Feasibility study on application of an artificial neural network for automatic design of a reactor core at the Kyoto University Critical Assembly. (January 2020)

• Record Type: Journal Article
• Authors: Kim, Song Hyun; Shin, Sung Gyun; Han, Sangsoo; Kim, Moo Hwan; Pyeon, Cheol Ho
• Is Part Of: Progress in nuclear energy. Volume 119(2020)
• Publication Date: 2020-01
• Subjects: KUCA -- Artificial neural network -- Reactor core design -- Research reactor -- Optimization; Nuclear energy -- Periodicals; Nuclear engineering -- Periodicals; 333.7924
• DOI: 10.1016/j.pnucene.2019.103183
• Languages: English
• ISSNs: 0149-1970

Abstract

Abstract: Designing reactor cores by means of an artificial neural network is a difficult challenge, because there are many variables in the core configuration. Especially, for designing a new type of reactor core with an artificial neural network, little (if any) previous data exists, and the appropriate number of results, such as multiplication factors and neutron fluxes, which require a large computational time for a single calculation, should be previously obtained for training the machine learning of the artificial neural network. This paper presents a feasibility study on the automatic design of a research reactor core (a simplified core based on the Kyoto University Critical Assembly) using an artificial neural network. By imitating conventional design procedure, a way to design the core is developed by means of the artificial neural network and automatic machine learning. After setting a design goal of the reactor core, the fuel assembly and core are designed by the proposed method and compared with those designed by conventional design procedure. The results reveal that the reactor core designed by the proposed method performs well and will, therefore, provide a clue to innovation in future reactor design with artificial intelligence. Highlights: A design method of KUCA core using AI is proposed. The feasibility of the core design based on AI is investigated. The feasibility analysis provided in this study will contribute to introduce AI into the field of the nuclear reactor design.