1. Study on flow regime prediction model for water-cooled reactor core based on machine learning algorithms.
- Author
-
Ma, Yichao, Kong, Dexiang, Zhang, Jing, Wang, Mingjun, Tian, Wenxi, Wu, Yingwei, Su, G.H., and Qiu, Suizheng
- Subjects
- *
WATER cooled reactors , *MACHINE learning , *NUCLEAR reactor cores , *PREDICTION models , *LIGHT water reactors , *ADVECTION - Abstract
• This study utilizes the increasing flow regime experimental data to extend the prediction range and improve the accuracy. • Integrated flow regime prediction models considering vertical and horizontal flow direction were developed and evaluated. • The MLP, RF, and SVM algorithms were utilized to develop the flow regime prediction model respectively. • The flow regime prediction model shows better performance than existing models in three directions respectively. The nuclear reactor core is the pivotal component in a nuclear power plant to generate and transfer heat, so accurate prediction of reactor core flow and heat transfer characteristics is a crucial problem for the reactor system design. Flow regime is closely related to the thermal–hydraulic characteristics of two-phase fluid. Still, flow regime models used in thermal–hydraulic calculation codes rely heavily on earlier experimental data, featuring a narrow application range and unsatisfactory accuracy. To extend the prediction range and improve the accuracy of flow regime prediction by making full use of the increasing experimental data, integrated flow regime prediction models considering vertical flow direction for light water reactor (LWR) and horizontal flow direction for Canadian Natural Deuterium Uranium (CANDU) were developed and evaluated based on three standard machine learning algorithms in this study. Firstly, the experimental data of horizontal and vertical flow collected from the literature was modified and preprocessed to be training data. Then, the multi-layer perceptron (MLP) algorithm, random forest (RF) algorithm, and support vector machine (SVM) algorithm were utilized to develop the flow regime prediction model respectively. The performance of the three models was evaluated and compared and the results showed that the flow regime prediction model based on RF was the optimal model with higher prediction and more efficiency than the other two models. Finally, the flow regime prediction model was compared with existing models in three directions respectively, which indicated that the range and accuracy of the model based on RF were superior to the existing models. This study provides an implantable and scalable approach for flow regime prediction, and the application range and accuracy of flow regime prediction can be continuously improved with updating training data. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF