滚动轴承振动与温度数据序列的灰关系分析与自助评估

Grey Relational Analysis and Bootstrap Evaluation on Data Series of Vibration and Temperature of Rolling Bearings

基于滚动轴承服役期间的性能数据序列,建立其振动性能与温度和温差的灰关联度分析模型,从而在轻载、常态和满载等不同工况下对滚动轴承性能进行分析和评估。对振动、温度和温差数据序列进行均值量纲一化和排序,基于灰关系分析法求解数据序列间的灰置信水平;由于传感器数量有限,为深入挖掘轴承性能数据信息,运用自助法对灰置信水平小数据样本进行抽样,并基于最大熵法求解自助样本的概率密度函数和灰置信水平的估计真值与估计区间。研究表明:在轻载、常态和满载工况下,QJS210三点角接触球轴承的振动与温度数据序列的灰置信水平的估计真值分别为78.32%,95.39%和91.58%,振动与温差数据序列的灰置信水平的估计真值分别为93.54%,82.09%和85.34%,间接表明轻载时轴承振动数据序列与温差数据序列之间的关系更亲密,在常态和满载时轴承振动数据序列与温度数据序列之间的关系更亲密。

Based on performance data series of rolling bearings in service,the grey relational analysis model is established for vibration performance,temperature and temperature difference,so as to analyze and evaluate performances of rolling bearings under different operating conditions such as light load,normal load and full load.Mean normalization and sequencing of data series of vibration,temperature and temperature difference are carried out.Based on grey relational analysis method,the grey confidence level between data series is solved.Due to limited number of sensors,the bootstrap method is used to sample small data samples of grey confidence level to deeply mine performance data information of the bearings.The probability density functions of bootstrap samples,estimated true values and estimated intervals of grey confidence level are solved based on maximum entropy method.The research shows that for QJS210 three point angular contact ball bearings under light load,normal load and full load conditions,the estimated true values of grey confidence level between data series of vibration and temperature are 78.32%,95.39% and 91.58% respectively,the estimated true values of grey confidence level between data series of vibration and temperature difference are 93.54%,82.09% and 85.34% respectively,which indirectly show that the vibration data series of the bearings is more closely related to temperature difference data series under light load,and the vibration data series of the bearings is more closely related to temperature data series under normal load and full load conditions.