Wang Hui
Mechanical bearings, fault diagnosis, convolutional neural network, gating unit, dual channel
Aiming at the problems of inefficiency, excessive cost, and excessive time-consumption that exist in the traditional manual inspection of mechanical bearing faults, this study designed a dual-channel feature fusion algorithm in combination with deep learning technology, and used the algorithm to build a mechanical bearing fault diagnosis model. Firstly, the role mechanism of convolutional neural network (CNN) and bidirectional gating loop unit is introduced, and then they are combined to design the upper and lower dual-channel feature detection algorithm. Then, on this basis, a bearing fault diagnosis model is built by combining the SENet feature fusion module, and its performance and application effects are tested. The results show that the final proposed fault diagnosis model has a good training effect, and when the iterations are 25, the model reaches the ideal stable state with a loss value of 0.48 and an accuracy of 0.91. In addition, the diagnostic accuracy of the proposed fault diagnostic model in the training dataset reaches up to 0.98, which is obviously more than that of the other comparative models. In conclusion, the bearing fault diagnosis model designed in this research not only shows better performance but also has certain application value, which can provide certain reference value for industrial equipment intelligence.
Important Links:
Go Back
