基于双视图故障特征提取的列控车载设备故障诊断方法

Fault Diagnosis Method for On-board Equipment of CTCS Based on Dual-view Fault Feature Extraction

从列车运行日志数据中充分准确地提取车载设备的故障特征,对于提高故障诊断效率、保障列车运行安全具有重要作用。针对故障特征提取中面临的故障模式分布不平衡、故障特征提取不充分、故障特征维度高等问题,提出基于双视图故障特征提取的列控车载设备故障诊断方法。首先在语义视图下增加类别比重因子改进互信息,调整故障特征空间,解决故障模式分布不平衡问题;然后在语序视图下考虑语序和句序的影响,使用句向量的分布记忆模型实现故障特征的充分提取,解决故障特征提取不充分问题;最后利用PCA方法对级联后的特征集合进行融合,解决故障特征维度高问题。使用铁路运营部门收集的300T型车载设备运行日志数据对本方法进行实验验证,实验结果表明:相比于两种传统的故障特征提取方法,本方法的F 1值分别增加了0.063、0.037,证明了本方法的有效性。

Extracting fault features fully and accurately from the operation log data of train control on-board equipment plays an important role in improving the efficiency of fault diagnosis and ensuring the safety of train operation.Aiming at the problems of unbalanced distribution of fault modes,insufficient extraction of fault features and high dimension of fault features in fault feature extraction,this paper proposed a fault diagnosis method for on-board equipment of CTCS based on dual-view fault feature extraction.Firstly,in the semantic view,the category proportion factor was added to improve mutual information,and the fault feature space was adjusted to solve the problem of unbalanced distribution of fault modes.Secondly,in the word order view,considering the influence of word order and sentence order,the distributed memory model of sentence vector was used to fully extract fault features to solve the problem of insufficient extraction of fault features.Finally,the PCA method was used to fuse the cascaded feature sets to solve the problem of high dimension of fault feature.By using the operation log data of 300T on-board equipment collected by the railway operation department,the experimental verification of this method was carried out.The results show that compared with the two traditional fault feature extraction methods,the evaluation index F 1-score of the proposed method is improved by 0.063 and 0.037 respectively,which proves the effectiveness of the method.