Seismic analysis of infinite pile groups in liquefiable soil

Soil Dynamics and Earthquake Engineering 24 (2004) 565 575 www.elsevier.com/locate/soildyn Seismic analysis of infinite pile groups in liquefiable soil Assaf Klar, Sam Frydman*, Rafael Baker Faculty of Civil & Environmental Engineering, Technion, Israel Institute of Technology, Haifa 32000, Israel Accepted 21 October 2003 Abstract Numerical analysis of an infinite pile group in a liquefiable soil was considered in order to investigate the influence of pile spacing on excess pore pressure distribution and liquefaction potential. It was found that an optimal pile spacing exists resulting in minimal excess pore pressure. It was also found that certain pile group configurations might reduce liquefaction potential, compared to free field conditions. It was observed that for closely spaced piles and low frequency of loading, pile spacing has little influence on the response of the superstructure. q 2003 Elsevier Ltd. All rights reserved. Keywords: Liquefaction; Piles; Lateral loads; Dynamic loads; Pile groups; Seismic response; Soil pile interaction; Non-linear 1. Introduction In the last decade, a number of analytical and experimental investigations have dealt with the behavior of soil pile interaction in liquefiable soils. Most of these investigations considered single piles, or small groups of piles. Most of the experimental studies focused on the extraction of p y curves from the seismic performance of the piles [1,2], while, the analytical work focused on simulating the experimental observations [3,4]. Kagawa [5] conducted an extensive analytical parametric study to examine the influence of different soil and loading conditions on the behavior of a single pile in liquefiable soil. In a separate paper, the authors [6] performed a parametric study to demonstrate the influence of flow characteristics on soil pile interaction for a single pile. Kagawa et al. [7] concluded, from shaking table tests, that excess pore pressure is affected significantly by redistribution and dissipation, and, in turn, affects the response of the soil pile system. The experimental results of Kagawa et al. [7] show that in most cases, excess pore pressures between the piles are higher than those at the same level far from the piles. Kagawa et al. suggested that this is a result of hindrance to dissipation due to piles reducing drainage. On the other hand, Sakajo et al. [8] performed shaking table tests on a pile group of 36 piles, and showed that existence of the pile group reduces the amount of excess pore pressure * Corresponding author. developed at any time, compared with the free field, and may even prevent liquefaction. These appear to be somewhat contradicting observations. Does the pile group configuration reduce excess pore pressure development, or does interference to the drainage prevent dissipation and thus contribute to an increase of excess pore pressure? The present paper addresses this question on the basis of a numerical parametric study. Two of the major factors controlling susceptibility to liquefaction, for soil with given relative density and permeability, are the drainage conditions and the loading intensity. For an infinite pile group, pile spacing has opposing effects on each of these two factors. As spacing between the piles decreases, excess pore pressures dissipate more slowly, and the potential for liquefaction is increased. However, as the distance between piles is decreased, each pile is subjected to less loading. Consequently, less pore pressure is generated, and the potential for liquefaction is, therefore, reduced. The purpose of this study is to investigate whether there is an optimal configuration of piles which reduces the potential for liquefaction. 3D numerical analysis of a large pile group requires large computational resources, and is almost unfeasible. However, numerical analysis of an infinite pile group requires as little computational effort as that required for a single pile. This is due to the identical behavior of each pile in the group. Therefore, it was decided to investigate the behavior of an infinite pile group, which is the limit case for a large pile group. The numerical analysis of an infinite pile group 0267-7261/$ - see front matter q 2003 Elsevier Ltd. All rights reserved. doi:10.1016/j.soildyn.2003.10.007

