The invention discloses a mechanical system time varying reliability evaluating method based on a dynamic Bayesian network. The mechanical system time varying reliability evaluating method comprises a first stage of determining model basic indexes, a second stage of structuring the structure of the Bayesian network and a third stage of updating a formula and the time varying reliability of a Monte Carlo simulation computer mechanical system according to Bayesian information. The mechanical system time varying reliability evaluating method has the advantages that a knowledge diagrammatic expression method is provided through the Bayesian network, directed diagrammatic expression can be carried out on the cause and effect probability relation between node variables, and the cause and effect probability relation can be used for uncertain knowledge expression, cause and effect reasoning, diagnosis reasoning and the like. The weak link of the reliability of the system can be effectively recognized through reasoning of the Bayesian network; the relation between components in the mechanical system becomes more visual and clear through diagrammatic display, the dynamic Bayesian network technology is applied to evaluation of the time varying reliability of the mechanical system, the multiple states and failure correlation of the mechanical system are analyzed, and a theoretical support is provided for improving the performance and the reliability of the mechanical system.