MULTI-OBJECTIVE DESIGN OPTIMIZATION OF REENTRY CAPSULE USING A MACHINE LEARNING-BASED AERO-THERMODYNAMIC SURROGATE MODEL

(Reentry capsule), (Deep-learning based model), (Machine learning), (Hypersonic aerothermodynamic) A B S T R A C TThis study presents a rapid and efficient analysis framework for the design of reentry vehicles operating in hypersonic flow environments, by proposing a multi-objective optimization method based on machine learning surrogate models.While computational fluid dynamics (CFD) provides highly accurate results, its high computational cost makes it impractical for early design phases where a wide range of flow conditions and geometric variations must be explored iteratively.To address this limitation, surrogate models based on Kriging and deep learning were developed using sample data generated from empirical aero-thermodynamic models that represent the aerodynamic and thermal characteristics of reentry vehicles.These surrogate models were then integrated with a genetic algorithm to enable rapid optimization.The results demonstrate that the surrogate models can reliably replace the empirical models and offer significant advantages in terms of computational efficiency, confirming their potential as practical design tools at the engineering level.