566 A. Klar et al. / Soil Dynamics and Earthquake Engineering 24 (2004) 565 575 layer are approximately 3.5 and 9.5 Hz. The E W record of the October 1, 1987 Whittier earthquake was chosen as the excitation input; the dominant frequency of that excitation is in the range of 5.5 7 Hz. This site and excitation input were used for the analysis of a single pile [6]. Fig. 2 shows the first 10 s of the input earthquake record. Since an infinite structure cannot rotate, the superstructure can be approximately represented as an equivalent mass located at the top of the pile. This equivalent mass, M; is given by M ¼ n b r f S 2 ð1þ Fig. 1. Unit cell of the infinite pile group problem. requires consideration of only a single pile, together with a periodic boundary condition. Such an approach was used by Law and Lam [9] for static loading. Fig. 1 shows a unit cell, ABCD, representing the infinite geometry. Considering the symmetric nature of the loading, only half of this cell (ABFE) is required. In this case, boundaries AB and EF should only prevent motion in their normal direction, whereas boundaries AE and BF are periodic boundaries, which are related to each other. Basically, the periodic boundary condition consists of interchanging forces/stresses between boundaries AE and BF. 2. Numerical analysis The numerical analyses were conducted for the case of piles embedded in a 15 m homogeneous sand layer of relative density Dr ¼ 60% and coefficient of permeability k ¼ 10 24 m=s; overlaying a rigid stratum. The water table is located at a depth of 1.5 m. A 0.5 m square cross section pile, with Young s Modulus 3 10 7 kpa, pinned at its tip, is considered. The first two natural frequencies of the sand where n b is the number of floors, r f is the mass of a floor per unit area and S is the distance between equally spaced piles. In order to allow an optimal investigation of different conditions, the analyses were chosen in such a way that most cases lie on three different characteristic lines, as shown in Fig. 3. The horizontal axis characterizes the building (i.e. the mass per unit area of the superstructure), and each vertical line characterizes a particular structure. The vertical axis represents the mass per pile and the sloped lines characterize the pile group geometry (i.e. spacing). The cases analyzed are indicated by circles. The numerical scheme is the same as that used by Klar et al. [6], but with a periodic boundary condition. This scheme utilizes the solution of plane strain problems by finite differences, in conjunction with coupling of the plane strain problems. More information about the technique is given in Refs. [6,10]. The constitutive relation adopted for the soil is an extension of the quasi-hysteretic model described by Muravskii and Frydman [11]. In this model, stiffness and damping depend on the weighted mean (with respect to time) of strain and strain rate. The original model was formulated in terms of a system consisting of one degree of freedom bodies, and it was extended in Ref. [6] to 3D problems. The model for pore pressure development and dissipation is detailed in Fig. 2. Acceleration bedrock input data.

A. Klar et al. / Soil Dynamics and Earthquake Engineering 24 (2004) 565 575 567 Ref. [6] and is essentially based on the fundamental model of Martin et al. [12]. 3. Results and discussion Fig. 3. Chosen analysis cases. 3.1. Soil structure interaction Fig. 4 shows contours of maximum acceleration at the pile head. Note that for normalized pile spacing, S=B; smaller than about 8, the maximum acceleration is almost constant along vertical lines, i.e. for constant mass per unit area, there is little influence of pile spacing on the motion of the superstructure. It was noted that, in general, the calculated acceleration time histories contained a single, high amplitude acceleration pulse. In order to eliminate the possibility that the results in Fig. 4 are associated with short duration acceleration peaks, the root mean square Fig. 5. Root mean square acceleration. acceleration was also calculated for the first 10 s of the motion. The root mean square is defined as sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi ðtd 1 a rms ¼ ½aðtÞŠ 2 dt ð2þ T d 0 where T d is integration time interval. The root mean square acceleration is sometimes used as a ground motion parameter, with T d taken equal to the strong motion duration. Fig. 5 shows contours of the root mean square acceleration, indicating a similar trend to that found for the maximum acceleration. In order to shed some light on this behavior, the nature of the stiffness of the infinite pile group is examined, using the concepts of interaction factors and superposition. The superposition method is implemented through a flexibility approach, in conjunction with interaction factors. For a linear, elastic soil pile system, one can write the following relation {U i } ¼ l s ½a i;j Š{P j } ð3þ where U i is the head displacement of pile i; l s is a single pile flexibility, P j is the force at the head of pile j and a ij is the interaction factor between pile i and pile j; defined as the ratio of the head displacement of pile i due to load on pile j to the displacement of pile j: For a rigid pile group cap, U ¼ U 1 ¼ ¼ U n ; resulting in the following expression for the pile group stiffness K group ¼ X n P j j¼1 U ¼ 1 l s X n X n 1 i;j ¼ k s Xn X n i¼1 j¼1 i¼1 j¼1 1 i;j ð4þ Fig. 4. Maximum acceleration at pile head. where k s is a single pile stiffness and 1 i;j are the elements of the inverse of matrix a: For an infinite pile group P 1 ¼ ¼ P i ¼ ¼ P 1 ; the reduced stiffness of a single pile, k s r may be defined as:

