LAUSR

Abstract The existing nucleate pool boiling correlations have theoretical footings and their usefulness is restricted by the failure to effectively account for the surface effect. To tackle this problem, a high-fidelity approach based on deep learning has been developed to predict the nucleate boiling surfaces subject to various surface roughness. The proposed model accounts for the effect of surface roughness, roughness fabrication method, surface material, surface inclination, saturation temperature, and pressure on the pool boiling performance of dielectric liquids, water, and refrigerants. The proposed method can accurately predict the boiling heat transfer performance of roughened surfaces by including the most influential surface characteristics, testing conditions, and liquid thermophysical properties into the architecture of the developed model. Correlation matrix identifies that heat flux, surface inclination, surface roughness, thermal conductivity of surface material, liquid saturation temperature, and pressure are the prime factors to affect the nucleate boiling heat transfer coefficient. Different neural networks (DNNs) are built and tested in order to find an optimal model based on an experimental dataset of 13000 data points. The final selected model can estimate the investigated parameter with a coefficient of determination (R2) = 0.994 and mean absolute error (MAE) = 0.65. The suggested method can be utilized to predict the boiling heat transfer performance of a variety of roughened surfaces subject to different working fluids and testing conditions.