The invention relates to a rolling bearing fault diagnosis method based on a vibration signal. A CEEMDAN algorithm is adopted to decompose the vibration signal, de-trending fluctuation analysis is carried out on an obtained intrinsic mode function, a scale function value of each IMF component is calculated, and a noise dominant IMF component is selected to carry out de-noising processing; the noise can be better removed, and the distortion degree of the signal is reduced; calculating correlation coefficients and kurtosis values of all orders of IMF components, selecting IMF components with relatively large correlation coefficients and kurtosis values to perform signal reconstruction, performing Hilbert envelope spectrum analysis on reconstructed signals, extracting fault feature frequency, introducing a grey wolf algorithm to optimize initial parameters of multi-scale permutation entropy, performing MPE value calculation on the reconstructed signals, selecting a proper MPE value to construct a rolling bearing fault feature set, and inputting a fault feature vector into the trained support vector machine to carry out rolling bearing fault recognition, so that the entropy discrimination degree is high, the constructed fault feature vector is better, and the recognition rate is higher.