A Support Vector Machine Based on Neural Dynamics Approach for Bearing Fault Classification

In modern industry, the complexity of monitoring equipment status has made it increasingly difficult to track the operation status of machines using human labor. Therefore, there is an urgent need to find a way to perform real-time intelligent equipment diagnosis. This paper aims to combine fault diagnosis with a neuro-dynamic approach to solve the fault diagnosis problem. The fitting of SVM parameters is converted into a QP problem, and a neural network solver is used to solve the QP problem. In this paper, we conduct three experiments to prove the convergence and classification performance of the model in the field of fault diagnosis. The first experiment shows that SVM can get optimal solutions quickly and has good convergence performance. The second experiment is based on the CWRU bearing dataset and verifies that our method can distinguish the health conditions and different faults. The current mainstream machine learning and deep learning methods are compared in the third experiment, and our method gets higher accuracy at 98.07% than other methods.