actualités

Free-flight experiments can be considered as a realistic source of aerodynamic data-based knowledge, since the behaviour of a projectile is observed under real flight conditions. However, the experimental framework introduces several sources of uncertainty: the initial flight conditions are unknown, and only an initial guess of aerodynamics of the projectile is available. Additionally, the nonlinear model structure used in identification of aerodynamic behaviour is different from the structures typically used in guidance and navigation activities, which necessitates switching between model structures---depending on the task at hand.

Such a setting lays the groundwork for the following main contributions developed in the thesis:

A novel model structure is proposed: the nonlinear model equations are transformed into a quasi-LPV structure in a non rolling reference frame. The new structure resembles the ones used in guidance and navigation activities, while preserving the projectile's nonlinear behaviour. It is also much faster in computation time for the case of spin-stabilised projectiles, making it more efficient for simulations.

Influence of the aforementioned uncertainties on the quasi-LPV model was assessed using global sensitivity analysis: it has allowed to determine which parameters of the model can be deemed non-identifiable, and to gain further insight into the physical behaviour of the quasi- LPV model..
These developments could be used in future safety analysis preceding the flight tests.