Fault Diagnosis Method for Heterogeneous Information Fusion of Permanent Magnet Generator Considering Classifier Performance and Weight of Evidence
Xun Yin, Xinyan Zhang, Shao-Ran Wang, Lulu Yang, Zhi-Wen Luo, Liwei Zhao · 2017
Aiming at the problem that the potential fault of the permanent magnet synchronous generator is difficult to be accurately identified, a potential fault diagnosis model based on probability output of multi-class support vector machine (SVM) and improved D-S evidence theory is proposed.Furthermore, the generator stator current and vibration characteristics are applied in the establishment of diagnostic model respectively and the failure probability based on the heterogeneous feature is obtained.Considering the difference of the fault characterization ability between the current evidence and the vibration evidence, as well as the generalization ability of SVM, the weight fusion model is established, and the output of the model is the final diagnosis criterion.