Bearing Performance Degradation Assessment Based On PKPCA And Logistic Regression Model

Rolling bearing is an important part of rotating machinery,and it is also the main force components.Its running state will affect the performance of the machine,so monitoring the state of the rolling bearing and predicting the remaining life of the rolling bearing can not only reduce the economic loss caused by the bearing failure,but also avoid the occurrence of major accidents.In this paper,we take the vibration signal of rolling bearing as the research emphasis and we use the method of MED-MEEMD to reduce the strong noise.The characteristic features of bearing are extracted from vibration signal which is interfered by strong noise.Then we use the method of Fast Kurtogram to find the best parameters of band pass filter and the envelope demodulation of filtered signal can find the fault type of bearing.The method of probabilistic kernel principal component analysis is used to select the features of the mixed domain which conclude the time domain features,frequency domain features and time frequency domain features.The selected features will be the covariate of logistic regression model to perform the degradation of bearings.The main content of this paper are as follows:(1)The research direction,research content and research significance of this subject are discussed.The basic structure and failure mechanisms of rolling bearings are introduced,and the fault characteristic frequencies of each part of rolling bearings are derived.Different types of fault tests such as outer ring,inner ring and rolling body are designed.The data acquisition and analysis system of rolling bearing vibration signal is compiled by the software of LabVIEW.The data collection equipments are provided by the National Instrument to realize the collection of data in real time and data preservation.The data analysis system can analyze the collected vibration signal in time domain or frequency domain,including the method of Cepstrum and envelope demodulation.(2)In order to accurately extract the fault characteristics of the rolling bearing in the strong noise environment,the method of Minimum Entropy Deconvolution-Modified Ensemble Empirical Mode Decomposition is proposed.The method of Minimum Entropy Deconvolution is used to reduce the noise of vibration signal and increase the kurtosis and the noise-signal ratio.Then use the method of MEEMD to decompose the signal and select the useful IMF to restructure signal.In order to verify the effect of noise reduction,the Fast Kurtogram is used to determine the parameters of the bandpass filter,and then use the method of envelope demodulation spectrum to determine the fault type of the rolling bearing.(3)Use the method of probabilistic kernel principal component analysis and logistic regression model to predict the residual life of rolling bearing.The high dimensional features are composed of the features of time domain,frequency domain and time frequency domain.Then use the method of probabilistic kernel principal component analysis to reduce the dimension of high dimensional features,then we can obtain the principal component which can describe the degradation of bearings.We use the obtained principal component as the covariate of logistic regression model to predict the residual life of bearing.