For those large-scale and complex mechanical products, considering economy and applicability, these products are impossible to be discarded as soon as they take place malfunction. Generally, they can be reused again and again after be repaired repeatedly, that is, they usually have multi-lifespan. Accordingly, it is very necessary to establish a kind of proper reliability evaluation mode for these complex repairable products. In this paper, the treatment procedure of the reliability test data for these repairable products is researched based on the repairable reliability theory, and the method of model identification about failure distributions law is proposed. In addition, how to reasonably determine or predict product lives by using random process model is also introduced. Finally, an analysis case for putting this mean into practice is expounded. This method for product reliability analysis is suitable to most large-scale and complex mechanical products.