Nonlinear circuit fault diagnosis based on simplified estimation of generalized frequency response function

Abstract

For the fault diagnosis of nonlinear analog circuits, a new method based on simplified estimation of generalized frequency response function and least squares support vector machine with hybrid kernel is proposed. The generalized frequency response functions of the nonlinear circuit are directly estimated through the time domain input and output, and the input signal is sinusoidal signal. Therefore, the Fourier transform of time-domain signal is avoided and the calculation amount for estimation is reduced. Using the amplitudes of generalized frequency response functions of analog circuit as feature data, a least square support vector machine classifier with hybrid kernel is constructed to diagnose the faults. A fault diagnosis simulation for a nonlinear analog circuit is carried out. The results show that the simplified method can accurately estimate the generalized frequency response function of the nonlinear circuit, and the proposed diagnosis method has a high accuracy.