PSFC hosts 1st IAEA workshop on Digital Engineering for Fusion Energy Research

Picture credits: Huihua Yang / PSFC

PSFC held the first IAEA Digital Engineering for Fusion Research Workshop, bringing together researchers from academia, national laboratories, and industry to discuss how digital engineering methodologies are transforming fusion energy research. The workshop focused on the integration of high-fidelity simulation, data-driven models, and digital twin technologies to enable more predictive, efficient, and robust approaches to fusion device design, operation, and control.

PSFC disruption group leader Dr. Cristina Rea served both as a member of the Programme Committee and as the local host of the workshop. The event was hosted at the Hacker Reactor in MIT’s iHQ and was simultaneously streamed online, enabling broad international participation and reinforcing its collaborative, multi-institutional scope. A poster session was also held at PSFC, providing an informal setting for in-depth scientific exchange across research groups.

Technical contributions covered diverse aspects of digital engineering for fusion, including high-fidelity multiphysics simulation, reduced-order and surrogate modeling, data assimilation, uncertainty quantification, and the development of digital twins for fusion systems. Participants represented a wide range of institutions from the United States, Europe, and Asia, spanning national laboratories, universities, and industry partners. Research contributions from PSFC spanned several groups, including the disruption group, the transport group, the LIBRA group, and the FESTIM team, highlighting PSFC’s broad engagement in digital engineering approaches for fusion energy research.

As part of the technical program, postdoctoral researcher Qiyun Cheng from the PSFC disruption group delivered a talk titled “A Cross-Machine and Parametric Neural Operator Surrogate Model for MHD Simulations”. The presentation introduced a fast, physics-consistent neural-operator-based surrogate model for nonlinear MHD dynamics that enables predictive modeling across multiple fusion devices and parameter regimes. By combining analytical geometry-normalized mappings with an equation-recast strategy for parametric extrapolation, the approach achieves cross-machine generalization and strong predictive capability using minimal training data. The resulting surrogate model can be deployed for rapid state prediction in plasma instability control systems or integrated into high-fidelity solvers as a preconditioner to accelerate large-scale MHD simulations.

The workshop highlighted the increasingly central role of digital engineering in fusion research, demonstrating how the tight integration of simulation, data, and reduced-order models can support predictive design, operational planning, and control of next-generation fusion devices. By convening experts across physics, applied mathematics, and industry, the event contributed to ongoing international efforts to accelerate fusion development through advanced computational methodologies.