小波奇异熵改进SOM神经网络在道岔故障诊断中的应用  

Application of SOM neural network optimized by wavelet singular entropy algorithm on turnout fault diagnosis

作　　者：秦小虎 Qin Xiaohu

机构地区：[1]重庆市轨道交通(集团)有限公司,重庆401120

出　　处：《现代城市轨道交通》2022年第2期23-27,共5页Modern Urban Transit

摘　　要：随着轨道交通技术的快速发展和智能算法的应用,用微机监测和人工处理的方法对道岔进行故障诊断存在效率低、及时性不够、准确性不足等问题,文章用 Meyer 小波分解原始数据,实现特征数据的选择和提取,再计算出相应的小波奇异熵作为神经网络的输入向量,加入训练数据到改进型的 SOM 神经网络中,从而实现对S700K 型转辙机的道岔故障诊断。通过提取 8 种典型的故障类型,在 MATLAB 中建立相应的模型并进行仿真,和目前普遍应用的 K-means 和 FCM 聚类算法进行时间和准确率方面的比较。仿真结果表明,在道岔故障诊断中,基于小波奇异熵和 SOM 神经网络的算法在时间和准确率方面具有明显优势。With the high-speed development of rail transit technology and application of intelligent algorithms,the way of diagnosing fault of turnout has such issues as low efficiency,insufficient timeliness and accuracy by using microcomputer monitoring date and manual analysis.This paper breaks down the original data by Meyer wavelets,which selects and extracts feature data.Then the corresponding wavelet singular entropy is calculated as an input vector of an SOM neural network,to which training data are added to the improved neural network,so as to realize the fault diagnosis of the turnout equipped with S700K switch machine.By extracting eight typical types of faults and establishing the model,this paper diagnoses such faults by simulation of MATLAB and it clearly finds that the algorithm based on wavelet singular entropy and SOM neural network saves time and improve accuracy when compared with K-means,FCM and other clustering algorithms.

关 键 词：轨道交通 道岔 故障诊断 特征提取 小波奇异熵 SOM 

分 类 号：U216.425[交通运输工程—道路与铁道工程] U284.92