Based on the deficiencies of traditional self-adaptive equalization, a new self-adaptive equalization method based on kernel recursive least-square(KRLS) is proposed. In the proposed method, the original nonlinear data are mapped to high-dimensional feature space by introducing kernel function. Then standard least-square algorithm is implemented in high-dimensional feature space. the proposed method is compared with the other self-adaptive equalization methods. The simulation results show that the proposed method is supiror to raditional self-adaptive equalization method. The proposed method can effectively eliminate the influence of transer channels. Finally the proposed method is applied to the extraction of weak inpact fault in the rotor system, the experiment results further verify the validity of the proposed method.