Sunteți pe pagina 1din 11

Non-linear analysis of pile groups Proc.

Instn
Civ. Engrs
Geotech. Engng,
1999, 137, April,
F. Basile, MEng 105115

Paper 11745
j A numerical code is presented for the pre- Pav average vertical load carried by each pile
diction of linear and non-linear behaviour of head Written discussion
axially and laterally loaded pile groups. Ptot applied vertical load on the cap closes 30 September 1999
p lateral load transfer
Although solutions based on nite element
rg radius of the circle of equivalent area to that Manuscript
modelling (FEM) and boundary element covered by the group received 9 June 1998;
modelling (BEM) of the soil continuum have s pile spacing (centre-to-centre) revised manuscript
been developed, the application of these smax maximum interaction spacing accepted 12 January 1999
methods is mainly restricted to linear elastic t axial load transfer
analysis and relatively small pile groups. In ftp g pile traction vector
particular, it has been shown that a rigorous fts g ctitious soil traction vector
BEM solution to large pile group problems t sc limiting bearing stress
is uneconomical and hence it was restricted t ss limiting shear stress
to small pile groups. In contrast, the pro- u cap horizontal displacement
fup g pile displacement vector
posed approach put forward here demon-
fus g soil displacement vector
strates that a complete BEM solution can be V tot applied vertical load on the cap
applied to large pile groups without pre- w cap vertical displacement
cluding its use in normal design. Bench- y lateral pile displacement
mark solutions in the linear and non-linear z axial pile displacement
range are presented, and comparisons of the empirical adhesion factor
results with some available published solu- cap rotation
tions are employed to assess the accuracy of s soil Poisson's ratio
the present approach. Finally, three case
studies are described which demonstrate the Introduction
applicability of the method to practical Many problems in pile group design are related
problems. to road bridges, often with adjacent embank-
ments, and to similar work on railways. In these
Keywords: foundations; geotechnical engi- circumstances horizontal loads can be large and
neering; piles & piling associated with a range of vertical loading
conditions. Traditional methods of design com-
monly treated the loads as equivalent eccentric
Notation and inclined forces applied to a rigid cap, which
fBg boundary condition vector in turn was supported by piles, modelled as
Cu undrained shear strength axially loaded elastic springs.
C uo undrained shear strength at ground level 2. The recent development of numerical
d pile diameter approaches, such as nite element modelling
Ep pile Young's modulus (FEM) and boundary element modelling (BEM),
Es soil Young's modulus in conjunction with the wide availability of
Eso soil Young's modulus at ground level
powerful desktop computers, has put more
G soil shear modulus
efcient methods of analysis at the disposal of
[Gp ] pile exibility matrix
[Gs ] soil exibility matrix the pile designer. Analytical techniques for pile
H horizontal load taken by the individual pile group response fall into two main categories:
head continuum-based approaches and load-transfer
Hav average horizontal load taken by each pile approaches. The latter category, based on the
head so-called Winkler idealization of soil, is attractive
Htot applied horizontal load on the cap in its exibility, enabling non-linear and
Ip second moment of area of pile section inhomogeneous soil conditions to be incorpo-
[K ] global stiffness matrix of the group rated easily (i.e. the t z or p y curve
KR group exibility factor
methods of analysis). The main drawback to this
kp ratio of applied vertical load to the settle-
approach is that no direct tests can be
ment of the group
L pile length conducted to establish forcedisplacement rela-
M tot applied moment on the cap tionships for that particular pile and soil type;
Francesco Basile,
N SPT value these curves must be back-calculated from the PhD Student,
Nc bearing capacity factor data obtained by conducting pile load tests. Depar tment of Civil
n number of piles in the group Thus, a signicant amount of engineering Engineering,
P vertical load acting on the individual pile judgement is needed when formulating these University of
head curves for site conditions which differ markedly Glasgow
105

Downloaded by [ University College London] on [21/09/16]. Copyright ICE Publishing, all rights reserved.
BASILE

