Europe’s fusion modellers advance AI and digital twins in Garching

ASDEX Upgrade’s vacuum vessel at IPP Garching, where E-TASC simulations target reliable plasma predictions for ITER and beyond.

(Image courtesy of Max Planck Institute for Plasma Physics)

More than 100 researchers gathered at the Max-Planck Institute for Plasma Physics in Garching last month for the second general meeting of E-TASC, EUROfusion’s programme for building the computational infrastructure behind fusion reactor development. The meeting reviewed progress across predictive simulation, high-performance computing, and software standardisation, and set priorities for the next EU research framework.

The technical focus landed squarely on verification, validation, and uncertainty quantification. VVUQ has become a central concern for the European fusion modelling community because simulations that perform well against existing experimental data still need to demonstrate reliable predictive behaviour at reactor scale – conditions no current machine can fully reproduce. Getting that right matters for ITER’s research plan and more so for the design decisions that will shape DEMO.

Alongside VVUQ, participants reviewed progress on what E-TASC calls the Digital Twin Environment, a multi-fidelity framework that integrates physics and engineering software with real-time data feeds and advanced visualisation tools. The aim is to make design iteration faster and more responsive to experimental results, rather than treating simulation and hardware development as sequential steps.

Underpinning both efforts is a push to standardise how data is managed and shared across the European programme. The meeting reinforced commitment to FAIR principles – findable, accessible, interoperable, reusable – and to structured data management built around the IMAS standard, a common format for plasma physics datasets that enables meaningful cross-machine comparisons. Without that foundation, the large-scale AI applications the community is moving toward have no clean data to work with.

On AI, the programme is pursuing surrogate modelling to reduce the computational cost of high-fidelity simulations, workflow automation, and AI-assisted code development. These are active workstreams rather than aspirations, and the structured data effort is what makes them viable at the scale the programme needs.

E-TASC’s priorities heading into the 10th EU Framework Programme include tighter integration between theory, simulation, and engineering, systematic VVUQ implementation across all projects, and expanded AI tooling. The meeting’s conclusions feed directly into planning for ITER and the longer development timeline for a fusion power plant.

Stay ahead in the fusion revolution explore more breakthroughs from leading innovators in clean energy technology.