568 A. Klar et al. / Soil Dynamics and Earthquake Engineering 24 (2004) 565 575 k s ri ¼ k s = X j a i;j ð5þ For a large group of piles, it is customary to use this reduced stiffness for the evolution of the group stiffness as follows K group ¼ nk s ri ¼ nks X n a i;j j¼1 ð6þ where i; in this case, takes the index of the middle pile of the outer row [13]. Eq. (4) is based on the assumption that all piles displace equally while Eq. (6) assumes that all piles are equally loaded. These two assumptions are generally incompatible, except in the limiting case of an infinite pile group in which both loads and displacements are equal for all piles. The above expressions hold both for dynamic and static problems, the difference between these cases being expressed in the values of the interaction factors. For an infinite pile group, the value of the group stiffness is meaningless since the number of piles is infinite. Therefore, a stiffness per unit area is defined. For equally spaced piles, this stiffness, k 1 ; is equal to the reduced stiffness of a single pile divided by the square of the distance between the piles: k 1 ¼ ks r s 2 ¼ k s s X ð7þ 2 a i;j j For static loading, which may be considered an approximation for loading at low frequencies, the approximation of Randolph [14] for the interaction factor for fixed head piles in homogeneous linear soil, a f is given as a f ¼ 0:6 2 p ð1 þ nþ E 1=7 p ð1 þ cos 2 cþ r 0 ð8þ ð1 þ 2nÞ E s R where E p and E s are Young s moduli for the pile and the soil, respectively, c is the angle between the line joining the pile centers and the direction of loading, r 0 is the radius of the pile and R is the distance between the piles. Considering the configuration shown in Fig. 6, the interaction factor is equal to 8 1; >< k ¼ l ¼ 0 a F ¼ 0:6 2 p ð1 þ nþ E 1=7 p 1 þ k2 ð1 þ 2nÞ >: otherwise E s k 2 þ l 2! r 0 p s ffiffiffiffiffiffiffiffi ; k 2 þ l 2 where k and l are column and row numbers, respectively. For very closely spaced piles, where a f may be greater than 0.5, Randolph suggested to replace it by 1 2 ð4a f Þ 21 to avoid overestimation of interaction. This correction was applied in the present calculations, where applicable. ð9þ Combining Eqs. (7) and (9) results in: k 1 ¼ ks r s 2 ¼ k s sr 0 0:6 2 1 þ 2n E 1=7 ð10þ p sum þ s 2 1 þ 0:75n sum ¼ X1 X 1 k¼21 l¼21 exclude k ¼ l ¼ 0 Fig. 6. Numbering of the piles. E s! 1 þ k2 k 2 þ l 2 1 p ffiffiffiffiffiffiffi ffi ; k 2 þ l 2 The series sum in Eq. (10) does not tend to a constant value. Consequently, use of Randolph s interaction factors results in a zero reduced stiffness of the piles for an infinite pile group. To overcome this problem of zero stiffness, the interaction of distant piles may be disregarded. Fig. 7 shows the variation of k 1 ; as calculated on the basis of Randolph s interaction factors, using the above suggestion, i.e. interaction between piles separated by a distance greater than R max is neglected. The figure shows the variation of k 1 Fig. 7. Variation of stiffness for unit area, n ¼ 0:25:

A. Klar et al. / Soil Dynamics and Earthquake Engineering 24 (2004) 565 575 569 Fig. 8. Maximum bending moment in the piles. as a function of the distance between the piles for different ratios of pile to soil stiffness and different values of R max : The discontinuities in the variation of k 1 with S=B are related to the discrete nature of the problem; as the spacing between the piles reaches a point, where piles are excluded from the analysis (i.e. when a pile distance is greater than R max ), a jump in the value of k 1 occurs. Fig. 7 shows that for small distances between piles, k 1 is almost constant. When dealing with seismic loading, it is customary to divide the soil pile interaction into kinematic interaction and inertial interaction. In the present case, the kinematic interaction is the response of the soil pile system to the seismic waves without the superstructure, while the inertial interaction is the response of the soil pile system to the inertial loading caused by the superstructure. This superposition of responses is exact only for linear systems, but it has become commonly applied, also, for non-linear systems. The above analyses refer to pile response to loading at the pile head, which is relevant to the inertial component of the response. Consequently, the constancy of k 1 in Fig. 7 indicates