Professional Documents
Culture Documents
Pile Group Analysis: A Study of Two Methods: by Harry G. Poulos, M. ASCE and Mark F. Randolph
Pile Group Analysis: A Study of Two Methods: by Harry G. Poulos, M. ASCE and Mark F. Randolph
ABSTRACT: A brief outline is given of the basis of analysis used in two com-
puter programs, PIGLET and DEFPIG, for analysis of pile group response. A
comparison of the capabilities of each program is presented and then a para-
metric study is made of a six-pile group to illustrate the similarities and differ-
ences between the two programs. Two series of field tests on groups of piles
are then analysed by DEFPIG and PIGLET, and it is shown that the predictions
from each are consistent and in generally good agreement with the field data.
The comparisons suggest that each program may be used with confidence for
practical problems, particularly in conjunction with single pile load test data.
INTRODUCTION
355
uses integral equation analysis to calculate the response of a single pile
to axial and lateral loading, and also to calculate interaction factors (in
the sense of Poulos (7)) between pairs of adjacent piles. The latter pro-
gram uses solutions in the form of algebraic expressions giving the re-
sponse of a single pile to axial, lateral, and torsional loading. These
expressions, appropriately modified to allow for interaction between piles,
have been developed from finite element studies of a pile embedded in
elastic soil. Both programs subsequently employ a form of frame analysis
in order to calculate the overall behavior of the pile group in terms of
the responses of the individual piles.
This paper gives a brief outline of the two methods of analysis, and
presents a parametric study of a simple 3 x 2 pile group, showing the
deflection and load influence factors predicted by the two programs for
various angles of rake of the corner piles in the group. The two pro-
grams are then used to analyze two series of field tests on groups of
piles (for one of which predictions are also available from the program
PGROUP (2)). It is hoped by this means to demonstrate the extent to
which widely different approaches to pile group analysis may yet yield
similar results. By implication this allows greater confidence to be put
in each program than might otherwise be warranted.
1. The interaction factors for axial and lateral deflections and rotation
are computed for a range of pile spacings, or alternatively, may be input
as data if appropriate values are available.
2. For the specified group configuration, the program determines the
spacings between all piles and interpolates the interaction factor data to
evaluate the axial and lateral interaction factors for all piles. For battered
piles, in order to simplify the computations, the interaction factors are
taken to be those for vertical piles at a spacing corresponding to the
distance between the piles at a depth of one-third of the pile length
below the surface (9).
3. For the average loads and moment in the piles, the axial move-
ment, lateral movement, and rotation of a single pile are either input as
data (if known or previously computed) or computed by the program.
In the latter case, allowance can be made for nonlinear pile response by
specifying limiting values of pile-soil stress (axial and lateral) at the var-
ious elements along the pile.
356
4. The following equation relating pile movements and pile loads is
then assembled:
[A]{P} = {5} (1)
in which {P} = vector of pile loads and moments; {8} = vector of pile
head deflections and rotations; and [A] = flexibility matrix of the group.
For a group of n piles, {P} and {8} are of order 3w while [A] is 3n x 3n.
The elements of [A] depend on the axial and lateral interaction factors,
the single pile responses, and the pile batter angles, and are detailed in
Poulos (9).
5. The equilibrium equations and pile-head boundary conditions are
incorporated into Eq. 1 and the resulting equation then solved. When
the group vertical load, horizontal load, and moment are specified, the
solution gives n vertical loads, n horizontal loads, n pile-head moments,
and the mean vertical deflection, horizontal deflection, and rotation of
the pile cap.
G = G0 + mz (2)
357
the following approximate solutions for the axial and torsional response
may be obtained.
Axial Response.—For an applied load V, the axial movement v of a pile
of length I, radius r0 and Young's modulus Ep (assuming the pile is solid)
is given by
16 Z tanh (t\l)
h 4ir p
.3 r0 yl . (4)
G/o4> t | 32 G, / tanhW)
3irpG r0 r\l
in which t)l = (8G;/Gp)1/2 Z/r0. The similarity of the aforementioned two
expressions is perhaps misleading since, for most practical cases, the
quantity r\l is much greater than the corresponding quantity pi, and is
such that tanh (iqZ) = 1. This allows Eq. 4 to be simplified. It has been
shown by Randolph (13) that, for piles which are of a length / such that,
1
>P (5)
r0 \GJ
simplification of Eq. 4 allows the torsional response to be expressed as
T G_
4.44 (6)
G/fo
in which Ge is an effective shear modulus for the soil, defined by
Ge = G0 + mle (7)
and W S (8)
rn \G.
As before, G is the average shear modulus over the effective length le
of the pile.
