SOILS AND FOUNDATIONS Vol. 39, No. 1, 97-111, Feb.

1999
Japanese Geotechnical Society

3-D ELASTO-PLASTIC FINITE ELEMENT ANALYSES OF PILE FOUNDATIONS

SUBJECTED TO LATERAL LOADING

AKIHIKO W AKAii), SHINGO GosEii) and KEIZO U GAl iii)

ABSTRACT
This paper is a study based on a number of model tests of free- or fixed-headed pile groups, each composed of nine
piles, subjected to lateral loading. The model piles made of aluminum were set in dry sand and laterally loaded. The
behaviors of the pile groups were then analyzed by the 3-D elasto-plastic finite element method (FEM), which
represents a realistic model to simulate the problem. This model includes elasto-plastic soil behavior with no-tension
characteristics, as well as thin frictional elements for slippage on the pile-soil interface. Parameter values for the sand
used in the analyses were determined from conventional triaxial compression tests. It is shown that the experimental
results can be precisely simulated by the analysis.
Furthermore, field tests of a prototype foundation of steel pile group (3 x 3 piles) subjected to lateral loading at its
footing were performed. The 3-D FE analyses were conducted to simulate the results. Parameter values for FEM have
been adjusted, using the back analysis technique, so as to simulate a load-displacement curve obtained from the single
pile loading test. As for the lateral load-displacement relationship and the bending strain distributions along each pile,
a good correlation between the experiments and the analyses can be seen in the case of the pile group.

Key words: coefficient of subgrade reaction, earth pressure, elasto-plasticity, field test, finite element method, horizon-
tal load, model test, pile group, sand, triaxial compression test (IGC: E4/E12)

et al., (1995b) conducted model tests on a group of nine

INTRODUCTION piles placed in sand and simulated the results in a reasona-
A large number of experiments have been previously bly exact manner. In addition to this analysis, some ap-
conducted on model or field pile groups subjected to later- proximate analyses have been completed by Maqtadir
alloading (e.g., Ogasawara et al., 1993; Okahara et al., and Desai (1986) and Shibata et al. (1988).
1993; Adachi et al., 1994; McVay et al., 1994 and 1995; Nonlinear bending moment-curvature relationships of
Kimura et al., 1995; Rollins et al., 1998) in order to eluci- concrete pile groups are properly considered in the anal-
date the ultimate state of the foundations during strong yses done by Adachi et al. (1994) and Kimura et al.
earthquakes. A long (or flexible) pile subjected to lateral (1995). However, they have applied the Drucker-Prager
loading has no definite limit state in the region below the criterion and the associated flow rule to the ground. Prob-
allowable displacement of the foundation required for lems accompanying such a constitutive model have been
the practical design of actual structures. Therefore, the indicated by Wakai et al. (1995a).
evaluation of the deformation of the ground around a The present research deals with FE analyses of pile
pile becomes very important. group and their comparison with the measurements of
At the present time, analytical approaches to the prob- model or field pile tests. Initially, in this paper, the
lems stated above are usually performed using the elasto- results of model tests for free- or fixed-headed pile
plastic assumption. However, few of them have been ana- groups subjected to lateral loading are analyzed by the 3-
lyzed using the 3-D elasto-plastic FEM, based on the D elasto-plastic FEM. The model piles were assumed to
supposition of realistic pile models and soil behavior be linearly elastic, and the sand was assumed to be an
(e.g., Brown and Shie 1990; Trochanis et al., 1991; elasto-perfectly plastic material with non-associative char-
Wakai and Ugai 1995a). Recent advances in computer acteristics, following the MC-DP model (Tanaka, 1992).
technology have made it possible to solve computational In addition to these properties, the no-tension analysis
problems with an enormous amount of freedom. Wakai (Zienkiewicz et al., 1968) is applied to the sand, in which

i) Assistant Professor, Department of Civil Engineering, Gunma University, Kiryu 376-8515.

ii) Deputy Director, Engineering Department, Katahira & Engineers International, Tokyo.
iii) Professor, ditto.
Manuscript was received for review on November 5, 1997.
Written discussions on this paper should be submitted before September 1, 1999 to the Japanese Geotechnical Society, Sugayama Bldg. 4F,
Kanda Awaji-cho, 2-23, Chiyoda-ku, Tokyo 101-0063, Japan. Upon request the closing date may be extended one month.

97

This is an Open Access article under the CC-BY-NC-ND license.

NII-Electronic Library Service
98 WAKAI ET AL.

the minimum principal stress is controlled as positive for analyses were conducted to simulate these results.
each element. Parameter values have been adjusted, using the back anal-
Parameter values for the sand used in the analyses ysis technique, so as to simulate the load-displacement
were determined from conventional triaxial compression curve measured in the in-situ single pile test. The elasto-
tests, because the ground was artificial and the distribu- perfectly plastic modelling used in this study was so sim-
tion of the density could be easily estimated. The physi- ple that it was easily applied to the back analysis tech-
cal parameters which can be obtained by the element nique. As for the lateral load-displacement relationship
tests, such as the Young's modulus E, the internal fric- and the bending strain distributions along each pile, a
tion angle cf> and the cohesion c, are used for the analyses. good correlation emerges between the experiments and
It is important that the general meaning of each the analyses; this can be seen in the pile group case.
parameter is clear. On the other hand, we often in- The main purpose of this paper is to point out that the
troduce such unclear parameters as the coefficient of sub- 3-D elasto-plastic FEM is very effective in the evaluation
grade reaction in the analyses, due to the convenience of of the behavior of a real foundation composed of pile
design. Its value should be changed according to the fixed group, and the appropriate constitutive model of soil and
conditions at the top of each pile. A serious problem its approprate parameters are necessary for the analyses.
arises when we analyze several types of pile foundations
with the same value parameters.
In the latter part of this paper, the field tests of a proto- CASES OF MODEL PILE TESTS AND ANALYSES
type foundation of steel pile group (3 x 3 piles) subjected Test Apparatus and Soil Description
to lateral loading at its footing are reported. The 3-D FE Figure 1 is a sketch of the test apparatus used showing

D : diameter of pile

Pile

Footing
2.50
Diameter of pile =50mm Weight(=8.83kN)
Thickness fo pile =1.5mm
0

Strain
gauge

E
E
0
0
l{)

ground

(a) Sketch of the test apparatus.

