Time-frequency complexity based remaining useful life (RUL) estimation for bearing faults

Reliable operation of electrical machines depends on the timely detection and diagnosis of faults as well as on prognosis, i.e. estimating the remaining useful life (RUL) of the components. Bearings are the most common components in rotary machines and usually constitute a large portion of the failure cases in these machines. Although there has been a lot of work in the study of bearing life failure mechanisms and modeling, the problem is still far from being solved. In this paper, we introduce a time-frequency feature extraction based method for estimating remaining useful life of bearings from vibration signals. The proposed approach extracts measures that quantify the complexity of time-frequency surfaces corresponding to vibration signals. The extracted features are then tracked through the life time of a bearing using curve fitting and Extended Kalman Filtering algorithms. The proposed methodology is tested on a publicly available bearing data set with known RULs.