A deep learning method for estimating the boiling heat transfer coefficient of porous surfaces

Owing to the high nucleation site density and relatively robust behavior, sintered coated surfaces are of great interest for thermal management via pool boiling in many industries/applications such as desalination, electronics cooling, petrochemical, and power sector. The coated surfaces have been extensively used to improve the performance of the pool boiling process over the years. Regardless of a large amount of experimental data on the pool boiling of coated surfaces, no accurate mathematical/empirical approaches have been developed to estimate the heat transfer coefficient of these surfaces. The present study develops an AI-based method to estimate the pool boiling heat transfer coefficient for coated porous surfaces. The proposed AI method can handle the complex nature of the coating characteristics such as porosity, coating thickness, and particle size. Via using deep neural networks, the proposed method is applicable for highly wetting fluids (dielectric liquids), refrigerants, and low-wetting liquid (water). Correlation matrix analysis confirms that porosity, coating thickness, particle size, wall superheat, and surface inclination as well as the thermophysical properties of the working fluids are the best independent variables to estimate the considered parameter. Different deep neural networks are designed and evaluated to find the optimized model in terms of its predictive accuracy by experimental data (373 points). The best model with an input layer, three hidden layers, and an output layer (11–30–15–1–1) was able to predict the heat transfer coefficient with overall R2 = 0.976 and (mean absolute error) MAE% = 5.74. The proposed approach is simple and can be employed to optimize the sintered coated surfaces for different cooling applications.