See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/272576968

Seismic Response of a Large Pile Group

Conference Paper · September 1998

CITATIONS READS

3 250

5 authors, including:

Rainer Zinn George Mylonakis

Stangenberg und Partner University of Bristol
12 PUBLICATIONS   10 CITATIONS    245 PUBLICATIONS   4,253 CITATIONS   

SEE PROFILE SEE PROFILE

Sissy Nikolaou
WSP
46 PUBLICATIONS   705 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Garini E., Gazetas G., and Anastasopoulos I. (2017) "Evidence of Significant Forward Rupture Directivity Aggravated by Soil Response in an MW6 Earthquake, and the
Effect on Monuments ", Earthquake Engineering and Structural Dynamics, DOI: 10.1002/eqe.2895 View project

School Children Safety to Natural Hazards in Developing Countries View project

All content following this page was uploaded by George Mylonakis on 08 June 2015.

The user has requested enhancement of the downloaded file.

Seismic response of a large pile group
George Gazetas
National Technical University, Athens, Greece
Peter Hess
Ministerium fur Finanzen und Energie, Kiel, Germany
Rainer Zinn
Stangenberg und Partner Ing., Bochum, Germany
George Mylonakis
City University of New York, USA
Aspasia Nikolaou
State University of New York at Buffalo, USA

Keywords: pile, pile group, interaction factor, wave propagation analysis

ABSTRACT: The dynamic stiffness, damping and seismic response of a large 230-pile group of a
major critical facility subjected to an earthquake with 0.11g peak ground acceleration are analyzed.
The seismic response of the foundation is obtained with due consideration of both kinematic and
inertial effects. Several different methods are being used in the paper to obtain the dynamic imped-
ances of the pile group, including rigorous and simplified methods. It is concluded that the overall
impedance of the group cannot be obtained by simply adding algebraically the impedances of the
eight individual subgroups); the simplified method of analysis (with properly obtained pile-to-pile
interaction factors) is the only feasible method for a realistic analysis of pile-group response.

1 INTRODUCTION

The paper refers to the study of the seismic response of a major critical facility founded on a large
mat supported on 230 piles. The geometry of the group is sketched in Figure 1 (plan) and Figure 2
(section). The 130 cm diameter pile pass through 9.5 meters of very soft saturated organic clay
61 m
38 m

Figure 1. Plan of the 230-pile configuration.

having an S-wave velocity Vs = 80m/s, mass density ρ = 1.5 Mg/m3 and Poisson's ratio νs = 0.49.
The piles are socketed in a stiff sand layer having Vs = 330 m/s into which they penetrate 6 meters.
The foundation transmits the loads of the superstructure through:
• the group of 230 piles, and
• the vertical sidewall of the embedded structure
No contact between the base of the foundation mat and the soil can be accounted for to transmit
shear or normal loads onto the soil.
(m)
+3
FILL
0

CLAY
- 6.5
Vs = 81.6 m/s, ρ = 1.5 Mg/m3
ν = 0.49, β = 5% h1 = 9.5 m

-16.0
SAND
- 22.0
Vs = 330.0 m/s, ρ = 2.0 Mg/m3
ν = 0.40, β = 5%
Figure 2. Cross-section of the foundation (not in scale) and the soil properties used.

Existing site response analyses have shown that the “effective” shear strains in the design earth-
quake are of the order of 10-3. With the “standard” curves of G = G(γ) for clay used in those analy-
ses, the effective shear modulus reduces from Gmax ≈12 MPa to about G ≈ 8 MPa. The latter value
corresponds to an S-wave velocity of about 73 m/s.

