Density and viscosity modeling of liquid adipates using neural network approaches

Abstract

Liquid Dialkylesters of adipic acid (adipates) have achieved prominence as alternative green solvents due to their special properties. To enhance their utilization, accurate thermophysical property data are required, especially for extended pressure ranges. This study presents the application of Artificial Neural Networks (ANNs) for predicting the densities and viscosities of liquid adipates over a wide range of temperatures and pressures. A substantial dataset, including 1145 viscosity and 891 density data points, was utilized across a broad range of temperature and pressure conditions. The dataset used for densities includes data for six different liquid adipates and covers a temperature range of 293.15–403.15 K and pressures up to 140 MPa, while the dataset used for viscosities encompasses data for four different liquid adipates, ranging from 293.15 K to 403.15 K in temperature and up to 65.62 MPa in pressure. Two ANN models were developed and fine-tuned, with two separate models created to predict the properties of liquid adipates. These models exhibited very good performance, achieving an Average Absolute Relative Deviation (AARD) of 0.028 % for density and 0.400 % for viscosity. The results from the ANN models were compared with semi-empirical models based on the equation of state and rough hard-sphere theory, showcasing the technique's potential as a powerful tool for characterizing thermophysical properties. As part of enhancing model reliability and robustness, various tests were conducted to validate the chosen model.