Researchers from the US Argonne National Laboratory have developed a “digital twin”
According to researchers, currentlythe neural network can diagnose and control the operation of the main components of a nuclear power plant in real time. Researchers are now training AI to make recommendations to service personnel.
The researchers note that although the system is based on data from the work of constructively new modern nuclear reactors, it can be easily adapted to the operation of any nuclear power plant.
Nuclear power plants produce a largeamount of electricity without polluting the planet. But such stations are expensive. First of all, because they require constant monitoring and maintenance to ensure a constant flow of energy and safe operation.
A typical nuclear power plant has hundreds or even thousands ofsensors that control the operation of the system. The work of checking each sensor, as well as the performance of system components such as valves, pumps, heat exchangers, is currently left to the personnel who walk the shop floor. This work can be replaced by a neural network, scientists say.
Operating and maintenance costsmaintenance is very important for nuclear units, which currently require a large number of crews and extensive maintenance. We believe that autonomous operation can help increase their profitability, as well as be useful in the deployment of advanced reactor concepts.
Roberto Ponchiroli, Chief Nuclear Engineer, Argonne National Laboratory
Read more:
On the pyramid in China found a portrait of the "king of ancestors". He ruled over 4,000 years ago
The length of the earth day is increasing. Scientists don't know why
Engineers have created a "perpetual motion machine" for portable atomic clocks