Data-Driven Modelling of Aeroengine systems Based on Variational Bayes and Adaptive Kalman Filter

This paper mainly develops a data-driven modeling approach for aeroengine systems. To describe how the system behaves dynamically, a state variable model is constructed, and it is further transformed such that the unknown parameters are treated as hidden states. As a result, the model parameters with their uncertainties, and the noise statistics can be jointly estimated by integrating an adaptive Kalman filter into the variational Bayesian (VB) framework. The efficacy of the proposed data-driven model is verified through comparison study with the component-level JT9D aeroengine model provided by NASA.