Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
a r t i c l e i n f o a b s t r a c t
Article history: The load response of a rock-socketed steel H-pile can be strongly influenced by the nonlinear interfacial
Received 1 June 2010 behavior between the grout and the steel H-pile, and between the pile and the rock mass. This paper
Received in revised form 22 November 2010 focuses on the load-transfer mechanism of the former interface through experimental push-out tests
Accepted 16 January 2011
and numerical simulation of the tests. The study is divided into two parts. In the first part, a series of
push-out tests have been carried out on four types of specimens (with studs + confinement (Case I); no
stud + confinement (Case II); no stud + no confinement (Case III); and studs + no confinement (Case
Keywords:
IV)). The second part is a numerical study based on three-dimensional finite element technique, which
Push-out test
Rock-socketed steel H-pile
takes into account possible damage and cracking in grout, as well as bond-slip along the interface. It is
Steel-grout interface shown that the numerical predictions of the four cases compare favorably with the corresponding test
Finite element technique results, including the load–displacement response and the development of cracks. Furthermore, paramet-
ric study has been carried out to investigate the influence of various factors, including the studs, the cas-
ing confinement, the grout fracture energy, and the dilation property of steel-grout interface. Lastly, some
implications, based on the test and numerical results, on the design of socketed steel H-piles are
discussed.
Ó 2011 Elsevier Ltd. All rights reserved.
0266-352X/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compgeo.2011.01.002
384 Y.D. Zhou et al. / Computers and Geotechnics 38 (2011) 383–392
the composite section. Moreover, three-dimensional finite element test, a compressive force was applied vertically on the top of
models are established to simulate the push-out tests and the re- H-section, and the force was increased gradually at a rate less than
sults are compared with the test results. Furthermore, a series of 0.5 kN/s. The loading rate indicates that the slippage rate at the top
parametric analyses have been undertaken for an in-depth study of the interface would not exceed 2 107 m/s, which was suffi-
on the influence of various factors, namely the studs on the flanges ciently low and can be taken as a quasi-static case as also approved
of H-section, the casing confinement, the grout fracture energy, as by former trial testings [9–11]. An automatic data logger was used
well as the interfacial dilation properties, on the behavior of spec- to record the readings of LVDT and the corresponding applied load
imens. Lastly, some implications, based on the numerical results, at every increment. As the maximum test force achievable from
on the design of socketed steel H-piles are discussed. this setup was 2000 kN, the test was terminated slightly below this
maximum force or failure of the specimen whichever came first.
Due to the limitation of the loading frame, the total height of
2. Push-out tests the test specimen has to be not more than 1.0 m. Herein the height
of the bonded portion was chosen as a half of the specimen, viz.
2.1. Experimental procedure 0.5 m. The diameter of the test section is the same as those of
the actual piles. In preparing the specimens, a 5 mm thick steel cas-
The arrangement of the push-out test is shown in Fig. 1. A steel ing, dimensioned as 0.5 m in height and 610 mm in external diam-
plate with a matching H-shape opening was placed underneath the eter, was placed and centered on a 30 mm thick steel template
specimen to allow the H-section to move downward freely during with an H-hole. The steel H-section of 1 m in length was then
the test. Four linear variable displacement transducers (LVDT) placed inside the casing and fitted into the H-hole of the steel tem-
were mounted on the H-section to measure the vertical displace- plate. Finally, the space between the H-section and the casing was
ments relative to the upper grout surface (Fig. 1a). During each filled with cement grout. All specimens were made and cured for
28 days in accordance with CS1: 1990 before delivering to the
laboratory for testing.
The H-sections (305 305 180 kg/m, Grade 55C structural
steel complying with BS4360), were prepared in a similar way as
that for in situ steel H-piles, whilst the steel casing was prepared
from those used as temporary support for the borehole during
construction. The cement grout was a mixed product of ordinary
Portland cement and water with a water/cement ratio of 0.45, also
Cebex 100 non-shrink admixture was applied to minimize the
shrinkage effect of the grout material. It should be noted that all
specimens were dry during the test and the influence of under-
ground water under practical conditions has not been included in
this study.