that the inertial response is independent of pile spacing for small pile spacing and static or low frequency loading. If the kinematic interaction is such that it is also independent of pile spacing in that range, then the total response will be independent of pile spacing, as was observed in Figs. 4 and 5 for closely spaced piles. To examine this, analysis of kinematic response of an infinite pile group was conducted for the considered earthquake excitation and pile geometry (i.e. no superstructure mass on the pile head). In these analyses, little influence of pile spacing was observed and in general, the motion of the pile head was fairly similar to that of the free surface, with about 5, 2 and 1% reduction in root mean acceleration value of that of the free field for normalized spacing of S=B ¼ 4; 7 and 13, respectively. These results support the trends observed in Figs. 4 and 5 based on superposition of kinematic and inertia responses. The similarity between the motion of the pile head and that of the free field surface is consistent with the observation of Fan et al. [15] who found that for low frequencies, the piles follow, almost exactly, the free field ground movement. It should be noted that for higher frequencies, the results of an infinite pile group kinematic response [16] are inconsistent with Fan et al. s [15] findings that the response of a pile group is similar to that of a single pile. Fig. 9. Free field excess pore pressure distribution.

570 A. Klar et al. / Soil Dynamics and Earthquake Engineering 24 (2004) 565 575 The values, or more accurately the trends, of both maximum acceleration and root mean acceleration in Figs. 4 and 5 are also of interest. If a linear elastic one degree of freedom system is considered, then the total response (kinematic þ inertial) due to kinematic loading is given by u ¼! v 2 1 v 2 1 2 U k e iv kt v2 k ð11þ where v k and U k are the frequency and amplitude p of the kinematic loading, respectively, and v 1 ¼ ffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffi kr=m s ¼ k 1 =n b r f is the natural frequency of the structurefoundation system. From Eq. (11), it can be seen that if the resonant frequency of the superstructure pile system, v 1 ; is less than that of the kinematic loading, a decrease in v 1 will lead to reduction in amplitude of motion (also acceleration and velocity), with a limit of zero motion. On the other hand, if v 1 is greater than v k ; an increase in v 1 results in reduction of motion amplitude, with a limit of U k : As can be seen in Figs. 4 and 5, for small pile spacing, the acceleration values decrease only with increasing superstructure mass ðn b r f Þ; while for greater pile spacing, they decrease also with an increase in pile spacing. Since the loading frequency is greater than the superstructure pile soil resonant frequency, the p behavior is directly related to the fashion in which v 1 ¼ ffiffiffiffiffiffiffiffiffi k 1 =n b r f changes. For closely spaced piles, k 1 is more or less constant (Fig. 7), as a result a reduction in v 1 is merely due to an increase in superstructure mass ðn b r f Þ: When spacing increases, k 1 also decreases (as seen in Fig. 7), leading to an even greater reduction in v 1 : This behavior is clearly seen in Figs. 4 and 5, where the contour lines curve to the left from vertical lines as the pile spacing becomes large. In the following section, it will be shown that in the same region in which k 1 is more or less constant, there is variation in excess pore pressure values. This may appear to be contradictory. However, the variation of excess pore pressure for these spacings is not sufficient to cause significant change of mechanical properties. Further discussion on the influence of excess pore pressure on the response of the soil pile system is given in Ref. [6] The maximum bending moments which develop in the piles are shown in Fig. 8. For a wide range of superstructure mass, the maximum bending moment is a function, only, of pile spacing. safety factor is equal to the inverse of r u ; i.e. the excess pore pressure ratio is an estimator for liquefaction potential.. In the present analyses, the free field excess pore pressure distribution was established as a reference value. Fig. 9 shows distribution of excess pore pressure, and the corresponding excess pore pressure ratio, r u ; at different times, in the free field for the earthquake and soil profile presented previously. Fig. 10a and b shows maximum values (in both time and space) of excess pore pressure ratio in the soil pile superstructure system. As can be seen from Fig. 10, there is a zone of minimum excess pore pressure ratio in the range of normalized spacing S=B ¼ 6 10; depending on the mass of the superstructure. For the conditions presently considered, this spacing is, therefore, optimal with regards to liquefaction mitigation. As the superstructure mass decreases from about 16 to 2 ton/m 2, the optimal normalized pile spacing increases linearly from about 6 to 10. It should be noted that in some cases, momentary liquefaction developed for large 3.2. Liquefaction susceptibility Liquefaction susceptibility may be quantified by referring to excess pore pressure ratio, r u ; defined as the ratio of excess pore pressure to the initial vertical effective stress. The factor of safety against liquefaction can be defined as the ratio of excess pore pressure required for initial liquefaction to the existing excess pore pressure. This Fig. 10. Excess pore pressure ratio.

