The AI model was trained on real data from previous fusion experiments and can predict tearing mode instabilities 300 milliseconds before they occur. This is a significant advancement as previous studies were only able to suppress these instabilities after they happened. While the model is still in the early stages of fine-tuning, the researchers are hopeful that it could be applied to other reactors to optimize the reaction or harvest energy from it.
Key takeaways:
- Researchers from Princeton University and its Princeton Plasma Physics Laboratory have developed an AI model that predicts and prevents plasma instabilities in nuclear fusion reactors.
- The AI model was trained on real data from previous fusion experiments and can predict tearing mode instabilities 300 milliseconds before they occur, providing enough time to control the plasma.
- The AI model was successfully tested on a real reactor, the DIII-D National Fusion Facility in San Diego, demonstrating its potential for practical application.
- While the model is still in the early stages of fine-tuning, the researchers are hopeful that it could be applied to other reactors to optimize the reaction or harvest energy from it.