Developing correlation for prediction of departure from nucleate boiling heat flux in subcooled flow boiling using artificial neural network-aided framework

Departure from nucleate boiling (DNB) in subcooled flow boiling leads to an abrupt rise in the heat transfer surface temperature within a short time. In light water reactors, knowledge of DNB heat flux is essential to prevent the catastrophic failure of heating elements. However, accurately predicting DNB heat flux remains challenging due to the underlying physical complexities. Existing empirical correlations, mechanistic models, and AI-aided models in the open literature continue to exhibit certain limitations, such as disagreements on the dominant DNB mechanism or the inherent lack of transparency in black-box AI frameworks. In this study, an artificial neural network (ANN) is utilized to analyze the interrelationships between DNB heat flux and eight influential dimensionless parameters gathered from extensive literature resources. Subsequently, a minimal number of the most critical dimensionless quantities were extracted to develop a novel empirical correlation. The results demonstrate that the proposed model possesses superior characteristics, including high robustness, mathematical simplicity, broad applicability across diverse operating conditions, and independence from thermal–hydraulic parameters.