Study Of Bearings Fault Diagnosis Method Based On EEMD Sample Entropy And Fuzzy Clustering

Bearing is one of the most widely used part in a variety of mechanical equipments, whose running conditions often has a directly impact on the performance of the whole machine. But it is also the most potential element liable to occure fault, thus bearing fault diagnosis has become an important branch in the field of mechanical fault diagnosis technology. When dealing with non-linear and non-stationary vibration signals, most of signal processing methods always have a variety of limitations, for they are non-self-adaptive. In this paper ensemble empirical mode decomposition, a self-adaptive method, was introduced particularly to analyse mechanical vibration signal. Also, a feature extraction method based on EEMD combinated with sample entropy was put forward, and it could realize to pick up signal complexity information to identify the fault. Besides, Gath-Geva fuzzy clustering algorithm was applied, due to the problem it is difficult to recognize mechanical fault pattern for its fuzziness.Firstly, the basic principles of empicical mode decomosition was researched, and this method was used to analyse signals. Then ensemble empirical mode decomposition was introduced, which was assisted by noise, in order to eliminate mode mixing problem. On the one hand, the signal could be decomposed according to the different time scales by EEMD, achieving the same purpose as EMD. On the other hand, the mode mixing problem could be suppressed in a certain extent.Secondly, owing to the complexity of dynamic characteristics of the machinery vibration signals reflect with the occurrence of the mechanical fault, the sample entropy was proposed to measure the complexity of the signal,which was influenced by its paraeters but was not sensitive to the length and loss of signal datas. At the same time the sample entropy had a good consistency. Therefore, the IMF component would be quantified by sample entropy to obtain the complexity information on the different frequency bands of the machinery. What’s more, the basis for fault diagnosis could be provided by building the sample entyopy into a feature vector. After that, as the fault diagnosis method, Gath-Geva fuzzy clustering algorithm was used to analysis feature vector above, by means of which, mechanical fault classification could be achieved accurately.Finally, the datasets of the rolling bearing fault from the Case Western Reserve University was taken as the experiment research object, and the experiment would be conducted from different signal types and different degree of injury of the rolling bearing. The results demonstrated that the feature extraction method based on the ensemble empirical mode decomposition and sample entropy is effective to portray rolling bearing fault in both cases. And Gath-Geva clustering algorithm has made good recognition effect in fault diagnosis.