Prediction of liquefaction induced lateral displacements using robust

Lateral spreading and flow failure are amongst the most destructive effects of liquefaction. estimation of the hazard of lateral spreading requires characterization of subsurface conditions, principally soil density, fine content, groundwater conditions, site topography and seismic characteristics. however, inaccuracies in the measurement or estimation of the influencing parameters have always been a major concern and, thus, various statistical approaches have been improvised to subdue the effect of such inaccuracies in the prediction of future events. very few empirical correlations consider the effect of uncertainties in predicting the extent of lateral movement. hence, in this article, an innovative approach, based on robust optimization, has been utilized to enumerate the effect of such uncertainties. in order to assess the merits of the proposed approach, a database containing 526 data points of liquefaction-induced lateral ground spreading case histories from eighteen different earthquakes, is used. the identification technique used in this article is based on the robust counterpart of the least squares problem, which is a second order cone problem, and is efficiently solved by the interior point method. a definition of uncertainty, based on the frobenius norm of the data, is introduced and examined against the correlation coefficients for various empirical models, including a new linear model, and, thereby, optimum values are determined. the results suggest that in comparison with al bawwab models, the robust method is a better pattern recognition tool for datasets with degrees of uncertainty. it is further shown that logarithmic correlations perform better in deterministic valuation, whereas, considering uncertainty, they give similar degrees of accuracy to the new linear model.