Assessment Methods Of Poor Information Of Performance Variation Of Rolling Bearings On Time Series For Unknown Distribution

Time series for unknown distribution is whose distribution does not be definite distribution type or can not be reflected by some distribution in statistics,its characteristics appears diversity and complexty,which belongs to main poor information theory.Performance of rolling bearings belongs to time series for unknown distribution,fusion theory is proposed to resolve evaluation problems of unknown probability distribution in the performance variation process of rolling bearings based on statistics and genetics principle in paper.(1)Considering effect of many factors for performance of rolling bearings and unsteady characteristics of test data,the fusion method of histogram and normal test of time series is proposed to analyze variation performanc of rolling bearings.For the studies of vibration and friction moment of rolling bearings,results show performance data of rolling bearings belongs to time series for unknown distribution,part performance data arises to vary and variation data lies in the both ends of order statistics on test data.(2)Based on modern statistical method,the method of combining with median estimate and Huber M estimate is proposed to resolve robust problems for test data for unknown distribution,and obtain genetic factors and varable factors.For the studies of friction moment and vibration data of rolling bearings,and obtain genetic factors and varable factors of rolling bearings,results show that robust of genetic factors is optum than the one of experiment data,so the combining method can effectively devide genetic factors and varable factors from time series.The method provides the determination method of significance level and boundary values of Huber M on statistics.(3)According to obtaining genetic factors and varable factors by the method of combining with median estimate and Huber M estimate process time series data,assessing system of rolling bearings can be constituted by working performance,variation performance and confidence level of time series data.For the analysis of friction moment and vibration of rolling bearings,working characteristics of rolling bearings appear the complexity and diversity.Parameter non-parameter fusion method is proposed to assess characteristics of time series for unknown distribution.Results show parameter assessments and non-parameter assessments can quantitative and qualitative analyze the data and change trend characteristics,choose a rolling bearing of optimal performance from small batch,and provide a choosing case on high precision bearings.(4)Combing with robust and chaos theory,proposed fusion method of robust and chaos methods analyzes dynamic characteristics of time series data for unknown distribution.For the studies for performance data of rolling bearings,analysis results show friction moment of the rolling bearing have the similar predictable cycle,strange attracting,relationship between median of physical space and correlation dimension of phase space is monotonically increasing;vibration of the rolling bearing have the same predictable cycle,strange attracting,relationship between median of physical space and correlation dimension of phase space is non-linear,non-monotonically.Fusion method can decrease sensitive of initial data,and increase reliability of data analysis,better disclose dynamic characteristics of time series.(5)Introducing the robust theory in order to evaluate the degradation of product performance,the method of performance variation proposed to assess performance degradation of time series data for unknown distribution.For the analysis of test data on different damaging diameters of groove raceway in inner ring for vibration effect of rolling bearings,results show that change trend of local intrinsic interval,variation rate,median and mean of different time phases are consistent,and agree with the test results,and regard as measurement on different damaging diameters of groove raceway in inner ring for vibration effect of rolling bearings,and variation rate can predict running state of rolling bearings.The method can effectively assess intensing experiment results of rolling bearings.(6)On the basis of modern statistical methods,the method of combing with robust and Bayesian theory proposed to resolve uncertainties problems of time series data on unknown distribution.For the analysis of vibration and friction moment of rolling bearings,results show that,in the running process of rolling bearings,variation performance of the confidence leve,mean and waving range appear deliversities and complexies,and assessment precision of the method is higher than one of classical statistics.Application meaning of the method lies in providing determination method of confidence level and boundary value of Huber M estimate,provides establishing method of prior distribution on Bayesian method of extracting robust data from time series data.Above methods regard robust process of test data as the center and take parameter-nonparameter fusion method constitute as the evaluation system of performance variation on poor information,provide new ideas for deep disclosing new characteristics and new mechanism of property variation of rolling bearings.Evaluation system allows the distribution,trend and confidence level of time series data to be unknown,and analyzes and resolves problems of poor information of rolling bearings.