2 METHOD OF ANALYSIS

Three different methods have been applied to obtain the dynamic impedances of the pile group:
(a) a rigorous semi-analytical method (Kaynia, 1982), which accounts for the layered soil profile and the
pile-to-pile interaction, including the "shadow"-forming of the in-between piles. This method, however,
requires extremely large computer capacity, and cannot handle the complete pile group. Thus, in order to
compute the overall impedances the 230-pile group was divided into eight (8) clusters of sub-groups.
Figure 3 shows these eight sub-groups; evidently the division is a “natural” one. Extensive comparisons
were performed between rigorous and simplified methods for the sub-groups.
(b) a rigorous superposition method, which generates rigorous interaction factors between any two piles in
the group and uses them (internally) to compute the impedances of the whole group. The shadow form-
ing by the in-between piles cannot be taken into account. This method also requires a large computer ca-
pacity, since for the 230 piles in the group the number of different distances between individual piles is
huge; linear interpolation introduces errors stemming from the oscillatory nature of both the real and the
imaginary parts of the interaction factors. For all these reasons, this method is only to validate the super-
position procedure and then use it with the simplified approach (c) described below.
(c) a simplified interaction-factor method, which uses the rigorous impedances of the single pile and semi-
analytically derived interaction factors between two piles at different spacings. These factors are ob-
tained using the aforementioned boundary-element-based interaction factors in conjunction with the
wave-interference analytical solution of Dobry & Gazetas (1988) and Mylonakis & Gazetas (1998a & b).
One of the main issues that this paper attempts to clarify is how to treat the interaction between
fairly distant piles (say spacings s = 15d - 30d). There is a perhaps justified suspicion that the elas-
 tic wave-propagation solutions produce interaction factors which approach zero only asymptoti-
cally --- i.e. too slow; they may thus overestimate (even slightly) the real but unknown interaction
factors.

F C1

A
D
B

3. The division of the pile group in eight (8) subgroups (A, B, C, C1, D, E, F, G).

Other potential effects that may cause a smaller actual pile-to-pile interaction than the one com-
puted by either the rigorous, (a) or the simplified, (b) and (c), solutions are: (i) the nonlinear soil re-
sponse (near the pile cap); (ii) the unavoidable variability in the geometric and material parameters.
Recent research on this subject (see Gohl, 1993) has shown that, indeed, both nonlinearity in the
soil and randomness in material and geometric parameters tends to reduce the strong interaction
produced by linear deterministic solutions, such as the ones used in this study.
To compensate for all the above effects in a simple way, results are presented with the interac-
tion between any two piles spaced at 20d or more are neglected.

3 INTERACTION FACTORS

Interaction factors for the actual profile, obtained with the rigorous analysis (a), are illustrated in
Figure 4 for the horizontal mode of vibration x (symbolized as Ax). The distances for which these
factors are calculated vary from the closest (s = 2m) to a medium large distance of (s =10d = 13m).
0.8 0.1
s = 13.0 m
0.6 s=2m 0.0
Real ( Ax )

Imag ( Ax )

0.4 2.6 m -0.1 6.5 m

3.9 m
0.2 -0.2
3.9 m
6.5 m
0.0 -0.3
2m
13.0 m 2.6 m
-0.2 -0.4
0 2 4 6 8 10 12 0 2 4 6 8 10 12
frequency f : Hz frequency f : Hz
Figure 4. Horizontal interaction factor Ax (θ=0o) for various pile distances.

An interesting finding is that below a frequency of about 2 Hz the horizontal interaction factors
Ax and Ay have vanishingly small imaginary parts, while their real parts exhibit a broad peak at f =
2 Hz. This behavior is similar to the behavior of flexibility functions of piles and shallow founda-
tions (Krishnan et al, 1983; Gazetas, 1983) and is a result of two phenomena: (i) resonance at the
 fundamental natural shear frequency of the stratum surrounding the pile (fs ≈ 2.10 Hz); (ii) no later-
ally propagating waves below fs.
Also notice an unusual but explainable behavior in the Ay interaction factors at large distances
s/d > 10: the amplitude |Ay| attains very large values for frequencies between 3 and 8 Hz (see Fig.
5), and in fact exceeds the amplitude for the smaller distance s/d = 5 for certain frequencies. Such
behavior is difficult to model analytically. The cross and rocking interaction factors are as usually
very small, and could be neglected for all but the closest piles (s < 5d).
0.8 0.6 0.4
| Ax | | Ay | | Az |
s=2m s=2m
0.6 0.3
Amplitude

0.4 s=2m
2.6 m 2.6 m
2.6 m
0.4 3.9 m 0.2 3.9 m
3.9 m
0.2 6.5 m 6.5 m
0.2
6.5 m 0.1
13.0 m 13.0 m
13.0 m
0.0 0.0 0.0
0 2 4 6 8 10 12 0 2 4 6 8 10 12 0 2 4 6 8 10 12

frequency f : Hz
Figure 5. Horizontal interaction factor Ax (θ=0o) for various pile distances.

