Aiming at the problem in expressing and dealing with uncertain information in reliability evaluation, DempsterShafer (DS) evidence theory is applied to modify Bayesian network. On the basis of the existing research, DS evidence theory and Bayesian network are introduced briefly, and conversion methods that convert fault trees to Bayesian network under uncertain conditions are proposed, including AND node, OR node, XOR node, NOT node and 2/3 VOTE node. Aiming at the difficulty in determining conditional probability values of multistate Bayesian network, a method based on DS evidence theory and analytic hierarchy process is proposed to analyze experts' experience. In the end, a missile control system is taken for example. Bayesian network model is established by referring to the fault tree, and DS evidence theory is used to determine the belief functions and plausibility functions of uncertain nodes and conditional probability values by fusing experts' information. Weak node in reliability design and distribution is pointed out by forward reasoning, backward reasoning and importance analysis of Bayesian network.