Multimodel ELM-Based Identification of an Aircraft Dynamics in the Entire Flight Envelope

The development of a multiple model-based identification algorithm is addressed in this paper for nonlinear modeling of a conventional aircraft in the entire flight envelope. The dynamic model of an aircraft varies significantly depending on changes in the flight condition of the air vehicle including the altitude and the equivalent air speed. Therefore, the conventional identification approaches for generating a single nonlinear model with time-invariant parameters cannot be used in the entire flight envelope of an aircraft. Accordingly, a multiple model-based approach using nonlinear autoregressive exogenous neural networks is introduced in this paper as a powerful tool in identifying complex nonlinear dynamic systems. Different methods of validity function determination are introduced in order to aggregate the separate local models into a single model. The obtained results show that the proposed approach using the extreme learning machine-based validity function determination method has a significant capability to predict the aircraft outputs in various flight conditions. Further, the proposed identification scheme can be used effectively as a precise multistep ahead predictor of nonlinear multivariable systems outputs.