Universal condensation heat transfer and pressure drop model and the role of machine learning techniques to improve predictive capabilities

Abstract

A novel, universal model is proposed to predict the condensation frictional pressure drop and heat transfer coefficient for horizontal microchannel and macrochannel flows. Over 4000 data points across the entire vapor quality range are collected with mass fluxes ranging from 50 – 800 kg m−2 s−1, reduced pressures from 0.03 – 0.96, and hydraulic diameters from 0.1 – 14.45 mm. The working fluids include synthetic refrigerants (R1234ze(E), R134a, R245fa, R404A, R410A), hydrocarbons (pentane and propane), and natural refrigerants (ammonia and carbon dioxide). Several models, including flow-regime based correlations and machine learning regression models (support vector regression, random forest, and artificial neural networks) are developed and compared with each other. Overall, machine learning algorithms, especially the random forest model, predict the data best for the pressure drop and heat transfer, with an absolute average deviation of about 4% for both models.