Due to the difficulty of reliability analysis of multistate systems, a new method based on Bayesian networks is proposed through an example. Reliability block diagram and logic operators are firstly established according to the hierarchy of structure and function of multistate systems, and the Bayesian networks is constructed based on the reliability block diagram, distribution of components and logic operators, then the analysis is performed. The paper shows that the new method based on Bayesian networks could model generic multistate systems and basic inference techniques on Bayesian networks may be used to obtain classical parameters in reliability analysis (i.e. probability of the top event or of any subsystem, etc.). Moreover, by using Bayesian networks, some additional information can be obtained at the analysis level which could be used to perform diagnosis and complex common cause problems can be handled conveniently. The comparison with traditional method is carried out by means of an example of radar system taken from literature.