1 INTRODUCTION 1.1 Context of the research Piles are often used in moderate to high seismic ar- eas to support structures (buildings and bridges) where the soil is inadequate to carry the load on its own. In these seismic areas, piles often pass through shallow loose and/or soft soil deposits and rest on competent end bearing soils. Post-earthquake recon- naissance work (Mexico City 1985; Kobe 1995) has shown that a large number of pile-supported build- ings built in layered soils suffered significant settle- ment and tilting and that in several cases pile dam- age has occurred close to interfaces separating layers with very different shear moduli. It is widely ac- knowledged that piles are affected by both the movement of the superstructure, i.e. inertial forces, and the kinematic bending moments induced by the surrounding soil. Recent building codes (Eurocode 8) include pile design provisions that account for the combined effect of both mechanisms. One of the challenges faced by the engineers lies in the predic- tion of the maximum bending moment in the pile at an interface having a sharp stiffness contrast. 1.2 Effect of soil conditions on pile response The effect of local soil conditions on the observed magnitude and patterns of seismic damage to build- ings have been studied extensively in the last four decades. A synergetic relationship between earth- quake engineering and soil dynamics research has developed, with soil geometric and stiffness charac- teristics becoming important parameters in the seis- mic design of structures. The shearing stress-strain behaviour of soils, in particular the shear modulus G (γ) and the damping ratio β (γ) were found to be the properties that affect most the dynamics of soil- structure interaction at small, medium and large strains γ. A large number of analytical and numerical meth- ods have been proposed for evaluating the dynamic lateral response of piles (simplified methods, Winkler foundation models, finite element/boundary element methods). 1.3 Research objectives This paper presents a set of experimental results from a program of dynamic pile testing carried out on the earthquake simulator at Bristol University. The research was carried out within the framework of the RELUIS (Rette di Laboratori Universitari Ingegneria Seismica) project. A small scale model pile was installed in a shear stack containing several granular material configurations. The shear stack was subjected to real seismic inputs while the free- field motion and the bending response of the pile were measured. Three classic theoretical models of soil-pile interaction (Dobry & O’Rourke 1983; My- lonakis et al. 1997 and Nikolaou et al. 2001) were employed in evaluating the soil-pile kinematic inter- action. A comparison was made between the ex- Physical modeling of kinematic pile-soil interaction under seismic conditions L. Dihoru, S. Bhattacharya, C.A. Taylor & D. Muir Wood, University of Bristol, Bristol, United Kingdom F. Moccia & A.L.Simonelli University of Sannio, Benevento, Italy G. Mylonakis University of Patras, Patras, Greece ABSTRACT: A series of shaking table tests were carried out to study the kinematic response of flexible piles in layered soil deposits under seismic excitation. These tests were carried out in a deformable shear stack where the dynamic responses of the pile and the free field were recorded for various seismic inputs, soil con- figurations and pile head boundary conditions. The pile bending moments were measured along the length of the pile using strain gauges. The bending moment profiles are compared with the predictions made by three theoretical models of kinematic pile-soil interaction: (a) Dobry & O’Rourke (1983); (b) Mylonakis et al. (1997) and (c) Nikolaou et al. (2001). This study showed that the theoretical models predicted the maximum kinematic pile response with a variable degree of success. The observed differences can be attributed to the limitation imposed by the idealizations in the respective model regarding the non-linear nature of the soil.