基于相空间的滚动轴承振动特征参数的乏信息分析

摘要

The original dynamics characteristics of roiling bearing vibration time series are restored by phase space reconstruction with chaos theory. The dynamical Bayesian posterior probability density function of bearing vibration characteristic parameter is constructed by maximum entropy bootstrap and Bayesian statistics in phase space. Finally, the methods about point estimation, interval estimation and trend estimation of characteristic parameter are proposed. The investigation shows that errors between inference results and experiment results are very small, and requirements in engineering are then satisfied. Therefore, the statistical characteristic of bearing vibration can be described using the proposed poor information mathematical model.%用混沌理论进行相空间重构,恢复轴承振动时间序列的原始动力学特征,然后在相空间里,用最大熵自助法并结合Bayes统计理论,建立轴承振动特征参数的动态Bayes后验概率密度函数,最后对特征参数进行点估计、区间估计及趋势估计等.试验研究表明,估计结果和试验结果在数值上很接近,满足工程要求.因此,用乏信息数学模型可以真实地评价轴承振动的统计特征.