Research On Gas Path Fault Diagnosis For Turbojet Engine

Gas path analysis for aero-engine fault diagnosis is an important approach to condition-basedmaintenance, maintenance cost reducing and the engine security. Gas path fault diagnosis technique isresearched for a turbojet engine in this thesis.A non-linear Component model is established using component-modeling method. Test data istrained by least square support vector machine used to compensate the model outputs. Individualdifferences for the same type engines and the distortion of engine component performance inlife-cycle will decrease the engine model accuracy. Genetic algorithm (GA) is presented and used tomodify components characteristics for improving the accuracy of non-linear model.Small perturbation method is used to set up an augmented state variable model based on theprevious non-linear model. Self-tuning adaptive model of turbojet engine is obtained based onKalman filter to estimate the engine’s performance change in real time. Constrained Kalman filter isresearched with the pre-knowledge in state variable constraints added to Kalman filter. Comparedwith the Kalman filter, the constrained Kalman filter improves the accuracy of engine’s performanceestimation.The way to choose measured parameters for fault diagnosis is analyzed with simulation. Themeasured parameters subset is acquired by the similarity analysis of different engine faults. Thelinearity of measured parameters and the distortion parameters of component performance areresearched. The sensors for fault diagnosis are determined. A diagnosis structure combined with aperiodic updating baseline model in life-cycle and a real time model is provided to demonstrate faultdiagnosis of turbojet engine, which can diagnose sensor faults on the gradually degraded engine.