Document Type : Research Paper

Authors

• javad vosoughi ¹
• Naser Vosoughi ²
• Ali Akbar Salehi ³

¹ (Department of Energy Engineering (Sharif University of Technology

² Chair, Dept. of Energy Engineering, SHARIF Univ. of Technology

³ vice president of the Academy of Sciences

Abstract

In performing reactor core neutron calculations, the main cost is cell calculations to generate macroscopic cross-sections of FAs. The development and use of a cross-section library based on the reactor type is the main procedure to reduce calculations cost. In some cases, there is no library to perform specific analyzes, so it is necessary to do cell calculations. Investigating FAs vibration neutron noise, or coolant asymmetric distribution around FAs due to bowing, are such cases. In this research by using an ANN (Artificial Neural Network) instead of high-volume cell calculations, the time for calculations was significantly reduced for the VVER-1000 reactor. In the criticality calculations for BOC, the calculation time was reduced from 42 minutes with 32 computational cores to 1.3 minutes. Also, the results were of acceptable accuracy so that the difference in the value obtained for the critical boric acid by using ANN compared to cell calculation as a reference was less than 1% and the power difference was less than 5%.

Keywords

• Artificial Neural Network
• Macroscopic Cross-Sections
• Neutron Calculations

Main Subjects

• SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS