Health Parameters Estimation Based on Hybrid Model For Turbo-shaft Engine

摘要

A hybrid model is proposed for health parameter estimation of turbo-shaft engine. The hybrid model is composed of the physical model based on Kalman filter and the empirical model based on neural networks. Gaussian mixture model (GMM) is used to cluster the flight data, on-board neural network training is realized in real-time. GMM not only reduces the data storage capacity, but also accelerates speed of neural network training after the end of the flight. Updated neural network model makes the Kalman filter expanding operation range. Simulation results show that the hybrid model estimate health parameters in flight envelope with satisfactory accuracy.