A Method Of Rolling Bearings Fault Diagnosis Based On Mathematical Morphology And Fuzzy C-means

With the development of modern industrial production, there are more and moreattention paid to machinery equipment fault diagnosis technology in recent years. Rollingbearing is one of the most important components of mechanical transmission system andits running condition will directly impact on the working conditions of the machine.Aiming at the noise in the rolling bearing vibration signals, the combination filterbased on morphology was introduced for de-noising. According to the complex features ofthe fault machinery such as non-stationary and non-linearity, a qualitative and quantitativeanalysis method was introduced for fault diagnosis based on multi-scale morphologyanalysis. For the problem that the fault pattern is fuzzy, the fuzzy center means clusteringalgorithm was introduced. These approaches are applied to the fault of rolling bearings.First, the fault mode and vibration mechanism of rolling bearings were summarized,and the noise reduction methods for vibrating signals were presented, which were thetraditional filter method, wavelet de-noising technology and empirical modedecomposition de-noising technology; and the conventional analytical methods forvibrating signals were presented, which were classified to the analysis in the time domain,frequency domain and so on.Secondly, in the light of the absence of the determined selection of the structureelement in the morphological combination filter, the shape, width and amplitude of thestructure element in the morphological combination filter were analyzed on the effect ofmorphological filter.Thirdly, the application of the multi-scale morphology was analyzed in the vibrationsignals. The characteristics of the fault signals were described by the fractal dimension andthe entropy of the multi-scale morphological spectrum. And the two characteristicparameters were introduced into the fuzzy c-means clustering algorithm as characteristicvector of clustering analysis, which would make preparations for fault patternsrecognition.Finally, the paper analyzed the datasets from the bearing faults datasets of Case Western Reserve University and the Baosteel1580SP measured data of rolling mill, andgave a conclusion. The results demonstrated that the morphological filter method iseffective to de-noising of the rolling bearing vibrating signals, the multi-scale morphologymethod is effective to analyze bearing faults in the various degrees qualitatively andquantitatively, which can be recognized by fuzzy center means clustering effectively.