Physical process failure plays a crucial role in the passive systems.Presently,the importance sampling Monte Carlo method is most widely used in the research of physical process failure probability,but it has to rely on the improved first-order second-moment(FOSM)method in searching of design point,and can not avoid the shortcomings of FOSM.The adaptive Monte Carlo method introduced in this paper chooses the important probabilistic distribution function by adaptive method,avoiding the shortcomings of FOSM,improving the sampling efficiency and the accuracy.Taking the passive system of nuclear desalination reactor(NDR) for an example,this paper evaluates the physical process failure probability by the adaptive Monte Carlo method and the presently used methods,and then compares and analyzes the adaptiveness of various methods on the passive systems.