摘要
多功能车辆总线(MVB)是列车的中枢神经,其可靠性对列车的安全稳定运行至关重要。在常见的MVB网络故障中,间歇性连接(IC)故障常在列车运行时的振动环境下间歇性出现,不仅难以诊断和定位,而且将导致网络丢包率和时延的增加,影响列车的运行安全。为此提出了一种MVB间歇性连接故障的定位方法。由于IC故障帧波形被打断后无法解码,因此通过训练稀疏自编码器对各网络设备的物理波形进行表征学习,并基于学习到的特征数据训练神经网络分类器,以识别IC故障帧的所属设备,进而建立IC故障代码、定位IC故障。实验结果表明,该方法可以有效提取MVB网络波形的内在特征,准确识别IC帧波形,有效定位IC故障。
The reliability of the Multifunction Vehicle Bus(MVB),as the central nerve of the train communication network,is of great importance to safe and stable operation of the train.Among common faults of MVB network,the intermittent connection(IC)fault often occurs intermittently in the vibration environment of train operation,making it difficult to diagnose and locate.IC fault will lead to an increase in packet loss and time delay of the network and affect the safe operation of the train.In this paper,an MVB IC fault location method was proposed.Since the broken IC frames can not be decoded,a sparse auto encoder was trained to learn the physical waveform of each network device.Based on these learned features,an artificial neural network classifier was trained to recognize the source devices of the IC fault frame accurately to establish the fault code and locate the IC fault point.The experimental results show that the proposed method can extract the internal features of MVB waveforms effectively,classify the fault frames accurately and locate the IC fault effectively.
作者
李召召
王立德
杨岳毅
申萍
LI Zhaozhao;WANG Lide;YANG Yueyi;SHEN Ping(School of Electrical Engineering, Beijing Jiaotong University, Beijing 100044,China)
出处
《铁道学报》
EI
CAS
CSCD
北大核心
2021年第1期107-114,共8页
Journal of the China Railway Society
基金
北京市自然科学基金(L171009)。
关键词
网络故障定位
特征学习
稀疏自编码器
神经网络
fault location of network
feature learning
sparse aotuencoder
neural network