基于电流源激励的BP神经网络模拟电路故障诊断

针对传统故障字典法对模拟电路故障诊断时存在的缺陷提出了新的故障字典法;将电流源激励下二端口网络输入端和输出端的电压增益比作为故障特征信息,在此基础上先直流测试,后利用BP神经网络交流测试;该方法充分考虑了电路元件的容差,减轻了BP神经网络诊断故障的负担,提高了故障诊断的速度、准确率以及故障覆盖率;利用MATLAB和PSPICE工具对该方法进行实例仿真,结果表明其能够实现快速、准确的故障定位。

New Method for Fault Diagnosis of Analog Circuit under Current Supply Based on BP Neural Network

A new kind of fault dictionary method for fault diagnosis of analog circuit is proposed to improve the traditional method,using the ratio of input-port voltage and output-port voltage under current power supply as fault character information in order to detecting faulty component more easily and accurately.The presented method contains direct current test and alternating current test with BP neural network,which can take the presence of component tolerance into account,decrease workload when detecting faults with BP neural network and advance fault diagnosis performance in terms of accuracy,speed and fault coverage rate.The simulated results obtained by MATLAB and PSPICE software illustrate that above method can identify faulty component quickly and precisely and diagnose whether the actual parameters are within tolerance ranges or corresponding components are faulty as well.