from the recorded eld tests. In addition, These limitations may be removed by simul-
disregard of continuity through the soil over- taneous consideration of all elements of all the
simplies the problem and makes it impossible piles within a group, that is, performing a
to nd a rational way to quantify the interaction `complete' analysis of the group. The computer
5
effects between piles in a group. program PGROUP is included in this category
3. In order to overcome these limitations, but it is restricted to linear analysis and problems
1
solutions based on nite element and boundary of small dimensions because of enormous
2
element modelling of the soil continuum have computational resources. Other computer codes,
6
been proposed. These solutions provide an such as PILGP1 and the one developed by
7
efcient means of retaining the essential aspects Chow, propose a `hybrid' approach in which
of pile interaction through the soil continuum load-transfer non-linear springs are used to
and hence a more realistic representation of the obtain the response of the individual piles and a
problem. Further, the mechanical characteristics continuum model is used to determine the
to be introduced into the model have now a effects of interaction between the piles.
clear physical meaning and they can be meas- 6. The main drawback of the application of
ured directly. FEM represents the most power- available numerical codes to realistic pile group
ful analytical approach to pile problems in which problems is that they are restricted to linear
a variety of constitutive soil models can be elastic analyses and hence misrepresent the
employed and such aspects as soil inhomo- forces in piles. In fact, it was found that corner
geneity and anisotropy can be considered. Finite piles in groups very often showed very high
element analyses are essential for clarifying the loading values when compared with the results
mechanism of load transfer from the pile to the from the traditional methods and when com-
surrounding soil, but, especially for pile groups, pared with the interior piles of a group. In
are not readily applicable to practical problems. consequence, engineers were frequently con-
The considerable effort of data preparation and fused. The combination of enhanced vertical and
the high computational cost (especially if non- horizontal loads per pile meant that costs of
linear soil behaviour is to be considered) render steel to cater for bending in piles were
such analyses uneconomical for most cases. By substantially increased. Importantly also, con-
contrast, BEM provides a complete problem siderations of the cracking of reinforced con-
solution in terms of boundary values only crete were introduced into the design of steel
(specically at the pilesoil interface), with reinforcement for piles via the UK Code BS
8
substantial savings in computing time and data 5400. Some deem this crack width considera-
preparation effort. tion to be inapplicable to piles but it is never-
4. Currently, a number of computer pro- theless still commonly applied and regarded as
grams are available, by which pile groups under being a necessary part of the application of the
general loading conditions (i.e. vertical and Code. Use of this Code further exacerbates the
horizontal loads and moments) may be analysed use of steel.
in order to give estimates of the deections and 7. The basic importance of non-linear group
3
load distributions among the piles. PIGLET is analysis systems to examine the distribution of
based on approximate analytical solutions for loadings on piles in groups is that it has the
single-pile response and for interaction between desirable effect of demonstrating a reduction of
two piles, in which linear elastic soil behaviour the corner loads in large groups in both the
4
is assumed; DEFPIG, based on a simplied horizontal and vertical senses. Therefore, a
BEM analysis and the use of interaction factors, reliable analysis program which is able to run
models soil non-linearity in an approximate on modern desktop computers is of considerable
manner but it needs a large amount of importance to industry because it can bring
computing time and hence it is limited to groups potential saving of materials and reduce costs
of 25 piles. for those who must pay for the works.
5. It should be emphasized that the inter- 8. For this purpose, the present work aims
action factor approach produces the following to develop a numerical procedure, implemented
limitations in the computer program PGROUPN, by which
pile groups of large dimensions subjected to any
(a) it ignores the stiffening effect of intervening combination of vertical loads, horizontal loads
piles in a group, thereby leading to an and moments may be investigated by means of
overestimation of interaction between piles a non-linear anlaysis. The proposed approach
(b) its use becomes questionable for cases in adopts an incremental procedure to include the
which not all the piles are identical effects of soil non-linearity while retaining
(c) it only gives the loads and bending moments continuity within the soil through a complete
at the pile heads, but not their distributions boundary element analysis of the group. The
along the piles; these may only be approxi- piles are assumed to be connected by a rigid
mated utilizing the single-pile solutions with free-standing cap (a reasonable assumption in
the corresponding pile head loads and most cases) which imposes the same head
bending moments. displacements and prevents differential head
106

Downloaded by [ University College London] on [21/09/16]. Copyright ICE Publishing, all rights reserved.
NON-LINEAR ANALYSIS
OF PILE GROUPS

rotation. Piles may have different geometries Soil deformation


(length, external and internal diameter, base 12. The indirect formulation of BEM in-
diameter) and may be inclined in the direction volves the integration of an appropriate elemen-
of horizontal loading. tary singular solution for the soil medium over
9. The main feature of PGROUPN lies in its the surface of the problem domain, that is, the
capability to adopt a complete BEM approach pilesoil interface. With reference to the present
while retaining a highly computationally efcient problem which involves an unloaded ground
10
code. The various strategies adopted for ach- surface, it is convenient to adopt Mindlin's
ieving efciency gains result in an economically solution for a point load within a homogeneous,
viable analysis even if non-linearity effects are isotropic elastic half-space. Interaction effects in
simulated in large pile groups. Among these is non-homogeneous soil may be approximated by
the option of making use of symmetries between using an averaging procedure. The soil defor-
piles which carry, even if approximately, equal mations at the pilesoil interface can be related
loads. If these symmetries are exploited, the size to the soil tractions by integration of the
of the global matrix of the group may be Mindlin's kernel, yielding
reduced with substantial savings in computa- fus g [Gs ]fts g (1)
tional costs. PGROUPN models the soil at the
interface as an elastic-perfectly-plastic material, where fus g is the soil displacement, fts g is the
which is assumed to behave linearly elastic at ctitious soil traction and [Gs ] is a exibility
small strain levels, but fails when the stresses at matrix of coefcients obtained from Mindlin's
the pilesoil interface reach certain limiting solution for the axial and lateral response. The
values (determined from a limit equilibrium off-diagonal exibility coefcients are evaluated
analysis). The only soil parameters required are by approximating the inuence of the continu-
the proles of the initial (tangent) Young's ously distributed loads by discrete point loads
modulus and the undrained shear strength. The applied at the location of the nodes. The singular
idealization of soil stiffness varying proportion- part of the diagonal terms of the [Gs ] matrix are
ally with depth (Gibson soil) is appropriate for calculated via analytical integration of the Mind-
9
most practical cases. Nevertheless, extension of lin functions, which has yielded very accurate
the model to handle layered soils will be an results with considerable saving of computing
11
object of future work. time. This is a signicant advance over
5
10. In the following sections, the BEM previous work where these have been inte-
solution incorporated in the PGROUPN program grated numerically, since these singular integrals
is presented, and its validity is examined by require considerable computing resources.
comparison with the response obtained from Further computational efciency is achieved by
alternative methods of analysis and eld tests. exploiting symmetries and similarities in forming
Further, in order to highlight the signicant single-pile and interaction exibility matrices.
inuence of non-linearity on load distribution This reduces the computation time and renders
among piles in a group, a parametric study in the analysis practical even for large pile groups.
the linear and non-linear range is presented.
Pile deformation
13. If the piles are assumed to act as simple
Method of analysis beam-columns which are xed at their heads to
11. The PGROUPN program is based on an the pile cap, the displacements and tractions
extension of the indirect boundary element over each element can be related to each other
5
algorithm by Banerjee and Driscoll which is via the elementary beam theory, yielding
further developed in order to make the analysis fup g [Gp ]ftp g fBg (2)
more numerically efcient and to include effects
of soil non-linearity by means of a stepwise where fup g is the pile displacement, ftp g is the
linear incremental procedure. The present pile traction, fBg is the pile displacement due to
analysis involves discretization of only the pile unit boundary displacements and rotation of the
soil interface into a number of cylindrical pile cap and [Gp ] is a matrix of coefcients
elements, while the base is represented by a obtained from BernoulliEuler beam theory.
circular (disc) element. The behaviour of each
element is considered at a node which is located Solution of the system
at the mid-height of the element on the centre- 14. Coupling equations (1) and (2) via
line of the pile. The soil and pile compliance compatibility and equilibrium constraints at the
equations are assembled for each element, and pilesoil interface, leads to the following system
compatibility and equilibrium are imposed at the of equations
interface. Given unit boundary conditions, that
ftp g [Gp Gs ]1 fBg (3)
is, pile group loads and moments, these equa-
tions are solved, thereby leading to the dis- where [Gp Gs ] is the global square matrix of
tribution of stresses, loads and moments in the the pile group.
piles for any loading condition. 15. By successively applying unit boundary
107

