Reproduce “Physics-Based Machine Learning Closures and Wall Models for Hypersonic Transition-Continuum Boundary Layer Predictions” on OpenResearch

About this paper

Ashish S. Nair, Narendra Singh, Marco Panesi, Justin Sirignano, Jonathan F. MacArt

The study introduces physics-constrained machine learning closures and wall models to enhance Navier-Stokes predictions for hypersonic transition-continuum boundary layers. This approach extends continuum solver accuracy into non-equilibrium regimes, yielding solutions 5-15 times faster than Direct Simulation Monte Carlo (DSMC) methods and with improved fidelity over traditional models.

Key insights

• Enforcing physical constraints, such as trace-free viscous stress for monatomic gases, significantly improves the accuracy and thermodynamic consistency of learned anisotropic transport models.
• Accurate modeling of near-wall non-equilibrium effects through data-driven wall models that capture non-Maxwellian velocity distributions is critical for correct predictions of wall heat flux and viscous stress.
• Parallel training of neural networks across a range of Knudsen and Mach numbers substantially enhances the model's ability to generalize and interpolate effectively across diverse flow conditions.

Summary by alphaXiv.