Bayesian approach to slope stability assessment by updating probability of failure treated as a random variable

An algorithm is proposed for updating the probability of slope failure on the basis of sequential testing using Bayesian probability theory. The approach is novel and computationally straightforward since, unlike other methods, updating of the probability of failure is carried out directly. This is because the probability is treated as a random variable whereas other methods have considered it as a purely deterministic value or as a function of shear strength parameters that are updated individually. The prior information required for the Bayesian analysis is obtained by a conventional probabilistic analysis. An exponential distribution is then used for the probability of failure, applying the principle of maximum entropy. This distribution is continually updated for individual slope cells, where repetitive excavation creates several such cells and a corresponding series of Bayesian estimates of the probability of failure is obtained. The approach is applicable to most simulations but there are limitations on its use. A more mathematically rigorous refinement using the beta distribution is currently under development