Fusion of Simulation and Experiment Data for Hypersonic Flow Field Prediction via Pre-Training and Fine-Tuning

TL;DR

This work addresses hypersonic flow field prediction over a compression ramp by fusing CFD simulations with experimental schlieren and pressure data through a pre-training and fine-tuning strategy. It compares encoder–decoder Vision Transformers and attention-based CNNs, showing that incorporating experimental data yields measurable improvements in pressure and density predictions and that the CNN architecture offers superior accuracy. A global stability analysis based on the high-fidelity model demonstrates the framework’s capacity to support engineering design while reducing computational cost. Furthermore, pre-training on CFD data and subsequent fine-tuning with schlieren images extend applicability to real-world conditions, reconstructing velocity and pressure fields beyond the initial parameter space. The results highlight a practical, transfer-learning-enabled pathway to accurate, efficient hypersonic flow modeling that couples simulation with measurement data.

Abstract

Accurate prediction of hypersonic flow fields over a compression ramp is critical for aerodynamic design but remains challenging due to the scarcity of experimental measurements such as velocity. This study systematically develops a data fusion framework to address this issue. In the first phase, a model trained solely on Computational Fluid Dynamics (CFD) data establishes a baseline for flow field prediction. The second phase demonstrates that enriching the training with both CFD and experimental data significantly enhances predictive accuracy: errors in pressure and density are reduced to 12.6% and 7.4%, respectively. This model also captures key flow features such as separation and reattachment shocks more distinctly. Physical analyses based on this improved model, including investigations into ramp angle effects and global stability analysis, confirm its utility for efficient design applications. In the third phase, a pre-trained model (using only CFD data) is successfully fine-tuned with experimental schlieren images, effectively reconstructing velocity fields and validating the transferability of the approach. This step-wise methodology demonstrates the effectiveness of combining simulation and experiment by pre-training and fine-tuning, offering a robust and efficient pathway for hypersonic flow modeling in real-world.

Figures (21)

• Figure 1: Schematic of fluid structures in a compression corner.
• Figure 2: (a): The experiment setup; (b): Geometry of the ramp configuration; (c): The schematic of the Ludwieg tube tunnel.
• Figure 3: The schematic of compression corner with flow simulation is composed of A, B, C three parts and the domain is enclosed by the red boundaries; The schlieren image only include B part; Green points: locations of Kulite pressure sensors. Unit: mm.
• Figure 4: Surface pressure coefficient comparison between DNS and experiment results of case 1, with $Ma$ = 4, $Re = 7.22\times 10^6$, $\alpha$ = 6°, $T_{\infty}$ = 70.7 K and $T_w /T_0$ = 0.99.
• Figure 5: The schlieren image before and after preprocessing for case 1, with $Ma$ = 4, $Re = 7.22\times 10^6$, $\alpha$ = 6°, $T_{\infty}$ = 70.7 K and $T_w /T_0$ = 0.99.