Alexander THEISS, Stefan HEIN, Alexander WAGNER, Paul HOFFMANN, Ana Teresa FAUSTINO
DOI Number: N/A
Conference number: HiSST-2025-075
The research presented in this paper represents a foundational first step toward the strategic objective of creating a powerful, multi-fidelity, and multi-modal surrogate tool that provides a holistic and computationally efficient yet physically sound solution for hypersonic transition prediction. This objective is driven by the fact that accurate prediction of laminar-turbulent transition is critical for hypersonic vehicle design. Yet established physics-based methods like Linear Stability Theory (LST) are computationally intensive and require significant expert intervention, hindering their use in automated design cycles. The development of surrogate models to overcome these limitations is particularly challenging for hypersonic boundary layers; for blunt-nosed vehicles, strong non-similar flow effects caused by the entropy layer render traditional, simplified profile parameterizations inadequate, complicating the surrogate modeling process. To address these challenges, this paper presents a methodology for creating high-fidelity surrogate models by training them exclusively on a comprehensive database from laminar simulations of a 7° blunt cone, focusing on the dominant second Mack mode instability. Two surrogate modeling
frameworks are employed and compared: a Radial Basis Function (RBF) interpolation model and an eXtreme Gradient Boosting (XGBoost) machine learning framework. The performance of both models is validated against flight test data from the MF-1 experiment, demonstrating excellent agreement with the N-factor envelopes derived from direct LST. Furthermore, a quantitative assessment of computational performance reveals a key advantage of the surrogate approach, with both models providing predictions significantly faster than direct LST and the XGBoost model being the computationally most performant.