On-board Real-time Adaptive Modeling Of Aero-engine

On-board a daptive real time engine model is the base of Integrated Flight / Propulsion Control, condition monitoring and diagnostic technology. Adaptive real time modeling for aero-engine is studied in this thesis.The adaptive modeling based on neural network is studied in this paper. Because the outputs of aero-engine will bias their nominal value in any case of off-nominal work, there will be a mapping model between the biases of aero-engine outputs and component deterioration. So the component deterioration can be got by designing some proper neural network mapping module. However, the problem of over fitting is often meet in application of neural network, so the adaptive modeling based on least square support vector regression is studied. With the help of artificial intelligence technology, least square support vector regression is used to build an adaptive model, which is validated in full flight envelop.Then the adaptive modeling based on model is studied. According to the engine component level model, state variable model augmented with component deterioration is built by least square fitness methods, then the Kalman filter is designed and the adaptive model which adds the output error feedback is built. Considering the requirement in real-time, baseline model is constructed with neural network method, and the deviation state-variable model is constructed with least square fitness methods. After designs Kalman filter, the aero-engine s adaptive simplified model is built.