DOI: 10.30518/jav.1972712 ISSN: 2587-1676

Constraint-Based Wing Planform Sizing of a Fixed-Wing UAV Using an AVL–ANN Surrogate and Artificial Bee Colony Optimization

Mustafa Çavdar, Reçep Kaş, Seda Arık Hatipoğlu, Mehmet Konar
Show PDF Cite
Choosing a wing for a fixed-wing unmanned aerial vehicle (UAV) is seldom about maximizing one aerodynamic figure of merit alone. The planform also has to stay stable and live within the limits of the propulsion system. This paper sizes the wing at the conceptual stage from two variables, the span (b) and the chord (c), through a constraint-based workflow. Each candidate first passes a fast feasibility screen for the stall limit, stall speed, wing loading, and power demand under a fixed mass model, a 50 km/h cruise, the motor power ceiling, and standard atmospheric conditions, with a further Reynolds-based realism filter. Wings that survive are analyzed in AVL, a Vortex Lattice Method (VLM) solver. For each one an operating point that meets the required lift is selected, and the lift-to-drag ratio (L/D) and the directional-stability derivative (Cnβ) are read at that point. A single feed-forward artificial neural network (ANN), a machine-learning surrogate mapping [b, c, AR, S] to [L/D, Cnβ], accelerates the search, after which an artificial bee colony (ABC) algorithm maximizes a multi-objective index built on that surrogate. The direct AVL trade-off and the ANN+ABC search point to the same wing, b = 1.90 m and c = 0.24 m (S = 0.456 m², AR = 7.92), giving L/D ≈ 16.61 and Cnβ ≈ 0.241 at α = 12°, with every feasibility constraint met. Under leave-one-out cross-validation the surrogate reproduces the AVL results with R² = 0.990 for L/D and R² = 0.998 for Cnβ, and the ABC search returns the same geometry in all 30 independent runs. A Monte-Carlo study shows, however, that the selected planform keeps only a 2–7% margin against wing loading, stall speed and installed power, and satisfies all four feasibility conditions simultaneously with a probability of 41.8% under the assumed parameter uncertainties. Directional stability and aerodynamic efficiency can therefore be balanced, and the most suitable planform identified in a systematic and repeatable way.

More from our Archive