Case e-2
(b) Suspected fixed-headed
single pile (Case e-2).

Fig. 1. Lateral loading tests of model pile foundations

NII-Electronic Library Service

 ANALYSES OF PILE FOUNDATIONS 99

a pile group loading test. The total horizontal load at the of fixed-headed series 2 piles (see Fig. 1(b)) were used.
top of piles was measured by a load cell and each horizon- These piles have enough distance from each other (pile
tal pile-head load was evaluated based on the gradient of spacing is 10D) to be a better simulation (Case e-2). It
bending moment along the pile. The bending moment is can be assumed that the front pile behaves like a fixed-
estimated by the bending component of axial strain meas- headed single pile without uplifting. The bottom of each
ured by strain gauges attached on the pile as seen in the pile was supported by sand in all cases.
leftmost picture in Fig. 1. It should be noted that a dead
load of 8.83 kN weight steel plates, was put on the foot- Finite Element Meshes
ing while the experiments of pile groups (Case e-3 and As shown in Table 2, each analysis has been tabulated
e-4; see Table 2) were conducted. This dead load is so as to correspond to each experimental case. Schematic
equivalent to the self weight of assumed superstructures diagrams of the finite element meshes, which consist of
supported by the pile group foundation. 20 node brick elements, are shown in Fig. 2. As seen in
The model pile is made of aluminum and has a 50 mm this figure, half of the domain was eliminated by taking
outer diameter and 1.5 mm thickness. The pile was set in advantage of the symmetry with respect to the center
sand to a depth of 1450 mm and laterally loaded at 5 em plane parallel to the loading direction. The plane of sym-
from the model ground surface. In pile group cases, the metry is a smooth boundary. Other side boundaries and
pile spacing is 2.5D, where D is the pile diameter. The the bottom are fixed. Only in the mesh of Case a-4 (fixed-
physical properties of the sand used in the tests are headed pile group) was the region below the bottom level
shown in Table 1. The model ground preparation consist- of each pile prepared, because it is estimated that a slight
ed of dry pluviation through a 2 mm sieve placed at the rotation of the footing would cause the thrusting of front
top of the test container. The average relative density Dr piles and the uplifting of back piles.
of the ground was 72%. Thin frictional brick elements were inserted between
Table 2 shows the details of the experiments and the the pile and the sand in order to consider slippage at the
analyses. The model tests for several cases of single pile pile-soil interface, although in the future it may be advisa-
and pile group (3 x 3) were performed, with two different ble to use the 3-D joint elements instead. The frictional
kinds of boundary conditions at each pile head, free or elements have c=O, c/>=25°, 1f1=0° and other material
fixed. In the free-headed pile group experiment (Case e-3), constants of these are the same as the sorrounding sand
each pile head was pinned at the footing so as to rotate layer.
freely. Since it was difficult to perform the experiment for In these analyses, the horizontal displacement of the
fixed-headed single pile strictly, a substitute experiment pile head was controlled and the calculations were done
up to 5 mm (0.1D). The iterative procedure in the anal-
yses is based on the modified Newton-Raphson's
Table 1. Physical properties of the sand
method.
Specific gravity Gs 2.71
Material Parameters in the Analysis
Average grain size Dso 0.232mm
The material parameters used in the analyses are sum-
Uniformity coefficient Uc 1.86 marized in Table 3. The pile was assumed to be linearly
Maximum void ratio emax 1.02 elastic, with elastic modulus E and Poisson's ratio v. The
sand was assumed to have non-associative characteris-
Minimum void ratio emin 0.63
tics, following the MC-DP model. The failure criterion
and the plastic potential of the MC-DP model are the
Mohr-Coulomb and the Drucker-Prager equations, re-
Table 2. Model tests and analyses
(a) Model tests spectively. Such a combination is useful for the improve-
ment of the convergence of the FE calculation. It is
No. Number of piles Rotation at each pile head known that, if the pure Mohr-Coulomb criterion with
e-1 single free the non-associated flow rule is employed, stress paths
tend to fall into singular points on the II plane in the prin-
e-2 single fixed
cipal stress space (Tanaka, 1992).
e-3 3X3 free The friction angle c/> of the sand was 42 ° based on the
e-4 3X3 fixed (by footing) triaxial tests. The dilatancy angle 1f1 of the sand was as-
sumed to be given by lfl=cp-30° (e.g., Tatsuoka, 1993).
(b) FE analyses
The elastic modulus E of the sand was assumed to be
No. Number of piles Rotation at each pile head proportional to the square root of the initial mean
single free
confining pressurep (=(at+a2+a3)/3). This was deter-
a-1
mined from the results of conventional triaxial compres-
a-2 single fixed sion tests under drained conditions as seen in Fig. 3,
a-3 3X3 free from which,
a-4 3X3 fixed (by footing)

NII-Electronic Library Service

100 WAKAI ET AL.

unit: mm

(Plan view)

(a) single pile

s~~~L fnct1onal pile
(Case a-1, a-2) elements

750

(b) a free-headed pile group

composed of 9 piles (Case a-3)
(b) a fixed-headed pile group
composed of 9 piles (Case a-4)

