Abstract: In view of problem that remaining useful life of complex machinical equipment was difficult to predict, an approach for remaining useful life prediction based on similarity was presented. Feature extraction method was established based on correlation analysis, reference remaining useful life and its weight can be determined by computing similarity of the relevant feature between sample data and prediction data. Finally, remaining useful life can be obtained by calculating weighted sum of the reference remaining useful life. Experiment results show that the proposed approach can effectively extract feature which can precisely reflect variation trend of the remaining useful life of bearing, and can more effectively predict remaining useful life of bearing with accuracy rate of 81.8%, and provides a scientific basis for life management of related equipment.