Reliability Evaluation of Regression Model for Estimating Co-seismic Landslide Displacement

Abstract

Seismic coefficient values coupled with minimum pseudo-static safety factors are still used for analysis, where selection of seismic coefficients relies on expertise and judgment. However, safety factor approach does not give any idea about the deformations and displacements that are expected to occur during earthquake loading. Displacements are mostly evaluated by equations based on yield acceleration of the slope and maximum acceleration of sliding mass. In deterministic calculation, input information is considered as constant amounts as in cohesion, internal friction angle, slope and other characteristics. Due to the fact that co-seismic slopes deformation is a function of seismic yield coefficient ($k_{\text{y}}$) and other effective parameters, with the use of statistical analysis, the statistical distribution and standard deviation of $k_{\text{y}}$ can be calculated. In the current state of the art, in order to study the changes in slope reliability the following steps were taken: analyzing the pseudo-static of slope and other factors having different mechanical and physical parameters, determining the relationship between seismic yield coefficient of slopes and other parameters with the use of genetic algorithm and revolutionary algorithm, simulation of the factors effective on $k_{\text{y}}$ with specific statistical distribution applying Monte Carlo simulation using random number generator, determining statistic distribution and standard deviation of seismic yield coefficient of slope and finally determining the slopes failure probability based on allowable co-seismic slope displacement.