The Thermophysical Properties of Imidazolium Based Ionic Liquids from Tao- Mason Equation of state and Artificial Neural Network

Ionic liquids (ILs) have been the focus of attention for many of their properties as potential substitutes for classic organic solvents in many different applications.In recent years, extensive studies have been done on the thermophysical properties of ionic liquids, such as density. Equations of state provide a way to predict density of these compounds. In addition, artificial neural networks (ANN) [1] have several unique characteristics and advantages for applications such as the prediction of physicochemical properties of ILs.In this study, first, synthesis of five imidazolium based ionic liquids have been done. Then improved Tao-Mason (TM) equation of state (EoS) [2] were applied to calculate density of ILs over a broad range of temperatures and pressures. Moreover, artificial neural network model was developed to predict density of ionic liquids. The weights and biases were randomly selected and hidden layers were regulated by tansig and purelin function. The multi-layer perceptron ANN with three layers was developed by random initial weights and biases. Also two activation functions were selected to regulating output of the network’s layers and the Levenberge-Marquardt algorithm was used to optimize the weights and bias of the layers. In this work, the parameters of temperature (in Kelvin), the pressure (in bulk), molecular mass and boiling temperature (in Kelvin) as the input of the artificial neural network and the ionic liquid density parameter as the objective of the ANN are considered. In this study 404 experimental data were used including:70% of the data (including 282 data) in the training set, 15% of the data (including 61 data) in the test set to examine the performance of the final network, 15% of the data (including 61 data) was used in the set to determine the validity of the network. To achieve the desired architecture, several networks were evaluated with different architectures. The performance of each network was evaluated by calculating its error in the training, evaluation and training stages.The density was estimated for 404 data from proposed TM EoS and ANN model for five imidazolium based ionic liquids over the range of the temperature (283.15 - 393.15 K) and a pressure range (1-1000 bar). The predicted results from Tao-Mason EoS and ANN were compared with experimental values. The overall average absolute deviation percent (AAD %) of five ionic liquids obtained 0.2% and 0.005% from Tao-Mason EoS and ANN respectively.The results show that the obtained densities from ANN model are in excellent agreement with the experimental data at various temperature and pressures.