3. Numerical model
Table 1
Summary of push-out test cases.
The double symmetry of the test specimens allows only one brief description of the damage-plastic model is given below,
quarter of the samples to be modeled. Fig. 2 presents the finite ele- whilst further details can be found in Lee and Fenves [13].
ment mesh for the specimens with or without studs. The H-section The constitutive equation of material with scalar damage vari-
and the grout body are modeled by 8-node solid elements, and a able takes the form:
layer of 3-D interface elements are defined between the steel and
grout elements to simulate possible slippage along the interface. r ¼ ð1 dÞDel0 : ðe epl Þ ¼ Del : ðe epl Þ ð1Þ
For the cases with studs welded on the flanges of H-section (Cases
I and IV), 8-node solid elements are used to model the studs where r is Cauchy stress tensor; d is the scalar stiffness degradation
(Fig. 2a). In Cases I and II, the outer confinement by steel casing variable; e is the strain tensor; epl is the plastic strain; Del0 denotes
is modeled by a layer of 4-node general-purpose shell elements. the initial (undamaged) elastic stiffness of the material, while
Since the bottom surfaces of both the grout body and the casing Del ¼ ð1 dÞDel0 is the degraded elastic stiffness tensor.
were vertically constrained by a steel plate underneath during The main two failure mechanisms of the grout consist of tensile
the push-out tests, it is assumed that there is no slippage along cracking and compression crushing. Two uni-axial damage vari-
the interface between the grout and the casing. ables, dt and dc, are introduced to represent the damage states in
The steel plate underneath the test specimen was simplified as tension (with subscript ‘‘t’’) and compression (with subscript ‘‘c’’),
a rigid surface constraining the vertical displacement of the bottom respectively. The damage variables are defined as independent
surface of the specimen; also a frictional contact interaction with a functions of two hardening (softening) variables ~epl ~pl
t and ec , respec-
friction coefficient of 0.57 was defined in between allowing lateral tively, which are referred to equivalent plastic strains in tension
displacement of the grout body. Along the double symmetric sur- and compression, respectively. From the stress–strain relations
faces of the numerical model, horizontal roller boundary condi- under uni-axial tension and compression loadings, two effective
tions are defined to constrain corresponding translational cohesion stresses, tensile and compressive, are defined as r t ¼
displacement. E0 ðet ~epl ~pl
t Þ; rc ¼ E0 ðec ec Þ, where E0 denotes the undamaged
elastic modulus. These variables control the evolution of the yield
3.2. Material models and parameters surface and the degradation of elastic stiffness variable.
The yield surface for the damage-plastic model is defined in
The steel H-section and the casing are assumed to be linear effective stress space and takes the form:
elastic. Compressive load tests were conducted to determine the
properties of the H-section. The average uni-axial compressive 1
F¼ þ bð~epl Þhr
3ap
ðq max i chr c ð~epl
max iÞ r c Þ ¼ 0
strength was determined to be 195.7 GPa. The relevant elastic 1a
parameters are given in Table 2. ð2Þ
The push-out test indicated that severe but localized damage
and fracture developed within the grout body during the push- in which q and p denote the Von Mises equivalent effective stress
out process. A damage-plastic constitutive model by Lee & Fenves and effective hydrostatic stress, respectively; a = (fb0/fc0 1)/(2fb0/
[13], capable of modeling concrete and other similar quasi-brittle fc0 1), where fb0 and fc0 denote the initial equi-biaxial compressive
materials, is used to simulate the behavior of the grout. A scalar strength and initial uni-axial compressive strength, respectively;
variable in the concept of damaged elasticity is applied in combi- b ¼ ð1 aÞr c ð~epl t ð~epl
c Þ=r t Þ ð1 þ aÞ, where rc and rt are the effective
nation with non-associated multi-hardening plasticity to describe tensile and compressive cohesion stresses, respectively; c is a
the irreversible damage that occurs during the fracture process. A parameter controlling the shape of yield surface in the deviatoric
Fig. 2. Finite element models of one quarter of the push-out test specimens (a–b).
