Surrogate Models for Aircraft Flight Control: Some Off-Line and Embedded Applications

The use of surrogate models is now very widespread in engineering activities to replace high-fidelity models, computation codes or simulators that are too complex or time consuming. This is especially the case in the aerospace field: although physical models are available, derived from aerodynamics, structural dynamics or flight mechanics, quite often they cannot be used just as they are. Hence, simplified representations need to be developed in order to achieve some tasks involving for instance optimization, parameter identification, embedded implementations, and so on. This paper illustrates some of these aspects in the field of aircraft flight control systems. After outlining the characteristics of the preferred surrogate models, the techniques that have been developed for constructing these models efficiently are presented. Then, three relevant offline and on-line applications are described, with the surrogate models being respectively used: to create parsimonious representations useful for building Linear Fractional Representations for analysis and design of control laws, as intermediate models in the modeling of rigid aircraft to facilitate the identification process of aerodynamic nonlinearities from flight tests, and to obtain embeddable models used for the virtual sensing of some flight parameters required to schedule the control/protection laws.