An analog circuit fault diagnosis method based on two dimensional convolutional neural network

Analog circuit fault diagnosis based on traditional machine learning method relies on complex signal processing technology and professional knowledge for fault feature extraction, and the process of fault diagnosis is complex. In view of this, an analog circuit fault diagnosis method based on two dimensional convolutional neural network (2D-CNN) was proposed. The original output voltage of the circuit under test was converted into fault gray image (fault gray image, FGI) as the input of 2D-CNN model. The convolution layer of the model was used to automatically extract the deep features of the fault, and the batch normalization (Batch Normalization, BN) layer was added to the model to regularize the data distribution, so as to alleviate the impact of data distribution offset. In the fault diagnosis experiments of Sallen-key bandpass filter circuit and Four-opamp biquad high-pass filter circuit, the fault diagnosis rates achieve by the proposed method are 100% and 99.46% respectively. The proposed method not only simplifies the fault diagnosis process, but also ensures the accuracy of fault diagnosis, and has strong generalization ability.