Modeling thermal conductivity of ionic liquids: A comparison between chemical structure and thermodynamic properties-based models

Abstract

Remarkable thermal properties of ionic liquids (ILs) such as high heat capacity and thermal conductivity have raised expectations to enhance their thermophysical properties for heat transfer applications. Despite various studies measuring the thermal conductivity of ILs, reliable models to predict this property are yet to be proposed. In this study, accurate models to estimate the thermal conductivity of the variety of ILs have been developed. To this end, the thermal conductivity was predicted by three models: (I) a simple group contribution method based on temperature, pressure, molecular weight, boiling temperature, and acentric factor; (II) a model based on thermodynamic properties, pressure, and temperature of ILs; and (III) a model based on chemical structure, pressure, and temperature. To develop model (I) a simple correlation was used and for development of models, (II), and (III), intelligent approaches comprising radial basis function (RBF) and multilayer perceptron (MLP) neural networks were implemented. Different optimization techniques including genetic algorithm (GA), gravitational search algorithm (GSA), scaled conjugate gradient (SCG), Bayesian regularization (BR), and Levenberg-Marquardt algorithm (LM) were utilized for optimization of the neural networks to accurately predict the thermal conductivity. To develop the models, 504 experimental data from 50 ILs over a wide range of pressure (100–20,000 KPa) and temperature (273–390 K) were used. The results reveal that although the empirical correlation does not predict the thermal conductivity of ILs accurately (mean absolute percent relative error (MAPRE) =12.26%), its intrinsic simplicity is yet valuable and is superior compared to the literature models. Model (II) (RBF-GSA with an MAPRE% of 1.051) shows better performance with more accurate predictions. Model (III) (RBF- GSA) also shows an acceptable accuracy (MAPRE% = 1.901). Finally, the results obtained by models developed in this study were compared with previous models/correlations/group contribution methods. The results revealed that the proposed models are more accurate with higher performance compared to other models in the literature.