This paper presents a Bayesian approach to determine characteristic values of the undrained shear strength S_u profile for geotechnical analysis and design, particularly those using probability-based design codes. The approach integrates systematically the prior knowledge (e.g., engineering judgment/local experience) and a limited number of project-specific liquidity index (LI) data under a Bayesian framework and transforms the integrated information into a large number, as many as needed, of equivalent samples of the S_u profile using Markov Chain Monte Carlo simulation (MCMCS). Then, conventional statistical analysis is carried out to estimate statistics of the S_u profile, and the characteristic values of the S_u profile is determined accordingly. Equations are derived for the proposed Bayesian approach, and the approach is illustrated through a set of real-life data. It is shown that the approach effectively tackles the difficulty in generating meaningful statistics and probability distributions of soil properties from a usually limited number of soil property data obtained during geotechnical site investigation.
展开▼
