Fault Diagnosis of Rolling Bearings Based on SVM and Improved D-S Evidence Theory for Multi-sensor Fusion

With the widespread application of rolling bearings in the industrial field, accurate diagnosis of rolling bearing faults is crucial for ensuring equipment reliability and improving production efficiency. Traditional fault diagnosis methods may face challenges in information fusion and uncertainty handling when dealing with multiple sensor information. A fault diagnosis method based on support vector machine (SVM) and improved D-S evidence theory is proposed, with rolling bearing fault diagnosis as the background, A corresponding fault diagnosis model is established to improve the accuracy and reliability of rolling bearing fault diagnosis. In the feature-level fusion, signal data obtained from different sensors are processed and feature extraction is performed. The obtained features are then combined to form a multidimensional feature vector. The fault features are used as inputs for SVM, and the output is used as the initial probability allocation value for the D-S evidence theory. In the decision-level fusion, an improved D-S evidence theory-based multi-sensor fusion algorithm is employed to enhance the fusion of initial probability allocation values. This algorithm comprehensively integrates and makes decisions on the diagnosis results from multiple sensors. Experimental results demonstrate that compared to single-sensor methods, the multi-sensor fusion method has significant advantages in rolling bearing fault diagnosis, it can effectively enhance the credibility of evidence, reduce uncertainty, and improve the accuracy of fault diagnosis and the robustness of fault diagnosis models.