Closed-form expressions were fitted to the above interaction functions. For instance, for the
horizontal lateral motion:

s
− 0.70  − (i + β s ) 3 π ( f − 1) s 
exp  
3
A x = A ( 0° ) ≈    
(1)
h 4 d  V
La 

4 IMPEDANCE FUNCTIONS

4.1 Impedance of a single pile

For the single d = 1.30 m pile, the results show that the dynamic stiffnesses are more or less con-
stant, independent of frequency. Their values are in accord with the closed-form expressions of
Gazetas (1991). For example:
0.21
 Ep 
Kz =E d  (2)
s E 
 s
Substituting for Es = 2 (1+νs) Gs = 2×1.49×9600 = 28600 kPa and Ep = 2.5 107 kPa,
4 0.21
K z = 28.6 × 1.30 × ( 2.5 × 10 / 28.6) = 0.15 GN / m (3)
which, indeed, compares well with the computed value of 0.17 GN/m. The underprediction is un-
derstandable in view of the presence of the stiff layer of sand, within the active length of the pile,
which the above expression ignores.
4.2 Subgroups
Results from the analysis of subgroup A are illustrated in Figure 6 (for the y direction). These
analyses had several objectives:
• to demonstrate that the superposition method is applicable to the problem,
• to investigate the degree of accuracy of the superposition method which uses simplified interaction fac-
tors,
• to get an idea of the sensitivity of the various impedance functions to the algebraic form and the exact
values of the interaction factors, and
• to explore the effects of several additional simplifications (such as, for example, using a cutoff spacing
scutoff = 20d beyond which interaction between two piles is ignored), which may lead to results closer to
reality than those of rigorous elastic methods
2 8
Real ( Ky ) : GN / m

Imag ( Ky ) : GN / m
6
0

4
-2

2
Simplified
-4
BEM
0
0 2 4 6 8 10 0 2 4 6 8 10
frequency f : Hz frequency f : Hz
Figure 6. Subgroup A : Lateral impedance Ky.

These objectives have been largely met with the performed comparative study. In Figure 6 the
overall agreement between rigorous and simplified results is satisfactory. Furthermore, the above
conclusions were justified by respective results for the other subgroups.

4.3 Impedance of the 230-pile group

4.3.1 Superposition of subgroup impedances

It could be intuitively believed that reasonably good estimates of the overall impedances of the
complete 230-pile group would be obtained by simply adding algebraically the impedances of the
eight individual subgroups, A through G. Such an addition would have been valid if the interaction
among piles of different subgroups were negligible.
It is turned out that this is not the case, since the piles are placed extremely closely to one an-
other (minimum spacing s = 1.54d). To prove that this is indeed the case, the impedances (i) of the
two neighboring subgroups A and E, and (ii) of the three neighboring subgroups E, F, and G were
combined and compared with the rigorous and the simplified methods. More specifically, both the
rigorous (BEM) and the simplified (interaction factors) method were used:
• to obtain separately the impedances of each constituting subgroup (A,E,F,G) and then add them up for
each frequency
• to obtain directly the impedances of each whole group, i.e. of group A+E and of group E+F+G

4.3.2 Results from the simplified method

The dynamic stiffness and damping for the whole pile group, obtained with the superposition
method (using the curve-fitted interaction factors) are presented in Figure 7 (for the y direction).
Two types of analysis were performed: one considering the interaction between all pile pairs,
and one imposing a cutoff pile-separation distance equal to 20d, beyond which all interaction fac-
tors were set equal to zero. It was found that, imposing the scutoff leads to an appreciable change in
the translational stiffnesses, but rather negligible differences in the rocking impedances. Such large
differences stem from the fact that there is a huge number of piles spaced at s > 20d, so that ac-
counting or not for their interaction makes a big difference.
Figure 8 is prepared to convince the reader of the above statement regarding the number of
piles, displaying the histogram of the number of pile pairs having s/d falling within a certain range.
 It is seen that there are about 8600 pile pairs within 20 < s/d < 25 alone! The total number of pile
pairs exceeding 20d is about 25000, nearly 1/2 of the total number of pile pairs in the facility.
10 80
Real ( Ky ) : GN / m

