In order to solve the problem of complexity and low-precision in spare parts requirement prediction in maintenance, un-repairable parts obeying normal distribution are taken as objects to analyze the method of predicting spare parts requirement. From the perspective of guarantee probability, the indexes related to spare parts are introduced and the guarantee probability model is established for units obeying normal distribution. Then, the number of spare parts required to meet the guaranteed probability is determined through inverse calculation and the spare parts obeying normal distribution can be determined. Through the case, the application conditions of the accurate prediction method and the approximate prediction method are obtained. The prediction results show that the approximate prediction method is suitable for determining the requirement of spare parts in a long time.