Fault diagnosis of analog circuit based on wavelet transform and neural network

Analog circuits need more effective fault diagnosis methods. In this study, the fault diagnosis method of analog circuits was studied. The fault feature vectors were extracted by a wavelet transform and then classified by a generalized regression neural network (GRNN). In order to improve the classification performance, a wolf pack algorithm (WPA) was used to optimize the GRNN, and a WPA-GRNN diagnosis algorithm was obtained. Then a simulation experiment was carried out taking a Sallen–Key bandpass filter as an example. It was found from the experimental results that the WPA could achieve the preset accuracy in the eighth iteration and had a good optimization effect. In the comparison between the GRNN, genetic algorithm (GA)-GRNN and WPA-GRNN, the WPA-GRNN had the highest diagnostic accuracy, and moreover it had high accuracy in diagnosing a single fault than multiple faults, short training time, smaller error, and an average accuracy rate of 91%. The experimental results prove the effectiveness of the WPA-GRNN in fault diagnosis of analog circuits, which can make some contributions to the further development of the fault diagnosis of analog circuits.