Physics-based Remaining Useful Life Prediction for Aircraft Engine Bearing Prognosis

Abstract

Aircraft engine bearing prognosis not only requires early detection of a bearing defect, but also the ability to predict bearing health conditions for all operational scenarios. This paper summarizes a physics-based remaining useful life (RUL) prediction method developed in the DARP A Engine System Prognosis (ESP) program. This investigation focuses on a typical roller bearing fault (or defect) on the outer raceway. Spall detection is based on the fusion of vibration and online oil debris sensors. Spall size estimation is derived from the amount of bearing debris chips that passed through the Oil Debris Monitor sensor. Subscale propagation tests were performed to generate the response surface of the spall propagation rate under various operating speeds and loads. A particle filter based approach was used to track the spall propagation rate and update the prediction according to newly calculated diagnostics information. The bearing spall propagation model outputs a RUL distribution, which is calculated based on future operating conditions and the time the spall size crossing the failure threshold. The developed RUL prediction method was validated using full- scale bearing spall tests. The comparison of model prediction and measured ground truth demonstrated that the developed model was able to predict the spall propagation rate accurately, and its prediction accuracy and confidence can be further improved by incorporating more diagnostics updates and/or increasing the confidence in the sensor data.