Abstract:  In order to dynamically predict the bearing vibration performance sequence, the bootstrap method and the least squares method were effectively integrated, and a dynamic prediction model of the bearing vibration performance sequence based on the bootstrap-least squares linear fitting was proposed. Firstly, 10 adjacent vibration data were simulated sampling by using the bootstrap method to form multiple sets of vibration side information at the current state. Then the multiple sets of vibration information were linear fitted by using the least squares method. Secondly, the probability density function, truth value and estimated interval of fitting coefficients a and c were obtained by the bootstrap-maximum entropy principle, and the truth value fitting and the interval fitting of rolling bearing vibration time series could be obtained. The dynamic prediction of the true value and interval of the vibration performance of rolling bearings were realized by constantly updating the 10 adjacent vibration data. Finally, the accuracy of the dynamic prediction model of the bearing vibration performance sequence was verified by using the vibration performance cases of a bearing in three service periods. The results indicate that the prediction value obtained using forecast models is in good agreement with the actual value, and the maximum prediction error is only 14.73%. The prediction interval difference is small and the accuracy is high. The dynamic prediction model of vibration performance sequence can better perform health monitoring and safety diagnosis for bearings applied in engineering practice.

Key words:  rolling bearing performance degradation; vibration signals; dynamic prediction model; self-help least square method; probability density function; prediction interval difference; bearing health monitoring; bearing early failure