386 Y.D. Zhou et al. / Computers and Geotechnics 38 (2011) 383–392
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
s ¼ c þ rn tan / ð4Þ
G ¼ ðmft0 tan wÞ2 þ q 2 p tan w ð3Þ where c is the interface cohesion, and / is the friction angle. The off-
diagonal terms in the elastic stiffness matrix are zero, hence no
dilatancy along the interface is considered in the elastic range.
where ft0 is the initial uni-axial tensile strength of grout; w is the The dilatancy is introduced after the failure criterion has been
dilation angle measured in the p q plane at high confining pres- reached. The flow potential function is of a similar form as Eq. (4)
sure; while m is an eccentricity of the plastic potential surface. with the friction angle replaced by the dilation angle wi. In order
The m value is chosen as 0.1 in the subsequent calculations, which for gradual mobilization of peak friction and dilation, as well as
implies that the flow potential tends to a straight line over a wide the strain softening behavior, wi is defined as a linear function of
range of confining pressure stress values. plastic shear strain magnitude eieff in the following manner:
The push-out test results indicated that tensile cracking is the For eieff 6 edil ,
dominant failure type in the grout, hence the relationship between
wi ðeieff Þ ¼ w0 ð1 eieff =edil Þ ð5Þ
the uni-axial tensile strength ft and ~epl t , which controls the evolu-
tion of failure surface, can greatly influence the initiation and and for eieff > edil ,
development of cracking failure. In order to minimize unreason-
wi ¼ 0 ð6Þ
able mesh sensitivity in the analysis, the Hillerborg’s fracture en- qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ergy approach [14] is adopted to control the degradation of ft, where eieff is defined as e2ix þ e2iy ; eix and eiy denote the two orthog-
i.e., the energy required to open a unit area of crack, Gf, is specified onal plastic shear strain components along the interface tangential
as a material constant. Moreover, a posteriori adaptive technique is directions, respectively. edil is the plastic shear strain magnitude at
applied during the analysis to ensure that the mesh size is fine en- which the interface does not represent the dilation effect any more,
ough and desired accuracy can be provided. In the following calcu- and w0 is the initial value of the dilation angle. The above approach
lations, exponential loss of strength is defined with respect to the bears some similarity with Leong and Randolph0 s model [17] for
cracking strain ~eck ~ck ~ck
t in the form of ft ¼ ft0 expðet =ef Þ, in which pile-rock interface, in which wi was defined to decrease exponen-
ck
~ef ¼ Gf =ft0 controls the shape of the softening curve. The equiva- tially with respect to the relative displacement along the interface.
lent tensile plastic strain ~epl t can be determined from ~eck t by Based on a series of push-off test results, Rabbat and Russell
~epl ~ck
t ¼ et dt =ð1 dt Þ ft =E0 . [18] recommended an average peak friction angle of about 30o
For identification of relevant parameters for the grout material, for the interface between steel plate and grout at a normal com-
cylindrical grout specimens with dimensions of U150 mm pressive stress level of 0.41 MPa. It should be noted that the recom-
300 mm were prepared for uni-axial compression test, and the mended friction value represents a mobilized angle of shear
average Young0 s modulus was determined to be 16.3 GPa, and resistance incorporating the dilation effect along the interface
the average Poisson0 s ratio was 0.27. The average compressive [19], i.e., /m, which is defined as follows:
strength of the cylindrical specimens was 40.0 MPa. Moreover, to
/m ¼ /c þ wi ð7Þ
determine the fracture energy of the grout material, the fracture
toughness test suggested by ISRM [15], which was originally in which /c denotes the critical state angle of shear resistance and
developed for measuring the fracture toughness of rock materials, wi is the dilation angle. Hoek [20] suggested that the friction angle
is adopted in this study. A series of fracture toughness tests on of grout itself falls within 17–20° for a range of water/cement ratio
similar grout material, with a bit larger water/cement ratio (1:2), of 0.41–0.52. It is generally accepted that a relatively lower friction
have been carried out and an average value of fracture energy parameter should be adopted for steel-grout interface as compared
Gf = 25 N/m is calibrated from the test results of load versus load to that for the grout material itself. From the reported test results,
point displacement curves. One can note that the fracture energy the present study takes /m = 30° and /c = 18° for the steel-grout
of grout is smaller than that of concrete, which is mostly in the interface. Reasonably a value of w0 = 12° can be inferred from above
range of 40 N/m–120 N/m. In the analyses, the uni-axial tensile Eq. (7). The parameter edil is determined through a series of sensitiv-
strength of the grout material is chosen as one tenth of the uni- ity analyses of the push-out tests, and a value of 500 le is found to
axial compressive strength, following commonly adopted assump- be appropriate for reproducing the test results.
