Change points estimation for equipment degradation process based on fuzzy c-means clustering

The working equipment would experience a degrading process until it runs to failure. Many works have been done to illustrate and describe this process as accurate as possible. In the real world applications, the monitored data are unlabelled data. In this paper, we aim to automatically assign degrading states to the monitored. To achieve this target, the change points between adjoining states should be evaluated. Instead of using the conventional hard clustering methods, we adopted the fuzzy c-means clustering, where the monitored data can be assigned to two adjoining states, to estimate the change points. Additionally, the degrading process of an intermittent working equipment is utilized to verify the effectiveness the evaluation method. The experiment results show that fuzzy c-means clustering is suitable for assigning state labels to the unlabeled data and evaluating the change point of involved equipment.