Downloaded by [ University College London] on [21/09/16]. Copyright ICE Publishing, all rights reserved.
BASILE

conditions, that is, unit vertical displacement, 18. Since the pile is assumed to remain
unit horizontal displacement and unit rotation, to elastic during loading, the pile matrix [Gp ] is
the pile cap, it is possible to obtain the system unaltered. It is possible to incorporate the
of vertical loads, horizontal loads and moments effects of soil yielding, in an approximate
acting on the cap that are necessary to equi- manner, by assuming that the interface beha-
librate the stresses developed in the piles. Thus, viour at an element is perfectly linear until the
the external vertical load V tot , horizontal load limiting stress (determined from equations (5)
H tot and moment M tot acting on the cap can be and (6)) is reached, specically bearing stress in
related to the corresponding vertical displace- the compression zone and shear stress in the
ment w, horizontal displacement u and rotation slip zone. For these elements which have gone
of the cap via plastic, no more increment in tractions is
8 9 8 9 permitted. This can be accomplished by apply-
<V = <w=
ing the external load incrementally so that the
H [K ] u (4)
: ; : ; soil matrix [Gs ] changes as one or more
M
elements reach the yield conditions. The re-
where the coefcients of the (3 3 3) [K ] matrix maining elastic elements are those given by the
are the equilibrating forces as discussed earlier. Mindlin's solution and there is no further
The [K ] matrix can be regarded as the global interaction through the soil between these
stiffness matrix of the pilesoil system which elements and the yielded elements in the pile
may be used as boundary conditions for the group. It may be noted that the manner in
superstructure analysis. By inverting this matrix, which the non-linear response of the pile group
it is possible to obtain the global exibility matrix is obtained is approximate. In fact, the yielding
of the pilesoil system and hence the vertical of an element introduces a discontinuity in the
displacement, the horizontal displacement and material property. Thus, the use of Mindlin's
rotation of the pile cap may be obtained for any solution is only approximate.
loading condition. In order to obtain the tractions 19. This procedure enforces load transfer to
acting on the piles for the prescribed loading the elastic elements in that the element which
conditions, pile tractions due to unit boundary has failed can take no additional load and the
conditions must be scaled using the cap dis- increase in load is thus redistributed between
placements and rotation obtained from equation the remaining elastic elements until all elements
(4). Finally, integrating the axial and transverse have failed. Thus, the non-linear response of the
tractions acting on the piles, yields the distribu- group may be obtained by casting the linear
tion of axial forces, shear forces and moments equations previously described in the incremen-
acting on each pile. tal form and solving using small load incre-
ments.
Limit stresses
16. The elastic equations previously de- Numerical results
scribed are valid only if the stress state in the 20. The validity of the proposed approach
soil does not violate the yield criteria. For has been veried by comparison with alternative
cohesive soils, the limiting bearing stress on the methods of analysis for piles and pile groups
compressive zone (i.e. the pile base for the axial embedded in a homogeneous soil layer. In the
response and the pile shaft for the lateral numerical solutions presented in the following
response) can be expressed as paragraphs, the soil Poisson's ratio is taken as
t sc Nc C u (5) 05, unless otherwise stated. As regards pile
discretization, the adopted height-to-diameter
while the limiting shear stress in the slip zone ratio of the pile shaft element is 2. For practical
(i.e. the pile shaft for the axial response) is taken purposes a coarser discretizing of the piles
as would be adequate. The computational work is
t ss C u (6) done on a Pentium 133 MHz, 16 Mb RAM and
the average CPU time for the simulations on the
where Nc is the bearing capacity factor, C u is the nine-pile group is about 1 s for the linear
undrained shear strength of the soil and is the analyses and 70 s for the non-linear analyses.
adhesion factor.
Axial response
Extension to non-linear soil behaviour 21. The accuracy of PGROUPN is rst eval-
17. The linear analysis previously described uated for a single pile in a homogeneous soil
can be extended to include non-linear effects by layer and compared with alternative numerical
generalizing an idea rst proposed by Poulos analyses. The input parameters are reported in
12
and Davis. In the present analysis, the Table 1. Two values of pile Young's modulus
boundary element equations used for the linear have been considered, 30 GPa and 30 000 GPa
response are solved incrementally while enfor- (the latter would be unrealistically stiff in
cing the conditions of yield, equilibrium and practice). Fig. 1 shows the computed pile head
compatibility at the pilesoil interface. loadsettlement response, together with the
108

