Predictive modeling for the boiling heat transfer coefficient of R1234yf inside a multiport minichannel tube

Abstract

Existing prediction models of flow boiling heat transfer coefficient, such as the well-known superposition, asymptotic, and flow pattern models, provide an applicable method to attain the closest to the true value of the heat transfer coefficient in specific ranges. In this study, heat transfer coefficient data are collected through an experimental study of R1234yf inside a multiport minichannel tube within a mass flux of 50–500 kg/m2s, heat flux of 3–12 kW/m2, saturation temperature of 6 °C, and vapor quality up to 1. The assessment of the heat transfer coefficient is conducted by comparing the heat transfer coefficient of each model with that of R1234yf. In addition, a machine-learning prediction model is proposed to improve the prediction accuracy of the heat transfer coefficient. A machine-learning method could provide an accurate prediction result for the heat transfer coefficient by feeding the program with a factor from heat transfer coefficient data (e.g., a dimensionless number). Therefore, an alternative prediction method could be applied to predict the heat transfer coefficient with the lowest error by providing the setting parameter that fits the pattern of heat transfer coefficient data. In addition, a heat transfer coefficient correlation is proposed to define the only-value result of the machine-learning model.