Semi-empirical correlations and an artificial neural network for liquid dynamic viscosity of low GWP refrigerants

n this work, simple semi-empirical correlations to describe the temperature and the pressure dependence of the dynamic viscosity of low GWP refrigerants, namely HydroFluoroOlefins (HFOs) and HydroChloroFluoroOlefins (HCFOs), in the liquid phase are presented. Firstly, the experimental liquid dynamic viscosity data available in scientific literature and databases were collected and statistically analyzed. From the data collected for low pressures, the Latini et al. (2002, 1990) correlation for the dynamic viscosity of liquid refrigerants in saturated conditions was re-fitted and constants expressly dedicated to the studied low GWP refrigerants were obtained. Then, the proposed temperature-dependent correlation was modified to represent liquid dynamic viscosity dependence on pressure. In addition, an artificial neural network was developed to predict the dependence of the liquid viscosity of the studied refrigerants on temperature and pressure. This model was trained, validated, and tested for the selected dataset. The results of the proposed correlations and the multi-layer perceptron neural network were compared with the liquid viscosity calculations provided by some of the most well-known literature correlations and REFPROP 10.0, proving the accuracy of the proposed models for engineering applications.