Condensation heat transfer and pressure drop of R1234yf/HFC mixtures inside small diameter channels

Hydrofluoroolefin/hydrofluorocarbon mixtures, such as R1234yf/HFC mixtures, have been proposed for the substitution of R134a in refrigeration and air-conditioning applications. Indeed, these fluids present good thermodynamic properties and low environmental impact. Here, the condensation process of two mixtures, R513A (R1234yf/R134a 56/44% by mass) and R516A (R1234yf/R152a/R134a 77.5/14/8.5% by mass), has been investigated inside two channels with inner diameters equal to 0.96 mm and 3.38 mm. Both R513A and R516A are azeotropic mixtures. Condensation heat transfer tests have been conducted at 40 degrees C saturation temperature and mass flux ranging between 40 kg m- 2 s- 1 and 600 kg m- 2 s- 1. In the present study, particular focus is put onto condensation at low mass flux (below 100 kg m- 2 s-1). Tests at low mass fluxes are very rare in the literature but of great relevance considering that chillers and heat pumps equipped with variable speed compressors work for most of their lifetime at partial loads, thus reducing the refrigerant mass flow rate in the heat exchangers and the mass flux in each channel as compared to the nominal working conditions. In the 3.38 mm diameter channel, the two-phase flow has been recorded using a high-speed camera and a comparison with two flow pattern maps is presented. The heat transfer coefficient data have been compared with the results of three semi-empirical models and with a specifically developed Artificial Neural Network model.

Automated summary

This study investigates the condensation process of two mixtures, R513A and R516A, in channels with different diameters. It is found that condensation at low mass flux is rare but important for chillers and heat pumps with variable speed compressors. The high-speed camera is used to observe the two-phase flow in the 3.38mm channel and compare it with flow pattern maps. Additionally, the heat transfer coefficient data are compared with three semi-empirical models and a specifically developed Artificial Neural Network model.