Research of Rolling Bearing Fault Diagnosis Method Based on IMF Entropy and Multi-sensor Networks Fusion

Xue-ping REN,Xiang-zhi XIN,Zhen PANG,Yi-tong XING,Jian-guo WANG
DOI: https://doi.org/10.13462/j.cnki.mmtamt.2015.06.022
2015-01-01
Abstract:A rolling bearing vibration signal can be decomposed into a number of intrinsic mode function ( IMF) adaptively according to its own scale with the empirical mode decomposition ( EMD) method. The information entropy can reflect the uncertainty degree of the system. The rolling bearing failure will change the information entropy. EMD space state feature entropy and IMF singular spectrum entropy are proposed by combining EMD and information entropy. In the multi-sensor system, primary diagnosis network of mul-tiple probabilistic neural networks was constructed and the feature vector of probabilistic neural network is EMD space state feature entropy and IMF singular spectrum entropy. Because output of accumulation layer of the probabilistic neural network are each kind of model of probability values, mass functions of D-S evi-dence theory can be built using these probability values. The fusion diagnosis of decision level proceeded by the D-S evidence theory. Using this method for rolling bearing fault pattern classification, the results of ex-periments show the feasibility and effectiveness of the method. Compared with the single sensor method in the diagnosis, it improved the accuracy of fault diagnosis.
What problem does this paper attempt to address?