Bearing Fault Diagnosis with Two-Stage Multi-Source Information Fusion Based on Improved Stacking Algorithm and D-S Evidence Theory

Considering the limitations of fault diagnosis performance of single sensor and single fault feature, a two-stage multi-source sensor information fusion method for bearing fault diagnosis is proposed. First, considering sample imbalance characteristics and multi-feature association, an improved Stacking algorithm is proposed to construct a one-dimensional residual feature fusion network. The multi-sensor features are trained and strengthened to achieve fault feature fusion. Then, in the view of the uncertainty of multi-source sensor information, the improved evidence combination rule considering the amount of evidence information and evidence credibility is proposed to achieve the fault decision fusion based on Dempster-Shafer evidence theory. The application on rolling bearing case shows that the proposed method achieve all above 0.98 on average on evaluation indices such as accuracy, precision, recall and f1 score under different working conditions, which shows excellent performance and good generalization compared with the other 7 methods.