 The proposed method uses improved local main decomposition, LMD, multi-scale permutation entropy, MPE, and hidden Markov model, HMM, to diagnose the faults of rolling bearings. It first decomposes the vibration signal into multiple components, then calculates the phase-space reconstruction of one component and sets the delay time and embedding dimension to determine the scale. Next, it extracts the features of the multi-scale permutation entropy, MPE. Finally, these features are used as HMM training and diagnosis. The experimental results demonstrate that this method can accurately identify the different faults of rolling bearings. This article was authored by Yang Degao, Francesco Vallejo, Mingli, and others.