Imag ( Ky ) : GN / m
0 60

-10 40

-20 20
Cutoff at s > 20 d
No cutoff
-30 0
0 2 4 6 8 10 0 2 4 6 8 10
frequency f : Hz frequency f : Hz
Figure 7. Lateral dynamic stiffness and damping of the 230-pile group obtained with a cutoff at pile distances
s > 20 d, and without cutoff.
12000
Total combinations of pile pairs

10000

8000

6000

4000

2000

0
0 5 10 15 20 25 30 35 40 45 50 55
dimensionless pile spacing s / d
Figure 8. Histogram of the number of pile pairs with s/d falling within a certain range.

5 CONTRIBUTION OF THE SIDE WALLS

The base of the foundation mat cannot transmit any appreciable action to the soil. In all likelihood
there is no contact between soil and foundation mat, since the piles (as end-bearing) are much
stiffer vertically than the surrounding soft clay. It is certain that over the years this clayey layer has
consolidated separated from the mat. Thus, only the sidewalls of the facility offer additional stiff-
ness and radiation damping to the foundation. Theoretical solutions to estimate the contributions of
sidewalls have been developed and published by Gazetas (1991) (perhaps the only solutions that
separate the effect of the sidewalls from the effect of the basemate).
The total stiffness and damping of the piles and sidewalls were obtained by:
K ≈ K + K (4)
total piles sidewalls

k (ω ) ≈k (5)
total piles

C ≈C +C (6)
total piles sidewalls
where K is the static stiffness, k(ω) is the dynamic stiffness coefficient, and C is the dashpot
(damping) coefficient.
 The first of the above equations is, of course, only an approximate one, since there may be an
amount of interaction between piles and the vertical loaded sidewalls. The addition considered
herein provides an upper-bound of the effect of the sidewalls.
Regarding the second equation, some evidence from rigorous solutions (Tyson & Kausel, 1983;
Gazetas, 1983) suggests with sufficient accuracy that:
k (ω ) ≈1 (7)
sidewalls
10 100

Imag ( Ky ) : GN / m
Real ( Ky ) : GN / m

80
0

60
-10
40

-20
Cutoff at s > 20 d 20
No cutoff
-30 0
0 2 4 6 8 10 0 2 4 6 8 10
frequency f : Hz frequency f : Hz
Figure 9. Lateral dynamic stiffness and damping of the 230-pile group (including the contribution of the
sidewalls) obtained with a cutoff at pile distances s > 20 d, and without cutoff.

Finally, the superposition of dashpots from the sidewalls and piles is an excellent approxima-
tion, as has been directly and indirectly shown by Gazetas & Tassoulas (1987) and Gazetas (1991).
Figure 9 compares the horizontal stiffnesses Ky obtained for the “no-cutoff” and “cutoff-20d”
assumptions. Both the 230-pile group and the sidewalls of the embedded structure are taken into
account. All impedances refer to the centroidal axis of the pile group and to massless foundation
mat and sidewalls. Notice the appreciable differences between the two approaches.

6 FOUNDATION INPUT MOTION

Detailed free-field response analyses have been performed using the strain-compatible equivalent-
linear iterative procedure embodied in the computer program SHAKE. The design ground motions
were deconvoluted to derive the base seismic motion at a depth of 100 m, as well as the seismic
motion at the level of the foundation. Furthermore, analyses for the kinematic response of the pile
group were performed. These analyses have shown that the engineering approximation of neglect-
ing modifications in the “effective input motion” due to the presence of the pile group are justified.

