DNS of Shock-Induced Instabilities Over Hypersonic Control Surfaces

PI Takahiko Toki, Purdue University
Co-PI Lian Duan, The Ohio State University
Award INCITE
Hours 150,000 Polaris Node-Hours | 315,000 Aurora Node-Hours
Year
Domain Engineering
Scalo Incite Graphic

Task 1: computational domain setup for DNS of a Mach 7.7 compression ramp, showing an instantaneous flow visualised by an iso-surface of Q-criterion. The iso-surface is colored by the streamwise velocity component, ranging from 0 to U8 (blue to red). L represents the length of the flat plate portion of the ramp. In (a) the numerical Schlieren visualization of the compression ramp. In (b-c) the instantaneous visualization by temperature contour normalized by wall temperature (Tw “ 293K). The yz plane shown is at streamwise location (b) x{L “ 2.0 and (c) x{L “ 2.3.

Project Summary

The INCITE proposal will generate a massive high-resolution simulation dataset of shock wave–boundary layer interactions using advanced supercomputers to better understand turbulence in hypersonic flows. The results will support improved turbulence models, hypersonic vehicle design, and future machine-learning–based CFD methods, with the data shared publicly for research use.

Project Description

The INCITE proposal aims to generate a Direct Numerical Simulation (DNS) dataset to study shock wave boundary layer interactions (SWBLI), starting with a ramp configuration and later the High-Speed Army Reference Vehicle (HARV). These simulations involve billions of grid points and millions of iterations, requiring top-tier supercomputers due to their computational demands. A highly scalable block-spectral code, upgraded for multi-platform use and tested on ALCF’s Polaris and Aurora systems, will perform the computations, demonstrating nearly perfect weak scaling up to 512 Aurora nodes—sufficient for the HARV’s fine mesh.

The resulting dataset will improve understanding of turbulence onset from shock waves on both simple and complex geometries, with implications for hypersonic vehicle design and thermal management. Analysis of subfilter-scale quantities will support the development of new LES and wall-modeled LES (WMLES) models, as wall-resolved LES remains unreliable under extreme flow conditions. The work builds on Dr. Scalo’s extensive experience in turbulence modeling and shock-capturing methods for hypersonic flows.

The dataset will be shared through publications, conferences, and a public SFTP server hosted by Dr. Scalo’s lab. It may also aid in developing or replacing CFD solvers with machine learning–based turbulence modeling methods.

Domains
Allocations

More Engineering Projects