The physics of this process is very complex – small changes can have serious consequences. “All this goes beyond our understanding, even with Google’s computer capabilities,” says Ted Boltz from the Google research division. Therefore, scientists combined the efforts of machine learning with the observations made by man, so that together they could find the right solution.
“We reduced the problem to a simple task: to find plasma states that experts in physics consider interesting, and not to break the machine,” Bolz said. “This is a classic example of cooperation between a person and a computer, which can be handled differently than together.”
The experiment proved that in this way the computer center of Tri Alpha works much faster, and the calculations that took months are accelerated to several hours. The algorithm discovered unexpected methods of plasma control, which allowed scientists to reduce energy losses by 50% and achieve an increase in the total energy of the plasma, which should reach the critical threshold in order for the synthesis to occur.
“The achievement of such results without the support of advanced computing takes years,” says Mihl Binderbauer, president and technical director of Tri Alpha Energy.
Recently, Tri Alpha announced the launch of a new Norman reactor, which is capable of operating at a temperature range of 50 to 70 million degrees Celsius. The amount of plasma that it can hold is 5 times that of the previous reactor.In June, scientists received the first plasma with it.