关键词: 故障诊断, 聚类分析, 数据挖掘, 道岔转换系统

Abstract: According to the characteristics of the current data of ZDJ9 switch machine, an improved density peaks clustering (DPC) algorithm is proposed to identify the outlier data. Firstly, the characteristic values which can characterize the running status of ZDJ9 switch machine are extracted to reduce the data dimension, thereby the computational complexity and the computing time are reduced. Then, the characteristic data is normalized such that all characteristic indexes are in the same order of magnitude, thus the dimension influences among different characteristic indexes are removed. Finally, an improved DPC cluster analysis algorithm is proposed and applied to the action current data of the switch machine. The outlier data is identified successfully and the efficiency of the algorithm is verified.

Key words: fault diagnosis, cluster analysis, data mining, turnout switching system