Prediction of remaining useful life (RUL) of rolling bearings is of great significance to ensure smooth operation of rotating machinery.Aiming at the problems of large fluctuation of prediction accuracy and low data utilization rate of time domain features,a rolling bearing RUL prediction method was proposed based on the fusion feature of time domain and spectral kurtosis and exponential degradation model.The features of signals extracted from time domain and spectrum kurtosis were smoothed and sorted based on monotone scale,and then the dominant features were selected to construct health indicators through principal component analysis (PCA).Then,the degradation point was determined by the 3σ criterion and the data were reprocessed.Finally,the RUL of bearing was predicted at every moment based on Bayesian theory and maximum likelihood function estimation of exponential degradation model parameters.The XJTU-SY dataset were used to verify the effectiveness of the proposed method.The results show that the proposed method can mine potential information of small sample data based on current observation bearings and accurately characterize the degradation process of non-stationary signals under strong noise background,thus improving the prediction accuracy of RUL.