Aiming at the problem that the weak fault features of rolling bearing signal is difficult to be extracted under strong background of noise,a feature extraction method for weak faults based on maximum correlated kurtosis deconvolution(MCKD) and complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN) is proposed.As the filtering effect of the MCKD method is affected by the filter length parameter,in order to adaptively achieve the best result of noise reduction, the variable step size grid search method is used to optimize the filter length of the MCKD.Firstly, the feature energy ratio (FER)) is used as the objective function to find the optimal filter length by the variable step size grid search method, the vibration signal is denoised by the adaptive MCKD algorithm. Then, the noise reduction signal is decomposed by CEEMDAN algorithm,and the Intrinsic mode function (IMF) with fault information is selected for signal reconstruction with the kurtosis criterion. Finally, the envelope spectrum analysis is performed on the reconstructed signal to extract fault features. Through simulation and experimental analysis, this method can effectively extract the weak fault characteristic information of rolling bearings.