Research On Key Technology Of EMU Maintenance Cost Based On Big Data

In recent years, with the rapid development of high-speed railway, the majority of EMU entered a period of repair and maintenance, and how to reduce the maintenance costs of EMU increasingly become the focus of research. In this paper, the maintenance cost data and operational data of EMU are mined to find out the deep associations between the maintenance costs and the operation status of EMU key components, which can provide a reference for formulating the maintenance strategies and reducing maintenance costs.With the continuous accumulation of the EMU maintenance cost data and operational data, the traditional single machine algorithm can’t do work well for the massive data. To solve the problem, this paper introduces the Hadoop computer cluster framework, adopting a distributed approach to deal with the huge maintenance and operation data of EMU. Traditional association rule mining algorithms can only support a single support degree. However, each attribute in the database is of different importance, and the single support can’t meet the requirements of the specific application. Therefore, this paper adopts the AMWARMS algorithm, which is a weighted association rules based on multiple supports. In this paper, we further improve the multi support of the algorithm, and optimize the operation efficiency of the algorithm, so that the algorithm can better adapt to the demand of the mass data of the EMU. Innovations in this paper are listed as follows:(1) The maintenance cost of the EMU is studied deeply. Through the analysis and investigation of the maintenance process of the key components of the EMU, the maintenance cost of the key parts is sorted out.(2) According to the actual needs of the EMU data, we improved the AMWARMS algorithm. Firstly, the theory of concept lattice is used to improve the multi support problem of the algorithm. Then the Tidset vertical data format is used to optimize the operation efficiency of the algorithm.(3) Experimental results show that the improved algorithm can significantly improve the operating efficiency of the original algorithm under different support, different number of nodes and different amount of data. The association rules between operation and maintenance data of the EMU was excavated by the improved MRAMWARMST algorithm, and the results were visualized.