Downloaded by [ University College London] on [21/09/16]. Copyright ICE Publishing, all rights reserved.
NON-LINEAR ANALYSIS
OF PILE GROUPS

Table 1. Parameters for the analysis reported in 18


Fig. 1 16 Ep 5 30000 GPa
Parameter Value 14 Ep 5 30 GPa
Pile length, L: m 30

Pile head load: MN


12
Pile diameter, d: m 075
Depth of soil layer: m 50 10
Pile Young's modulus, Ep : GPa 30, 30 000
Young's modulus, Es : MPa 1056 8
Poisson's ratio, s 049 6
Limiting shear stress, t ss : kPa 220 Jardine et al.14
4 PGROUPN
2 Poulos13
results from a boundary element analysis by
13 0
Poulos, which adopts an elastic-plastic con- 0 5 10 15 20
tinuum-based interface model (similar to Head settlement: mm
PGROUPN), and from a nite element analysis by
14
Jardine et al., which involves the use of a non- Fig. 1. Comparison of loadsettlement curve in single pile
linear soil model in which the (initial tangent)
Young's modulus decreases markedly as the
axial strain level increases.
22. It is worth noting that, for the more 12
compressible (and realistic) pile, the approx-
L /d 5 25
imate method to account for non-linearity
Normalized group stiffness: k p /(sGn0 5)

10 Ep /G 5 1000
.

employed by Poulos and PGROUPN is capable of s /d 5 2.5


predicting a very similar loadsettlement re-
sponse to that obtained from the FEM analysis 8
which utilizes a proper non-linear constitutive
model of soil behaviour. For the stiffer pile, the
6
agreement between the curves is not as close,
the greatest potential for differences being for
loads of more than one-half of the ultimate. It is 4
clear that, for very stiff piles, the details of the DEFPIG GRUPPALO
pilesoil interface model have a greater inu- PGROUP PIGLET Limiting stiffness
2
ence on the loadsettlement response than for PGROUPN
more compressible piles.
23. In order to investigate the effectiveness 0
0 5 10 15 20
of the proposed approach in the linear elastic
Square root of number of piles in group
range, Fig. 2 compares PGROUPN results with
those obtained by some of the computer codes (a)
15
mentioned earlier. Results are expressed in
termspof
the normalized group stiffness 9
kp =( n sG) of square groups of piles at different L /d 5 25
spacings (where kp is the ratio of the total 8 Ep /G 5 1000
Normalized group stiffness: k p /(sGn0 5)

vertical load acting on the group to the average


.

s /d 5 5
settlement of the group, n is the number of 7
piles in the group, s is the pile spacing and G is 6
the soil shear modulus). It has been assumed
that interaction effects between piles become 5
insignicant for pile spacing greater than a
16
limiting value smax , which can be dened as 4
Limiting stiffness
smax 2:5L(1 s ) rg (7) 3
DEFPIG GRUPPALO
where, for rectangular pile group congurations, 2
PGROUP PIGLET
rg may be taken as the radius of the circle of
1 PGROUPN
equivalent area to that covered by the pile group.
24. It may be seen that DEFPIG and PIGLET 0
approaches give divergent results, while a rea- 0 5 10 15 20
sonable agreement is obtained between Square root of number of piles in group
PGROUPN and the computer program (b)
17
GRUPPALO, which is based on the interaction
factor method. Results from PGROUPN are in Fig. 2. Comparison of different pile group analysis methods for: (a)
excellent agreement with the rigorous BEM s=d 2 :5; (b) s=d 5
109

Downloaded by [ University College London] on [21/09/16]. Copyright ICE Publishing, all rights reserved.
BASILE

solution of PGROUP, but the latter is limited to 1.6 s


groups of 8 3 8 piles, due to the magnitude of 1.4
1
Pile 1
computer resources required to analyse larger 3
1.2 2
groups. In contrast, PGROUP took about 30 CPUs
for the 20 3 20 pile group, considering the 1.0
symmetry of the pile arrangement. This obser-

