Parametric and nonparametric system identification of an experimental turbojet engine

This paper focuses on parametric and nonparametric system identification of an experimental turbojet engine. The input-output data of the jet engine is first acquired through an experimental investigation. Then, the obtained data are used for black-box modeling the jet engine. Two nonlinear identification approaches namely parametric and nonparametric methods are considered. An extensive investigation is carried out to obtain a suitable nonlinear structure for the parametric model. The nonparametric model identification is implemented using neural networks (NNs). Therefore, an appropriate configuration for the NN model is presented. Finally, model validity tests based on statistical measures and output prediction are carried out. It is demonstrated that the models obtained characterize the dynamic behavior of the system well over finite and infinite prediction horizons. Furthermore, the superiority of the nonparametric model compared to the parametric one is demonstrated.