Mechanical fault diagnosis based on the sinc convolution combined with the attention mechanism

Traditional fault diagnosis methods face challenges in handling non-stationary signals and high noise environment, and they are difficult to adapt to complex and changeable fault modes. This paper innovatively presents a one-dimensional fault diagnosis model combining Sinc convolution with attention mechanism. The model cleverly uses Sinc convolution to capture the frequency domain characteristics of the signal, and flexibly allocates the weight by the attention mechanism, so as to achieve a significant improvement in the accuracy and interpretability of fault diagnosis. Experimental, the model is excellent in diagnostic efficiency and noise resistance: the average accuracy is 99.76%, and the peak accuracy is 100%.