基于小波包分解与BP神经网络的制动系统电磁阀故障诊断研究

Research on Solenoid Valve of Brake System Fault Diagnosis Based on Wavelet Packet Decomposition and BP Neural Network

针对轨道车辆制动控制系统中电磁阀故障识别困难的问题,提出了一种基于小波包分解与BP神经网络的电磁阀故障诊断方法。该方法首先将原始信号进行小波包分解;然后对分解后的小波包系数进行特征提取;最后采用BP神经网络算法对不同故障提取出的特征进行诊断。为了验证该方法的有效性,基于Simulink进行了仿真验证,仿真结果表明:该方法通过了仿真验证,从理论上证明了该方法可行。在此基础上,进一步开展了实物验证,通过对比真实的正常电磁阀、阀芯卡滞电磁阀、弹簧失效电磁阀,实物验证结果从工程实现方面进一步说明了文中提出方法的有效性,能够实现制动系统电磁阀故障诊断。

Aiming at the difficulty in identifying solenoid valve faults in rail transit brake control system,a solenoid valve fault diagnosis method is proposed based on wavelet packet decomposition and BP neural network.This method first decomposes the original signal by wavelet packet decomposition;then,performs feature extraction on the decomposed wavelet packet coefficients;finally,uses the BP neural network algorithm to diagnose the features extracted from different faults.To approve this method,simulated validation is firstly implemented based on Simulink system.Results of simulated validation shows that this method has been approved theoretically.Based on result of simulated validation,we further carry out diagnostics validation on entities of solenoid valve.In this experiment,we completely compare differences among three kinds of solenoid valve,including normal solenoid valves,abnormal solenoid valves which valve-spools are stuck,and abnormal solenoid valves which springs lose efficacy.Results of validation on entities approve the effectiveness of this method from aspect of engineering.