Abstract: Grade-life was used to describe rolling bearing-s service life, which means the entire service life is divided into four stages: good bearing condition, initial defect condition, damaged bearing condition and failure coming condition, and an assessment model was presented for bearing-s Grade-life. Signal feature extraction and pattern recognition algorithm are keys to construct the model. Vibration signals of the rolling bearing were analyzed, and the wavelet packet analysis theory was adopted to extract the Grade-life characteristics. Through signal decomposition and single reconfiguration with wavelet package to extract feature in every frequency bands effectively, and energy of various frequency bands acting as the life feature vector was input into the improved BP neural network to realize the mapping between the Grade-life vector and the Grade-life of rolling bearing, and the model in establishing the identification model by using bearing test stand run-to-failure data as learning samples was employed. The validity and creditability of model has been demonstrated by bearing test stand dates.