Lateral Displacement of Liquefaction Induced Ground Using Least Square Support Vector Machine

Introduction

An earthquake can lead to various types of hazards like Tsunami, ground shaking and liquefactions of granular soil deposits. The hazard is generally more prominent in terms of lateral ground displacement due to liquefaction causing damages to major infrastructures such as buildings, bridges, pipe lines, shore line utilities etc. When the surface slope is mild, a common mode of failure is lateral spreading with surface displacements that can be in the order of several meters. Hence, to evaluate the impact of liquefaction, seismic hazard assessments often require estimates of lateral ground deformations due to lateral spreading. Liquefaction-induced lateral spreading occurs on mild slopes of 0.3 to 5% underlain by loose sands and a shallow water table (Bartlett & Youd, 1995). The geologic conditions conducive to lateral spreading (gentle surface slope, shallow water table, and liquefiable cohesionless soils) are frequently found along streams and other waterfronts in recent alluvial or deltaic deposits, as well as in loosely-placed, saturated, sandy fills. The magnitudes of displacements in a lateral spread are controlled by the degree of shear strength loss in the liquefied soil, boundary conditions around the slide, static and dynamic shear forces acting on the mass of moving soil and the length of time for which the driving forces exceed the resisting forces (Bartlett & Youd, 1995). The liquefaction induced displacement is a highly complex system. The factors such as earthquake, ground slope, thickness of liquefied layer, lowest SPT value and average fine contents contribute to the displacement the most. Based on the above principles, different methods like discrete and finite element models, simplified analytical models and empirical models are currently in use for predicting the lateral spread due to liquefaction.

Newmark’s sliding block analysis for prediction of horizontal settlement due to earthquake does not consider liquefaction effect of the soil. Hence a modified Newmark’s theory is used for analysis of prediction of displacement of earth slopes (Seed et al., 2003). For the prediction of co-seismic and post seismic displacements, based on changed soil parameters, different discrete and finite element models have been proposed by Seed et al. (2003). In case of discrete system the soil is assumed as rigid body and in finite element methods it is difficult to model soil at high strain values. A rigorous model of this problem needs consideration of dynamic and three dimensional effects as well as the anisotropic and heterogeneous nature of liquefiable soil deposits. However, liquefiable sediments are often highly variable over short distances. Numerical modeling techniques that require complete, three dimensional representations of the subsurface soils are viable options only for the less common analyses of critical structures subjected to lateral spreading. Due to difficulties in obtaining representative, “undisturbed” test samples from the in situ deposit, development of accurate constitutive modeling of a liquefiable is a complex problem. In seismic hazard assessments of lifeline networks, the risks due to lateral spreading must be evaluated at sufficient number of sites covering large geographic areas. Hence, in such cases, comprehensive modeling of the geology in every potential lateral spread may not be feasible or it may be quite difficult.