Fig. 2. Discretized 3D meshes for FEM

Table 3. Material parameters used in the analyses

Young's modulus Poisson's Cohesion Friction angle Dilatancy angle Unit weight Coefficient of earth
Material
E (MPa) ratio v c (MPa) ¢ (deg) If/ (deg) y (kN/m 3 ) pressure K 0
pile 6.86 X 104 0.345 large - - 26.4 -

sand Eq. (1) 0.4 0 42 12 15.3 0.5

pile-soil interface Eq.(l) 0.4 0 25 0 15.3 0.5
footing large 0.345 large - - 26.4 -

of the pile-soil interface was found to be equal to 25°,

E=Eo (P (1) based on the results of the plain slippage tests on an alu-
~Po minum plate and the sand. The dilatancy angle of the in-
was obtained, where Eo= 19.6 MPa, Po=98 kPa. terface was assumed to be zero. In the analysis, the no-
It should be noted that we used the secant elastic modu- tension analysis (Zienkiewicz et al., 1968) was applied to
lus E 50 measured in the tests in Eq. (1). The friction angle the sand, in which the minimum principal stress was kept

NII-Electronic Library Service

 ANALYSES OF PILE FOUNDATIONS 101

- - - Observed
-Calculated

• Eso- P
0 <1>-P

1
0
50 100 150
/ Initial mean confining pressure p (kPa)

estimated p at the bottom

of a pile in experiments
20
Fig. 3. Results of the triaxial tests and parabolic fitting curve for their
E 50 - p relations Displacement (mm)
(a) Load-displacement curves for each case

positive for each element.

Ci
a> - - - Observed
~
Q)
-Calculated
OBSERVED AND CALCULATED RESULTS OF c,
c
MODEL TESTS ro
c
0
Load-Displacement Relationships
Figure 4(a) shows the relationships between the ob- ~0::
,
served and calculated horizontal load and the displace-
ment of pile head for single and 3 x 3 pile group cases.
10 20
The experimental and calculated curves correlate well
with each other. Since computations take some time to Displacement (mm)
complete, the calculated curves are only shown below the
(b) Rotation angle of footing vs. displacement
displacement of 5 mm (O.lD) as represented in Fig. 4(a).
(Group, Fixed-headed case)
In Fig. 4(a) it can be seen that the load for the fixed-head-
ed cases is much greater than that found in the corre- Fig. 4. Load-displacement curves for each case, and the rotation of
sponding free-headed cases. The load for the pile group footing for fixed-headed pile group case
case is smaller than the number of piles in the pile group
times the lateral load of a single pile. Therefore, the pile
group effect is very significant and the FE calculations are
highly effective at estimating these values. In the analysis,
the pile group efficiency was estimated to be 0.6-0.7 when
the displacement reached O.lD.
Figure 4(b) is a graphical representation of the relation-
ship between the rotation angle of the footing and the
horizontal displacement for the fixed-headed pile group
case (Case e-4). The measurements and the analytical
results agree with each other. It was found that the lateral
loading causes a slight rotation of the footing. The inter-
Fig. 5. Assigned pile name in the pile groups
action between horizontal and vertical movement of the
piles, that is, thrusting of front piles and the uplifting of
back piles, appears in this case. On the other hand, for
the free-headed pile group case (Case e-3}, the rotation of can be seen. It seems that it is caused by a difference of
the footing was very small. stress-strain characteristics in the region of small strain.
Figure 5 is a diagramatic representation of the assigned In the analyses, soil behaves as a linear elastic body be-
pile names in the pile groups, which will be used in the fol- fore failure, although the actual soil in the experiments
lowing part of this paper. Figures 6(a) and (b) show the shows a nonlinear characteristics from immediately after
load distribution ratio of each row for the 3 x 3 pile the loading starts.
group case. As for the initial part of each curve, a differ- On the other hand, a part from the initial part of each
ent tendency between the experiments and the analyses curve, the piles in the front row carry a substantially

NII-Electronic Library Service

102 WAKAI ET AL.

greater load than the piles in the middle and back row. In
---- Fixed-headed - - Fixed-headed
- Free-headed - Free-headed both the experiments and the analyses, the load distribu-
tion ratio of each row was found to be independent of
'----==~-'-'Front
the fixed conditions of each pile head as seen in Figs. 6(a)
and (b), with the exception of the initial part of each
curve during the experiments.
Figure 6(c) shows calculated results of the relationship
between the load distribution ratio of each pile and the
displacement for the fixed-headed pile group case. It can
10 20 10 20 be seen that, in the first half of each curve, the loads act-
Displacement (mm) Displacement (mm)
ing on the outer piles (F-F, M-F, B-F) are greater than
(a) Observed (b) Calculated those acting on the inner piles (F-B, M-B, B-B) when
three piles in the same row are compared with each other.
2~--~--~~----~--~~--~--~ After a large deformation occurs, however, the difference
of load distribution ratio between inner piles and outer
piles becomes very small. Almost all of the tendencies dis-
cussed above can also be seen in the results for the free-
headed pile group case.
In such an elasto-perfectly plastic analysis, the ground
behaves as a linearly elastic material before it yields.
- Therefore, it is an inevitable consequence that the load of
the piles in the front row (F-F, B-F) shows agreement
with those in the back row (B-F, B-B) at the initial part of
the curves, which corresponds with the theoretical solu-
tion based on linear elasticity.

