Abstract: The problems of high redundancy of vibration signal and low accuracy of classification in the degradation stage exist in AC contactor contact system degradation, therefore, a method based on Spearman's rank correlation coefficient (SRCC) and fuzzy C-means clustering (FCM) is proposed to divide the degradation phase of the contact system. Firstly, the vibration signals of the opening and closing stages of the contact system throughout the life cycle are obtained through the full life test of the AC contactor, and the time domain and frequency domain characteristics reflecting the degradation state of the contacts are extracted from the results.Secondly, SRCC and principal component analysis are used to select feature and reduce feature dimensionality, and to construct performance degradation indicators. Thirdly, unsupervised learning (UL) FCM is introduced as a theoretical basis for dividing the contact system into 2 to 5 degradation stages. Finally, silhouette coefficient (SC), Calinski Harabasz score(CHS) and Dacies Bouldin index (DBI) are selected to evaluate the clustering effect internally, and five supervised classifiers are used to verify the accuracy of this method. The results show that the internal evaluation is the best when the contact system is divided into three degradation stages, the average recognition accuracy of five classifiers for the three degradation stages can reach 95.54%, and the degradation stage division is the most reasonable. The proposed method can be adopted to effectively solve the problem of signal redundancy and low accuracy of phase division, which can accurately realize the division of degradation phase of AC contactor contact system.