Machine learning analysis for condensation flow heat transfer in mini/micro-channels

Miniature condensers have emerged as an efficient solution for thermal management of compact high-power devices due to their exceptional heat dissipation capability. However, accurate prediction of heat transfer coefficient(HTC) remains challenging due to complex flow and thermal behaviors in two-phase heat transfer. This study employed explainable machine learning to develop condensation HTC prediction models in mini/micro channels. A multidimensional feature database containing 4003 experimental data points across 19 fluids in hydraulic diameter 0.1mm≤D ≤ 4.8 mm was constructed. Four machine learning models, including Artificial Neural Network (ANN), Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Support Vector Regression (SVR), were developed utilizing the database to explore their potential in predicting condensation HTC. The models were validated through internal and external datasets, with comparison against six traditional correlations. The SHapley Additive exPlanations (SHAP) method was subsequently applied to explain the XGBoost prediction mechanism. Results demonstrate all machine learning models achieved satisfactory performance compared to traditional correlations, with XGBoost exhibiting optimal accuracy and generalization. It attained a coefficient of determination (R²) of 0.993 and a mean absolute relative deviation (MARD) of 3.6 % across the database, with strong generalization even for new fluid datas. SHAP explanation revealed Froude number and dimensionless vapor velocity were critical features, while the influence of features such as thermal conductivity and mass flux on the model's prediction aligned with the trend of physical laws and experimental results, effectively enhancing predictive rationality of "black-box" models. This work shows machine learning's significant potential for two-phase heat transfer prediction, providing an efficient predictive tool for mini/micro-channel condenser design.