Bending Strain Distribution along Each Pile

I I
Figures 7 and 8 show the bending strain distributions
2 4 6 along each pile for the single and the 3 x 3 pile group
Displacement (mm) cases, respectively. 6 represents the horizontal displace-
ment of the pile head. In Fig. 8, the bending strain distri-
Fig. 6. Load distribution ratio of each row for 3 x 3 pile group case butions for two typical piles F-F and B-B are illustrated.
A good correlation between the experiments and the ana-
lyses can be seen in Figs. 7 and 8, except that the ob-

3=5.0

'E
§.
"'C
as
~ 500
~
·c.
Q)
:5
E
.g
~ 1000
c: 3: Displacement (mm)
as
3: Displacement (mm) iii
---- Observed i5 ---- Observed
- Calculated - Calculated

15
1500 0 .9~ooo 1000

Bending strain (1 0-6) Bending strain (1 0-6)

(a) Free-headed case (b) Fixed-headed case

Fig. 7. Bending strain distributions along the single piles

NII-Electronic Library Service

 ANALYSES OF PILE FOUNDATIONS 103

o: Displacement (mm) o: Displacement (mm)

- -•- Observed (F-F) 1000 - • - Observed (F-F)
- Calculated (F-F) - Calculated (F-F)
--o- Observed (B-B) - o- Observed (B-B)
--o- Calculated (B-B) --o- Calculated (B-B) f
15
~?200 -800 -400 0 400

Bending strain (1 0-6) Bending strain (10-6)

(a) Free-headed case (b) Fixed-headed case

Fig. 8. Bending strain distributions along each pile in the 3 x 3 pile group

ca
a..
~

~ 40
0
tsca
~
OJ
"'C
~
C)
..c
::J
en

x : Depth (mm) x : Depth (mm)

0 0
Horizontal displacement y (mm) Horizontal displacement y (mm)

(a) Free-headed case (b) Fixed-headed case

(Calculated)

Fig. 9. Subgrade reactions along the single piles

served bending strains near the top of each pile are a little pletely free) and Case e-2 (completely fixed). In the pile
smaller than the calculated ones for the fixed-headed pile group case, the piles in the back row tend to be pulled out
group case (Fig. 8(b)). vertically, causing a rotation of the footing as seen in Fig.
Comparing them to the case for the fixed-headed single 4(b). Therefore, to consider the vertical resistance of
pile (Fig. 7(b)), it was found that both the observed and each pile is also important for the evaluation of the later-
the calculated bending strains near the top of the pile al behavior of the pile group.
group were much smaller than the ones for the single pile
case. The boundary condition at the top of each pile for Horizontal Subgrade Reaction along Each Pile
the fixed-headed pile group case, therefore, is thought to Figure 9 represents the relationship between the calcu-
be intermediate between the conditions of Case e-1 (com- lated subgrade reaction p and the horizontal displace-

NII-Electronic Library Service

104 WAKAI ET AL.

ment y of the single piles. The curves are plotted at sever- These results suggest that the lateral resistance of the
al depths along the piles. In the free-headed case the piles in the back row does not decrease proportionally,
experimental values are also shown, which are consistent compared to the piles in the front row.
with the calculated ones. Each value of the sub grade reac- The numerical expression of the subgrade reaction p
tion is evaluated by differentiating the bending moment derived from a function composed of variables x, y, and
twice with respect to the depth x, by making use of the the material constants relating to the deformation and
beam theory. It should be noted that the bending mo- failure, is useful. To simplify such a relationship, the
ment distribution has been approximated by the seventh function of p is assumed to be .
order polynomial equation before the differentiation.
The experimental values for the fixed-headed case are not p=k* ( -X )m( -y )n (2)
shown in the figure because the bending moment distribu- D D
tion could not be approximated accurately by the poly- where k*, m and n are the constants. Dis a pile diameter.
nomial equation. In this experimental case, the space be- m, n, (x/ D) and (y/ D) have no dimensions. If we sub-
tween each gauge should have been changed as it became stitute m=O and n= 1 in Eq. (2), it leads to the equation
more minute and direct differentiation should have been proposed by Chang (1937), where k* I D(=k) is usually
carried out. called the coefficient of sub grade reaction. Since k*, m
In Fig. 9 it can be seen that the subgrade reaction in- and n seem to be dependent on the material constants
creases as depth increases. This is because the Young's relating to the deformation and failure, they should be de-
modulus of the ground was assumed to be proportional termined statistically based on a large number of experi-
to the square root of the depth in the analysis, and the ments.
strength increased with an increase in depth due to the The values of k*, m and n in cases a-1, a-2, a-3 and a-4
effect of the internal friction angle of the sand. The are shown in Table 4. The values in the front row of the
values of p for the fixed-headed case are larger than those pile group cases (a-3 and a-4) are found to be relatively
for the free-headed case after a large deformation occurs, similar to the single pile case, while the values in the back
while the values during a small deformation agree well row have a much smaller k* when compared to the single
each other. pile cases. Generally, the larger the estimated value of m,
Figure 10 is a representation of the calculated p-y the smaller the value of n.
relationship of each pile in the 3 x 3 fixed-headed pile
group. Comparing them to the single piles, the subgrade Calculated Earth Pressure Acting on the Piles
reaction for the pile group is much smaller than the one In order to investigate the mobilization of the sub-
for the single pile. This is due to the interaction between grade reaction p, the distribution of the earth pressure
neighboring piles. In addition, the curves of Pile B-B around each pile was calculated. Figure 11 shows the
have a characteristic tendency toward strain-softening at horizontal earth pressure acting on the single piles. Here
shallow depths (curves at x =50 or 100 mm in the figures). 'P is the angle between the radial direction through each

80~~-~,~-~,~-~,~-~,~-~~

ro-
o..
~

~ 40 -
0
tsco x=300. x=200.
~ x=100. -
Q) x=300.
"0 x=50. x=200.
~
0>
.c
::J
~ x=100.
(/)
x=50.

x: Depth (mm) r- x: Depth (mm) -

I I I I I

0 2 4 0 2 4
Horizontal displacement y (mm) Horizontal displacement y (mm)

