基于模糊认知图模型的轨道电路故障诊断

Track circuit fault diagnosis based on fuzzy cognitive map model

针对目前人工分析监测数据进行ZPW-2000A轨道电路故障判别存在判别效率低、判别周期较长和对数据分析所依赖的人员经验程度高的缺陷,将模糊认知图的概念引入到ZPW-2000A轨道电路的故障判别中。通过分析10类故障产生的机理,结合现场经验,对不同类型的故障数据进行分析,得到故障特征参数,构建出基于特征参数与故障类别的模糊认知图分类器。在此过程中利用自适应遗传算法完成FCM权重的学习。研究结果表明:根据监测数据用ZPW-2000A轨道电路故障判别的FCMCM判别方法是有效可行的,较人工分析监测数据判别故障的方式具有较高的故障识别率和较短的判别时间。

In view of the disadvantages of low diagnostic efficiency,long cycle time of diagnosis and highdependence on the experiences of data analyzers of ZPW-2000A track circuit fault distinguishes with manualwork according to monitoring data,the concept of Fuzzy Cognitive Map was introduced into fault diagnosis forZPW-2000A track circuit.Based on the mechanism analysis of ten kinds of faults with the combination of thepresent day experiences,and the analysis of different types of faults data to obtain faults feature parameters,theFuzzy Cognitive Map classifier based on fault types and feature parameters were establish,and to complete FCMweights learning in terms of the adaptive genetic algorithm during the process.Computer simulations showed thatthe monitoring based FCMCM diagnosis method proposed in this paper for ZPW-2000A track circuit is effectiveand feasible,and was compared with manual analysis monitoring data to discriminate failure and was found to bea high failure recognizer within a short diagnosis time.