Skip to main content

Using Neural Networks in Atomic Energy Thermophysical Problems (Review)

Abstract

The prospects for the implementation of modern digital technologies based on artificial neural networks (ANN) in the nuclear facility safety analysis are considered and the review of works on this topic is presented. The review includes publications in which neural networks are used in experimental studies and their result processing in such areas as heat transfer and hydrodynamics in single-phase media, in phase transformations, and hydrogen safety. Computational methods of machine learning based on artificial neural networks are successfully used to identify the relationship between the input and output system parameters without building complex mathematical computational models as well as for systems in which processes are characterized by nonlinearity and data uncertainty. However, a unified approach to the substantiation and assessment of the applicability of such methods has not yet been formed. As a result of the analysis, the need is shown for special requirements in order to assess the applicability of these technologies and substantiation (verification and validation) of calculation models based on neural networks.

Year of Publication
2020
Journal
Thermal Engineering
Volume
67
Start Page
497
Issue
8
Number of Pages
497-508+
Date Published
Aug
Type of Article
Review
ISBN Number
0040-6015 1555-6301
Accession Number
WOS:000755756700001
DOI
10.1134/s0040601520080108
Short Title
Using Neural Networks in Atomic Energy Thermophysical Problems (Review)
Alternate Journal
Therm Eng+