DNS of Wall-Bounded Magnetohydrodynamic Turbulence at High Reynolds Number

PI Myoungkyu Lee, University of Alabama
Lee Incite 2023

Denoising with DeepImagePrior: DNS + white noise.

Project Summary

The high-fidelity data generated with this INCITE allocation will reveal the spectral behaviors of turbulent kinetic energy (TKE) as functions of fluid speed, the strength and direction of the magnetic field, and the wall-normal distances.

Project Description

Understanding the fundamental physics of wall-bounded magnetohydrodynamic (MHD) turbulence is the key to developing multiple engineering applications, such as liquid metal blankets in nuclear fusion reactors. The study of such applications is challenging because operating conditions are often at a high Reynolds number. Therefore, either very high-fidelity experimental measurement systems or leadership-scalecomputing systems are required.

This project will perform direct numerical simulations (DNS) of wall-bounded MHD turbulence at high Reynolds numbers. Since DNS requires resolving the entire spectrum of length and time scales of turbulent flows, it is impractical to use for complicated scenarios. Instead, the researchers will study incompressible canonical channel flow at different flow speeds and magnetic field strengths & directions in both transient and statistically stationary states.

Primarily, this project will provide the life-cycle of TKE from its production and dissipation at different length scales while displaying the influence of large-scale motion on the near-wall flows in MHD turbulence. Additionally, the resulting high-fidelity data will be useable for developing and improving reduced-order models, such as RANS (Reynolds-averaged Navier-Stokes) models, subgrid stress models for large eddy simulation, and machine learning surrogates.