Evaluation Of Performance Variation Of Rolling Bearing Based On Poor Information

Rolling bearings are one of the important components in mechanical transmission systems,and their performances play major role in the safe operation of mechanical equipment.The performance indicators of rolling bearings mainly include vibration,friction torque,temperature rise,life and so on.Performance variation/degradation refers to the deterioration of rolling bearing performance which is gradually difficult to maintain at the best level with the change of service conditions and service time.The signs of performance degradation are often hidden in the performance data series of rolling bearings.How to extract useful information to characterize the performance variation of rolling bearings from the performance data series,so as to avoid bearing or mechanical equipment failure caused by the cumulative excess of performance degradation factors,has become an urgent problem to be solved.The performance variation of rolling bearings is assessed in this paper by using the relevant knowledge of the poor information theory,which mainly includes three parts:1.In view of the deficiencies of peak factor,pulse factor and kurtosis factor in the monitoring of rolling bearing running condition,a new method to evaluate the variation of vibration performance of rolling bearing was proposed based on the fuzzyset theory.Firstly,the original data sequence of the rolling bearing vibration was divided into the sample sequences,and then the intrinsic sequence was selected to calculate the standard deviation of each sample sequence and the absolute correlation degree,the gray confidence level and the closeness degree relative to the intrinsic sequence as new sample values.Then a fuzzy similarity matrix made up of these values processed in order of magnitude and by the linear mapping formula was further transformed into a fuzzy equivalence matrix via transitive closure method and the evaluation of the variation of vibration performance of rolling bearing was made by comparing equivalence coefficients in the fuzzy equivalence matrix with the set threshold(?=0.5).2.In the case of small samples with unknown probability distribution,the fuzzy equivalence relation and bootstrap maximum entropy models were proposed and the variation process of vibration performance of rolling bearing was analyzed by variation probability.Firstly,the original data of rolling bearing vibration acceleration was divided into the samples,and then the intrinsic sample was selected to calculatethe fuzzy equivalent coefficients between the samples.Then the bootstrap maximum entropy model was used to establish the probability density function of each sample and the variation probability of each sample relative to the intrinsic sample was calculated by the intersection method.Thus,the relation curve between the fuzzy equivalent coefficients and variation probabilities was established to realize the monitoring of the variation process.The experimental investigation shows that the variation probability curve presents nonlinear upward trend like a deck chair with the increase of wear scar diameter,which corresponds to the three stages of rolling bearing wear,namely the initial running stage,the normal performance degradation stage and the performance deterioration stage.3.Based on the gray confidence level,bootstrap-least square method and maximum entropy principle,the mathematical model is established and applied to the dynamic prediction of friction torque performance reliability of the satellite momentum wheel bearings.Firstly,the original data of friction torque are grouped into samples,and the intrinsic sample is selected.Then,a new method to calculate the variation intensity of each sample by the gray confidence level is proposed,and the actual value of the reliability of each sample is obtained.Then,the closest 5 variation intensities are integrated into the bootstrap-least square linear fitting to obtain the fitting coefficients,and then the prediction value and upper and lower intervals of the next sample variation intensity are obtained by the maximum entropy principle.By the closest 5 variation intensities continuously updated,the prediction value and upper and lower intervals of each sample reliability are obtained,and the dynamic prediction of friction torque performance reliability of rolling bearing is finally realized.The experimental results show that the reliability prediction errors are less than 4.1% at constant speed and 9.4% at variable speed.Poor information theory is the effective integration of grey system theory,fuzzy set theory,information entropy principle,bootstrap principle,bayesian principle and even classical statistical knowledge.Therefore,when evaluating the performance variation of rolling bearing with poor information theory knowledge,it often needs the ingenious combination of all kinds of theoretical knowledge,never relying solely on a theoretical knowledge,which also reflects the complexity and variability of practical engineering problems.