The operation condition of rolling bearing performance is describled using time series of the friction torque current signal, which are segmented to establish intrinsic sequences. Based on grey relation, each section of the friction torque signal is sorted to match the intrinsic sequence, as a result, the grey confidence level is acquired. The grey confidence level is used to determine the extent of stability on bearing performance. Via bootstrap resampling for the segmented data, the probability density function is calculated by using the maximum entropy method, and estimated interval is obtained according to the corresponding grey confidence level. Relying on the counting process, the raw information of variation intensity is simulated. The reliability function is constructed with the Poisson process to realtime monitor the reliability evolution of rolling bearing. Simulation cases and experimental test show that the proposed model can truly monitor the stability and reliability of bearing running performance, and effectively deal with the time series with strong fluctuation and varied trend.