Gas Path Health Parameters Filtering Estimation For Turbo-fan Engine

Fault diagnosis for aero-engine is the core part of engine health management system, which isalso an important prerequisite to reduce maintenance cost and ensure flight safety. Based on thecomponent-level model of a turbofan engine, gas path component health parameters estimation isrealized from linear and nonlinear filtering.First, engine state variable model is established based on the component-level model. Linearkalman filter is designed to estimate the gas path component health parameters online and theabove-idle self-tuning on-board real-time engine model is constructed. A simplified startup model andafterburner model is established, the full state simulation of on-board engine model is realized.Considering the low accuracy and the strong coupling of linear kalman filter applied to enginegas path component health parameters estimation, EKF and UKF are introduced for gas pathcomponent health parameters estimation. According to the strong nonlinearity of engine and the actualnon-Gaussian noise, particle filter is proposed, which uses EKF to update particles and generate theimportant density function, therefor, the filter can avoid the particle degeneracy to some extent and thegas path health parameters estimation is realized in non-Gaussian noise with a better accuracy.On this basis, three improved particle filters are proposed. Considering the large computationalburden of particle filter, adaptive particle filter is proposed, which can reduce the computationalburden and ensure the filtering accuracy. The strong track particle filter can enhance the ability oftracking abrupt states and improve the stability and convergence of the filter, which also can ensurethe completion of the parameters estimation within a sampling period. The particle filter which adjuststhe important weights based on the neural network can effectively avoid the particle degeneracy ofnormal particle filter and the poor particle phenomenon of resample based on weights fission andparticle adjustment, thus the health parameters estimation accuracy is improved.Finally, particle filter is tested on the rapid prototyping platform. The simulations andverifications of adaptive particle filter and strong track particle filter with hardware in loop arerealized in the engine health management platform based on the LabVIEW environment and theembedded controller CRIO.