基于数据的轨道电路故障诊断的混合算法

提出一种基于数据的神经网络混合算法故障诊断网络,用于轨道电路的故障诊断.考虑铁路信号需求,设计出符合神经网络训练快速性和有效性要求的BP-LM-PSO-GA混合算法,就是将轨道电路复杂网络分解设计为许多小的神经网络组态,通过综合这些小的神经网络诊断结论,得出最终结果,以解决单独设计神经网络带来的运算量问题;然后以广泛使用的ZPW-2000A型轨道电路为例,验证了该算法网络训练的快速性及故障诊断的有效性.最后给出了该诊断网络对轨道电路的诊断步骤.仿真结果表明该诊断网络具有可行性和有效性,为轨道电路故障诊断的应用提出了一条新途径.

Study of data-based hybrid algorithm on track circuit fault diagnosis

A neural fault diagnosis network based on data for railway track circuit is presented in this paper.Firstly,BP-LM-PSO-GA hybrid algorithm which meets the requirements of speed and effectiveness in neural network training is designed considering railway application.Secondly,taking dominant ZPW-2000A-type track circuit for example,the rapidity and effectiveness of fault diagnosis are analyzed for fault diagnosis.Thirdly,track circuit is divided into many small neural networks so that the conclusion can be drawn by combining the results of these small neural networks.Finally,the steps of diagnosis for track circuit network are described.Simulation results show that this diagnostic network for track circuit is feasible and effective.This kind of new approach in fault diagnosis for track circuit reveals a promising solution.