Fusion Digital Twin Tokamak Enabled by AI-Machine Learning

Fusion Digital Twin Tokamak Enabled by AI-Machine Learning William Tang1, Tom Gibbs2, John Stone2, Eliot Feibush1, Peter Messmer2, Xishuo Wei3, Ge Dong1, Zhihong Lin3 1Princeton University/Princeton Plasma Physics Laboratory, Princeton, New Jersey, 08543, USA; 2NVIDIA Corporation, USA, 3UC Irvine, Irvine, California 92717, To help realize the Fusion Energy goal (e.g., April 2022 US DOE Bold Decadal Vision, for delivery of a Fusion Pilot Plant by 2035), modern tools which integrate engineering and supply chain data with science models that are significantly advanced with Artificial Intelligence (AI) enabled Machine Learning methods will be required. We have accordingly introduced an R&D effort with promising early progress on the delivery of a realistic “Digital Twin Tokamak” that has been accelerated by rapid advances in AI/Deep Learning for providing innovative near-real-time simulators accompanied by technological capabilities such as NVIDIA Omniverse™ — an open computing platform for building and operating applications capable of connecting to leading scientific computing visualization software. Working with the CAD files for the General Atomics (GA)/DIII-D tokamak in the US and incorporating recent near-real-time simulators for equilibrium evolution as well as for AI/HPC (High-Performance Computing)-enabled real-time dynamics, Princeton University has collaborated with GA and NVIDIA in using Omniverse to connect and accelerate state-of-the-art, synthetic, near-real-time HPC simulators to model magnetic fusion energy (MFE) devices and associated control systems. The aim is to deliver an interactive scientific digital twin of an advanced MFE tokamak that enables near-real-time simulation workflows built with Omniverse to help open doors to new capabilities for generating clean power for a better future.