基于方法组合的轨道电路故障诊断模型

为了解决单一方法对25 Hz相敏轨道电路的故障诊断精度偏低等问题,提出基于最优权值的多方法组合故障诊断模型。首先通过模糊综合评判、灰色关联分析和BP(Back propagation)神经网络3种不同的诊断方法,对轨道电路进行初步故障诊断,然后根据各诊断方法的误判率计算出对应的最优权值,最后对各方法的诊断输出进行最优权值加权平均得到综合诊断的输出结果,确定故障类型。诊断结果表明:组合诊断模型有效地提高了轨道电路故障诊断的准确度,并证明了组合诊断模型的诊断准确度随着诊断方法的数量增加而提高。

Fault Diagnosis Model of Track Circuit Based on Method Combination

In order to solve the problem of low accuracy in fault diagnosis of 25 Hz phase-sensitive track circuits, a multi method combination diagnosis model based on optimal weight is proposed. First of all, three kinds of different diagnostic methods, including fuzzy comprehensive evaluation, grey relational analysis and BP neural network, are used for preliminary fault diagnosis of track circuits. Then the corresponding optimal weight is calculated according to the misjudgment rate of each diagnosis method. Finally, the results of the comprehensive diagnosis are obtained by optimal weighted average of the diagnostic output of each method so as to determine the type of fault. The diagnosis results show that the combined diagnosis model can effectively improve the accuracy of 25 Hz phase-sensitive track circuit fault diagnosis, and prove that the diagnostic accuracy of the combined diagnosis model improves with the increase of the number of diagnostic methods.