Using Artificial Intelligence to face nuclear fusion challenges
Yoeri Poels developed data driven tools to improve control and insight in fusion experiments, supporting clean energy progress.
Understanding and controlling fusion energy is one of the most complex challenges in modern science. Experiments produce massive amounts of data. The behavior of plasma, an extremely hot electrically charged gas, is notoriously difficult to predict. his complexity makes it hard to run experiments efficiently, interpret results, or prevent sudden failures that can damage equipment.
To help tackle this problem, PhD researcher Yoeri Poels developed smart data driven tools that assist scientists in analyzing and controlling fusion experiments. His work supports the safe and efficient development of fusion energy, a clean and potentially limitless power source for the future. He defended his thesis on Monday June 30.
Smarter tools for understanding fusion plasma
Fusion experiments often use a device called a , which uses powerful magnetic fields to hold extremely hot plasma in place and allow fusion reactions to happen safely.
By analyzing large amounts of data from tokamak fusion experiments, used machine learning to create faster simulation models that save time.
He developed more robust monitoring systems capable of detecting subtle changes in the plasma as they happen.
Additionally, he introduced new methods to recognize and study dangerous plasma instabilities, helping scientists prevent equipment damage and improve control within the tokamak.
These innovations combined support safer and more efficient fusion energy research.
Three practical challenges
In his thesis, Poels explored how artificial intelligence, specifically machine learning, can support fusion research. He developed new methods to address three important challenges:
- Faster plasma simulations
Fusion experiments often rely on detailed computer simulations, but these can take a long time to run. Poels created fast data based simulation tools that learn from past results. These tools are not meant to fully replace traditional simulations, but they can support quicker studies when time or computing power is limited. - Monitoring energy performance
Keeping energy well confined inside the plasma is essential for good fusion results. Poels built a tool that automatically detects how well the plasma is performing, even if some measurements are missing or faulty. It also tells scientists how confident it is in its predictions, which is important for real time decision making. - Understanding disruptions
Sometimes fusion experiments end in sudden disruptions that can damage equipment. These events are still not well understood. Poels used machine learning to find simplified patterns in large datasets, helping researchers better spot warning signs and analyze past disruptions.
Smart data tools accelerate fusion energy development
Yoeri Poels developed advanced machine learning tools to accelerate simulations, enable real-time monitoring, and improve prediction of disruptions in tokamak fusion experiments.
His work enhances the safety and efficiency of fusion research, contributing to the advancement of clean and sustainable fusion energy.
PhD researcher Yoeri Poels. Photo: Vincent van den Hoogen
Supporting the future of fusion
Poels鈥檚 work shows how combining smart data tools with fusion research can help solve some of the toughest problems scientists face in this field.
By improving how researchers understand and control the hot plasma inside tokamak reactors, his innovations make experiments safer and more efficient.
These advances bring us closer to making fusion energy a clean, reliable, and unlimited power source for the future.
-
Supervisors
Vlado Menkovski, Olivier Sauter (external)
Written by
More on AI and Data Science
Latest news
