Fault Prediction of Automotive Bearings Based on Weibull Distribution

Abstract

The existing prediction of automotive bearing faults has problems such as short prediction time and large errors. This paper takes Weibull distribution as the theoretical basis. Firstly, Weibull parameters of under different operating states are calculated, and the optimal parameter estimation method is determined by goodness of fit analysis. Secondly, each sub-series are calculated to determine the feasibility of Weibull parameters to characterize the evolution of performance. Finally, the vibration time subseries in the stable interval of the test is analyzed, and the bearing fault prediction is realized by parameter change. The results show that the maximum likelihood method has the highest accuracy. There is a high goodness of fit between Weibull probability density function and the actual vibration time series when the bearing is running normally. Bearing performance evolution is consistent with Weibull parameter change, and bearing performance evolution can be analyzed by Weibull parameter change. In this way, bearing faults can be detected 44 minutes in advance.