Gas Path Fault Diagnosis For Turbojet Engine Based On Nonlinear Model

Aircraft engine fault diagnosis is a key technololgy to the propulsion system prognostic health management. The engine fault diagnosis system plays an important role in achieving condition-based maintenance, flight safety, usage cost reducing, and device reliability improvement. This dissertation focuses on turbojet engine gas path fault diagnosis system based on nonlinear filtering estimate techniques. It mainly involves turbojet engine modeling and model correction, nonlinear filters estimation accuracy and computational efforts, degradation and abrupt faults diagnosis, and rapid prototype experiments evaluation. The major work and contributions are as follows:(1) The nonlinear component level model of a turbojet engine is studied and the correction approaches are present. Overall framework of general simulation performance model for turbojet engine is designed, and the class codes of components, algorithms, and interfaces are compiled in Visual C++ software. The turbojet engine component level model from idle to maximum condition in whole flight envelope is established. Due to the differences between the model outputs and testbed data, the model outputs compensation based on least square support vector regression(LSSVR) is presented. Component characteristics correcting factors are introduced and denoted as the individual change of mass flow and efficiency, which are optimized by the GA to obtain corrected component characteristic maps. The steady and dynamic model accuracies are improved as the corrected maps utilized into turbojet engine model.(2) The extended Kalman filter(EKF) and its improved algorithms are studied and applied to turbojet engine gas path fault diagnosis. The component health parameters are defined based on typical engine component fault modes, and fault coefficients are designed to simulate gas path faults. Considering how to use the prior knowledge of performance degradation to state estimation, two novel algorithms, i.e. probability density function truncation EKF and gain projection EKF, are proposed to health parameters estimation. To the underdeterminate estimation problem that the measurements number is less than the health parameters number, a novel underdeterminate EKF method is proposed. The transform matrix is introduced and used in the underdeterminate EKF, which broads the application scopes of the EKF on gas path fault diagnosis.(3) Turbojet engine life cycle gas path fault diagnosis in terms of component performance normal degradation is necessary to discuss. A dynamic integrated diagnostic architecture based on the underdeterminate EKF is proposed, which is derived from the linear diagnostic frame. Two real time engine models with different performance update cycles are used, and fault diagnosis and isolation logic for engine component and sensor failure is designed. The digital simulations show that engine model in this architecture has satisfied capability to track component normal degradation, and the component abrupt fault and sensor fault are recognized over the engine life course.(4) Due to the engine characteristics of strong nonlinear and complex noise polluted, particle filters are studied for turbojet engine gas path fault diagnosis. A novel filters named quantum particle swarm optimization particle filter(QPSO-PF) is proposed, and particles usage abilities are enhanced to improve filter response velocity of abrupt faults and reduce the computational time-consuming. Another improved PF algorithm involved particle depot is proposed, and the particles near the real state on the abrupt point are selected from the depot. A scheme to restrain particle drift is presented when some estimated state coupling is taken into account. From the aspects of diagnostic accuracy and computational efforts, comparisons of three PF algorithms for turbojet gas path fault diagnosis are given.(5) A rapid prototyping evaluation method for turbojet engine gas path fault diagnosis system is proposed based on analysis of diagnostic functional requirements. The modularization route is adopted to construct turbojet engine diagnostic rapid prototyping platform, which is based on virtual instrument language LabVIEW and rapid prototype NI CompactRIO. The platform mainly includes system monitoring workstation, engine simulator, signal interface unit, and rapid prototype diagnostic unit. The diagnostic unit is designed with MathScript, sub VI and DLL, and the designed algorithms can be downloaded to the platform for evaluation. The experiments of turbojet engine gas path fault diagnosis on rapid prototyping platform are carried out and the effectiveness of the proposed methods is validated.