Abstract: Generally, the test data for a specific site are very limited. However, the Bayes method can overcome the limitation of small sample size of test data. This paper aims to apply the Bayesian updating with structural reliability methods (BUS) to update the reliability of soil slopes and estimate statistics of soil properties with limited site-investigation data. The effectiveness of the proposed approach for slope reliability is demonstrated by an infinite soil slope with direct shear test. The effects of prior information of soil properties including sample size of test data, marginal probability distribution and likelihood function on the slope reliability updating are systematically investigated. The results indicate that the BUS method can accurately estimate the statistics of soil properties and update the slope reliability incorporating limited site information and the effects of probability distribution and likelihood function. The probability distribution of soil properties has a significant effect on the updated slope reliability results (e.g., posterior mean and standard deviation, updated probability of failure). It will lead to conservative estimates of the slope reliability results when the normal or lognormal distributions are used. In contrast, the likelihood function has a slight influence on the updated slope reliability. In addition, the uncertainties of soil properties and updated probability of failure decrease as the sample size of test data increases, but they will change slightly once the sample size reaches a certain value.

Key words: slope reliability, Bayesian updating, subset simulation, prior information, probability distribution