Analog Circuit Fault Diagnosis Based on FFT-CNN-LSTM

Abstract

To improve the performance of analog circuit fault diagnosis, an ensemble fault diagnosis method combining fast Fourier transform (FFT), convolutional neural network (CNN) and long and short-term memory (LSTM) is proposed. First, FFT is used to convert data to the frequency domain. Then special zone features are obtained by CNN network. Finally LSTM network is used to complete the fault diagnosis of the analog circuit. Experiment on CSTV analog circuit shows that FFT-CNN-LSTM can be used to improve the quality of analog circuit fault diagnosis.