Considering that there are often many variables correlated to each other in complex system, copula functions are introduced into reliability analysis. While with the increase of variables, the dimension of copula function will be high. For the above problems, Bayesian network (BN) is introduced into reliability analysis and a copula Bayesian network (CBN) model is proposed. Taking driving system of new rail mounted container gantry (NRMG) as an example, Gaussian copula and t-copula are used and compared in the process of generating correlation data. Additionally, in order to construct local copula function between parent and descendent nodes, Gumbel and Clayton copulas are compared based on Akaike information criterion. Finally, the reliability inference based on CBN model is performed. Compared with BN model, the rationality of CBN model is verified, and the results show that the reliability calculated by CBN model is smaller, which will put forward higher requirements for the driving system.