Abstract: A fault tree (FT) model can be converted to a Bayesian network (BN) by mapping relationships. This paper describes a risk probability computing method using a Bayesian net work model that is based on the group judgment of experts and the fuzzy theory, and presents an application of the method to analysis on the risk of earth rock-fill dam break associated with multi-state and fuzzy basic events. A case study showed that this new method can calculate the joint probability distribution of a dam break event and the posterior probability of those events on every level of the system, and thus obtain an importance sequence of the multi-state and fuzzy basic events using the values of posterior probabilities calculated. It expands the application scope of system risk calculation methods to more in-depth analysis, very helpful not just for risk assessment of earth rock-fill dam breaks but for analysis of failure events in other engineering fields. Thus, this new approach would be a useful guide for examination of complex multiple level, multi-factor, multi-state problems.