Abstract

Abstract

An integrated bearing prognostics method for remaining useful life prediction



Nowadays, in order to improve the productivity and quality, more and more resources are invested in maintenance. In order to improve the reliability of an engineering system, accurate predictions of the remaining useful lifetime of the equipment and its key parts are required. Bearing plays an important role in the rotating machines. The purpose of using a bearing is to reduce rotational friction and support the load imposed on it in radial and axial directions.
The common types of bearing defects include damage in rolling elements, inner and outer races, etc. In this thesis, we focus on the spall propagation caused by rolling contact fatigue. The existing bearing prognosis methods are either model-based or data driven. In this thesis, we develop an integrated bearing prognostics method, which utilizes both physical models and condition monitoring data. In the physical model part, a Hertz contact model is used to analyze the stress developed from the contact point between two curved surfaces which are pressed together, the ball and the deep groove. Based on Paris’ law, a damage propagation model is used to describe the spall propagation process. It is difficult to measure a defect size when the machines are running. Therefore, online data is obtained and processed to transform raw signals into useful information. In this thesis, the uncertainty factors are considered, including material uncertainty, model uncertainty and measurement error. A Bayesian method is used to update the distribution of this uncertainty factor by fusing the condition monitoring data, to achieve updated predictions of remaining useful life.
Finally, two sets of data are used to verify and validate the proposed integrated bearing prognostics method. The first set of data includes a group of simulated bearing degradation histories. The second set of data were collected from lab experiments conducted using the Bearing Prognostics Simulator. These examples demonstrated the effectiveness of the proposed method.
The key contribution of this thesis is the development of an integrated bearing prognostics method, where the uncertain model parameters are updated using the collected condition monitoring data, while the existing bearing prognostics methods are either model-based or data driven. Both the development of the method and the experimental validation are significant contributions to the field of bearing prognostics.