Scientists have created an AI capable of preventing malfunctions in thermonuclear reactors

Scientists who work on fusion machines called tokamaks are at increased risk

failures during experiments to maximize the power of thermonuclear fusion. Such experiments make it possible to create thermonuclear fusion on Earth, which powers the sun and stars.

Thus, scientists need to know exactlylimits of the tokamak in order to use it at maximum power, but without creating a threat to life and health. This capability is crucial for the largest nuclear fusion facility, ITER, which will open in France in the coming years.

Synthesis reactions combine light elements intoa plasma form - a hot charged state of matter - to generate a huge amount of energy. To avoid the sudden release of energy stored in the plasma, the researchers used artificial intelligence that could warn scientists in time about reaching the upper limit of tokamak power, after which an explosion could follow.

During the training process, the neural network received data aboutthousands of previous experiments on fusion reactors, as well as the rules for working with them and the incidents that occurred during the experiments.

“Surprisingly, a relatively simple machine learning model was able to accurately predict the complex behavior of fusion plasmas.”

Yichen Fu, lead author of the study

Previously physicistshave learnedcontrol the nucleus of an atom using an electric field. An error during the experiment helped make this discovery.