update

Author: delcmo

paper/paper.md

@@ -50,7 +50,7 @@ The demand for high-performance computational fluid dynamics and multiphysics so
 
 Fusion energy development draws from a wide range of disciplines to describe design and to develop a functioning system. One challenging engineering task is to develop the fusion core component known as a blanket. Because this component surrounds the burning plasma and must absorb almost all of the power from nuclear reactions, it must breed fuel, provide nuclear shielding, and energy deposition. Molten salt (MS) is a primary choice for cooling the blanket. A “salt blanket” in fusion energy is a layer of molten salt surrounding the fusion plasma. The molten salt acts as both a coolant and a material for neutron absorption, which is essential in fusion reactions. It absorbs the high-energy neutrons produced by fusion, reducing the wear on reactor components and converting some of the energy into heat for electricity generation. Molten salts have low electrical and thermal conductivity experiencing lesser electromagnetic forces but are still turbulent. Heat transfer degradation in a MS flow caused by the reduction of turbulence by a magnetic field is a possible limitation for MS blanket [@Smolentsev01042005].
 
-Two approaches are commonly adopted to model MS flows exposed to a magnetic field: high-fidelity modeling (large eddy simulation [LES] or direct numerical simulation [DNS]) and Reynolds-averaged Navier-Stokes (RANS) turbulence models. LESs is a powerful tool that can resolve turbulences at temporal and spatial scales and are required to fully understand the behavior and to create accurate closure models. The design of a blanket with LES is not possible because of current HPC limitations. Design optimization often requires multiple simulation runs to investigate performance under various conditions. The main technique that reduces the computational requirements of the analysis is RANS turbulence model. This approach filters out the instantaneous velocity component, and the influence of the turbulence is modeled solely by the closure models. The modeling of turbulence is a complex problem, and many turbulence models are available as described in the literature [@Chen_2022][@Menter1992ImprovedTK], albeit with many limitations [@10.1023/a:1022818327584]. Furthermore, these models are not readily applicable to the MHD flows and would require modifications [@Smolentsev2002] because MHD effects introduce additional terms in the turbulence balance equations.
+Two approaches are commonly adopted to model MS flows exposed to a magnetic field: high-fidelity modeling (large eddy simulation (LES) or direct numerical simulation (DNS)) and Reynolds-averaged Navier-Stokes (RANS) turbulence models. LESs can resolve turbulences at temporal and spatial scales to the expense of large HPC resources. The design of a blanket with LES is not possible because of current HPC limitations. Design optimization often requires multiple simulation runs to investigate performance under various conditions. The main technique that reduces the computational requirements of the analysis is RANS turbulence model. This approach filters out the instantaneous velocity component, and the influence of the turbulence is modeled solely by the closure models. The modeling of turbulence is a complex problem, and many turbulence models are available as described in the literature [@Chen_2022][@Menter1992ImprovedTK], albeit with many limitations [@10.1023/a:1022818327584]. Furthermore, these models are not readily applicable to the MHD flows and would require modifications [@Smolentsev2002] because MHD effects introduce additional terms in the turbulence balance equations.
 
 VERTEX-CFD is a new open-source package designed to address the aforementioned challenges by leveraging and integrating artificial intelligence and machine learning (AI&ML) tools to enhance current turbulence models from high-fidelity datasets, and relying on a robust multi-physics solver that scales on HPC platforms.