Experimental results and analysis of gearbox bearing based on Sanc and 1D-CNN

In order to verify the advantages and effectiveness of intelligent fault diagnosis of gearbox bearing based on Sanc and 1d-cnn, it is compared with SVM, RNN and 1d-cnn without Sanc. In order to ensure the comparability of different methods, 1d-cnn and RNN use the same optimizer parameter configuration, cost function and activation function as the proposed method. SVM uses Gaussian kernel function, and penalty coefficient and other parameters use default values. Each method is tested for 10 times, and then the average value is taken. The results are shown in the figure.

It can be seen from the figure that:

① Compared with other models, the accuracy of the proposed method is 99%, which is nearly 5% higher than that of using 1d-cnn directly. The reason is that Sanc eliminates the influence of strong interference signals such as gears;

② The accuracy of 1d-cnn and RNN is 94% and 93% respectively. The difference between them is not very big. The former mainly extracts the signal features in space, while the latter extracts the signal features in time;

③ For complex classification problems, the performance of traditional machine learning method SVM is obviously worse than that of deep learning method, and its accuracy rate is 75%. The reason is that for complex classification problems, the traditional shallow feature machine learning algorithm has limited feature extraction ability and can not accurately represent the mapping relationship between data;

To sum up, the proposed method has more advantages and effectiveness than other methods.