2nd China International Youth Conference on Electrical Engineering (CIYCEE 2021)

     This paper presents a new hierarchical fault diagnosis method of analog circuit based on back propagation (BP) neural network group. This method realizes soft fault and hard fault diagnosis of analog circuit. To improve the accuracy of automatic diagnosis of this method, a multi-feature parameter fusion preprocessing scheme is presented. The Fast Fourier Transform (FFT) is used to calculate the DC component, fundamental amplitude, fundamental phase angle, amplitude of second harmonic, phase angle of second harmonic and distortion of the time domain output signal，these feature data are used as the input of BP neural network group. The fault diagnosis performance of this method is verified by the simulation and test of the triode single-stage amplifier circuit. The results show that this method performs well on fault diagnose. The accuracy of simulation diagnosis is up to 95%, and the accuracy of test is up to 90%.