(a) Pile F-8 (b) Pile 8-8

(Calculated) (Calculated)

Fig. 10. Subgrade reactions along each pile in the 3 x 3 fixed-headed pile group

NII-Electronic Library Service

 ANALYSES OF PILE FOUNDATIONS 105

Table 4. Values of k*, m and n in Eq. (2)

pile name assigned k* (kPa)

No. m n
as Figure 5
a-1 - 106. 0.788 0.652
a-2 - 168. 0.749 0.744
F-B 192. 0.789 0.946
M-B 91.4 0.865 0.897
B-B 29.2 0.842 0.642
a-3
F-F 172. 0.747 0.854
M-F 89.7 0.801 0.820
B-F 38.6 0.762 0.628
F-B 89.4 0.835 0.874
M-B 11.1 1.132 0.568
B-B 1.52 1.487 0.251
E' 75
a-4
F-F 274. 0.768 0.841 s
M-F 0.707 0.761 ~
138.
~
B-F 70.7 0.674 0.579 ::J
!/)

-g 400
::J
e
Cl
Q)
:5
E' 0-10 E
.g
s
Q)
u
~
::J
!/)
'0
---
c
::J
e
Cl 'P: Direction angle (b) Pile B-B
Q)
:5 o: Displacement (mm)
Fig. 12. Calculated horizontal component of earth pressure acting on
E ~ -o=O.O
.g --- o=0.5 each pile in the 3 x 3 fixed-headed pile group
Load ___ o= .0
~ 2
c - -o=5.o

acting on the other sides (at IJ'=60°, 120° and 180°) of

(a) Free-headed case the single piles is found to be very small. The directions
of the horizontal earth pressure are reversed at a certain
depth. This suggests that a reversing point of the deflec-
75 tion of the piles exists at this depth.
--- ----- ---)
Figure 12 is a graphical representation of the distribu-
tions of the earth pressure along each pile in the 3 x 3
------- fixed-headed pile group. There is relatively large pressure
------------ 'P: Direction angle
acting on the obliquely directed front side (at IJ'= 60°) of
o: Displacement (mm) Pile F-B, compared to the corresponding single pile. This
~ -o=O.O is due to the interaction between the neighboring piles. It
--- o=0.5 is also noteworthy that, for Pile F-B, the pressure acting
Load ___ o= .0
2
- -o=5.o on the back sides of the pile (at IJ'= 120° and 180°) grows
in such a direction as to actually push the pile. On the
other hand, for Pile B-B, it can be seen that the distribu-
(b) Fixed-headed case tions are relatively flat and the depth of the reversing
Fig. 11. Calculated horizontal component of earth pressure acting on
point is deeper than that of the single pile cases. These
the single piles results suggest that the difference of p-y relationship be-
tween Piles F-B and B-B, shown in Fig. 10, is caused by
the interaction between the neighboring piles. The p-y
relationship is strongly dependent on pile location within
point and the loading direction and b represents the pile- the group. Further studies are necessary to clarify the
head displacement. From Fig. 11, it can be seen that most cause or mechanism of these phenomena.
of the soil resistance is concentrated on the front side (at
IJ'=0°) of the piles. When comparing the fixed- and free- Calculated Shear Strain Distributions in the Ground
headed cases, the region in which the soil resistance is The magnitude of the mobilized strength of the sand
mobilized is significantly enlarged with increased depth can be related to the magnitude of shear strain in the
in the fixed-headed case. The horizontal earth pressure ground. Figure 13 shows the calculated shear strain distri-

NII-Electronic Library Service

106 WAKAI ET AL.

butions in the ground when the pile-head displacement

loading direction
reaches 5.0 mm. The horizontal load is directed to the
right in the figure. Figures 13(a) and (b) depict the distri-
bution of the free-headed and fixed-headed single pile
case, respectively. Figure 13(c) shows the 3 x 3 fixed-head-
ed pile group case. The failure region in front of the sin-
gle pile case is larger in the fixed-headed case than in the
free-headed case in both horizontal and vertical plane.
The reason for this is that the passive earth pressure act-
ing on the front of the fixed-headed single pile is a greater
magnitude. As can be seen from Fig. 13(c), the strain dis-
tribution of the fixed-headed pile group case shows that
the failure region has extended, surrounding the pile
group. This suggests that the sand between the piles tends
(a) Free-headed single pile (b) Fixed-headed single pile to provide little resistance to the movement of the sur-
rounding piles. This phenomenon causes the pile group
effect on the load-displacement relationship as discussed
above.

CASES OF FIELD TESTS OF A PROTOTYPE

Octahedral
FOUNDATION AND THEIR ANALYSES
shear strain Test Apparatus and Soil Description
0.010 The field tests of a prototype foundation of steel pile
group (3 x 3 piles) subjected to lateral loading at its foot-
ing were done by Saito et al. (1993). Figure 14(a) shows
0.005 the pile embedding (14.5 m below the ground surface)
related to a ground cross-sectional view. The symbols St
through to slO shown in the figure indicate the location of
0 each strain gauge used.
The soil constants used in the initial analysis are indi-
cated in Fig. 14(b). These values will be adjusted later, us-
ing the back analysis technique. The reason such an ana-
lytical method was used is that the soil constants from
(c) 3 x 3 fixed-headed pile group the element tests represent the soil properties before pil-
Fig. 13. Calculated octahedral shear strain distributions in the ing, and they seem to change due to subsequent pile driv-
ground, when the displacement d equals 5.0 mm ing and compacting processes.