A. Klar et al. / Soil Dynamics and Earthquake Engineering 24 (2004) 565 575 571 pile spacings. However, these cases were not included in Fig. 10, since the change of maximum excess pore pressure close to liquefaction is not a continuous one, and a sudden jump occurs in maximum excess pore pressure when going from a stable state to the state of liquefaction. This excess pore pressure threshold is discussed in Ref. [6]. In the analyses, which were conducted here, the highest pore pressure ratio, which was obtained without reaching the unstable liquefaction state, was 0.75. One may, consequently, refer to the contour line of 0.75 in Fig. 10 as the boundary between stable and liquefaction states. A striking result obtained on the basis of Figs. 9 and 10 is that in some of the analyses, the maximum excess pore pressure value was smaller than that of the free field. In the pile group configuration, the maximum excess pore pressure ratio in some cases reached a value as low as 0.45 (Fig. 10) compared to about 0.55, which is the maximum pore pressure ratio in time and space in the free field (Fig. 9). Fig. 10b shows how the excess pore pressure developed in the soil depends on both loading and drainage conditions. Optimal pile spacing, resulting in minimum r u ; is in the range S=B ¼ 6 10: r u increases for both closer spacing, due to disturbance to drainage, and to greater spacing, due to higher inertial loading resulting from greater superstructure mass per pile. To obtain more insight into this phenomenon, the development of excess pore pressure ratio with time is considered. Fig. 11 shows development of excess pore pressure for normalized spacing of 6.9, for n b r f of 8 and 16 ton/m 2, for four different depths, Z=B ¼ 5; 10, 15, 20 and for different distances, d; from the pile along a line connecting the piles (parallel to the direction of loading). It should be noted that the analysis of kinematic response (i.e. zero mass of structure), discussed previously yielded excess pore pressure values only slightly less than those of the free field, with a maximum deviation of less than 10%; Fig. 11. Development of excess pore pressure ratio, n b r f ¼ 8; 16; S=B ¼ 6:9:

572 A. Klar et al. / Soil Dynamics and Earthquake Engineering 24 (2004) 565 575 this maximum deviation was obtained for the smallest spacing considered S=B ¼ 4: The difference from free field values seen in Fig. 11 is, therefore, mainly due to superstructure inertia effects. The results, shown in Fig. 11 are actually consistent with the experimental results of both Kagawa et al. [7] and Sakajo et al. [8], which appeared to be contradictory, as pointed out in the introduction. As seen in Fig. 11, at shallow depth, the pore pressure value is greater than that of the free field, similar to the behavior observed in most of Kagawa et al. s experiments in which measurement were all made at shallow depth. At greater depth, pore pressures are, in general, smaller than those of the free field, similar to the behavior noted by Sakajo et al. A decrease of excess pore pressure at greater depth would not occur for a single pile, since excess pore pressure would equalize with that of the free field, due to horizontal flow. In an infinite pile group configuration, however, there is no flow between adjacent unit cells, and this prevents the tendency for pore pressure to reach free field values. To supplement the comparisons between the two cases shown in Fig. 11, Fig. 12 presents the variation in bending moment with time, at different depths. It is seen that at the beginning of the strong motion phase of the earthquake (about 3.5 s), the bending moments in the pile increase. The development of excess pore pressure is, as expected, seen to be directly related to the intensity of loading which is expressed in the magnitude of bending moment. It is of interest to consider the effect of preventing horizontal motion of the heads of the piles. This is equivalent to considering the behavior of a superstructure with infinite mass. Fig. 13 shows the development of Fig. 12. Bending moment.

