Experimental investigation on flow boiling characteristics of R1233zd(E) in a parallel mini-channel heat sink for the application in battery thermal management

Abstract Aiming for the application of refrigerant two-phase cooling on battery thermal management, this study presented an experimental investigation on the flow boiling characteristics of R1233zd(E) in a parallel mini-channel heat sink, which consisted of 21 channels with the length of 140 mm and the cross section of 1.5 × 1.5 mm2. Under the saturation pressure of 125 kPa, the boiling curve of R1233zd(E) and the heat transfer coefficient were analyzed under different heat fluxes (5-100 kW/m2), mass fluxes (150-600 kg/m2 s) and inlet subcoolings (2.5-8 K). In addition, the flow pattern was captured using high-speed camera. The results showed that the onset of nucleate boiling was delayed with the increase of refrigerant mass flux and inlet subcooling, while the transition of heat transfer was verified by flow pattern visualization. The measured wall superheat was compared with present correlations, showing that the predicted value was always smaller than the experimental results. Flow boiling on local points showed that the nucleate boiling was triggered later at the channel located at the center axis of the heat sink, meanwhile the approach of critical heat flux was more obvious at channel outlet. Due to the graduate transition of flow boiling mechanism, refrigerant mass flux had limited influence on the local heat transfer coefficient when the heat flux was lower than 30 kW/m2, while enhancement on heat transfer could be observed with larger mass flux when the heat flux exceeded 30 kW/m2. The local experimental data of heat transfer coefficient were compared with 6 correlations from the literatures. It was found that the best prediction was presented by Liu and Winterton model with a MAE of 23.4%, while Kim and Mudawar model showed best agreement under different mass fluxes. A simple modification was conducted on Kim and Mudawar model, and the calibrated model presented a better prediction accuracy with the MAE of 17.6% and 83.2% of total data within the error of ±30%.