Explainable Aerodynamic Design Framework for Tandem-Wing UAV Based on BO-xRFM

Tandem-wing aircraft boasts broad application prospects in low-speed flight scenarios due to its merits including compact structure, high lift-to-drag ratio and high stability. Nevertheless, the multi-parameter coupling effect between the front and rear wings brings considerable difficulties to the aerodynamic optimization design of such aircraft. Current relevant studies mostly concentrate on the exploration of aerodynamic mechanisms, while the practical design optimization methods are plagued by shortcomings such as poor pertinence, low efficiency, high computational cost and insufficient integration of advanced technologies, which are hard to meet the demands of practical engineering.To tackle these problems, this paper proposes a rapid aerodynamic design method for explainable xRFM-based tandem-wing unmanned aerial vehicle (UAV) with Bayesian optimization (BO), by integrating machine learning and explainable artificial intelligence (XAI) technologies. Firstly, model simplification and parametric modeling are performed on the tandem-wing configuration, nine key aerodynamic configuration parameters are selected as design variables, and Latin hypercube sampling (LHS) is adopted to split the dataset into training set, validation set and test set. Secondly, the vortex lattice method embedded in OpenVSP software is utilized to calculate the lift coefficient (CL) and drag coefficient (CD) at specific angles of attack. Subsequently, an xRFM surrogate model is constructed, hyperparameter optimization is accomplished through Bayesian optimization and compared with other machine learning algorithms. Finally, the Shapley Additive Explanations (SHAP) method is applied to quantify the contribution of each design variable, explore the law of parameter interaction, and verify the effectiveness of the method via local interpretation. The results indicate that this method can effectively cut down the computational cost in the initial design stage, achieve efficient and accurate optimization of the tandem-wing aerodynamic configuration, and solve the "black-box" problem of machine learning models, providing reliable technical support and theoretical reference for the aerodynamic design of tandem-wing aircraft.