c' 4> ' w' r or(r') E;o v

(kPa) (deg) (deg) (kN/m3) (kPa)
!1111111111

2. 7m J G.W.L.
2. 7m A1 16.7 34.4 0
14.1 (8.34)
[Ave. 12.7]
1370 0.3

1. 6m A2 0 34.4 0 (8.63) 14200 0.3

4.3m

1. 4m A3 27.5 34.4 0 (8.63) 16200 0.3

5. 7m

1. 3m A4 74.6 0 0 (7.36) 2600 0.49

7.0m

3.0m A5 8.44 40.5 0 (8.83) 19600 0.3

14.5m

8.44 40.5 0 (9.81) 96100 0.3

Nvalue

(a) Ground cross-sectional view (b) Investigated soil constants for the analyses (initial values)

Fig. 14. Soil conditions at the testing site

NII-Electronic Library Service

 ANALYSES OF PILE FOUNDATIONS 107

In the analyses, the Poisson's ratio v was assumed to the simplified piles are the same as each other. Because of
be that shown in Fig. 14(b). The secant elastic modulus this simplification, the value of yield stress for steel was
Eso was assumed to be given by 28N (the Japan Highway changed by an equivalent value, 188.8 MPa. The footing
Design Manual (1990)), where the Nvalue was given by a was made of reinforced concrete.
standard penetration test. The distribution of theN value
is shown in Fig. 14(a). Here c, ¢, If/ were determined Finite Element Meshes
from the results of the element tests y is the wet unit Schematic diagrams of the finite element meshes used
weight andy' is the submerged unit weight. Since no ac- are shown in Fig. 16. Half of the domain was eliminated
tual laboratory data was available for the layer De-l, the because of the symmetry concerning the center plane
values for the layer Dc-2 were used instead as the value of parallel to the loading direction. In the mesh for the pile
the layer De-l. group, the region below the bottom level of each pile was
Figure 15 is the schematic representation of the field prepared. Thin frictional brick elements were inserted be-
tests. The experiments conducted by Saito et al. were tween the pile and the soil to consider slippage at the pile-
foundations of single pile and pile group (3 x 3 piles) sub- soil interface. The wall friction angle for these elements
jected to lateral loading. For the pile group case, the bot- was assumed to be 25 o. The no tension analysis was ap-
tom of the footing and the loading point are located at plied to the layers Al and A2 (Fig. 14(b)).
0.30 m and 1.30 m above the ground surface, respec- The assigned pile names in the pile group are the same
tively. Each pile head was fixed by the footing of each as in the model tests as shown in Fig. 5. However, the
other so as not to rotate freely. The spacing between each symmetry with respect to the center plane parallel to the
pile wad 2.5D. The dead load of 1.96 MN weight acted loading direction cannot be found in the experimental
on the footing while the pile group experiment was con- results. Thus, as will be described in Fig. 20, when we esti-
ducted. This dead load is equivalent to the self weight of mate the pile-head load of each pile in the outer columns
superstructures supported by the foundation. For the sin- (the piles F-F, M-F and B-F), we use the averaged values
gle pile case, the loading point was 0.50 m above the of two piles each located at the symmetrical position.
ground surface. In this case, the pile head could be rotat-
ed freely.
In a similar way to the model tests, the total horizontal
load was measured by a load cell and each horizontal
pile-head load was evaluated based on the gradient of
bending moment. The bending moment is. estimated by
the bending component of axial strain along the pile. The
pile is made of steel and has a 318.5 mm outer diameter
and 6.9 mm thickness. The strain gauges were protected
from damage, using the covered channels. In the follow-
ing analyses, however, the cross-section of the pile will be (Plan view)
simplified (Gose et al., 1997) to simplify the FE meshes.
The thickness and the bending rigidity of the original and ~II]~
soil
frictional pile
elements
Diameter of pile =318.5mm

Pile
Footing

Displacement
gauge

14.50

~
3.20

unit: m

(a) single pile (b) a fixed-headed pile group (a) single pile (b) a fixed-headed pile group
composed of 9 piles composed of 9 piles

Fig. 15. Schematic of the field tests Fig. 16. Discretized 3-D FE meshes to simulate the field tests

NII-Electronic Library Service

108 WAKAI ET AL.

is the horizontal displacement of the pile head. There is a

OBSERVED AND CALCULATED RESULTS OF good correlation between the experiments and the anal-
FIELD TESTS yses as can be seen in this figure.
Results for Single Pile Case
Figure 17 shows the observed and calculated relation- Results for Pile Group Case
ship between the horizontal load and the displacement of Figure 19 is a graphical representation of the observed
pile head for single pile case. As was previously stated, and calculated relationships between the horizontal load
the soil constants used in the initial analysis have been ad- and the displacement of pile head for the pile group. A
justed, using the back analysis technique, so as to simu- good correlation between the experiments and the anal-
late a load-displacement curve obtained from the single yses can be seen in this figure. It was found that a back
pile test. analysis technique was effective in determining
Both the original and the adjusted analytical curves are parameters which can then be applied in the analyses of
shown in this figure. Since the experimental loading path other problems at the same site.
is cyclic, only the experimental skeleton curve should be Figure 20 represents the observed and calculated load
compared to the analytical curves. The experimental distribution ratio of each pile for the pile group. The ra-
skeleton curve and the adjusted calculated curve corre- tio is normalized against the load of the pile F-F. The
late well with each other. comparisons are made at each level of load. As for the
The changes in the soil constants are as follows:
1) The Young's modulus for the saturated sand layers
(layers A2, A3 and A5) was increased to 1. 8 times the
initial values. This was to take into consideration the
densification of sand due to the pile driving.
2) tan ¢ for the sand layers was increased to 1.2 times the
g
"'0
initial values. 1f1 for the sand layers was determined <U
<D
.c
from lfl=¢-30°. This was also in consideration of j!2
the sand densification. ·a.
<D
3) The Young's modulus for layer A1 was changed to :5
8829 kPa. This was because this layer is unsaturated. E
4) c and ¢were changed to 29.4 kPa and 20°, respec-
e
.._
<D
0
tively. This was also because this layer is unsatu- c o: Displacement (mm)
rated. ~
Such a procedure for adjusting each soil constant has i5 12
- - -Observed (field test)
little theoretical foundation. More reasonable algorithms - Calculated
for back analysis should be developed in the future. The
adjusted constants were used in the following analyses 0 200 400 600 800
for the single pile and the pile group case.
Bending strain (10-6)
Figure 18 is a representation of the observed and calcu-
lated bending strain distributions along the single pile. l5 Fig. 18. Bending strain distributions along the single pile

