Data-driven stochastic identification of a UAV under varying flight and structural health states

This paper presents the assessment of a stochastic data-driven global modelling framework for data-driven state awareness under varying flight (airspeed) and structural (damage) states. The proposed framework utilizes stochastic Vector-dependent Functionally Pooled (VFP) models that compound data pertaining to different flight states to develop a single “global” model representing the system dynamics. In this case, the system identification procedure is implemented for an Unmanned Aerial Vehicle (UAV) modeled in ASWING, which is a low-fidelity aeroelastic software for aerody- namic, structural, and control-response analysis of aircraft with flexible wings and fuselages. A series of dynamic simulations is performed for flight states corresponding to varying airspeed and damage size. Simulated signals are used for the stochastic system identification process involving both parametric and non-parametric analyses. The identified VFP model accurately represents the system dynamics and outperforms its non-parametric counterparts. The modelling method is shown to be effective and accurate in identifying the aeroelastic response for an expanded range of flight and structural states using sample data obtained within that range.