| The goal of this research is to address the issues that must be resolved for a rigorous reliability analysis of soil liquefaction. As performance-based earthquake engineering concepts are gaining wider acceptance, it is essential to estimate accurately the probability of liquefaction. The first order reliability method (FORM) is one of several rigorous and robust techniques that may be used for estimating the probability of liquefaction. Rigorous reliability analysis using FORM requires knowledge of both model and parameter uncertainties, where model uncertainty refers to the uncertainty of the adopted limit state model and parameter uncertainty refers to the uncertainty associated with the input parameters that are required by this limit state model. In this study, the focus is on stress-based simplified methods, as they are generally the methods of choice for liquefaction evaluation by the practicing geotechnical engineer. In the context of reliability analysis, the boundary curves used in these simplified methods are the limit state models. Where the existing simplified methods tend to be conservative, the uncertainty associated with any of these limit state models has never been characterized. In this dissertation study, the uncertainty of three limit state models that are based on three respective in situ tests, namely, Standard Penetration Test (SPT), Cone Penetration Test (CPT), and shear wave velocity (Vs) measurements are characterized within the framework of the FORM.; A methodology is developed in this study to characterize the uncertainty of each of the three limit state models. This methodology is based on the premise that a fully calibrated Bayesian mapping function is available that can produce an accurate estimate of the probability of liquefaction under a set of given conditions. In the case of reliability analysis of soil liquefaction, this premise stands as the databases of liquefaction/no-liquefaction case histories are available, and the Bayesian mapping functions based on SPT, CPT, and Vs can be developed and calibrated. In fact, the robustness of the Bayesian mapping function approach is demonstrated in this study. Following the developed methodology and the procedures, the uncertainty of each of the three limit state models is characterized within the framework of the FORM.; The results obtained in this study demonstrate that the uncertainty of a limit state model may be represented with a random variable, c1, that can be characterized with two statistical parameters, the mean (μ c1), and the coefficient of variation (COV). (Abstract shortened by UMI.)... |
