Probabilistic Characterization of Site-Specific Correlation between Geotechnical Parameters Using Limited Site Observation Data

摘要

Characterization of correlation between geotechnical parameters is important in engineering practice, particularly in probabilistic assessment and design. However, geotechnical parameters data obtained from field or laboratory tests of a site are usually limited and insufficient to provide a meaningful joint probability distribution of geotechnical parameters or to quantify their correlation. To address this challenge, Markov Chain Monte Carlo (MCMC) simulation-based Bayesian approaches are developed, for probabilistic characterization of site-specific joint probability distribution of geotechnical parameters and quantification of their correlation, using limited site observation data. Consider, for example, two geotechnical parameters X and Y with a correlation coefficient, ρ_(XY) between them. In this study, two ways of modelling ρ_(XY) are presented, which depend on availability of existing empirical model between X and Y. The proposed approaches probabilistically integrates the limited site-specific observation data pairs of X and Y with prior knowledge under a Bayesian framework. The integrated knowledge is transformed into a large number of X and Y sample pairs using MCMC simulation. Using the generated X and Y sample pairs, ρ_(XY) between X and Y is estimated and the marginal distributions of X and Y are evaluated. The approaches are illustrated and validated using real geotechnical data.