Multifidelity Data Fusion Method for Aerodynamic Heating Prediction of Hypersonic Vehicles

Accurate prediction of aerodynamic heating remains a critical challenge in hypersonic vehicle design. Current methodologies for acquiring heating data including computational fluid dynamics (CFD) simulations and wind-tunnel experiments involve inherent tradeoffs between measurement accuracy, spatial resolution, and implementation costs. Conventional multifidelity modeling approaches fail to ensure both comprehensive surface prediction improvements and reliable extrapolation to novel flight conditions due to experimental data scarcity. This study presents an augmented data fusion framework that synergizes data enhancement strategies with an innovative multifidelity U-NET (MF-UNet) architecture. The proposed framework achieves dual advancements through expanding limited experimental datasets via systematic data augmentation to enhance extrapolation robustness, coupled with developing an optimized multifidelity mapping mechanism via MF-UNet for high-precision heat-flux reconstruction. Validation on a hypersonic double-ellipsoid configuration demonstrates the framework’s effectiveness in integrating three sparse experimental datasets to generate complete surface heating profiles. Experimental results show significant improvements over baseline CFD solutions, achieving surface-averaged normalized root-mean-square error below 3% as compared with experimental measurements. The methodology successfully addresses challenges in multifidelity integration while enabling full-surface extrapolation, providing an engineering-viable solution for aerodynamic heating analysis.