- - - Observed (field test) - - - Observed (field test)

-Calculated -Calculated

~~
2000 , ... -, ~
~
' ,
,,
~
I

z~
\ I
~ \I I
, I

'
~'
\1
' I
I
I
I
I<
I
I ,I
I

~ II ~
~' I
"'0 I
,
II
J
"'0 ,'
' II
I
I
I
I I '
cu cu /:
I
I
I
I
I
I
I
I
I

0 0 '
''
I
..J ..J ~~
I I
,'' I
I
~
' I
' I I
'
' '
I

I
I I ' I
I
I
I
I
I
I
I; I
I

''
I I
I I I I I
I I
I I I I I'
I

.~:'~" '
I
I
' ' I
' ''
''
I I I
I I
I

''
I I 'I
I
I I 'I
I I
I
I
,, 'I
I;
,' ' ,'
I I
' , , ~

,f''
~
I ' I

~/~~:~
~

''
I
, ~
, ~

,,
I ~
I I ~

, _.... ..., I ;
I'
' I ~
~
I ~ ' , ~

I~' ~
~
~
I'

100 200 100 200

Displacement (mm) Displacement (mm)

Fig. 17. Load-displacement curves of the single pile Fig. 19. Load-displacement curves of the pile group

NII-Electronic Library Service

 ANALYSES OF PILE FOUNDATIONS 109

1.2 1.2
0 0
-~
~
§ 0.8 !J!::::.--.,.::::::,-....:-__:~===~--=---A § 0.8
~
·.g
..0 0.6 ..0 0.6
-~
~ :.0 0.4
:.0 0.4
""0 ""0
co co
.3 0.2 .3 0.2
_..__Calculated I Inner column
0 0

Front row Middle row Back row Front row Middle row Back row

(a) Load = 392kN (b) Load = 785kN

1.2 1.2
0
-~
§ 0.8 ~-~-.::--~~=------------1
·.g
..0 0.6
·~
:.0 0.4 1-------------=:~,---------1
""0
co
.3 0.2

0 0
Front row Middle row Back row Front row Middle row Back row

(c) Load= 1180kN (d) Load= 1570kN

Fig. 20. Load distribution ratio of each pile

piles F-F, M-F and B-F in the experiments, the averaged

values of two piles each located at the symmetrical posi- CONCLUSIONS
tion were used. As the total load is increased, the front The conclusions obtained from this study are summa-
piles (F-F and F-B) carry more load than the piles in the rized as follows:
middle and back row. Such a general tendency can be (1) The behaviors of the model pile groups were ana-
well simulated by FEM. After a large deformation oc- lyzed by the 3-D elasto-plastic FEM. The sand was
curs, the difference of the load distribution ratio between assumed to be an elasto-perfectly plastic material
the pile in the outer and inner columns becomes relatively with non-associative characteristics, following the
large for the experimental case, compared to the analyti- MC-DP model. Parameter values for the sand were
cal results. The reason for this is not clear. determined from conventional triaxial compression
Figure 21 shows the bending strain distributions along tests. It is shown that the experimental results can
the piles in each row for the pile group case. Each value be simulated by the FEM.
has been averaged among three piles in the same row (2) The load for the pile group case is smaller than the
(e.g., the pile F-FX2 and F-B). There is a good correla- number of piles in the pile group times the lateral
tion between the experiments and the analyses as can be load of a single pile. Therefore, the pile group effect
seen in the figure. In addition, as the displacement is very significant.
becomes larger, a difference between the experiments and (3) The load distribution ratio of each row was found
the analyses also becomes larger. Since the analytical to be independent of the fixed conditions of each
values around 5.0 m depth are a little smaller than the ac- pile head.
tual experimental values, the Young's modulus for the (4) In the pile group case, the piles in the back row tend
layer A3 (depth between 4.3 to 5.7 m) should be changed to be pulled out vertically, causing a rotation of the
by a larger value. footing. Therefore, to consider the vertical
It was shown that not only model tests but also proto- resistance of each pile is also important for the
type tests for pile group foundations can be simulated by evaluation of the lateral behavior of the pile group.
the 3D elasto-plastic FEM presented in this paper. (5) The values of subgrade reaction for the fixed-head-
ed case are larger than those for the free-headed
case after a large deformation occurs, while the

