This paper presents the identification of the aerodynamic model of the "Flying-V", a novel aircraft configuration. The aerodynamic model is estimated using flight test data from a 4.6\% sub-scale model. The dataset includes longitudinal and lateral-directional maneuvers performed by both the pilot and the autopilot to excite the aircraft dynamic modes. The so-called Two-Step Method is used to decouple and simplify the aerodynamic identification problem; the state estimation step is performed by an Iterated Extended Kalman Filter, and the parameter-estimation step using ordinary least squares. A stepwise regression technique and previous knowledge from wind-tunnel tests are combined to select the model structure, and the identified model is validated using a third of the gathered data. The estimated models are parsimonious and considered adequate in terms of model fit, with a maximum relative Root Mean Square Error of 10% for the worst validation case. For the considered location of the center of gravity and flight conditions, the estimated aerodynamic derivatives confirm that the aircraft is longitudinally stable, both statically and dynamically; and that it is also laterally and directionally statically stable. The analysis of the dynamic modes of the sub-scale model showed stable short period and roll subsidence modes, a lightly damped Dutch roll mode, and lightly damped/unstable phugoid and spiral modes.

The aerodynamic model identification of a novel aircraft configuration known as the “Flying V” is presented. A global longitudinal aerodynamic model is estimated using static wind tunnel data from a 4.6% sub-scale model. The aerodynamic model structure, unknown a priori, is determined from the data using a modified stepwise regression technique. Orthogonal polynomial models using Multivariate Orthogonal Functions and non-orthogonal spline models in the angle-of-attack dimension are defined for the estimation of the measured aerodynamic coefficients. The estimated models are validated against a partition of the data not used for the estimation, which shows that an adequate model fit and good prediction capabilities are attained. Spline models achieve better results in terms of fitting and show a better matching between the estimation and validation data. All estimated models are considered adequate, with a maximum relative Root Mean Square error below 8% for the polynomial models and below 3% for the spline models.