Highly accurate interpretable bearing fault diagnosis based on SHAP-RFE with Bayesian optimization support vector machines

Abstract Early diagnosis of bearing faults is the key to ensuring the stable operation of rotating equipment. To address the problem of low diagnostic accuracy and interpretability of the rolling bearing model, this paper proposes a highly interpretable bearing fault diagnosis method based on SHAP-RFE and an improved support vector machine. The core is based on SHapley Additive ExPlanations (SHAP) and Recursive Feature Elimination (RFE), which proposed the SHAP-RFE feature selection method. The method can not only simplify the feature matrix and thus simplify the model but also improve the accuracy and generalization; at the same time, this paper uses the Bayesian algorithm to optimize the hyper-parameters of the support vector machine to improve the classification accuracy and the use of SHAP to explain the Support Vector Machine (SVM). The experimental validation of the bearing vibration dataset from Case Western Reserve University and Xi’an Jiaotong University shows that the average accuracy of the bearing diagnostic model is over 98.9%, with good stability and generalization as well as interpretability, and it can effectively extract the best fault features and accurately diagnose and classify them.