Based on the reliability theory, liquefaction probability calculation involving inequality and/or equality constraints becomes a constrained optimization problem. The genetic algorithms (GAs) are popular approaches to solve constrained optimization because of their simplicity and ease of implementation. This paper aims to develop an evaluation method of liquefaction probability (PL) via GAs. The total 226 CPT-based case histories of liquefaction and non-liquefaction are collected from Juang et al. (2003) to calculate the PL. The analysis results show the GAs model is suitable for liquefaction probability problem. The relation between factor of safety (FS) and PL based on the cases studied is obtained: PL = 1/(1 + FS9.117).
Earthquake-induced soil liquefaction in loose sand layers often causes settlement and tilting of buildings. Structural failure caused by liquefaction has been observed in many earthquakes (e.g., the 1964 Niigata, Japan earthquake, the 1995 Hyogoken-Nambu, Japan earthquake, the 1999 Kocaeli, Turkey earthquake, and the 1999 Chi- Chi, Taiwan earthquake) (Ishihara, 1993; Earthquake Engineering Research Institute, 2000; Ku et al., 2004). Many empirical methods for assessing liquefaction hazard have been developed based on analysis of case histories of liquefaction and non-liquefaction around the world. The probabilistic evaluation of liquefaction is desirable because many key influence factors and environmental statues, such as earthquake force, soil properties and groundwater for earthquake-induced soil liquefaction are variability. Several approaches based on Hasofer-Lind reliability method are available for evaluating liquefaction probability (e.g. Haldar and Tang, 1979; Juang et al., 2000; Lee et al., 2007). In the Hasofer - Lind method, generally known as the advanced first order second moment (AFOSM) method, the reliability index RI is defined as the minimum distance from mean values of variables to the boundary of the failure region, in the vector direction of directional standard deviations (σ) (Low, 1997). The AFOSM relies on a constrained optimization scheme to determine the reliability index involving inequality and/or equality constraints.