Prediction of Critical Heat Flux of Wire-Wrapped Rod Bundle Based on Artificial Neural Network

Abstract Accurate prediction for the critical heat flux of rod bundle assemblies is important for reactor safety assessment. However, the existing prediction methods fail to unify their generalization and accuracy. The theoretical prediction models are unable to provide a satisfactory estimation. while the empirical methods that have been proposed for several decades could give results precisely, their suitable conditions are strictly limited, causing the endless development of different correlations for every specific rod bundle design. An exploration of new methods has been undertaken to provide a better method for critical heat flux prediction based on an artificial neural network since neural networks have gained recent attention for their outstanding performance in predicting complex problems. This study discusses the prediction performance of a traditional polynomial fitting method, a neural network method, and a hybrid method combined with empirical correlations and a fully connected neural network regression model. The analysis concludes that when artificial neural networks replace empirical models in prediction tasks, achieving equivalent prediction accuracy is easily accomplished; however, there is a need for improvement in error characteristics. Besides, the hybrid method was validated as a superior approach, demonstrating its effectiveness in enhancing prediction performance across various input variable regions. The ratio of predicted values to true values is well-maintained, indicating that this proposed method yields improved error characteristics, and consequently, more accurate critical heat flux estimations. The research will help improve the feasibility of the application of machine learning algorithms such as neural networks in the nuclear energy industry.