.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
.png)
Minisymposium Presentation
Developing Turbulence Models with Mhd for Fusion Engineering
Presenter
Dr. Katarzyna (Kate) Borowiec is a R&D Associate Staff Member – System and Data Analytics Engineer in the Nuclear Energy and Fuel Cycle Division at Oak Ridge National Laboratory (ORNL). She received her PhD (2021) and MS (2017) in Nuclear, Plasma and Radiological Engineering from University of Illinois at Urbana-Champaign. She has extensive experience with thermal-hydraulics modeling of the nuclear reactor systems. She was involved in many projects in the area of validation and verification, uncertainty quantification and sensitivity analysis, advancing development of statistical methods for nuclear reactor safety applications. Her research takes advantage of reduced order modeling, data analysis, machine learning and aims at combining data-driven approaches with physics-based modeling. Her research interests include software development, high-performance computing, computational fluid dynamics, machine learning and data-driven modeling approaches.
Description
The fusion device optimization will require accurate reduced order models to explore design space and identify a conceptual design. Existing system codes used for fusion applications rely on scaling laws that do not have desired accuracy. There is a need for knowledge transfer from high-fidelity to low-fidelity models producing computationally efficient and accurate reduced order representation. The AI/ML approaches are a perfect candidate for this task. One of the specific challenges of the blanket design in magnetic confinement fusion are MHD effects present in the coolant and/or breeder. The MHD effects have significant influence on pressure drop and heat transfer both being extremely influential in system design. These effects are also difficult to model often requiring direct numerical simulations and fine mesh resolution to capture steep wall gradients. Fortunately, this analysis can be replaced with lower-fidelity approaches such as RANS with appropriate turbulence models that capture MHD effects on the mean flow. However, the 3D turbulence models for MHD flows are not available. As part of the VERTEX project, we have developed an AI/ML model with LES database that can capture the influence of the MHD effects on turbulent flow showcasing a successful knowledge transfer from high-fidelity to low-fidelity models.