7 INTERNAL FORCES IN PILES

7.1 Internal forces due to the inertial loads of the superstructure

A series of dynamic analyses were performed to obtain floor response spectra at the various levels
of the major critical facility. In those analyses, an artificial earthquake time history (Fig. 10) was
used as base excitation of the elastically supported structure. Two different supporting springs and
dashpots were considered, corresponding to the two sets of impedance curves, with and without
cutoff frequency, respectively.
Time histories of total shear force and overturning moment transmitted onto the foundation
were obtained for each of the 6 (six) cases studied, for both directions, x and y. Table 1 summarizes
the findings of that study in the form of peak values of shear force and overturning moment.
 0.15 0.4

Spectral acceleration : g
Ground acceleration : g

0.10 β = 10%
0.3 β = 5%
0.05

0.00 0.2

-0.05
0.1
-0.10
artificial motion
-0.15 0.0
0 2 4 6 8 10 12 0.0 0.5 1.0 1.5 2.0
time t : sec period T : sec

Figure 10. The 0.1g artificial base motion and its 5% and 10% damped response spectra.

The results of that study also showed that the fundamental period of the structure-foundation
system is about 1 second, and that the influence of the higher modes is negligible.

Table 1. Maximum structural reaction transmitted to foundation (units: Q in MN and M in MN m).

Impedance curves x direction y direction
Qx Mry Qy Mrx
“cutoff”, S > 20d 87.4 2230 93.9 2300
no cutoff 73.8 1960 85.7 2160

2.0
Pile shear force Fy / total shear force (%)

f = 1 Hz

1.6

1.2
P / 230
total
0.8

0.4

0.0

0 20 40 60 80 100 120 140 160 180 200 220 240

pile number

Figure 11. Distribution of the total shear foundation force on each pile at frequency f = 1 Hz.

Figure 11 displays for the 230 piles their share of the total shear force at the fundamental fre-
quency of the system. In the sequel, the response of two particular piles is examined:
• Pile 227 represents the perimetric piles that carry a very large part of the load (about 1% of the total
shear, compared to the average of 0.43%)
• Pile 40 represents the central piles that carry a very small part of the load about 0.2% of the total shear).
For piles 227 and 40 the variation with frequency of the shear pile-head forces is presented in Fig-
ure 12. The change with frequency of the load taken by the two piles is thus clearly seen.
For the seismic excitation, the time histories of shear forces Q at the head of these two piles are
portrayed in Figure 13. In addition to the significant differences in amplitudes, it is important to
notice the differences in the frequency content of the forces in pile 227 versus pile 40.
 6e-3 0.02
PILE 40 (CENTRAL) PILE 227 (CORNER)
Qy / total shearforce

Qy / total shearforce
real

3e-3 0.00

imaginary

0 -0.02
0 5 10 0 5 10
frequency f : Hz frequency f : Hz
Figure 12. Time history of pile-head shear force of piles 40 and 227. Notice the differences in magnitude and
frequency content of the response for the two piles.
150 1000
PILE 40 (CENTRAL) PILE 227 (CORNER)
shear force Qy : kN

shear force Qy : kN

0 0

-150 -1000
0 7 14 0 7 14
time : s time : s
Figure 13. Envelopes of the peak pile kinematic bending moment with depth for the time- and the frequency-
domain values of moment
7.2 Kinematic bending moments of piles
Results have been obtained for the bending moments of a single fixed-head pile, due to vertically
propagating shear waves that are consistent with the base input acceleration history.
0

-4
depth : m

-8

-12
steady-state (at f = f1)
time domain
-16
0 1 2 3 4 5 6 7
bending moment : MN m

Figure 14. Envelopes of the peak pile kinematic bending moment with depth.

Figure 14 summarizes the above results in the form of the distribution with depth of the peak (in
the time domain) values of moment, and the largest (in the frequency domain) values of the mo-
ment-to-acceleration transfer functions. The latter occur at, or about, the fundamental frequency of
the soil deposit.
 It is worth noting that the largest peak values occur at the head of the pile (due to the restriction
of rotation by the rigid base slab) and at the interface between the two soil layers, at a depth of 9.50
m, (stemming from the great difference in soil stiffnesses). Also note the difference between the
time-domain and the steady-state plots (by a factor of the order of 4.5). A study on these effects can
be found in Nikolaou et al (1995).

