Dynamic performance simulation and control of an aeroengine by using NARX models

Dynamic aeroengine model plays a key role in the design of engine control systems. Moreover, modelling of the engine using performance simulations is an important step in the design process in order to reduce costs, decrease risks and shortening development period. Parameters such as engine spool speeds, vibration, oil temperature, exhaust gas temperature, and fuel flow are often used to estimate performance in gas turbine engines. In this study, two artificial neural network methods were used for the prediction, under transient operations, of one of the most important engine parameters, the Exhaust Gas Temperature (EGT). The data used for model training are time series datasets of several different flight missions, which have been created using a gas path analysis, and that allow to simulate the engine transient behaviour. The study faces the challenge of setting up a robust and reliable Nonlinear Input-Output (NIO) and a Nonlinear AutoRegressive with eXog nous inputs (NARX) models, by means of a good selection of training. At the end of the study, two network that predicts the engine EGT in transient operations with the smallest error have been identified.