Gas-Path Fault Diagnosis For Aero-Engine Based On Sliding Mode Observers

Gas-path diagnostic technique is an important way to improve flight safety and reliability,and reduce maintenance costs for aero-engines.In view of the problem that the traditional linear Kalman Filter(LKF)responses in a sluggish manner in handling rapid shift faults,and lacks robustness in application of gas-path fault diagnostics,this paper investigates the field of gas-path fault diagnostics focusing on a high by-pass ratio turbofan engine by using sliding mode observers,including the fault diagnostic of gas-path components and sensors,performance degrading estimations,on-line realization of diagnosis approaches,and sensor fault tolerant control.A robust sliding mode state estimation approach considering the impact of fault derivatives is proposed,and aero-engine gas-path fault diagnostic is realized.To enhance the robustness of state estimation and reduce observer’s chattering,a robust sliding mode state estimator based on pseudo sliding modes is designed,and the estimation result is entirely decoupled with uncertainties which satisfy the matching condition,thus robustness is strengthened;considering fault derivatives could affect estimating dynamics harmfully,by treating fault derivatives as disturbances,the observer gains are solved by optimizing the norm of the transfer function from fault derivative to estimated states,so that the response speed of the observer is improved.Compared with LKF,the proposed method reacts quicker in rapid shift fault cases,and performs a better state estimation accuracy concerning uncertainties that satisfy the matching condition.To avoid the restriction of the matching condition and the impact of fault derivatives in sliding mode state estimator,an approach of robust unknown input reconstruction sliding mode observer based on super-twisting is proposed,and the robustness and convergence speed of aero-engine gas-path fault diagnostic are further improved.Faults are treated as unknown inputs,thus the influence of fault derivatives to state estimation in sliding mode state estimator is avoided;using the state information in output vector of the component fault model,a reduced-order observer is designed,so that the observer for component fault diagnosis is structurally simplified;Considering any arbitrary boundary uncertainties in cases that the amount of sensors is limited,the transforming matrix is constructed to create freedom for robust design,and an unknown input reconstruction sliding mode observer is designed based on Linear Matrix Inequalities method;viewing the problem of chattering,approaches of nonlinear switching term with multi-step iteration and variable-gain coefficient are investigated based on super-twisting theory,so that observing chattering is resisted.Compared by LKF and sliding mode state estimator,the proposed method is quicker and more robust against any arbitrary uncertainties but boundary.Aiming at the integrated diagnostic system involving performance degradations,component faults and sensor faults during aero-engine life-cycle,the degradation tracking approaches separately served for component degradations and sensor soft faults are devised based on sliding mode observers and fussy logics,and degradations tracking in gas-path is realized;using the degradation estimation results,a gas-path fault diagnostic approach with fault model updating is proposed for degraded aero-engines,so that the degradation influences to fault diagnosis is eliminated;taking the information of fault reconstruction signal,the methods of fault identification and fault diagnostic crossed revision are proposed,and fault diagnostics of gas-path components and sensors during life-cycle are realized;for off-design conditions,the baseline model reflecting the dynamic output of the standard engine is established by adopting similarity transformation corrected by BP neural network,and using the unknown input reconstruction sliding mode observer,the fault diagnostics both at steady state and transient state in cruise flight zone are realized.Aiming at the problem that the rapid shifts of control signal in the sensor fault tolerant control approaches would cause transient over-temperature and over-speed in aero-engines,a sensor fault tolerant control scheme is proposed based on the state estimation by sliding mode observer.The reduced mode is designed to be the state estimation error,and the sensor fault can be fully decoupled with the reduced mode;by replacing the physical sensor signal with state estimation in feedback control,the undesired dynamics caused by reconstructed signal offset are avoided,and good control performance can be retained in rapid shift sensor fault scenarios.