ANALOG CIRCUIT FAULT DIAGNOSIS METHODS BASED ON RBF NEURAL NETWORK

摘要

There are many difficulties in diagnosis of analog circuit faults, and effective methods are needed to make breakthroughs. In this study, fault feature vectors are obtained by the Haar wavelet decomposition, and the radial basis function (RBF) neural network was improved by particle swarm optimization (PSO) algorithm. The signals generated when the circuit was in normal state and when eight kinds of faults occurred were collected by the Butterworth low-pass filter circuit. In total 450 sets of data were obtained, including 350 sets of training samples and 100 sets of experimental samples. It was found that the training time of the PSO-RBF was shorter, the error was smaller, and the accuracy rate was 96%, indicating the high reliability and broad application prospects of the algorithm.