Simulation Framework for Remaining Useful Life Prediction on Ball Bearings in Electromechanical Flight Control Actuators

Abstract

Deep learning-based approaches have shown promising results for remaining useful life (RUL) prediction for components such as ball bearings of electro-mechanical flight control actuators (EMA). In order to use those approaches, it can be challenging to generate the required amount of high-quality labeled vibration data under variable operating conditions. A way to cover this demand is the use of simulation data. This paper proposes a simulation framework for RUL prediction on ball bearings of an EMA, defining interfaces and dependencies between all data stages forming a requirement basis for the simulation model. The load and movement of the bearing is derived from the operating conditions and applied in the test rig operation as well as in the simulation model. The test rig data provides a database for generating the simulation model. At the same time, the deep learning-based RUL prediction yields requirements for the simulation data. Predicting the degradation trend requires data samples that include vibration data, health indicator labels and an uncertainty quantification. The presented approach accounts for simulation data requirements from the perspective of both the operating case and the RUL prediction. The developed framework is applied to the short stroke movements of an EMA during the cruise flight phase and can subsequently be used as a basis for the development of simulation models.