This paper presents a degradation-level assessment and failure prognostics methodology for degrading systems. The proposed methodology consists of offline and online phases. In the offline phase, different time-domain health indicators (HIs) are extracted and the best indicator of degradation is selected by filter-based methods. Then, a degradation model is defined and its parameters are estimated using the selected HI. In the online phase, the k-means clustering is utilized to detect a change(s) in the system’s health state and to trigger failure prognostics for remaining useful life (RUL) prediction. The degradation model parameters are updated as new data are available, and the RUL is predicted iteratively. The proposed methodology is implemented on point machine sliding chair degradation using in-field condition monitoring (CM) data. The results show that the methodology can be effectively used in machine degradation-level assessment and in online RUL predictions.