P/Pav
vation is of great signicance because it 0.8 Pile 2

demonstrates the applicability of the complete 0.6


BEM to large pile groups, whereas previous PGROUP & PGROUPN
0.4 (linear)
work (i.e. PGROUP) was restricted to small Poulos and Davis19
groups of piles. 0.2 Pile 3 PGROUPN (non-linear)
25. It is worthy of note that, for very large
0
pile groups, where the ratio of pile group width 0 2 4 6 8 10
to pile length becomes much greater than unity, s /d
the group stiffness should approach that of a
shallow foundation. This would correspond to a Fig. 3. Comparison of axial load distribution to individual piles in 3 3 3
18
limiting stiffness of about 45 as indicated in pile group
Fig. 2.
26. Figure 3 shows the effects of non-
Table 2. Parameters for the analysis reported in
linearity on the axial load distribution between a
Fig. 3
group of 3 3 3 piles embedded in a homoge-
neous soil layer. The input parameters of the Parameter Value
analysis are given in Table 2. The load
Pile length, L: m 25
distribution is expressed in terms of the ratio of Pile diameter, d: m 1
load on pile to the average pile load in the Pile Young's modulus, Ep : GPa 25
group (P=Pav ), and is plotted against the Young's modulus, Es : MPa 25
normalized pile spacing (s=d). In the linear Undrained shear strength, C u : kPa 25
elastic range, the load distribution predicted by Adhesion factor, 05
PGROUPN compares favourably with those pre- Applied axial load on the group, Ptot : MN 8
dicted by PGROUP and the interaction factor
19
approach of Poulos and Davis. It may be
observed that the solutions of PGROUP and
PGROUPN (in the linear range) are almost moment of area of pile section), which corre-
indistinguishable to plotting accuracy. The non- sponds to relatively exible piles. In order to
linear response, as predicted by PGROUPN, has analyse piles commonly encountered in practice,
been obtained by applying a total load on the Fig. 5 adopts a stiffness ratio (K R ) of 104 . The
group of 8 MN, which is about two-thirds of the details of the input parameters used in the
ultimate load. It is evident that consideration of solutions shown in Figs 4 and 5 are reported in
the non-linear response yields a reduction in the Table 3.
load concentration at the corner piles (pile 1) 28. In the linear elastic range, the PGROUPN
and a more even load distribution. Clearly, the solutions compare favourably with PGROUP (the
higher the load level, the higher the reduction two solutions are indistinguishable graphically
in the load concentration at the corner piles on this scale), while the agreement with the Fig. 4. Comparison of
20
obtained by the non-linear analysis. hybrid method of Leung and Chow, in which lateral load
the individual pile response is modelled using distribution in xed
Lateral response load-transfer ( p- y) curves and the group inter- head 3 3 3 pile group
27. The predicted lateral load distribution to action is determined based on Mindlin's solu- (K R 105 )
the individual piles in xed-head pile groups is
examined by comparison with alternative meth-
ods, and the inuence of soil non-linearity is 1.6 s
discussed. It should be emphasized that the 1.4
1
numerical simulations presented later take no Pile 1 H tot
2
account of possible failure by yielding of the pile 1.2
section, that is, the pile is assumed to remain 1.0
elastic. Figs 4 and 5 show the lateral load
H/Hav

0.8
distribution (where H is the load acting on the
individual pile head and Hav is the average load 0.6 Poulos and Davis19
acting on each pile head) as a function of the Pile 2 Leung and Chow20
0.4
normalized pile spacing (s=d) for square groups PGROUP & PGROUPN
(linear)
of piles embedded in a homogeneous soil layer. 0.2
PGROUPN (non-linear)
Fig. 4 shows the distribution of lateral load for a 0
pile group having a stiffness ratio K R 0 2 4 6 8 10
Ep I p =Es L 4 of 105 (where I p is the second s /d

110

Downloaded by [ University College London] on [21/09/16]. Copyright ICE Publishing, all rights reserved.
NON-LINEAR ANALYSIS
OF PILE GROUPS

tion, is reasonable except for closely spaced pile 1.4 s


groups. Instead, some discrepancies are ob- 1
19 1.2 Pile 1
served with the analysis of Poulos and Davis H tot
2
in which the two-pile interaction factors are 1.0
merely superimposed, thereby ignoring the
stiffening effect of piles within the soil mass. 0.8

H/Hav
This procedure results in an overestimation of 0.6 Pile 2
group interaction and hence in a marked non- PGROUP & PGROUPN
uniformity of load distribution to the individual 0.4 (linear)

piles. In addition, it is worth noting that non- PGROUPN (non-linear)


