Due to the large amount of noise in the AC contactors monitoring signal, the degradation trend cannot be clearly characterized, so the traditional method cannot divide the similar and overlapping degradation states effectively when dividing the degradation stages. In view of the above problems, A degradation stage division method for AC contacter contact system based on Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise -Permutation Entropy -Wavelet Threshold (ICEEMDAN-PE-WT)and Change point detection was proposedin this paper. Firstly, the feature parameters were decomposed by ICEEMDAN to obtain a series of IMF components. Then, PE was used to analyze the complexity of IMFs, and the filtered IMFs were denoised by wavelet threshold, and the signal was reconstructed.Finally, the change point detection algorithm was used to divide the degradation stage, and six degradation states of the AC contactor were obtained.