基于PLSA和SVM的道岔故障特征提取与诊断方法研究

Research on Fault Feature Extraction and Diagnosis of Railway Switches Based on PLSA and SVM

铁路局和电务段长期以来保留的道岔故障记录是非常宝贵的数据,对道岔故障类型统计、故障特征分析、故障诊断和故障预测有非常好的参考作用,但这些数据往往保存格式多样,难以直接利用。本文提出基于主题模型PLSA和支持向量机SVM的道岔设备故障特征提取与诊断方法。通过分词算法将故障文档表达在词项特征空间中;采用主题模型算法提取主题特征,并将故障文档表达在主题特征空间上;以SVM算法构造诊断器实现道岔设备的故障诊断。利用中国铁路广州局集团有限公司道岔故障记录的真实数据,对提出的算法有效性进行验证。实验表明,提出的算法能有效实现道岔设备故障诊断,对现场维护有一定的指导意义。

Railway switch fault records kept by railway bureaus and communication and signal districts are very valuable data on switch fault type statistics,fault feature analysis,fault diagnosis and fault prediction.However,these data are often stored in various formats and are difficult to use directly.This paper presented a fault feature extraction and diagnosis method for railway switches based on topic model PLSA and support vector machine（SVM）.First of all,the word segmentation algorithm was used to express the fault documentin the feature space of the word.Next,the topic model algorithm was used to extract the theme features,and to express the fault document in the theme feature space.Finally,SVM algorithm was used to construct diagnostor to realize the fault diagnosis of switch equipment.The real data of switch fault records of China Railway Guangzhou Group Co.,Ltd.were used to verify the validity of the proposed algorithm.The experimental results show that the proposed algorithm can effectively realize the fault diagnosis of railway switches,and provides certain guiding significance for field maintenance.