A Data-Driven Approach for Bearing Fault Prognostics

Bearings are one of the critical components widely used in rotary machines. Bearing failure can be catastrophic and may lead to a lengthy downtime of systems for maintenance. Bearing fault prognostics can help reduce the cost for maintenance and avoid catastrophic failures of the systems. This paper proposes a new data-driven approach for bearing fault prognostics, which is based on the Kolmogorov-Smirnov test, self-organizing map, and unscented Kalman filter (UKF). The proposed approach has two steps. The first step is to detect bearing's degradation process by learning the historical data and the second step is to predict the remaining useful life (RUL) with the aid of a degradation model and the UKF. The proposed approach is validated by bearing's life data obtained from a run-to-failure experiment. Results show that the proposed approach can detect the bearing degradation process successfully and predict the RUL effectively.

I. Introduction

Bearings are critical to many rotary machines such as wind turbines, automobiles, and high-speed trains. Unfortunately, bearing failure is one of the most common failure modes in these machines [1]–[3]. The failure of bearings will lead to a malfunction or catastrophic failure of their host systems and result in a lengthy and costly downtime [4]. Thus, there has been a growing interest in prognostics and health management (PHM) for bearings. By doing so, the health condition of bearings can be estimated and predicted online. The prediction of the remaining useful life (RUL) of bearings can save time and money for maintenance and help extend the system's lifespan and reduce life-cycle costs [5], [6].