US EditionPublished • loading... • Updated
British breakthrough brings nuclear fusion a step closer
GyroSwin AI model reduces fusion plasma simulation from days to seconds, enabling rapid design iteration and cost reduction, UKAEA says millions of simulations are needed for fusion plant development.
- Researchers in Britain and Austria developed GyroSwin, an AI surrogate from UKAEA, Johannes Kepler University and Emmi AI that produces plasma-simulation results in seconds, replacing days-long supercomputer calculations.
- Engineers must heat plasma to around 100,000,000C and confine it with magnetic fields in tokamaks, while predicting turbulence requires five-dimensional simulations to track complex particle motion.
- Using extensive simulation archives, researchers trained GyroSwin on a training archive of detailed simulations, enabling it to reproduce physical flow structures shaping plasma turbulence.
- Design teams face millions of plasma simulations, Rob Akers, director of computing programmes at the UKAEA, said, and trained surrogates help answer many `what if` questions quickly, guiding work on MAST Upgrade and STEP.
- Experts caution that the AI surrogate models must still prove reliable beyond current experiments, and supercomputers remain essential to generate training data, but if dependable, surrogates could accelerate fusion R&D and enable near-carbon-free fusion power with less radioactive waste.
Insights by Ground AI
5 Articles
5 Articles
geeky-gadgets.comReposted by 
BizToc
BizTocThe Nuclear Fusion Breakthrough Scientists Once Said Was Impossible
What if the key to solving the world’s energy crisis was hidden in a process so ambitious, so technically demanding, that even experts once dismissed it as science fiction? For decades, nuclear fusion—the same process that powers the stars, has been heralded as the ultimate clean energy solution. Yet, the challenges of harnessing it on […] The post The Nuclear Fusion Breakthrough Scientists Once Said Was Impossible appeared first on Geeky Gadget…
Coverage Details
Total News Sources5
Leaning Left0Leaning Right1Center2Last UpdatedBias Distribution67% Center
Bias Distribution
- 67% of the sources are Center
67% Center
C 67%
R 33%
Factuality
To view factuality data please Upgrade to Premium
