Case history based soil liquefaction models correlating cyclic resistance ratio (CRR) with engineering index properties are usually formulated using field cyclic stress ratio (CSR) data representing seismic demand. CSR data either exceeds or falls short of soil capacity CRR, depending on the occurrence or non-occurrence of liquefaction. The concept of data censoring, which captures this excess or shortfall, is introduced to model such data. Survival analysis regression methods, explicitly dealing with censored seismic demand, are applied to correlate CRR as a function of standard penetration resistance (N 1)60, fines content (FC) and earthquake magnitude (M w). Unlike logistic regression which is conditioned with only the binary data outcomes (survival and failure), the survival analysis regression is conditioned on binary outcomes coupled with the censored seismic demand. Cox’s proportional hazards semi-parametric model is formulated to identify the relative risks for various earthquake and soil covariates, explain their influence on liquefaction, select the important covariates for soil liquefaction modelling and build risk scores to compare liquefaction potential for different sites. Finally, soil liquefaction models based on survival analysis parametric regression are developed to evaluate the factor of safety and probability of liquefaction.
