Integrating statistical characterization of multivariate parameters in the reliability assessment of liquefaction dynamics

Abstract

Uncertainty caused by the inherent variability of soil properties assumes a pivotal role in the prediction of soil dynamics. The primary objective of this study is to establish a methodological framework for quantifying statistical estimation errors that are associated with geotechnical investigation aspects (such as investigation number and location) within the context of liquefaction analysis. This objective presents a significant challenge due to the considerable number of analytical parameters involved in the targeted liquefaction analysis, with a substantial portion of them remaining unobservable. Within this research, we assemble sets of parameters from historical cases that have been analyzed by experts, thereby elucidating the statistical attributes inherent in these analytical parameter sets through fundamental statistical techniques. This encompasses the scrutiny of correlation structures among parameters and their subsequent classification into distinct clusters. Furthermore, we endeavor to integrate the statistical attributes of analytical parameters into the liquefaction numerical analysis through the application of simple Bayesian estimation.Subsequently, through the utilization of a basic one-dimensional vertical seismic response analysis as an illustrative example, we investigate the propagation of uncertainty into the seismic response analysis. Building upon the previously discussed findings, we provide evidence that the statistical characteristics of analytical parameters, along with the observational components in analytical parameters, give rise to substantial variations in the dynamics of liquefaction.