A. Klar et al. / Soil Dynamics and Earthquake Engineering 24 (2004) 565 575 573 Fig. 13. Development of excess pore pressures for infinitely heavy structure, S=B ¼ 6:9: excess pore pressures with time for this case, indicating that excess pore pressure values are smaller than those shown in Fig. 11. The influence of pile spacing on the development of excess pore pressures is shown in Fig. 14, for a superstructure of n b r f ¼ 8 ton/m 2 and normalized spacings S=B ¼ 4 and 10 (the additional case of S=B ¼ 6:9 is shown in Fig. 11). It is interesting to consider the effect of the distance between piles on the dissipation of excess pore pressure. For closely spaced piles ðs=b ¼ 4Þ; the pore pressure curve in the vicinity of the pile ðd=b ¼ 0:5Þ coincides with that mid-distance between the piles ðd=s ¼ 0:5Þ; implying that there is no dissipation of excess pore pressure in the horizontal direction. However, for normalized pile spacing S=B ¼ 10; a significant deviation between pore pressure in the vicinity of the pile and that midway between piles ðd=s ¼ 0:5Þ is observed at shallow depth. The peak excess pore pressures for normalized spacing S=B ¼ 10 dissipate rather quickly compared to the peaks for S=B ¼ 4; and those shown previously for S=B ¼ 6:9: This is probably partly due to horizontal dissipation, which occurs with large spacing. The influence of pile spacing on horizontal drainage is best seen in a plot of flow vectors. Fig. 15 shows plots of flow vectors for n b r f ¼ 8 ton/m 2 and S=B ¼ 4; 6.9, and 10 at 7 s. At this time, excess pore pressures in the three cases reach their maximum values. Fig. 15 shows clearly that for normalized spacing S=B ¼ 10; the flow has a relatively large horizontal component, whereas, for S=B ¼ 4 and 6.9, the vertical component is predominant. As can be seen, the horizontal flow is depressed as it approaches the boundaries of the unit cell, indicating the inability of water to flow from one unit cell to another in an infinite pile group. It is also evident from these plots that any significant reduction in excess pore pressure for closely spaced piles is associated, mostly, with vertical flow. 4. Summary and conclusion Analyses of infinite pile groups were conducted in order to investigate the kinetic response of a large pile group and to examine whether an optimal configuration exists which minimizes liquefaction potential for such a system. Fig. 14. Development of excess pore pressure ratio, n b r f ¼ 8 S=B ¼ 4; 10.