To evaluate the response of piles to lateral loading, Randolph (14) has
fitted the results of a parametric study using finite element analysis, by
simple algebraic expressions giving the ground level deflection h and
rotation 8, due to applied lateral load H and moment M. These expres-
sions are,
358
KG,.
h = 0.27 H + 0.3 M (9)
?fic
I = 2rr (11)
359
TABLE 2.—Program Statistics for DEFPIG and PIGLET
Feature DEFPIG PIGLET
(1) (2) (3)
Lines of code (excluding com-
ment lines) -2,200 -750
Number of segments 10 11
Storage of r variables (for piles) ~24«2 ~18M 2
CPU run time for 3 X 2 group (1) Full analysis: 15 min 24 sec 2.7 sec
(2) With interaction factors and
single pile responses input:
7.7 sec
V H M
Esa hsd Esd
V r _H_ _M_
®~Esd2hv+Esd2lm +
Esd3lm (12C)
For symmetric groups the deflection influence factors IvH, IVM, hv,
361
jf/d=25
s/d = 3
~d
Ep/E5 = 10 3
v.=0-5
(a)
|-s + s-J
1 "" •' 1
J
2 3 -----
f _ 4 5 6a-. ,-.-
y
(b)
*1 *2 3*'_
y
(c)
FIG. 1.—Pile Group Analyzed: (a) Elevation; (ft) Single Rake Plan; and (c) Double
Rake Plan
and 1^ will be zero. From the reciprocal theorem, Im and IhM will be
equal. For a typical pile with l/d = 25, K = Ep/Es = 103 and vs = 0.5, the
values of IvV, Im, and Im calculated by DEFPIG and by PIGLET are in
reasonably good agreement, the largest difference being for Im where
the value calculated by PIGLET is some 15% smaller than that calculated
by DEFPIG.
Corresponding interaction factors (following the notation of Poulos
(7)), for pairs of piles at different spacings are tabulated in Table 3. There
is a tendency for the factors calculated by PIGLET to be higher than
those calculated by DEFPIG, particularly at close spacings, but the agree-
ment is generally reasonably good, and the calculated stiffnesses for a
group of piles should agree to within about 20%.
Performance of 3 x 2 Pile Group.—The pile group chosen for the par-
ametric study is shown in Fig. 1. Analyses were performed using
DEFPIG and PIGLET for the case where the corner piles were battered
in the x-z plane at angles (4) of 0°, 7.5°, and 15° (see Fig. lb). Analyses
were also performed using PIGLET where the piles were battered in two
directions as shown in Fig. lc. These latter analyses were undertaken
in order to investigate the effect of the assumption that it is sufficient
to consider a single plane of horizontal loading and batter at any one
time (an assumption adopted in the program PGROUP as well as in
DEFPIG).
Figure 2 shows the deflection influence factors (as defined by Eqs. 12)
362
006 012
004
0-02
75
0004
0 0 0 2 -5
IhM,
IeH 0
\
7
-0002
PIGLET
— DEFPIG
PIGLET - Double rake
FIG. 2.—Variation of Group Deflection Influence Factors with Angle of Rake i|/
CQv CaH
01 0-1
I
0 75 15 0
>l>°
7-5 15 0
<l>° PIGLET
DEFPIG
~8%). However, the rake in the y-z plane introduced additional lateral
loads and moments; for example, the total moment on pile 6 due to a
unit vertical load on the group increased from 0.084d to (0.0702 +
0.1142)1'2 d = 0.134rf, an increase of 60%. Thus, some care is evidently
needed when idealizing a general pile group as one where piles only
rake in a single plane.
deflected
pile shape
456 in
1. Single pile elastic solutions from both PIGLET and DEFPIG for the
pile deflection at the lower support were obtained taking into account
the rotational restraint of the loading system. From plots of soil modulus
vs. deflection per unit load, the appropriate soil modulus was deter-
mined from the static single pile test data for five different load levels,
ranging from 1-5 tons. Table 4 gives these modulus values so deter-
mined, and shows very close agreement between the values from the
DEFPIG and PIGLET analyses. These modulus values range between
365
TABLE 4.—Backfigured Soil Modulus Values from Single Pile Test, from Ref. 5
Backfigured Soil Modulus, Es, in
Kips Per Square Inch
Load level, in kips DEFPIG PIGLET
(1) (2) (3)
1 0.68 0.72
2 0.55 0.59
3 0.46 0.49
4 0.35 0.39
5 0.28 0.30
Note: 1 kip = 4.54 kN; 1 kip/sq in. = 6.9 MPa.
about 150 c„ -350 cu (cu = undrained shear strength) and such values are
consistent with previous experience (1,8).
2. The behavior of the five- and ten-pile groups at each of the above
load levels (i.e. for the same load per pile in the group) was determined
from the programs, using the appropriate backfigured soil modulus. The
loading support system was assumed to apply the same relative restraint
to the group piles (between the free-head and fixed-head cases) as to
the single piles.
The results of the analyses for each group are compared with the
measured behavior in Fig. 5, in which the group deflection ratio Rp is
defined as the ratio of the group lateral deflection to the single pile lat-
eral deflection at the same load per pile. Figure 5 shows that the DEFPIG
predictions of group deflection ratio are lower than those from PIGLET.
