A hash mapping method using cell vectors in Monte Carlo code RMC. (15th September 2021)

• Record Type: Journal Article
• Authors: Guo, Xiaoyu; Shen, Pengfei; Li, Kaiwen; Huang, Shanfang; Liang, Jingang; Wang, Kan
• Is Part Of: Annals of nuclear energy. Volume 160(2021)
• Publication Date: 2021-09-15
• Subjects: Cell vector -- RMC -- Hash -- Shift; Nuclear energy -- Periodicals; Nuclear engineering -- Periodicals; 621.4805
• DOI: 10.1016/j.anucene.2021.108395
• Languages: English
• ISSNs: 0306-4549

Abstract

Highlights: A hash mapping method is proposed to match CSG cells with distributed parameters for MC codes. A shift hash function is developed to convert cell vectors to hash values without hash collisions. A reconversion method converts the hash values to the corresponding cell vectors. Noticeable calculation acceleration and memory savings are achieved. Abstract: As high-resolution nuclear reactor simulations have become increasingly popular, the repeated structure geometry employed in Monte Carlo (MC) transport codes requires modification in order to represent nonrepetitive parameters, such as the temperature distribution and material densities in burnup. To address this issue, a hash mapping method using cell vectors is developed to build a one-to-one relationship between the cells and the parameters. In this method, the shift hash function is proposed to calculate the hash values; this function is intuitive and does not lead to hash collisions. Then, the 〈 cell vector → hash value → cell parameter 〉 mapping relationship is established to rapidly obtain the cell information. A reconversion method for the shift hash value is also developed to convert the hash value back to the cell vector. The hash mapping method is implemented in an MC code (RMC) and is integrated with the burnup, thermal–hydraulic feedback, and neighbor cell capabilities. A comparison between the hash mapping method and the usual methods shows a clear efficiency enhancement, with a speedup ranging from 1.06 to 1.64. Implementing the reconversion method saves over 100 megabytes of memory for the neighbor cell capability.