基于改进的DPC算法的ZDJ9转辙机异常数据挖掘  

Outlier Data Mining of ZDJ9 Switch Machine Based on Improved DPC Algorithm

作　　者：李智宇 Li Zhiyu(CRSC Research&Design Institute Group Co.,Ltd.,Beijing 100070,China)

机构地区：[1]北京全路通信信号研究设计院集团有限公司,北京100070

出　　处：《铁路通信信号工程技术》2022年第5期1-6,共6页Railway Signalling & Communication Engineering

基　　金：中国国家铁路集团2021年科研计划项目(P2021G014)。

摘　　要：针对ZDJ9转辙机动作电流数据的特点,提出一种基于改进的密度峰值聚类(Density Peaks Clustering,DPC)算法的数据挖掘方法来识别异常数据。通过提取能有效表征转辙机运行状态的特征值来降低数据维度,从而降低运算复杂度,减少计算时间;对特征数据进行归一化处理,使得各特征指标处于同一数量级,消除不同特征指标之间量纲的影响;提出一种改进的DPC算法,并将其应用于ZDJ9转辙机动作电流数据,成功识别出异常数据,验证该算法的有效性。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 infiuences 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.

关 键 词：故障诊断 聚类分析 数据挖掘 道岔转换系统 

分 类 号：U213.6[交通运输工程—道路与铁道工程]