Data-driven models in fusion exhaust: AI methods and perspectives

• DOI: 10.1088/1741-4326/ad5a1d
• Publication Date: 2024-06-20T22:22:21Z
• Authors: S. Wiesen; S. Dasbach; A. Kit; A.E. Jaervinen; A. Gillgren; A. Ho; A. Panera; D. Reiser; M. Brenzke; Y. Poels; E. Westerhof; V. Menkovski; G.F. Derks; P. Strand
• Keywords: Computer and information sciences; Modeling; modeling; QC770-798; Exhaust; Physical sciences; Computational Mathematics; machine learning; Nuclear and particle physics. Atomic energy. Radioactivity; exhaust; Subatomic Physics; Computer Science; Machine learning; SDG 7 - Affordable and Clean Energy; AI methods

Abstract A review is given on the highlights of a scatter-shot approach developing machine-learning methods and artificial neural networks based fast predictors for application to fusion exhaust. The aim enable facilitate optimized improved modeling allowing more flexible integration physics models in light extrapolations towards future devices. project encompasses various research objectives: (a) developments surrogate model power & particle exhaust plants; (b) assessments time-dependent phenomena plasma-edge; (c) feasibility studies micro–macro discovery plasma-facing components surface morphology durability; (d) enhancements pedestal databases through interpolators generators exploiting uncertainty quantification. Presented results demonstrate useful applications intelligence schemes, enabling an unprecedented combination both accurate simulation.