0.2
uniformity of load distribution becomes more
signicant as K R increases (and thus Es de- 0
creases). 0 2 4 6 8 10
29. Consideration of soil non-linearity (as s /d
predicted by PGROUPN) leads to a further Fig. 5. Comparison of lateral load distribution in xed head 3 3 3 pile
reduction in the load concentration at the corner group (K R 104 )
piles and hence a more even load distribution.
The amount of this reduction will depend on the
load level.
Table 3. Parameters for the analyses reported in Figs 4 and 5
Application to published eld studies Parameter Fig. 4 Fig. 5
30. Further application of the PGROUPN
Pile length, L: m 25 25
analysis is examined by comparison with the
Pile diameter, d: m 1 1
axial and exural response of full-scale piles and Pile Young's modulus, Ep : GPa 25 25
pile groups. Three well-known case histories Young's modulus, Es : MPa 314 314
have been identied. These have been already Undrained shear strength, C u : kPa 100 50
used by other authors as a benchmark to assess Adhesion factor, 05 05
the accuracy of methods of pile group analysis. Applied lateral load on the group, Htot : MN 10 10
Stiffness ratio, K R 105 104
Axial response
31. The single-pile response is compared
with an axial load test on concrete piles in
21
alluvium reported by Gurtowski and Wu. The Table 4. Parameters for the analysis reported in
13
input parameters used by Poulos have been Fig. 6
adopted directly, as shown in Table 4. These are
based on a prole of standard penetration test Parameter Value
(SPT)N-values increasing linearly with depth Pile length, L: m 30
from zero at ground surface to 60 at the pile Pile diameter, d: m 063
base level. It is worth noting that the PGROUPN Pile Young's modulus, Ep : GPa 35
non-linear analysis has been described in terms Young's modulus, Es : MPa 4N
of the undrained shear strength of the soil Poisson's ratio, s 03
(refer to equations (5) and (6)); however, in this Limiting shear stress, t ss : kPa 2N
test, the values of the limiting shear stress and End-bearing pressure, t sc : MPa 04N
end-bearing pressure have been input directly.
Fig. 6 compares the pile head loadsettlement
response measured by Gurtowski and Wu with 6
those predicted by PGROUPN and the BEM
analysis (with elasto-perfectly-plastic soil model) 5
by Poulos. A generally good agreement is
obtained between numerical analyses and eld
Pile head load: MN

4
test data.
22
32. O'Neill et al. reported the results of
3
axial loading tests on single piles and pile
groups driven into stiff overconsolidated clay. Measured21
Nine of the piles were arranged in a 3 3 3 2
Poulos13
conguration with centre-to-centre spacing
s 3d. The steel pipe piles were connected to 1
PGROUPN
a massive reinforced concrete cap with a
clearance of 09 m from the groundline. This 0
23 0 5 10 15 20 25
test was analysed by Chow by means of a
Head settlement: mm
hybrid method in which the response of the
individual piles is modelled using the load- Fig. 6. Comparison between predicted and measured loadsettlement
transfer (t-z) method and the interaction be- behaviour in single pile
111

Downloaded by [ University College London] on [21/09/16]. Copyright ICE Publishing, all rights reserved.
BASILE

tween the piles is effected using Mindlin's 700


solution. In order to ensure consistency be-
tween analyses, the set of soil properties 600
employed by Chow is used in this study as
follows: the approximate linearly increasing 500
prole of the (initial tangent) soil modulus is
deduced from the cross-hole data of Kraft et 400

Load: kN
24
al. , while the shaft resistance is given by
back-analysis of the single pile tests. The 300 Measured22
parameters used in the analysis are reported in
Chow23
Table 5. 200
33. Figures 7 and 8 show a very good PGROUPN
agreement between the computed and measured
100
loadsettlement behaviour of the reference
single pile, the four-pile subgroup (piles 2) and
0
the nine-pile group. It therefore appears that the 0 1 2 3 4 5 6 7 8
initial tangent soil modulus, derived from seis- Settlement: mm
mic cross-hole data, may be successfully em-
ployed in the prediction of the pile settlement, Fig. 7. Comparison between predicted and measured loadsettlement
thereby conrming the ndings of Mandolini behaviour in single pile
17
and Viggiani. The computed load distribution
with depth among the piles is compared with 6 s s
the measured values for the nine-pile group at a
9-pile group
working load of 258 MN, as shown in Fig. 9. It 5 1 2 1
is worth noting that, even at a working load 2 3 2
4
level, the mild non-linearity in the computed
Load: MN

1 2 1
solution results in a reduction of the load 3
concentration at the outer piles and hence in a
Measured22
more uniform load distribution. This improves 2 4-pile subgroup
on the agreement between the computed and PGROUPN
the measured values. 1
Chow23
0
Lateral response 0 2 4 6 8 10
34. Field tests were conducted by Matlock et Settlement: mm
25
al. on a single-pile, ve-pile and ten-pile
circular groups embedded in soft clay. All piles Fig. 8. Comparison of predicted and measured load-settlement behaviour in
were made of steel and the centre-to-centre four-pile subgroup and nine-pile group
spacing was 34 and 18 pile diameters for the
ve-pile and ten-pile groups, respectively. Pile- strength is deduced from an in situ vane-shear
head deections were enforced at two elevations device, while an empirical correlation factor of
(one at 023 m above the groundline and the Es =C u 150 is adopted for the Young's mod-
other near the pile top) by a special loading ulus of soft clay. The input parameters are
device to simulate pile-head restraints typical of shown in Table 5.
offshore structures. In this study, the set of soil 35. Figures 10 and 11 compare the meas-
parameters used follows that reported by ured load-deection response for the single pile,
26
Bogard and Matlock and employed by Leung the ve-pile and ten-pile groups with those
20
and Chow, as shown later: an approximate predicted by PGROUPN and the hybrid approach
linearly increasing prole of the undrained shear of Leung and Chow (described earlier). It has

Table 5. Parameters for the analyses reported in Figs 711


Parameter Figs 79 Figs 1011

Embedded pile length, L: m 131 116


Pile external diameter, d: m 02740 01680
Pile internal diameter: m 02554 0.1538
Centre-to-centre pile spacing, s: m 08220 05712, 03024
Young's modulus of steel piles, Ep : GPa 210 210
Young's modulus at ground level, Eso : MPa 144 1575
Rate of increase of Young's modulus with depth: MPa=m 236 039
Poisson's ratio, s 05 05
Undrained shear strength at ground level, C uo : kPa 479 105
Rate of increase of undrained shear strength with depth: kPa=m 146 26
Adhesion factor, 034 050

