Machine learning prediction for thermal–hydraulic parameters of semicircular-fin fuel bundle in lead–bismuth fast reactor

Qian Li, Siwei Cai, Shengcai Zhang, XueChen Liu, Nianmei Zhang, Chen Hu, Xian Zeng, Yulong Mao, Weihua Cai,
Machine learning prediction for thermal–hydraulic parameters of semicircular-fin fuel bundle in lead–bismuth fast reactor,
Annals of Nuclear Energy,
Volume 224,
2025,
111696,
ISSN 0306-4549,
https://doi.org/10.1016/j.anucene.2025.111696.
(https://www.sciencedirect.com/science/article/pii/S0306454925005134)
Abstract: In this study, a machine learning approach is used to predict the thermal–hydraulic parameters of lead–bismuth eutectic flow in a new-style semicircular-fin rod bundles. With numerical simulation data, a novel micro segment method was used to extract data from 42 sub-channels and 19 fuel rods in the fluid domain and establish a database with 23,313 data points. Four machine learning models are used to predict the h and ΔP by normalizing the input parameters with model hyperparameter optimization. The results show that all of the four machine learning models have good prediction accuracy, with the error of less than 5% and MAPE were all within 1%. The performance of ANN model is better than that of the other three models in predicting new cases. It indicates that ANN model has a high accuracy in predicting thermal parameters under new cases, which verifies the applicability of the machine learning prediction method for multiple cases. This study confirmed the advantages of machine learning in predicting complex regular parameters, and proposed a new method for flow and heat transfer parameters prediction in the development of a subchannel analysis program for semicircular-fin rod bundles.
Keywords: Semicircular-fin rod bundles; Lead-bismuth fast reactor; Flow and heat transfer; Machine learning; Artificial neural network