Deep Learning models as an approach to nuclear fuel irradiation processes in Pressurized Water Reactors

Victor Casas Molina, Augusto Hernandez Solis, Ivan Merino Rodriguez, Pablo Romojaro

Research outputpeer-review

Abstract

ANICCA is the nuclear fuel cycle code developed by the Belgian Nuclear Research Centre (SCK CEN). Nuclear Fuel Cycle codes are of special importance for the assessment of the scenarios and the study of nuclear reactor fleet deployment and decommissioning. In said studies, the flow and inventory of Spent Nuclear Fuel (SNF) are of paramount importance, which are calculated through the irradiation module. In this work, a new approach to the irradiation module is presented. The approach is based on two direct neural networks which predict the final isotopic inventory in the SNF by using the initial fuel composition and the discharge burnup as inputs. These neural networks have been trained in Keras by a database produced with SERPENT2 continuous energy Monte Carlo transport code. Said models are dedicated to two of the most common nuclear fuel technologies for pressurized water reactors: UOX and MOX. Results showed a nice agreement between the new and the classical approach. At the same time, a quicker response in simulations was reported, especially for complex scenarios that involve multi-recycled fuel strategies (known as closed cycle). Thanks to the new method the prebuilt libraries needed in the previous module can be avoided, and so are the simplifications brought by the use of these.

Original languageEnglish
Title of host publication2022 41st International Conference of the Chilean Computer Science Society, SCCC 2022
PublisherIEEE - Institute of Electrical and Electronics Engineers
Number of pages8
ISBN (Electronic)978-1-6654-5674-6
ISBN (Print)978-1-6654-5675-3
DOIs
StatePublished - 2022
Event2022 - SCCC : 41st International Conference of the Chilean Computer Science Society - Universidad San Sebastián, Santiago
Duration: 21 Nov 202225 Nov 2022

Publication series

NameProceedings - International Conference of the Chilean Computer Science Society, SCCC
Volume2022-November
ISSN (Print)1522-4902

Conference

Conference2022 - SCCC
Abbreviated titleSCCC
Country/TerritoryChile
CitySantiago
Period2022-11-212022-11-25

ASJC Scopus subject areas

  • General Engineering
  • General Computer Science

Cite this