Hypothesis Testing Of Rolling Bearing Performance Based On Poor Information Process

The dynamic performance which contains vibration and friction torque is the mostimportant in the rolling bearing performance. The vibration can accelerate wear andfatigued damage of bearing components, and friction torque is an important indicatorof evaluating the operational sensitivity of bearing. Current research shows that rollingbearing vibration and friction torque with nonlinear characteristics, which belongs toan information poor system with unknown probability distributions and trends, thiscounteracts hypothesis testing for the rolling bearing vibration and friction torque asthe time series, therefore, how to analyze correctly the complex vibration and frictiontorque characteristics as the time series has become a problem to be solved.Firstly, raw data of the rolling bearing performance is section processed as timeseries, secondly, the probability density function of the prior sample and the currentsample are established by the maximum-entropy bootstrap method, thirdly, thedynamical Bayesian probability density function of the rolling bearing characteristicparameters is constructed by the Bayesian statistical theory, finally, according toposterior probability density function, we can calculate the posterior expectations,posterior variance ratio and posterior overlap ratio of characteristic parameters. So far,the hypothesis testing of information poor mathematical model is constructedcompletely, according to this we can effectively realize status evaluation analysis andjudgment of the rolling bearing performance characteristic parameters in time series.By analysis and hypothesis testing of rolling bearing performance experimental data,the results show that, the posterior expectations of the rolling bearing performancecharacteristic parameters are in the different interval range under different workingconditions. The posterior overlap ratio changes with time series and chosensignificance level of0.1, when the posterior overlap ratio is more than0.9, the bearingis in the stationary time series under the current period. When the posterior overlap ratio is less than0.9, the bearing is in the nonstationary time series under the currentperiod.The posterior overlap ratio perfect fuses posterior expectation and posterior varianceratio, not only caring about the dispersion degree of the bearing performance data inthe current time period, but also considering the overall trend of the bearingperformance data in the current time period. When posterior expectation is closer tothe posterior expectation of priori samples, and the posterior variance ratio is closer to1, the stability of the bearing in the current time period is better, when posteriorexpectation is deviate from the posterior expectation of priori samples, and theposterior variance ratio is deviate from1, the stability of the bearing in the currenttime period is worse.There are a lot of similarities and trend of greater consistency among the posteriorexpectation, the posterior variance ratio and the posterior overlap ratio of two timeseries methods, it shows that the method based on the phase space of time series withless data can accurately fit the method based on time series, it is employed to recoverthe original dynamic characteristics of rolling bearing performance as time series.Elementary theories and methods studied in this paper are of the versatility andfeasibility. The investigation shows that the proposed hypothesis testing mathematicalmodel is confirmed to be reasonable and feasible with the analysis and judgment ofrolling bearing performance experimental data based on time series. The error betweenthe analysis result and the actual result is very small and requirements in engineeringare satisfied, thus it lays a new theoretical foundation for the analysis and judgment ofrolling bearing performance experimental data based on time series.