Using High Performance Computing for Liquefaction Hazard Assessment with Statistical Soil Models

Abstract

Conventional methods for liquefaction assessment using engineering indices such as Factor of safety against Liquefaction (FL) tend to overestimate liquefaction hazards. The soil dynamics analysis-based assessment with automatic modeling is more rational and robust. Soil properties are known for large uncertainties. Rather than deterministic soil models, statistical models for soil parameters should be considered. With automatic modeling, a large number of statistic models can be generated without difficulty. The problem becomes how to assess liquefaction hazard with statistic models in an efficient way. Using high performance computing, we develop an efficient liquefaction assessment method for statistical modeling of soils. A high parallel efficiency can be achieved and a large number of statical models of the order of 104 can be simulated within a reasonable time span. The method developed in this paper can be used as an efficient tool for unravelling critical parameters of soil liquefaction.