tion. Also a dilation angle w = 30o is chosen for the grout material Rabbat and Russell [18] also suggested that the cohesive
[13,16]. A summary of the input parameters for the grout is also strength between rolled steel plate and cast-in-place grout can
given in Table 2. be at a negligibly small value of about 50 kPa. The degradation of
Y.D. Zhou et al. / Computers and Geotechnics 38 (2011) 383–392 387
c value with respect to the plastic shear displacement is neglected 4.1. Influence of studs
and maintained as a constant of 50 kPa in this study. A summary of
the above mechanical parameters for the steel-grout interface is Generally, one can note that for either case with casing confine-
listed in Table 2. ment or not, there is a significant increase in load capacity if studs
are welded onto the flanges of H-section. The abrupt change in the
3.4. Analysis programme slope of each curve was due to the cracking in grout. As shown in
Fig. 3a, the load–displacement curves in Case II quickly level off
Based on the above finite element models and the calibrated with a head displacement greater than 0.5 mm, whilst the studs
parameters (Table 2), a series of numerical analyses have been car- in Case I was capable of giving rise to a continuous increase of axial
ried out to simulate the behavior of the four types of specimens in load with respect to the head displacement. Similar observations
the push-out test. Furthermore, parametric study has been carried can be also found from the two cases without casing confinement
out on various factors, including (1) confinement by varying the (Fig. 3b). Furthermore, the test results demonstrate that the studs
thickness of the casing (0–7 mm as well as infinite), (2) the grout contribute to a noticeable increase of the incipient stiffness. For in-
fracture energy Gf (from 15 to 45 N/m as well as infinite), and (3) stance, the incipient stiffness dropped by 35% in Case III as com-
the initial dilation angle w0 of the steel-grout interface (from 8° pared to the results in Case IV, and a gradual softening behavior
to 14°). Typical simulation results are given in the following sec- was shown in the former case after severe cracks were mobilized
tions, and comparisons are also made with corresponding test in the grout portion.
results. Based on the simulation results, a comparison of the axial load
distribution in H-section, normalized by the corresponding peak
4. Results and discussions load of each case, is shown in Fig. 4. It is clearly illustrated that
the presence of studs significantly influences the distribution of ax-
The load–displacements of the four types of specimens obtained ial load in H-section. Without stud, a smooth decay of axial force
from the tests are shown in Fig. 3. It can be seen that for each case, with depth is observed (Cases II and III), whereas the axial load
the above finite element models are capable of reproducing the shows abrupt changes at the studs (Cases I and IV).
load–displacement response up to the maximum load. It should The stud reinforcement responsible for the increase in load
be noted that the analyses of Cases II and III stopped at a displace- capacity originates from two perspectives. Firstly, the mechanical
ment of 0.16 mm owing to the convergence problem as the grout interlock caused by each stud introduces a side-bearing function,
body was broken into blocks after the formation of cracks running i.e., a resisting force against tangential slippage acting upon the
through the whole depth of the specimen. Nevertheless, the pre- underside of each stud is given by the grout in continuous contact.
dictions of the push-out capacity of these two cases compare well Secondly, the installation of studs roughens the steel-grout
with corresponding test results. Moreover, the predicted crack pat- interface as surface asperities that are unlikely to shear off, which
terns in grout are in agreement with the test observations and it equivalently leads to an increase in the asperity angle of the inter-
will be discussed in the subsequent sections. The comparison dem- face. Consequently, the interfacial shear strength and dilatancy
onstrates the capability of the above model for simulating the properties can be considerably influenced, as featured by the mobi-
push-out behavior of the specimens. lization of a larger mobilized friction angle /m (Eq. (7)).
