Axle fault prognostics of electric multiple units based on improved Apriori algorithm

Abstract

To effectively reduce the occurrence rate of axle faults of electric multiple units (EMUs), in this study, classical Apriori algorithm is improved based on Apache Hadoop big data and applied to prediction studies of axle faults of EMUs. First, for deficiencies of the classical Apriori algorithm, the improved Apriori algorithm that is constrained by business experience is proposed under the MapReduce framework. Second, based on the improved Apriori algorithm put forward in this paper, in-depth data mining of information of one of the railway bureaus, such as EMU status, fault warning, and maintenance history, is conducted. The axle fault of the EMUs is then predicted through the obtained association rules. Finally, experimental results indicate that the proposed algorithm reaches an accuracy of 67% and its operation efficiency in experimental environments is improved by 50% when compared with the operation efficiency of the classical Apriori algorithm.