The feed forward neural network is used to predict the bearing life distribution in this paper,and puts forward the maintenance decision model based on the maintenance window. The mean and variance of the bearing life distribution are predicted by using neural network,and the life distribution of the bearing is obtained. A maintenance decision model based on the maintenance window is proposed and compared with the threshold value maintenance model. The relationship between different maintenance window position and optimal maintenance point is analyzed. The model is verified by using the data of the whole life of bearing,it is concluded that the maintenance decision model based on the maintenance window is better than the threshold limit maintenance window is better than the threshold limit maintenance model in a certain time interval by calculate the maintenance cost rate of the two models.