112

Downloaded by [ University College London] on [21/09/16]. Copyright ICE Publishing, all rights reserved.
NON-LINEAR ANALYSIS
OF PILE GROUPS

Load/average pile head load 30


0 0.5 1.0 1.5
0 25

2 20

Pile shear: kN
Depth below pile cap level: m

4 15

6 10
PGROUPN

8 5 Measured25
Fig. 10. Comparison
between predicted and
0 measured lateral load-
10 0 5 10 15 20 25 30 deection response in
Lower support deflection: mm single pile
12

14 30
(a)

0 0.2 0.4 0.6 0.8 1.0 25


0
Average pile shear: kN

2 20
Depth below pile cap level: m

4 15

6
10
Measured25
8
Leung and Chow20
5
10 PGROUPN
0
12 0 10 20 30 40 50
Lower support deflection: mm
14 (a)
(b)

30
0 0.2 0.4 0.6 0.8 1.0
0
25
2
Average pile shear: kN
Depth below pile cap level: m

20
4

6 15

8
10
Measured25
10
5
Leung and Chow20 Fig. 11. Comparison
12
between predicted and
PGROUPN
measured lateral load-
14
0 deection response in
0 20 40 60 80
Measured22 pile groups for: (a)
Lower support deflection: mm
ve-pile group; and (b)
PGROUPN (non-linear) (b) ten-pile group
PGROUPN (linear)

(c)
been found that about 100 load increments are
Fig. 9. Comparison between computed and generally sufcient to achieve convergence of
measured axial load distribution in nine-pile the PGROUPN solution process. It is worth noting
group for: (a) corner pile; (b) mid-side pile; and that, due to the effect of pilesoilpile inter-
(c) centre pile action, the response of the pile group becomes
113

Downloaded by [ University College London] on [21/09/16]. Copyright ICE Publishing, all rights reserved.
BASILE

more linear with increasing the number of piles proposed method, taking into account the
in the group. This trend has already been continuous nature of pilesoil interaction, re-
observed in axially loaded pile groups. Reason- moves the uncertainty of empirical load-transfer
ably good agreement with the measured values approaches and provides a simple design tool
is achieved for the single pile and the ve-pile based on conventional soil parameters.
group, while the PGROUPN response for the ten-
pile group is too stiff at high load levels. Such Acknowledgements
discrepancies (also noted in the solution of 39. The author is greatly indebted to Dr W.
Leung and Chow) may be partially attributed to G. K. Fleming of Kvaerner Cementation Found-
overlapping of failure zones and gap formation ations Ltd whose encouragement, concern and
behind the closely spaced piles. This `shadow- insight have been crucial to the development of
ing' effect, which leads to increased group this project. Special thanks also go to Professor
deections, cannot be readily reproduced in the S. J. Wheeler and Professor N. Bicanic, Uni-
model. However, nite element analyses by versity of Glasgow, for their support and super-
27
Brown and Shie and model tests by Cox et vision over the course of this work.
28
al. show that shadowing effects rapidly de-
crease as the spacing between piles increases
and become insignicant for spacing greater
References
than about six pile diameters. 1. OTTAVIANI M. Three-dimensional nite element
analysis of vertically loaded pile groups. Geotech-
Conclusions nique, 1975, 25, No 2, 159174.
36. The application of available numerical 2. BUTTERFIELD R. and BANERJEE P. K. The elastic
methods to realistic pile group problems is analysis of compressible piles and pile groups.
decient because these are mainly restricted to Geotechnique, 1971, 21, No. 1, 4360.
linear elastic analyses and relatively small 3. RANDOLPH M. F. PIGLET, a Computer Program for
groups of piles. The assumption of a linear the Analysis and Design of Pile Groups. Report GEO
87036, University of Western Australia, Perth,
elastic soil model is unrealistic for most soils
1987.
and the choice of a suitable secant soil modulus
4. POULOS H. G. User's Guide to Program DEFPIG
is by no means straightforward. In addition, for Deformation Analysis of Pile Groups. School of Civil
pile group systems designed according to the Engineering, University of Sidney, 1980.
deformation-based approach, and hence charac- 5. BANERJEE P. K. and DRISCOLL R. M. Program for
terized by a low safety factor, consideration of the Analysis of Pile Groups of any Geometry
soil non-linearity becomes compulsory. Further, Subjected to Horizontal and Vertical Loads and
current methods suffer from limitations imposed Moments, PGROUP, (2.1). Department of Transport,
by the number of piles in the group and the HECB/B/7, London, 1978.
global dimension of the problem which renders 6. O'NEILL M. W., GHAZZALY O. I. and HA H. B.
Analysis of three-dimensional pile groups with non-
such analyses computationally inefcient.
linear soil response and pilesoilpile interaction.
37. In this paper, an attempt at removing
Proceedings of the 9th Offshore Technology Con-
these limitations has been made. A highly ference, Houston, 1977, 245256.
efcient code (called PGROUPN) has been 7. CHOW Y. K. Three-dimensional analysis of pile
presented for the linear and non-linear analysis groups. Journal of Geotechnical Engineering,
of pile groups under general loading conditions, American Society of Civil Engineers, 1987, 113, No.
by which problems of large dimensions may be 6, 637651.
investigated with an ordinary desktop computer. 8. BRITISH STANDARDS INSTITUTION. BS 5400. Code of
The proposed approach, based on the complete Practice for Fatigue: Steel, Concrete and Composite
BEM, simulates the effects of soil non-linearity Bridges. BSI, London, 1994, part 4, table 1, p. 9.
9. FLEMING W. G. K., WELTMAN A. J., RANDOLPH M.
by means of a stepwise linear incremental
F. and ELSON W. K. Piling Engineering. Blackie
procedure which requires specication of only
Academic & Professional, Glasgow, 1992, 2nd edn.
one additional soil parameter, the undrained 10. MINDLIN R. D. Force at a point in the interior of a
shear strength distribution with depth. This semi-innite solid. Physics, 1936, 7, 195202.
parameter is routinely measured in soils inves- 11. BASILE F. Non-linear Analysis of Pile Groups under
tigation. General Loading Conditions. Forthcoming PhD
38. It has been shown that the PGROUPN thesis, University of Glasgow, 1999.
solution avoids exaggeration of pile loads at 12. POULOS H. G. and DAVIS E. H. The settlement
group extremities which is common with linear behaviour of single axially loaded incompressible
elastic models, and reduces consequent high piles and piers. Geotechnique, 1968, 18, No. 3,
351371.
loads and moments. The general good agree-
13. POULOS H. G. Pile behaviourtheory and applica-
ment with the results of eld tests and
tion. Geotechnique, 1989, 39, No. 3, 365415.
alternative numerical analyses lends condence 14. JARDINE R. J., POTTS D. M., FOURIE A. B. and
in the applicability of the method to practical BURLAND J. B. Studies of the inuence of non-
problems, offering the prospect of tangible linear stressstrain characteristics in soilstruc-
improvements in design procedures and worth- ture interaction. Geotechnique, 1986, 36, No. 3,
while savings in construction costs. Further, the 377-396.
114