This difference arises primarily because of the difference in the lateral
interaction factors computed by each program (see Table 3). The theo-
retical value of Rp increases with increasing load level because of the
decrease in soil modulus and the consequent increase in relative stiffness
of the pile. The value of Rp from the tests also increases with increasing
load level, and is generally bounded by the computed values.
o Measured
Computed by DEFPIG
Computed by PIGLET
(a) (b)
366
-100 "200
Bending Moment at Lower Bending Moment at Lower
Support (kip in) Support (kip in)
(a) (b)
Measured
Computed by DEFPIG
Computed by PIGLET
0 -100 -200
Bendiing Moment at Lower
Support (kip in)
(c)
0 10 10 2-0 10 20
Lower Support Lower Support Lower Support
Deflection (in) Deflection (in) Deflection (in)
(a) W (c)
For the single pile, five-pile group and ten-pile group, Fig. 6 compares
the measured bending moments at the lower support with those cal-
culated from DEFPIG and PIGLET, making the aforementioned as-
sumption regarding the restraint provided by the support system. Both
programs give almost identical results for the single pile and are in very
good agreement with the measured values. For the five- and ten-pile
groups, the DEFPIG values are again in good agreement with the mea-
367
TABLE 5.—iackfigured Values of Nh = d EJdz, from Ref. 4
Program Nh, in pounds per cubic inch
(1) (2)
DEFPIG 81
PIGLET 46
PGROUP 40
Note: 1 lb/in.3 = 271.5 kPa/m.
1. The theoretical single pile response was fitted to the measured re-
sponse by adjusting the values and distributions of soil Young's mod-
ulus Es and pile-soil yield pressure Py until satisfactory agreement was
obtained. As shown in Fig. 7a, a good overall fit was obtained by using
a constant value of Es = 0.55 kips/sq in. (3,790 kPa) and a Py distribution
which varied from 0.5 lb/sq in. (65 kPa) at the surface to 25 lb/sq in.
368
(172 kPa) at a depth of 45 in. (1.15 m) and remained constant thereafter.
2. Using the previously mentioned parameters, a fixed-head analysis
of the single pile was carried out in order to determine the ratio of the
fixed-head deflection to that for the restraint imposed by the experi-
mental setup. This ratio was found to decrease with load level, from
0.84 at a load of 1 kip (4.5 kN) to about 0.54 at a load level of 5 kips
(22.7 kN).
3. The five and ten pile groups were analyzed using the aforemen-
tioned values of Es and Py, assuming fixed-head conditions at the pile
head.
4. The group deflections thus computed were multiplied by the ratios
determined in Step 2 in order to obtain the final theoretical values.
369
of modulus with depth Nh was determined by fitting the theoretical de-
flection for an eight-pile group with vertical piles to the measured de-
flection. The values of Nh thus obtained are shown in Table 5, together
with the value obtained by Banerjee (1) who also analyzed these tests
using the program PGROUP which employs a boundary element anal-
ysis. It will be observed that there is a considerable difference in the
backfigured values of Nn, reflecting the differences in the elastic solu-
tions from the three programs for a nonhomogeneous soil mass. The
value of Nh from DEFPIG is about twice the value given by PGROUP
since the analysis employed by DEFPIG produces a "softer" lateral re-
sponse for a given modulus than the other two analyses.
The computed response of the groups from the three different anal-
yses is shown in Table 6, together with the measured response. Both
the PIGLET and DEFPIG solutions assume a rigid connection between
piles and cap, but allow for the possibility of cap rotation. Two sets of
results are given: (1) The horizontal deflection for a horizontal load of
5 tons per pile; and (2) the load per pile for a horizontal deflection of
0.25 inches (6.5 mm). As the three analyses used here are purely elastic,
the second set of results can be obtained directly from the first if the
same soil modulus is used for each.
Table 6 indicates that, for the case of a load of 5 tons/pile, the three
theories generally give quite similar results but all tend to over-predict
the group deflection. This is due largely to the fact that the real load-
deflection response of the groups is nonlinear and the soil modulus has
been obtained by fitting to the group with the softest response (i.e. the
largest deflection). If the lateral load for a group deflection of 0.25 in.
(6.3 mm) is considered, the agreement between measurement and the-
ory is better. Certainly, it can be seen that, with the exception of Test
7, the theoretical analyses are consistent with the field measurements,
indicating the beneficial effects of pile batter on lateral stiffness of the
group. It is not clear why the predictions of stiffness from DEFPIG and
PIGLET for Test 7 are so much smaller than either the PGROUP pre-
diction or the measured value, although it could be partly due to the
effect of pile batter on the interaction factors, which is not properly ac-
counted for in either PIGLET or DEFPIG. This effect may be more sig-
nificant when all the piles are battered in the same direction.
The predicted lateral deflections due to a vertical load of 20 tons (181.6
kN) per pile are in good agreement with the measurements and the val-
ues predicted by PGROUP.
CONCLUSIONS
ACKNOWLEDGMENTS
The work described in this paper forms part of a research project into
the behavior of offshore foundations being carried out at the University
of Sydney. The project is supported by a grant from the Australian Re-
search Grants Committee. The second writer was a visiting lecturer in
the School of Civil Engineering, the University of Sydney, and received
support from the Post-Graduate Civil Engineering Foundation of the
University of Sydney.
APPENDIX.—REFERENCES
372