基于BP神经网络的模拟电路故障诊断

摘要

提出了一种基于小波多层分解和BP神经网络相结合的模拟电路故障诊断方法。该方法利用了多层小波分解优异的时频特性来提取故障特征参数，进行能量特征提取、归一化，并结合BP网络强大的非线性分类能力和快速的收敛特性构造了一种既能用于诊断单故障，又能诊断多故障的模型。以ITC’97标准电路中的CTSV滤波电路为诊断实例进行了仿真实验仿真，结果表明该方法比传统BP网络方法的学习收敛速度快得多。%A analogous circuit fault diagnosis method based on multilevel wavelet decomposition and BP neural network is proposed in this paper. The method makes full use of the time-frequency characteristic of multilevel wavelet decomposition to ex-tract the fault feature parameters for the energy feature extraction and normalization,and to structure a diagnosis model for diag-nosing not only single fault but also multiple faults in combination with strong ability of nonlinear classification and fast conver-gence of BP neural network. Simulation experiments on a CTSV(continuous-time state-variable)filtering circuit in ITC’97 stan-dard circuit were carried out.The simulation results show that the BP wavelet neural network has faster convergence speed than conventional BP network.