基于WNN算法的BTM故障诊断方法  

BTM Fault Diagnosis Method Based on WNN Algorithm

作　　者：程剑锋 王心仪 夏凯 CHENG Jianfeng;WANG Xinyi;XIA Kai(Signal&Communication Research Institute,China Academy of Railway Sciences Corporation Limited,Beijing 100081,China;Postgraduate Department,China Academy of Railway Sciences,Beijing 100081,China)

机构地区：[1]中国铁道科学研究院集团有限公司通信信号研究所,北京100081 [2]中国铁道科学研究院,北京100081

出　　处：《中国铁路》2023年第10期83-90,共8页China Railway

基　　金：国家自然科学联合基金项目(U1934222)。

摘　　要：高效、准确的故障定位技术是列车安全运行的重要保证。针对列车超速防护系统(ATP)车载设备故障分析存在复杂性高、依赖专家经验等问题,提出将小波神经网络(Wavelet Neural Network,WNN)算法应用于车载设备故障诊断的方法。针对车载设备中的应答器传输模块(Balise Transmission Module,BTM),首先根据经常发生的故障类型,匹配ATP中相应的故障日志语句;然后建立网络结构,利用小波理论修正网络的权值与参数;最后结合WNN算法精准地分析和预测故障。选取BTM单元的100组故障数据作为样本进行仿真实验,并与BP神经网络、GA-BP神经网络以及SVM算法进行对比。实验结果表明:通过小波算法优化神经网络的测试样本平均绝对误差降低至6.917%,相关系数提高到97.402%,该算法在高速铁路列控车载设备故障分析方面有较高的准确性。Efficient and accurate fault location technology is an important guarantee for safe operation of trains.Aiming at the problems of high complexity and reliance on expert experience in fault analysis of vehicle equipment of automatic train protection(ATP),this paper proposes a method to apply Wavelet Neural Network(WNN)algorithm to fault diagnosis of vehicle equipment.For the Balise Transmission Module(BTM)in the vehicle equipment,firstly match the corresponding fault log statements in ATP according to the frequent fault types;then establish a network structure and correct the weights and parameters of the network by using wavelet theory;finally accurately analyze and predict faults in combination with WNN algorithm.100 groups of fault data of BTM unit are selected as samples for simulation experiments,and compared with BP neural network,GA-BP neural network and SVM algorithm.The experimental results show that the average absolute error of test samples of neural network optimized by wavelet algorithm is reduced to 6.917%,and the correlation coefficient is increased to 97.402%.This algorithm has high accuracy in fault analysis of vehicle equipment for HSR train control.

关 键 词：高速铁路 车载设备 列车超速防护系统(ATP) 应答器传输模块 故障诊断 WNN算法 

分 类 号：U284.92[交通运输工程—交通信息工程及控制]