NUST MISIS and the All-Russian Research Institute of Automatics named after Dukhov have developed an innovative approach using neural networks to predict the appearance of defects in steel structures of nuclear power plants (NPPs) under the influence of fast neutron flux. This was reported by the university's press service.
In the process of reactor operation, the cladding of fuel elements inside it gradually "swells" under the influence of radiation and negatively affects its strength and durability. This reduces the service life of fuel rod claddings and other structural elements made of heat-resistant steel.
To train our model, we considered dozens of materials that can "swell" up to 50%. As a result, we can predict "swelling" with high accuracy.
The new method allows predicting the effect of neutron irradiation on the structure of steel structures in various temperature ranges and material compositions with an accuracy of about 98.9%.
To create the neural network, scientists analyzed data from more than 1,100 irradiated steel samples and used them to train the algorithm. The calculation results helped to identify the influence of various impurities and alloying materials on the radiation resistance of austenitic steel. This data will help specialists in the nuclear industry to select the optimal compositions and structures of materials for fuel rod claddings and other reactor components.
Earlier, www1.ru reported that a series of installations for modeling accidents at nuclear power plants were created in Novosibirsk.
Read materials on the topic:
Reactor shaft parts installed at the 4th power unit of the Akkuyu NPP
