Fault Diagnosis of Circuits Based on IHHO-MA1DCNN

Aiming at the problem of nonlinearity and tolerance in analog circuits, an analog circuit fault diagnosis method based on the improved Harris Hawk optimization algorithm and the multi-scale one-dimensional attentional convolutional neural network (IHHO-MA1DCNN) is proposed. Firstly, the MA1DCNN with three channels is built to extract the features of different receptive fields and enrich the feature information. Secondly, an attention mechanism is introduced to optimize feature weight allocation and enhance the expression of key features. Thirdly, the Harris Hawk optimization (HHO) algorithm is improved to optimize network parameters so as to build an optimal fault diagnosis model. This improvement results in an enhanced diagnostic capability for the MA1DCNN. Finally, two typical filter circuits, the Sallen-Key bandpass filter circuit and the double-T filter circuit, are chosen for fault diagnosis experiments. The results reveal that, in comparison to other models, the proposed method offers superior fault identification capability. The average fault diagnosis accuracy of the two circuits reaches 99.15%.