Aeroengine Performance Health Forecast Based On Fusion Filtering Methods

Performance health prediction of aero-engine is a critical issue for improving flight safety and reducing maintenance cost.In recent years,engine health prediction has become a hotspot in the field of performance and control of flight propulsion system.This paper focuses on the approach of state estimation based on fusion filters to enhance aero-engine health performance prediction capability.To deal with the inherent problems of conventional centralized Kalman filters such as heavy computational effort on one processor and poor sensor-fault tolerance,a fusion filtering is proposed for gas path health parameters estimation.According to different fusion strategies,the forward and feedback structures of fusion Kalman filter are designed and analyzed.Then the fusion filtering architecture is extended to nonlinear situation,and the equivalence of fusion extended Kalman filter in feedback mode and centralized extended Kalman filter is mathematically proved.For aero-engine health estimation,the linear Kalman filter(LKF)is simple and with a little computational consumption,while the extended Kalman filter(EKF)achieves better accuracy.Giving consideration of the advantages of these two filtering algorithms,a hybrid fusion methodology to integrate the LKF and EKF is presented.The structure of hybrid federated filter based on LKF and EKF is designed,and the equivalence of this novel fusion filter and centralized EKF is mathematically proved.Then the performance health prediction capabilities for turbofan engine are systematically analyzed.Since the turbofan engine is strong nonlinear system,the unscented Kalman filter(UKF)is introduced into the hybrid filtering architecture.A nonlinear hybrid federated filter relied on the EKF and UKF is designed.This hybrid federated filter avoids the direct linearization error in EKF algorithm,and enhances the estimation accuracy by utilizing UKF to handle the sensor measurements with strong nonlinearity.In order to solve the inconsistency of output periods of local filters,the asynchronous fusion mechanism is discussed in the nonlinear hybrid federated filters.Finally,the proposed two hybrid federated filters are examined on the semi-physical simulation platform.The experiments on semi-physical hardware in loop simulation reveals the validity of these involved filtering methodology.