NII-Electronic Library Service

110 WAKAI ET AL.

fixed-headed case.
(7) In the fixed-headed pile group case, there is rela-
tively large earth pressure acting on the obliquely
directed front side of the pile in the front row, com-
pared to the corresponding single pile. The earth
pressure acting on the back sides of the pile grows
in such a direction as to actually push the pile. On
the other hand, for the pile in the back row, the dis-
tributions are relatively flat. The distributions of
earth pressure are strongly dependent on the pile lo-
cation within the group.
(8) The back analysis technique is effective in determin-
ing parameters which can then be applied in the ana-
lyses of other problems at the same site. However,
0 1000
the procedure for back analysis presented in this
Bending strain (10-6)
paper has little theoretical foundation and more
(a) Front row (average) reasonable algorithms should be developed in the fu-
ture.
(9) Not only model tests but also prototype tests for
§: pile group foundations can be simulated by the 3D
"0 elasto-plastic FEM presented in this paper.
ro
Q)
.J::
~
·c.. REFERENCES
(J)
:5
E 1) Adachi, T., Kimura, M. and Zhang, F. (1994): "Analyses on ulti-
.g mate behavior of lateral loading cast-in-place concrete piles by 3-
dimensional elasto-plastic FEM," Proc. of the 8th Int. Conf. on
Computer Methods and Advances in Geomechanics, Morgantown,
USA, pp. 2279-2284.
2) Brown, D. A. and Shie, C. F. (1990): "Numerical experiments into
group effects on the response of piles to lateral loading," Com-
puters and Geotechnics, Vol. 10, pp. 211-230.
-1000 0 1000 3) Chang, Y. L. (1937): Discussion on "Lateral pile loading tests" by
Bending strain (10-6) Feagin, Trans. ASCE, Paper No. 1959, pp. 272-278.
4) Gose, S., Ugai, K. and Ochiai, H. (1997): "The behavior of pile
(b) Middle row (average)
foundation using three-dimensional elasto-plastic FEM," Proc. of
the 6th Int. Symp. on Numerical Models in Geomechanics, Mon-
treal, Canada, pp. 539-544.
5) Japan Highway Design Manual (1990) (in Japanese).
§: 6) Kimura, M. (1993): "Fundamental studies on group pile behavior
"0 subjected to lateral loading," Doctral thesis submitted to Kyoto
ro
Q)
.J:: Univ. (in Japanese) .
~ 7) Kimura, M., Adachi, T., Kamei, H. and Zhang, F. (1995): "3-D
·c.. finite element analyses of the ultimate behavior of laterally loaded
(J)
:5 cast-in-place concrete piles," Proc. of the 15th Int. Symp. on Nu-
E merical Models in Geomechanics, Davos, Switzerland, pp. 589-
.g
594.
8) Maqtadir, A. and Desai, C. S. (1986): "Three-dimensional analysis
of a pile-group foundation," Int. J. of Numerical and Analytical
Methods in Geomechanics, Vol. 10, pp. 41-58.
9) McVay, M., Bloomquist, D., Vanderlinde, D. and Clausen, J.
(1994): "Centrifuge modelling of laterally loaded pile groups in
sands," Geotech. Testing J., Vol. 17, No.2, pp. 129-137.
10) McVay, M., Casper, R. and Shang, T. (1995): "Lateral response of
Bending strain (10-6) three-row groups in loose to dense sands at 3D and 5D pile spac-
(c) Back row (average) ing," Proc. of ASCE, Vol. 121, GT5, pp. 436-441.
11) Ogasawara, M., Kikkawa, H., Saito, A. and Gose, S. (1993): "In-
Fig. 21. Bending strain distributions along the piles in each row situ test of pile foundation," Proc. of the Int. Symp. on Limit
State Design, Vol. 1.
12) Okahara, M., Kimura, Y., Takagi, S. and Ohori, H. (1993):
values during a small deformation agree well each ''Horizontal loading tests of group piles and its simulation analy-
other. sis," Structural Engineering Letters, JSCE, Vol. 39A, pp. 1371-
(6) When comparing the fixed- and free-headed cases, 1384 (in Japanese).
13) Rollins, K. M., Peterson, K. T. and Weaver, T. J. (1998): "Lateral
the region in which the soil resistance is mobilized, load behavior of full-scale pile group in clay," Proc. of ASCE,
is significantly enlarged with increased depth, in the

NII-Electronic Library Service

 ANALYSES OF PILE FOUNDATIONS 111

Vol. 124, GT6, pp. 468-478. Chapter 2.4, pp. 57-81 (in Japanese).
14) Saito, A., Hanko, M., Gose, S. and Yi, F. (1993): "A study of pile 18) Trochanis, A. M., Bielak, J. and Christiano, P. (1991): "Three-
foundation behavior at large-scale horizontal deformation," Struc- dimensional nonlinear study of piles," Proc. of ASCE, Vol. 117,
tural Engineering Letters, JSCE, Vol. 39A, pp. 1395-1407 (in GT3, pp. 429-447.
Japanese). 19) Wakai, A. and Ugai, K. (1995a): "Model tests and analyses on
15) Shibata, T., Yashima, A., Kimura, M. and Fukada, H. (1988): lateral loading behavior of single piles,'' Proc. of the Japanese Soci-
"Analysis of laterally loaded piles by quasi-three-dimensional ety of Civil Engineers, No. 517, III-31, pp. 159-168 (in Japanese).
finite element method," The 6th Int. Conf. on Numerical and Ana- 20) Wakai, A., Ugai, K. and Gose, S. (1995b): "The 3-D FE analysis
lytical Methods in Geomechanics, Vol. 2, pp. 1051-1058. of model group piles embedded in sand," Proc. of the 15th Int.
16) Tanaka, T. (1992): "Deformation and stability analysis by finite ele- Symp. on Numerical Models in Geomechanics, Davos, Switzer-
ment method,'' in Principle of Soil mechanics (First Revision), The land, pp. 613-618.
Japanese Geotechnical Society (in Japanese). 21) Zienkiewicz, 0. C., Valliappan, S. and King, I. P. (1968): "Stress
17) Tatsuoka, F. (1993): "Relationships between Stress and Strain In- analysis of rock as a 'no tension' material," Geotechnique, Vol.
crements," in Introduction to Strength of Soils and Failure of 18, pp. 56-6fj.
Ground (Third Revision), The Japanese Geotechnical Society,

NII-Electronic Library Service