Aiming at the problem that the traditional convolutional neural network has low accuracy and poor performance in feature extraction in gearbox fault diagnosis, a SincNet combined with attention mechanism method for gearbox fault diagnosis was proposed. First, use the parameterized Sinc function to design the filter and obtain the Sinc convolutional layer, Sinc convolutional layer replace the first convolutional layer of traditional CNN to construct the SincNet network structure, Extract the characteristic information of the input data. Then, combined attention Mechanism with Softmax enhances characteristic information. Finally, the gearbox fault data set was used to verify the proposed method. The results show that the average diagnostic accuracy of the proposed method is 99.68%, which is higher than that of the comparison method. In addition, the method can accurately locate the recognition information in the input data and better understand the feature extraction process of neural network through the visual analysis of the feature map, which provides a reference for the feature extraction process of mechanical vibration signals.