ANN-based prediction of critical heat flux in cryogenic pool boiling

• An artificial neural network model is created to predict the critical heat flux in cryogenic fluid pool boiling. • The ANN model is trained with 590 datasets from the literature and self-conducted experiments. • The model can achieve a mean absolute error of 15.72% and a determination coefficient of 0.964. • Shapley additive explanations method is used to analyze how key factors affect the model prediction. Pool boiling critical heat flux (CHF) of cryogenic fluid is a crucial parameter for ensuring the safety and operational efficiency of application systems such as superconducting cooling, aerospace propulsion, and cryogenic transportation. To improve the accuracy of CHF prediction, this study developed an artificial neural network (ANN) predictive model which was trained and tested by 590 datasets from the literatures and self-conducted experiments. A SHAP (Shapley additive explanations) method was employed to analyze the influence pattern of key factors on the CHF. The model’s mean absolute percentage error (MAPE) is 15.72%, and the coefficient of determination (R 2 ) is 0.964, demonstrating better prediction performance compared to traditional theoretical models. Input parameters of the model include physical property parameter group, pressure ratio, orientation angle, subcooling degree, boiling surface area, and thermal conductivity. The physical property parameter group M-value shows a significant positive correlation with the CHF. The pressure ratio shows a nonlinear trend, with the CHF increasing first and then decreasing as the pressure ratio increases. In contrast, the boiling surface area shows no significant correlation with the CHF. This predictive modeling study provides an improved approach for the high-accuracy prediction of CHF in cryogenic fluid, along with a comprehensive analysis of the factors influencing the CHF prediction accuracy, which offers valuable insights for optimizing prediction models in this field.