1800 1800
1500 1500
1200 1200
Case I
Load (kN)
Load (kN)
900 900
600 600
Case IV
Case II
300 300
Case III
0 0
0.0 0.2 0.4 0.6 0.8 0.0 0.2 0.4 0.6 0.8
Displacement (mm) Displacement (mm)
(a) With casing (b) No casing
(Case I: with studs, and CaseII: no stud) (Case III: no stud, and Case IV : with studs)
Fig. 3. Influence of stud measure on load–displacement curves (a–b) (Fill area: test bounds; marked curves: numerical predictions).
388 Y.D. Zhou et al. / Computers and Geotechnics 38 (2011) 383–392
Normalized force
0 0.2 0.4 0.6 0.8 1
0
0.5
1800 1800
h=7mm
h= h=3mm h=
1200 1200
Load (kN)
h=1mm
Load (kN)
900 900
h=7mm
h=5mm
Case I h=3mm
(with studs
600 and casing) 600
h=0 Case II
Case IV (no stud
(with studs with casing)
300 300 h=1mm
no casing) Case III
(no stud
h=0 no casing)
0 0
0.0 0.2 0.4 0.6 0.8 0.0 0.2 0.4 0.6 0.8
Displacement (mm) Displacement (mm)
(a) With stud measure (b) no stud measure
Fig. 6. Influence of different casing thicknesses on load–displacement curves (a–b) (Fill area: test bounds; marked curves: predictions).
h=3mm
h=0
0.2 0.2 h=1mm
Depth (m)
Depth (m)
h=
h=
0.3 0.3
0.4 0.4
0.5 0.5
(a) with stud measure (b) no stud measure
Fig. 7. Influence of different casing thicknesses on the axial load distribution in H-section (a–b) (all curves correspond to the results at a top displacement of 0.8 mm except
the curves at h = 0, which shows the results at peak load).
agreement with the test observations (Fig. 5). Moreover, it is 4.3. Influence of grout fracture property
shown by the modeling results that increasing the casing
confinement could be beneficial for retarding the growth and Fig. 9 presents the load-settlement results based on a series of
development of cracking in the surrounding grout. Take the case analyses by varying the grout fracture energy parameter (Gf)
without stud for instance, with the casing thickness h increased within the range of 15 N/m–45 N/m. Other parameters are taken
from 0 to 7 mm, the top settlement at which a through longitudinal as the calibrated ones. The response of an idealized case with
crack is formed in the grout presents a 67% increase from 0.15 mm non-crackable grout material representing the upper bound, i.e.,
to 0.25 mm. Gf = 1, is also carried out. Cases II and III specimens excluding
390 Y.D. Zhou et al. / Computers and Geotechnics 38 (2011) 383–392
The above test and simulation results give better insight into
the behavior of grout encased H-section specimens when subjected
to axial loading. Even though the dimensions, embedment ratio,
and working conditions of an in situ socketed H-pile are different
from those of the push-out tests, the findings can still bear some
implications on the design of rock-socketed steel H-pile.
The results in Figs. 3 and 6 demonstrate the effectiveness of
studs in increasing the stiffness and bearing capacity of load-
settlement response. It is inferred from the above results that
installing of studs at the pile top, even though the confinement
from shallow soil layers could be small at that location, is still
Fig. 8. Typical crack patterns smeared in grout elements as indicated by tensile
capable of improving the load-settlement response.
damage factor (dt) by the push-out process (a–b).
