In the process of product reliability design, usually, the distribution of reliability index is not completed in one time. Different design stages have different reliability allocation methods. With the deepening of the design process, the results of the preliminary distribution may be unsatisfactory, so it's necessary to re-allocate the reliability index in middle and late stages. Prior to the index allocation, all the minimal cut sets are classified by a certain algorithm, and then the selected minimum cut sets to be allocated are distributed by using an artificial neural network (ANN). The ultimate goal is to allocate the reliability index to the minimal cut sets or elementary events of fault tree. By comparing the initial reliability before distribution and the reliability after distribution, we can find out the weak points in product design and then improve the design, which examines the results of FMECA (failure mode, effects and criticism analysis) to some extent.