bg
News
16:11, 16 August 2026
views
16

Kurchatov Institute Develops Neural Network to Optimize Nuclear Safety

The technology could help prevent incidents at nuclear power plants.

Photo: nrcki.ru

Specialists at the National Research Centre «Kurchatov Institute» have developed a neural network that analyzes thermal-hydraulic parameters to model severe accidents at nuclear power plants. According to the developers, the task was particularly challenging: the model must accurately calculate changes in the neutron multiplication factor — a key parameter governing the chain reaction in uranium fission. This requires taking thousands of different parameters into account and substantial computing power.

The neural network can accumulate “experience” and use it to identify potentially hazardous nuclear states that might previously have been overlooked. This makes it possible to perform calculations more quickly and take the necessary measures to prevent incidents.

“The artificial neural network (ANN) model we developed is designed for a baseline state in which the fuel assemblies have not yet been damaged. We found that creating and training a single universal ‘large’ ANN model capable of working with all baseline states would require enormous resources. We therefore decided to develop separate models for the corresponding baseline states,” said Alexander Glazkov, a research scientist at the Department for Nuclear Safety Analysis of Nuclear Energy Facilities at the National Research Centre «Kurchatov Institute».
like
heart
fun
wow
sad
angry
Latest news
Important
Recommended
previous
next