The present study also sheds light on the significance of con-
finement by surrounding soil and rock layers. The confinement
the effect of stud reinforcement were chosen in this study. Gener- condition given by the steel casing in the push-out tests can be
ally it can be observed that the variation of Gf in the selected range simplified as thin-wall cylinder under internal pressure. According
presents only limited influence on the load response of push-out to the elastic theory [21], the normal confinement stiffness pro-
specimens. However, the upper bound analysis results with infinite duced by a casing subjected to uniform radial pressure is given by
Gf value demonstrate considerable potential for raising the push-
out capacity, which can be achieved if the fracture resistance of kr ¼ Es h=r 2 ð8Þ
surrounding cementitious material can be increased to a greater
in which Es is the Young0 s modulus of the casing; h and r denote the
extent.
thickness and the internal radius, respectively. On the other hand,
With confinement (Case II), only a 5% rise in peak load from
the confining condition of a pile socketed into soil or rock can be
640 kN to 670 kN is induced with Gf increased from 15 N/m to
simplified as a borehole of the same diameter surrounded by
45 N/m, and the curves (Fig. 9a) exhibit fairly flat plateaus at a set-
semi-infinite solid mass, and the produced confinement stiffness
tlement greater than 0.5 mm. As the grout fracture resistance in-
under uniform internal pressure is defined by
creases, the curves become nonlinear at a greater settlement and
therefore a larger push-out force is required. Moreover, it is found kr ¼ Er =ð1 þ lr Þr ð9Þ
that with Gf value increased from 15 N/m to 45 N/m, the top settle-
ment at which a through longitudinal crack is formed in the grout in which Er and lr denote the deformation modulus and Poisson0 s
presents a 53% increase from 0.15 mm to 0.23 mm. ratio of surrounding rock mass, respectively.
Y.D. Zhou et al. / Computers and Geotechnics 38 (2011) 383–392 391
1200 1000
1000
800
800
Gf =
600
Gf =
Load (kN)
Load (kN)
600
Gf =45N/m
Gf =35N/m
400
400
Gf =45N/m
Gf =25N/m
200 Gf =35N/m
200 Gf =15N/m Gf =25N/m
Gf =15N/m
0 0
0.0 0.2 0.4 0.6 0.8 0.0 0.1 0.2 0.3 0.4 0.5
Displacement (mm) Displacement (mm)
(a) Case II (no stud with casing) (b) Case III (no stud no casing)
Fig. 9. Influence of different fracture energy values (Gf) of grout material on predicted load–displacement curves (a–b).
2000 800
ψ 0 =14º
1600
ψ 0 =14º
600 ψ0 =12º
ψ 0 =12º
1200 ψ 0 =10º
Load (kN)
Load (kN)
400
ψ0 =10º ψ 0 =8º
800
ψ0 =8º
200
400
0 0
0.0 0.2 0.4 0.6 0.8 0.0 0.2 0.4 0.6 0.8
Displacement (mm) Displacement (mm)
(a) Case I (with studs and casing) (b) Case II (no stud with casing)
Fig. 10. Influence of different dilation angles (w0) of steel-grout interface on predicted load–displacement curves (a–b).
From the above definitions, the normal confinement stiffness confinement effect, and in turn, the load capacity and top
provided by a 5 mm-thick casing (kr = 10.5 MPa/mm), is approxi- settlement of a socketed H-pile.
mately equivalent to the confining effect by a rock mass with The sensitivity analysis results shown in Figs. 9 and 10 also
Er = 4.2 GPa and lr = 0.3. Based on the rock mass rating system illustrate the significance of increasing the grout fracture property
[22], rocks with such modulus can be considered as ‘‘fair’’ rock. Gf and the dilation angle w0of steel-grout interface. The results in
The results in Figs. 6 and 7 imply that even for the same ‘‘fair’’ rock Fig. 9 infer that increasing the grout Gf parameter can improve
socket, a small variation of Er can strongly influence the the capability of H-pile against cracking. Besides, the confinement
392 Y.D. Zhou et al. / Computers and Geotechnics 38 (2011) 383–392