Probability information extraction and degradation analysis for performance time series of bearings

CHANG Zhen1,2, WEI Jian-bo1,2, PING Xiao-ming1,2, CAO Mao-lai1,2, WANG Er-hua3
(1.Hangzhou Bearing Test & Research Center Co.,Ltd, Hangzhou 310022, China; 2.Machinery Industry Bearing
Quality Inspection Center (Hangzhou), Hangzhou 310022, China; 3.Department of Intelligent
Equipment, Changzhou College of Information Technology, Changzhou 213164, China)

Abstract: Aiming at the problem of probability information extraction and degradation analysis of bearing performance time series, three performance time series of bearing vibration, temperature and friction torque were researched, and probability density function and degradation index quantification scheme were proposed based on bootstrapmaximum entropy method and fuzzy equivalent relation. The probability density function of the training group of bearing performance time series was established by using the bootstrapmaximum entropy principle, and the accuracy of above proposed model was proved according to the frequency of the verification group falling into the function interval. Then, the fuzzy equivalence relation of bearing performance signals was extracted by the fuzzyset theory, and the bearing degradation characteristics was estimated by combining the 0.5 threshold parameter. The results indicate that the interval false positive rate of bearing vibration and temperature is only 2% and 4%, and the minimum degradation coefficient is 0.600 and 0.609 respectively, indicating that the probability information obtained is accurate, and the service condition of bearings is good. The interval false positive rate of bearing friction torque is 66%, and the minimum degradation coefficient is 0.477, indicating that the verification group has occurred significant variation, and the corresponding bearing shows obvious degradation signs.

Key words: bearing performance; friction torque; probability density; degradation analysis