Downloaded by [ University College London] on [21/09/16]. Copyright ICE Publishing, all rights reserved.
NON-LINEAR ANALYSIS
OF PILE GROUPS

15. RANDOLPH M. F. Design methods for pile groups 23. CHOW Y. K. Analysis of vertically loaded pile
and piled rafts. Proceedings of the 13th Interna- groups. International Journal for Numerical and
tional Conference on Soil Mechanics Foundation Analytical Methods in Geomechanics, 1986, 10, No.
Engineering, New Delhi, 1994, 5, 6182. 1, 5972.
16. RANDOLPH M. F. and WROTH C. P. An analysis of 24. KRAFT L. M., RAY R. P and KAGAWA T. Theoretical
the vertical deformation of pile groups. Geotech- t-z curves. Journal of Geotechnical Engineering,
nique, 1979, 29, No. 4, 423439. ASCE, 1981, 107, No. GT11, 15431561.
17. MANDOLINI A. and VIGGIANI C. Settlement of piled 25. MATLOCK H., INGRAM W. B., KELLEY A. E. and
foundations. Geotechnique, 1997, 47, No. 4, BOGARD D. Field tests of the lateral-load behaviour
791816. of pile groups in soft clay. Proceedings of the 12th
18. FRASER R. A. and WARDLE L. J. Numerical analysis Offshore Technology Conference, Houston, 1980,
of rectangular rafts on layered foundations. Geo- 163174.
technique, 1976, 26, No. 4, 613630. 26. BOGARD D. and MATLOCK H. Procedures for
19. POULOS H. G. and DAVIS E. H. Pile Foundation analysis of laterally loaded pile groups in soft clay.
Analysis and Design. Wiley, New York, 1980. Proceedings of a Conference on Geotechnical Practice
20. LEUNG C. F. and CHOW Y. K. Response of pile in Offshore Engineering, Austin, 1983, 499535.
groups subjected to lateral loads. International 27. BROWN D. A. and SHIE C. F. Numerical experi-
Journal for Numerical and Analytical Methods in ments into group effects on the response of piles
Geomechanics, 1987, 11, No. 3, 307314. to lateral loading. Computers and Geotechnics, 1990,
21. GURTOWSKI T. M. and WU M. J. Compression load 10, No. 4, 211230.
tests on concrete piles in alluvium. Analysis and 28. COX W. R., DIXON D. A. and MURPHY B. S.
Design of Pile Foundations (Meyer J. R. (ed.)). Lateral load tests on 254-mm (1-in.) diameter
American Society of Civil Engineers, 1984, pp. piles in very soft clay in side-by-side and in-line
138153. groups. In Laterally Loaded Deep Foundations:
22. O'NEILL M. W., HAWKINS R. A. and MAHAR L. J. Analysis and Performance, ASTM STP 835 (Langer
Load transfer mechanism in piles and pile groups. J. A., Mosley E. T. and Thompson C. D. (eds)).
Journal of Geotechnical Engineering, ASCE, 1982, American So-ciety for Testing and Materials, 1984,
108, No. GT12, 16051623. pp. 122139.

Please email, fax or post your discussion contributions to the publisher: email:
ttjournals@ice.org.uk; Fax: 0171 538 9620; or post to Terri Harding, Journals Department,
Thomas Telford Limited, Thomas Telford House, 1 Heron Quay, London E14 4JD.

115

Downloaded by [ University College London] on [21/09/16]. Copyright ICE Publishing, all rights reserved.

S-ar putea să vă placă și