8 CONCLUSIONS

The elastic wave-propagation solutions produce interaction factors that approach zero only as-
ymptotically. The effect of interaction between distant piles (s>15d) is significant for large pile
groups such as the 230-pile foundation of the studied facility. It was found that imposing a cutoff
distance at s = 20d, leads to a change in the translational stiffness of the group of the order of 25%
at low frequencies. This is because, the total number of pile pairs exceeding 20d is about 25000;
nearly half of the total number of pairs in the group. Unfortunately, very little is known about the
effect of the small variations of soil mechanical characteristics on such “distant-pile” interaction.

9 REFERENCES

Dobry, R. & Gazetas, G. (1988). “Simple method for dynamic stiffness and damping of floating pile
groups”, Geotechnique, Vol. 38, No. 4, pp. 557-574
Finn, L. & Gohl, B.W. (1992). “Response of model pile groups to strong shaking”, Piles Under Dynamic
Loads, Geotech. Special Publ. No. 34, ASCE, S. Prakash, ed., pp. 27-55
Gazetas, G. & Tassoulas, J.L. (1987). “Horizontal damping of arbitrarily-shaped embedded foundations”,
Jnl of Geotech. Engng ASCE, Vol. 113, No. 5, pp. 458-475
Gazetas, G. (1983). “Analysis of machine foundation vibrations: state of the art”, Soil Dyn. Earthq. Eng.,
Vol. 2., No. 1, pp. 2-41
Gazetas, G. (1991). “Foundation vibrations”, Foundation Engineering Handbook 2nd edition, Van
Nostrand Reinholds, pp. 553-593
Gazetas, G., Fan, K., Kaynia, A., & Kausel, E. (1991). “Dynamic interaction factors for floating pile
groups”, Jnl Geotechn. Engng ASCE, Vol. 117, pp. 1531-1548
Gohl, 1993 Gohl, W.B. (1993). “Response of pile foundations to earthquake shaking - general aspects of
behaviour and design methodologies”, Seismic soil-structure interaction seminar
Kaynia, A.M. (1982). “Dynamic stiffness and seismic response of pile groups”, Research Report R82-03,
Massachusetts Inst. of Technology
Michaelides, O., Gazetas, G., Bouckovalas, G., & Chrysikou, E. (1998). “Approximate nonlinear dy-
namic axial response of piles”, Geotechnique, Vol. 48, No. 1, pp. 33-53
Mylonakis, G., Nikolaou, A., and Gazetas, G. (1997). “Soil-Pile-Bridge Seismic Interaction: Kinematic
and Inertial Effects. Part I: Soft Soil” Earthq. Engng & Struct. Dynamics, Vol. 26, 1997, pp. 337-359
Mylonakis, G. & Gazetas, G. (1998). “Settlement and additional internal forces of grouped piles in lay-
ered soil”, Geotechnique, Vol. 48, No. 1, pp. 55-72
Mylonakis, G. & Gazetas, G. (1998). “Vertical vibrations and additional distress of grouped piles in lay-
ered soil”, Soils & Foundations, Vol. 38, No. 1, pp. 1-14
Nikolaou, A., Mylonakis, G., and Gazetas, G. (1995). “Kinematic Bending Moments in Seismically
Stressed Piles”, Report NCEER-95-0022, National Center for Earthquake Engineering Research, State Univ.
of New York, Buffalo, NY
Novak, M. (1991). “Piles under dynamic loads: State of the art”, Proceedings, 2nd International Confer-
ence on Recent Advances in Geotechnical Earthquake Engineering and Soil Dynamics, St. Louis, Vol. 3, pp.
2433-2456
Schnabel, P.B., Lysmer, J & Seed,, H. B. (1972). “SHAKE: A computer program for earthquake re-
sponse analysis of horizontally layered sites”, Report EERC 72-12, University of California, Berkeley
Tyson, T.R. & Kausel, E. (1983). “Dynamic analysis of axisymmetric pile groups”, Research Report R83-
07, Massachusetts Inst. of Technology
Waas, G. & Hartmann, H.G. (1984). “Seismic Analysis of Pile Foundations Including Pile-Soil-Pile In-
teraction”, 8th Int. Conference of Earthquake Engineering, San Francisco, Vol. 5, pp. 55-62

View publication stats