Aiming at the problem of rolling bearing fault severity recognition,a fault severity recognition method of rolling bearing based on compensation distance evaluation technique（ CDET） and grey relational analysis（ GRA） is proposed. By utilizing the compensation distance evaluation technique to choose features sensitive to fault severity from the high dimensional feature set,and weights such sensitive features according to their sensitive degree. Then,by using the sensitive feature subset,a standard degradation pattern matrix is built,and the grey relational analysis is applied for fault severity recognition in accordance with correlation degree index. The results show that the proposed method can effectively recognize different rolling bearing fault severity,and at the same time,the calculation complexity is reduced.