574 A. Klar et al. / Soil Dynamics and Earthquake Engineering 24 (2004) 565 575 Fig. 15. Flow vector at 7 s, n b r f ¼ 8; S=B ¼ 4; 6.9, 10. The seismic response of the superstructure was found to be relatively unaffected by the spacing between the piles for relatively close piles (approximately S=B, 8). This finding was supported by an analytical solution for low frequency conditions. For the earthquake excitation considered, there is an optimal pile spacing, which lowers the potential for liquefaction. In general, pore pressure values may be higher or lower than that of the free field, depending on the spacing between the piles, the superstructure mass and the depth being considered. When considering maximum pore pressure ratio in space, which is an indication of liquefaction potential, it was observed that there is a distinct region in which it was lower than that of the free field. This region is dependent both on the pile spacing and the superstructure mass. The piles, by themselves, have little influence on the excess pore pressure, i.e. without a superstructure, excess pore pressures are similar to those of the free field. These results suggest that use of piles alone would not significantly contribute to liquefaction mitigation, although they may prevent its destructive effects. For large pile spacing, it was shown that horizontal flow participates in the reduction of excess pore pressure at shallow depth. It must be emphasized that the above observations were obtained from analysis of one particular earthquake input, and one soil profile. Therefore, further study should be carried out for other input condition before these observations are generalized. Acknowledgements The research described in this paper is being supported by the Israel Ministry of Housing and Construction, through the National Building Research Station at the Technion. References [1] Wilson DW. Soil-pile-superstructure interaction in liquefying sand and soft clay. PhD Dissertation. University of California at Davis, Center for Geotechnical Modeling, Report No. UCD/CGM-98/04; 1998. [2] Tokimatsu K, Zusuki H, Suzuki Y. Back-calculated p y relation of liquefied soils from large shaking table tests. Fourth International Conference on Recent Advances in Geotechnical Earthquake Enginerring and Soil Dynamic; 2001. Paper No. 6.24. [3] Boulanger RW, Curras JC, Kutter BL, Wilson DW, Abghari A. Seismic soil pile structure interaction experiments and analyses. J Geotech Geoenviron Engng 1999;125(9):750 9. [4] Finn WDL, Thavaraj T, Wison DW, Boulanger RW, Kutter BL. Seismic analysis of piles and piles groups in liquefiable sand. Seventh International Conference on Numerical Models in Geomechanics, Numog; 1999. p. 287 92.

A. Klar et al. / Soil Dynamics and Earthquake Engineering 24 (2004) 565 575 575 [5] Kagawa T. Effect of liquefaction on lateral pile response. ASCE Geotech Spec Pub, Piles Under Dyn Loads 1992;34:207 23. [6] Klar A, Baker R, Frydman S. Seismic soil pile interaction in liquefiable soil. Soil Dyn Earthq Engng this issue doi: 10.1016/ j.soildyn.2003.10.006. [7] Kagawa T, Minowa C, Abe A, Oda S. Shaking-table tests on and analyses of piles in liquefying sands. Earthquake Geotechnical Engineering Proceedings of the First International Conference, IS- Tokyo 95; 1995. p. 699 704. [8] Sakajo S, Chai JC, Nakajima K, Maeda M. Effect of group pile on liquefaction resistance of sandy ground. Earthquake Geotechnical Engineering Proceedings of the First International Conference, IS- Tokyo 95; 1995. p. 755 60. [9] Law HK, Lam IP. Application of periodic boundary for large pile group. J Geotech Geoenviron Engng 2001;127(GT10): 889 92. [10] Klar A, Frydman S. Three-dimensional analysis of lateral pile response using two-dimensional explicit numerical scheme. J Geotech Geoenviron Engng, ASCE 2002;128(GT9):775 84. [11] Muravskii G, Frydman S. Site response analysis using a non-linear hysteretic model. Soil Dyn Earthq Engng 1998;17:227 38. [12] Martin GR, Finn LWD, Seed HB. Fundamentals of liquefaction under cyclic loading. J Geotech Engng ASCE 1975;101(GT5):423 38. [13] Novak M, El Sharnouby B. Evaluation of dynamic experiments on pile group. J Geotech Engng 1983;110(GT6):738 75. [14] Randolph MF. The response of flexible piles to lateral loading. Geotechnique 1981;31(2):247 59. [15] Fan K, Gazetas G, Kaynia A, Kausel E, Ahmad S. Kinematic seismic response of single piles and pile groups. J Geotech Engng 1991; 117(GT12):1861 79. [16] Klar A. Model studies of seismic behavior of piles in sand. PhD Thesis, Technion Israel Institue of Technology, 2003.