To improve the scientificalness of equipment maintenance decision-making, an optimal replacement strategy of non-repairable equipment was realized by establishing the optimal maintenance decision model of integrated equipment RUL(remaining useful lifetime) prediction data and uncertain failure threshold. The nonlinear Wiener process was used to establish the degradation model of the equipment and the maximum likelihood estimation was used to determine the parameter estimation of the degradation model. A new estimation method for the distribution coefficient of uncertainty failure threshold based on the EM(expectation maximization) algorithm was proposed. By introducing the virtual failure threshold data, the synchronous iterative update of the distribution coefficient for failure threshold was realized. Based on the concept of the first hitting time, the RUL probability density function of equipment with considering the uncertain failure threshold was derived. The decision model was built on the basis of the renewal-reward theory to determine the optimal replacement occasion. The example analysis shows that the failure threshold of the equipment has an important impact on the result of the maintenance decision. This conclusion fully takes into consideration of the fact that the uncertainty of the equipment failure threshold can not only improve the accuracy of the RUL prediction but also can effectively reduce the life cycle cost of the equipment.