WNN Tolerance Fault Diagnosis for Analog Circuits Based on Wavelet Packet Transform Features

Abstract

For the sample construction of Wavelet Neural Network (WNN) fault diagnosis modeling for analog circuits, a feature extraction method based on Wavelet Packet Transform (WPT) is proposed. The simulation results on Sallen-Key band-pass filter show that, compared with Multi-Resolution Transform, WPT has better feature extraction ability for the tolerance faults of analog circuits, and can improve the convergence rate and diagnosis correct rate of WNN.