摘要
ZPW-2000A型轨道电路是我国铁路信号系统中广泛使用的设备,针对ZPW-2000A型轨道电路发送器、接收器故障排查程序复杂、效率低的问题,提出一种基于声谱分析的故障诊断方法,实现非接触故障诊断。首先,通过梅尔频率倒谱系数和小波包分析对采集的轨道电路发送器、接收器故障时的声音信号进行特征提取,获得多维特征矩阵;然后,利用支持向量机和随机森林作为分类器,将故障诊断转化为多分类问题,实现发送器与接收器的故障分类。研究结果表明,以支持向量机作为分类器的平均准确率为89.4%,随机森林作为分类器的平均准确率为95.4%,可以实现故障的准确识别。
ZPW-2000A track circuit is widely used in China’s railway signal system.In order to solve the problems of complex troubleshooting procedures and low efficiency of ZPW-2000A track circuit transmitter and receiver,a fault diagnosis method based on sound spectral analysis is proposed to realize non-contact fault diagnosis.Firstly,through Mel-Frequency Cepstral Coefficients and wavelet packet analysis are used to extracted the characteristics of the collected track circuit transmitter and receiver fault sound signals to obtain a multi-dimensional feature matrix.Then,support vector machine and random forest are adopted as classifiers,transforming the fault diagnosis into a multi-classification problem to realize the fault classification of transmitter and receiver.The results show that the average accuracy of support vector machine as classifier is 89.4%,and the average accuracy of random forest as classifier is 95.4%,which could realize the accuracy recognition.
作者
庞茂盛
陈大山
邹劲柏
谢鲲
陈文
黄宇轩
PANG Mao-sheng;CHEN Da-shan;ZOU Jin-bai;XIE Kun;CHEN Wen;HUANG Yu-xuan(Faculty of Railway Rransportation,Shanghai Institute of Technology,Shanghai 201418,China)
出处
《信息技术》
2023年第2期41-45,51,共6页
Information Technology
基金
上海市科委地方院校能力建设项目(20090503100)
上海应用技术大学引进人才项目(10120K216055-A06)
上海应用技术大学中青年教师科技人才发展基金项目(10120K209043-A06)。
关键词
轨道电路
特征提取
故障诊断
支持向量机
随机森林
track circuit
feature extraction
fault diagnosis
support vector machine
random forest