Probabilistic evaluation of sand liquefaction based on Logistic regression model

A simple model for evaluating liquefaction probability using standard penetration test (SPT) data is developed based on logistic regression analysis of 200 sets case histories of 23 earthquakes. The proposed model uses corrected blow count (N160cs) and cyclic stress ratio (CSR) as input parameters. When liquefaction probability level is equal to 50 % as a limit state, the formulation of cyclic resistance ratio (CRR) was proposed with an exponent and the proposed model's reliability for liquefaction and non-liquefaction estimation of saturated sands is 85.71 % and 76.14 %, respectively. Comparisons of the proposed probability model with previously proposed deterministic and probabilistic approaches are performed to demonstrate the slightly conservation, because the new test data and modified coefficients are used in this analysis, so some sources of uncertainty and variance are eliminated. In order to correlate deterministic approach and probability approach, the relationship between the factor of safety against liquefaction and liquefaction probability is proposed. Five case records are further analyzed to demonstrate the feasibility and reliability of the proposed method.