steel pipe-jacking.
that of pure cylindrical shells. The global buckling load
 L. Zhen et al. / Engineering Structures 132 (2017) 139151
increases as the pipe slenderness ratio decreases, and it gradu-
ally turns into the local buckling mode as the coefficient of sub-
grade reaction (k) increases, which indicates that the Winkler
foundation prevents the occurrence of global buckling of cylin-
drical shells.
 For buckling of cylindrical shells embedded in the Pasternak
foundation, the local buckling load is enhanced compared with
pure cylindrical shells and cylindrical shells embedded in the
Winkler foundation. Moreover, the greater shear modulus (G)
and the smaller inner diameter result in the greater local buck-
ling load within the scope of the relatively small slenderness
ratio (L/q). However, the change of the shear modulus (G) has
little impact on the global buckling. The global buckling hardly
occurs with the increasing of inner diameter if the diameter-
thickness ratio (d/t) keeps constant. It reveals that the shear
foundation effect plays an important role in strengthening the
buckling capacity of cylindrical shells embedded in the Paster-
nak foundation. The buckling mode also changes accordingly.
The Pasternak foundation is closer to the situation in practice,
and the design analysis based on FSM provides an efficient yet
accurate method (a 2-D model) to simulate steel jacking pipes
embedded in the Pasternak foundation.
 The modified design can be proposed based on the FSM analy-
sis. Compared with the traditional design, the new design based
on FSM and considering the Pasternak foundation plays a full
role on the effect of the elastic foundation in buckling of steel
pipe-jacking. Thus, the present study offers the meaningful ref-
erence for more safe and reasonable design and construction of
steel jacking pipes.
Acknowledgments
The authors would like to thank for the financial support from
the National Natural Science Foundation of China (NSFC Grant
Nos. 51478265 and 51678360).
References
[1] Zhen L, Chen JJ, Qiao P, Wang JH. Analysis and remedial treatment of a steel
pipe-jacking accident in complex underground environment. Eng Struct
2014;59:2109.
[2] Forrestal MJ, Herrmann G. Buckling of a long cylindrical shell surrounded by an
elastic medium. Int J Solids Struct 1965;1(3):297309.
[3] Fok S. Analysis of the buckling of long cylindrical shells embedded in an elastic
medium using the energy method. J Strain Anal Eng Des 2002;37(5):37583.
[4] Luscher U. Buckling of soil-surrounded tubes. J Soil Mech Found Div 1966;92
(6):21128.
151
[5] Duns CS, Butterfield R. Flexible buried cylinders: Part IIIbuckling behaviour.
Int J Rock Mech Mining Sci Geomech Abstr 1971;8(6):61327. Pergamon.
[6] Yun H, Kyriakides S. On the beam and shell modes of buckling of buried
pipelines. Soil Dyn Earthq Eng 1990;9(4):17993.
[7] Cheney JA. Local buckling of tubes in elastic continuum. J Eng Mech 1991;117
(1):20516.
[8] Muc A. On the contact of cylindrical shells with an elastic or rigid foundation.
Contact loading and local effects in thin-walled plated and shell
structures. Springer; 1992. p. 3441.
[9] Moore ID, Haggag A, Selig ET. Buckling strength of flexible cylinders with
nonuniform elastic support. Int J Solids Struct 1994;31(22):304158.
[10] Kang J, Parker F, Yoo CH. Soil-structure interaction for deeply buried
corrugated steel pipes Part I: Embankment installation. Eng Struct 2008;30
(2):38492.
[11] Mandal P, Calladine CR. Buckling of thin cylindrical shells under axial
compression. Int J Solids Struct 2000;37(33):450925.
[12] Sheng G, Wang X. Thermal vibration, buckling and dynamic stability of
functionally graded cylindrical shells embedded in an elastic medium. J Reinf
Plast Compos 2008;27(2):11734.
[13] Bagherizadeh E, Kiani Y, Eslami MR. Mechanical buckling of functionally
graded material cylindrical shells surrounded by Pasternak elastic foundation.
Compos Struct 2011;93(11):306371.
[14] Shen HS. Postbuckling of axially-loaded laminated cylindrical shells
surrounded by an elastic medium. Mech Adv Mater Struct 2013;20(2):13050.
[15] Li ZM, Qiao PZ. Buckling and postbuckling of anisotropic laminated cylindrical
shells under combined external pressure and axial compression in thermal
environments. Compos Struct 2015;119:70926.
[16] Zhang L, Xiang Y, Wei GW. Vibration analysis of cylindrical shells by a local
adaptive differential quadrature method. Int J Mech Sci 2006;48(10):112638.
[17] Civalek , Grses M. Free vibration analysis of rotating cylindrical shells using
discrete singular convolution technique. Int J Pressure Vessels Pip 2009;86
(10):67783.
[18] Demir , Mercan K, Civalek . Determination of critical buckling loads of
isotropic, FGM and laminated truncated conical panel. Compos Pt B-Eng
2016;94:110.
[19] Liew KM, Ng TY, Zhao X. Free vibration analysis of conical shells via the
element-free kp-Ritz method. J Sound Vib 2005;281(35):62745.
[20] Liew KM, Lee YY, Ng TY, Zhao X. Dynamic stability analysis of composite
laminated cylindrical panels via the mesh-free kp-Ritz method. Int J Mech Sci
2007;49:115665.
[21] Zhao X, Yang Y, Liew KM. Geometrically nonlinear analysis of cylindrical shells
using the element-free kp-Ritz method. Eng Anal Boundary Elem 2007;31
(9):78392.
[22] Cheung YK. Finite strip method in structural analysis. Publication of Pergamon
Press Incorporated; 1976.
[23] Cheung YK, Tham LG. The finite strip method. CRC Press; 1997.
[24] Reddy JN. Mechanics of laminated plates and shells: theory and analysis. CRC
Press; 2004.
[25] Xiang Y, Kitipornchai S, Liew KM. Buckling and vibration of thick laminates on
Pasternak foundations. J Eng Mech 1996;122(1):5463.
[26] Dawe DJ, Roufaeil OL. Buckling of rectangular Mindlin plates. Comput Struct
1982;15(4):46171.
[27] Schmidt H. Stability of steel shell structures: general report. J Constr Steel Res
2000;55(1):15981.
[28] Batdorf S. A simplified method of elastic-stability analysis for thin cylindrical
shells No. NACA-TN-1342. Langley Field (VA): National Advisory Committee
on Aeronautics, Langley Aeronautical Laboratory; 1947.
[29] Rotter JM. Shell structures: the new European standard and current research
needs. Thin-Wall Struct 1998;31(1):323.