Scientists, Amnesty International, the success of the success of the nuclear fusion – and it already works

Artificial intelligence gives a big boost to one of the largest nuclear fusion facilities in the world – but perhaps not the way you think.

In research published today in sciencesThe scientists at the Lawrence Levermore National Laboratory informed how the newly developed deep learning model predicted 2022 fusion experience In the national ignition facility (NIF). The model, which was appointed by 74 %, outperforms the ignition in that experiment, is outperforming traditional computing methods by covering more parameters with more accurately.

“What we are enthusiastic about in this model is the ability to make options explicitly for future experiences that increase the possibility of success every time,” a co -author of the study Kelly Hombird Tell Gizmodo during a video call. Humbird, who leads the NIF cognitive simulation group in NIF, added, given the amount of lands that we must cover, “so that a large and installed facility like NIF can only” make a few twenty of the attempts to ignite a year-even in fact a lot, given the amount of lands that we have to cover. ” Self -deficiency incense program.

Currently, nuclear power plants work to nuclear fission, which embodies the energy resulting from the division of heavy atoms, such as uranium. The researchers eventually want to shift towards nuclear fusion, a process that combines light hydrogen atoms to export huge amounts of energy. Fusion produces more energy and does not create harmful secondary products, and therefore the presence of integration as a reliable energy source that would benefit greatly from the transfer of our society to sustainable energy. Although the field has made some Promising progressand The consensus is that we are still far from implementing the nuclear fusion on a commercial scale.

NIF fusion experiences are driven by laser. First, the laser heats a golden cylinder called hohlraum, from which the flow of strong X -ray is emitted. Extreme temperatures are pressed on fuel granules that contain dotirium and striteium, which are hydrogen analogues used in fusion experiments. In the perfect scenario, this leads to the appearance of blood reactions to the production of more energy than the laser consumes.

Hombeard said that computer simulations cannot reliably predict all physics in this process. This is partly because the symbols are often simplified so that they are “accountable to the opposition”, but simulation itself can also make some errors. She added that even if you have taken all kinds of precautions, it still takes days until computers ended through the symbol.

Hombeard said the achievement of nuclear integration is similar to the expansion of a tall mountain. Computer simulation operations are similar to a “incomplete” map that researchers are supposed to learn how to reach peak – but this map can be a miracle of mistakes that may or not to be the product of research design. Meanwhile, the watch is suspended, and researchers must quickly decide whether they will take the height on that day and the tools they will use. Of course, every “height”, or attempt to ignite, burns a huge hole in the budget.

Thus, the Humbird team began searching for the maps drawing, collecting together “pre -collected NIF data, high -resolution physics simulation, and knowing the subject experts” to build a comprehensive data collection. After that, they downloaded data to modern giant computers, which conducted a statistical analysis of more than 30 million CPU watches.

“What we have found mainly is the distribution of things that are wrong (in) NIF,” Homber explained. “All the different methods in which we noticed the collapses. Sometimes the laser does not divorce how you asked for it. Sometimes your goal has defects that can cause things well.”

The model allows researchers to determine the effectiveness of their experimental design in a proactive manner, providing them with great time and money. Humbird uses the form to evaluate its own design from 2022 experimentWhich is precisely described by the specified operating results. In particular, Hombeard was happy with a vision that the subsequent amendments to the model physics increased its predictions from 50 to 70 %.

For Humbird, the strength of the new model is that it accepts and repeats the faults of the real world – whether it is a defect in the tool or search design or just a ridiculous trick of nature. At the same time, it is remembered that although rapid progress is exciting, things often take a lot of time and will lead to complete failure.

“People have been working on fusion for decades … We should not be pressured for times when things do not succeed,” Hombeard said. “The fact that we sometimes get 1 megapot from the return instead of two should not bother us, because not long ago, we haven’t only got 10 kilograms. It is a big step forward to search, and we hope that it will be a big step forward for clean energy in the future.”