CVAE-Based Surface Flow Prediction for Separately Modeled Missile Body and Fin

The growing demand for advanced missile systems has intensified the need for accurate and efficient aerodynamic prediction methods. Traditional techniques, including semi-empirical models such as missile DATCOM, Computational Fluid Dynamics (CFD), and wind tunnel experiments, often involve trade-offs between accuracy, computational cost, and scalability. Recent advances in data-driven modeling, particularly those employing deep neural networks, offer a compelling alternative by enabling direct aerodynamic predictions from data. Among these, Conditional Variational Autoencoders (CVAEs) have demonstrated promise as generative models capable of capturing complex flow patterns through structured latent representations. This study presents a CVAE-based framework for predicting surface aerodynamic distributions of a missile configuration composed of a slender axisymmetric body and attached fins. To isolate the aerodynamic characteristics of each component, separate CVAE models are developed for the body and the fins, allowing for high-resolution flow reconstruction under varying flow and geometry conditions. The integration of body-fin interactions into a unified model is left for future work. The results highlight the potential of component-wise CVAE models in achieving accurate and efficient aerodynamic predictions for missile design applications.