Can Artificial Intelligence run nuclear reactors?



The Chernobyl nuclear power plant today (REUTERS/Gleb Garanich REFILE-CORRECTING SPELLING OF CITY)
The Chernobyl nuclear power plant today (REUTERS/Gleb Garanich REFILE-CORRECTING SPELLING OF CITY)

The images of what happened in Chernobyl that the homonymous series shows us, remembering that it is a fiction built on real events, leaves several topics for analysis. One of them is the interpretation of the accident in the moments after. There you see the Shift Chief, Anatoly Diatlovin a control room with hundreds of readings from different instruments, surrounded by his half dozen assistants.

When the alarms start to sound, Diatlov worries about two things: lowering the control rods, in order to “turn off” the nuclear reaction, and cooling the core. His assistants, whom he sends to assess the situation and manually lower the bars, tell him that “there is no longer a reactor, there is no core”, but Diatlov does not believe them, and insists on his priorities, which at that point were unrealizable. .

Remembering again that it is a fiction, the scene shows a real fact: the difficulty of interpreting a complex situation that occurs when one has the temporal measurement of hundreds or thousands of variables, but many of them are outside their usual ranges at the same time. We can also remember that in the accident that occurred in Three Mile Island, in the United States, without the catastrophic consequences of Chernobyl, dozens of people struggled in the control room to understand what was happening.

The abandoned town of Pripyat where the Chenobyl nuclear accident occurred, in Ukraine (REUTERS / Gleb Garanich / File Photo / File Photo)
The abandoned town of Pripyat where the Chenobyl nuclear accident occurred, in Ukraine (REUTERS / Gleb Garanich / File Photo / File Photo)

Could Artificial Intelligence do better? Could it, perhaps not in a first step, operate a nuclear reactor autonomously, but assist the operator, particularly in the evaluation of incidental situations?

Well, many people believe so and there are important developments in this regard. As of 2015, approximately, academic studies began to skyrocket and papers published by researchers from prestigious universities, such as MIT and Michigan, while the International Atomic Energy Agency took note and organized seminars on the subject. And that development was reaching the main companies in the sector.

Artificial Intelligence applications that are already in use or are about to be used include: maintenance optimization, automatic analysis of inspection and monitoring data, neutron calculation, accident prevention. The first case is common to other industries, such as the oil industry, in which the analysis of the data of each machine with techniques of machine learning allows a diagnosis to be made for each one in real time, instead of subjecting them to maintenance every certain number of hours of operation.

The impact of a good calculation is important in the economics of nuclear reactors, since a more precise exchange of fuel leads to a better utilization and, with it, cost reduction.

The following examples, on the other hand, are already specific to the nuclear industry. Indeed, the calculation of the neutron flux both inside a nuclear reactor and in spent fuel pools has a long tradition; in fact, Henry Fermi He made those calculations in 1942 when the first man-made nuclear reactor at the University of Chicago went critical. Currently we can find the resolution of this problem with neural networks and other similar techniques. The impact of a good calculation is important in the economics of nuclear reactors, since a more precise exchange of fuel leads to better utilization and, with it, cost reduction.

But perhaps the strongest impact would be the one mentioned at the beginning, the analysis and interpretation of incidental or accidental situations and, perhaps, the handling of such situations, at some later stage. At present, there are already studies that show good performance on a laboratory scale to detect some significant events such as the loss of coolant from a reactor, and progress is being made with the digital twinssimulators that assist operators in real time.

Finally, the highly automated operation of nuclear reactors would be indispensable for their use in space propulsion, where a team of operators and engineers cannot be counted on to attend to the reactor, which, for example, would significantly reduce the travel time of manned missions loving You.

Thus, nuclear energy, artificial intelligence and space exploration would open a new stage for humanity.

KEEP READING:





Source link


Leave a Reply

Your email address will not be published.