Rolling Bearings Fault Diagnosis Based On Chaos Fractal And Fuzzy Clustering

With the development of science and technology, mechanical equipment is becoming more and more complicated and automatical, and there are more and more attention paid to fault diagnosis technology. Aiming at complex features of the fault machinery such as nonstationary and nonlinearity, a qualitative and quantitative analysis method for fault diagnosis based on chaos and fractal is introduced. For the problem that is difficult to recognize fault pattern in mechanical equipment,a new method for fault diagnosis based on fuzzy maximum likelihood estimates (FMLE) clustering algorithm is introduced. These approaches are applied to detect the fault of rolling bearings and the result is given.Firstly, the structure, classification, vibration mechanism and fault mode of rolling bearings were summarized, and the conventional analytical methods for vibrating signals were presented, which were classified to the analysis in the time domain, frequency domain, and time frequency domain and so on.Secondly, on the basis of the analysis of the existing phase space reconstruction methods'limitations, a method based upon the differential entropy for determining both embedding dimension and delay time was proposed, and a method based on phase space reconstruction called delay vector variance (DVV) was introduced in fault diagnosis. DVV algorithm was very sensitive to nonlinear signal, the qualitative description about nonlinearity of the fault signals were given by DVV plot or DVV scatter diagram, then different faults were analyzed quantitatiely by multifractal detrended fluctuation analysis, and multifractal spectrum parameters were extracted as new criterions to diagnose machinery faults, which would make preparations for fault patterns recognition.Thirdly, in view of the traditional Fuzzy C-Means (FCM) clustering algorithm was only suited to spherical-shape distribution datasets, a distance norm based on the fuzzy maximum likelihood estimates was introduced, which suited to datasets with different shape and size, density, then the different faults in rotating machinery were detected automatically.Finally, after the analysis of the bearing faults datasets from Case Western Reserve University, the results demonstrate that the presented method is effective to analyze bearing faults in various degrees qualitatively and quantitatively, which can be recognized by fuzzy clustering effectively...