摘要
列车测速测距设备是列车运行控制系统的重要组成部分,也是故障率较高的设备之一.针对测速测距设备故障诊断自动化程度低的问题,提出一种基于LSTM(Long Short-Term Memory)神经网络的列车测速测距设备故障诊断方法,利用自适应最小支持度的加权Apriori算法从测速测距设备的时间序列中提取与故障分类关联度高的时间序列,构建故障数据集;利用LSTM神经网络对故障进行分类对比,测试结果表明:在以时间序列为故障特征的条件下,LSTM分类效果优于全卷积神经网络(Fully Convolutional Neural Networks,FCNN)与循环神经网络(Recurrent Neural Network,RNN).
Train speed and distance measuring equipment is an important part of train control system,and it is also one of the equipment with high failure rate.Aiming at the problem of low automation of fault diagnosis of speed measuring and ranging equipment,a fault diagnosis method of train speed measuring and ranging equipment based on LSTM neural network is proposed.The time series with high correlation with fault classification is extracted from the time series of speed measuring and ranging equipment by using the weighted Apriori algorithm of adaptive minimum support,and the fault data set is constructed.Then,LSTM neural network is used for fault diagnosis.Comparison results show that LSTM is better than FCNN full convolution neural network and RNN cyclic neural network under the condition of time series as fault feature.
作者
付文秀
李弘扬
靳东明
FU Wenxiu;LI Hongyang;JIN Dongming(School of Electronic and Information Engineering,Beijing Jiaotong University,Beijing 100044,China)
出处
《北京交通大学学报》
CAS
CSCD
北大核心
2020年第2期9-16,共8页
JOURNAL OF BEIJING JIAOTONG UNIVERSITY
基金
国家自然科学基金(61673049)。
关键词
车载测速测距设备
故障诊断
LSTM神经网络
APRIORI算法
on-board vehicle velocity measuring and ranging equipment
fault diagnosis
LSTM neural network
Apriori algorithm