Abstract: A two-stage classification method is proposed for engineering machinery spare parts aiming at the characteristics of randomness,diversity and complexity.In the first stage,the spare parts are divided into two categories according to the stability of the service spare demand time series.In the second stage,combined with the factors such as the value,service,time and other factors of the spare parts classification,the Rough Set (RS) theory and Self-Organizing Map (SOM) neural network are combined to design the RS-SOM clustering model.The index data are discretized by using fuzzy c-means clustering algorithm.Then the improved matrix algorithm is used to reduce the index set.In the kernel based SOM model,the training process is improved by introducing rough set theory.Finally,the clustering results of engineering machinery spare parts are obtained.Data experiments show that compared with the method of the ABC classification method and the traditional SOM clustering method,the classification result is better and it can provide a more reliable basis for the selection of the spare parts forecasting method and inventory strategy.

Key words: engineering machinery spare parts, two-stage classification method, requirement time series, rough set, self organizing map neural network