Probabilistic Characterization of Undrained Shear Strength Using Liquidity Index

This paper develops Bayesian approaches for probabilistic characterization of the undrained shear strength Su of clay using limited project-specific liquidity index (LI) data. The proposed Bayesian approaches include two major components: a Bayesian model comparison approach that selects a model for probabilistic characterization of Su among several candidates and a Bayesian equivalent sample approach that employs Markov Chain Monte Carlo simulation (MCMCS) to generate a large number, as many as needed, of equivalent samples of the Su profile based on both prior knowledge (e.g., engineering judgment/local experience) and project-specific LI data. Conventional statistical analysis of the equivalent samples is then carried out to estimate statistics (e.g., mean and standard deviation) of the Su profile. The proposed approaches are illustrated and validated using the LI data simulated from a virtual clay site in a series of sensitivity studies. The results show that the proposed approaches properly identify the correct model for probabilistic characterization of Su and provide reasonable statistics of the Su profile based on prior knowledge and limited project-specific LI data. The accuracy of the estimated statistics improves as the number of test data used in the proposed approaches increases.