Refrigerant blends and pure refrigerants have wide applicability in thermal engineering. One of the critical parameters in the design and evaluation of thermal equipment is the heat transfer coefficient, which can be difficult to determine for refrigerants that undergo phase change within the equipment. For pure refrigerants, classical experimental relations developed by Gungor and Winterton (GW87) are known to exhibit errors around 15% on average, and reaching more than 40% in some cases. For refrigerant blends larger uncertainties are expected due to a complex number of factors such as nucleate boiling degradation, particularly when using functional forms previously developed for pure refrigerants. This work provides a comprehensive experimental study on the determination of heat transfer coefficients for R-22, R-134a, and the predefined refrigerant blends R-404A and R-407C. Genetic optimization is used to obtain more accurate semi empirical relations based on the classical GW87 correlation, and results of the optimization analysis show large improvement for pure refrigerants. The use of a degradation factor in the optimized correlation for R-407C allows for substantial error reduction for refrigerant blends. (C) 2015 Elsevier Ltd. All rights reserved.