Fault Diagnosis Method of Hydropower Units Based on Integrated Information Fusion Technology

A diagnosis method based on integrated information fusion which combining neural network and Dempster-Shafer evidential theory is presented in this paper. In the method, vibration-testing data of hydropower units is processed through several sub-neural networks and the output result of each sub-neural network is used as the corresponding BPA (basic probability assignment) function that is hard extremely to be obtained. Whereafter, the more accurate and comprehensive diagnosis result can be obtained by fusion diagnosis. Diagnosis example shows that, using information fusion of multi-symptom domains, the belief function of fault target increases markedly, and uncertainty of diagnosis decreases obviously, as a result, the reliability of diagnosis can be greatly improved.