Fault diagnosis method of integration of neural networks based on Dempster-Shafer evidential theory
MA Xiao-jiang · Systems engineering and electronics · 2004
The basic probability assignment method and the rules of diagnosis decision-making are given, and then a new method of fault diagnosis based on integration of neural networks is presented, based on Dempster-Shafer evidential theory. Taking the gearbox bearing fault diagnosis for example, the implementing process of this method is elaborated in details. The results indicate that through information fusion of multi-fault features the reliablilty of diagnosis result is improved evidently, and the uncertainty decreases markedly. The effectiveness of the method is proved fully.