A hybrid smart modeling approach for estimation of pure ionic liquids viscosity

Abstract

Good knowledge about the physicochemical properties of ionic liquids (ILs) at different temperature and pressure is very important, especially in the field of green chemistry and chemical engineering. As a transport property, viscosity plays a crucial role in the process of mass transfer and consequently it is of great importance in all chemical processes. In this study application of a novel approach namely, committee machine intelligent system (CMIS) for modeling the viscosity of pure ILs is investigated. To this end, 579 experimental viscosity data points related to the 23 different ILs at elevated temperatures and pressures were gathered from the previously open literature for the purpose of this study. Except temperature and pressure, to discriminate among various ILs, the physical properties of ILs were considered as inputs of the CMIS model. CMIS paradigm uses four intelligent models multilayer perceptron networks (MLP-ANN), radial basis function networks optimized by genetic algorithm (GA-RBF), least square support vector machine optimized by coupled simulated annealing (CSA-LSSVM), and adaptive neuro inference system optimized by conjugate hybrid and PSO algorithm (CHPSO-ANFIS) in its algorithm to enhance the preciseness of the predictions. Reliability and robustness of each models and also CMIS were determined using statistical and graphical methods. Although constructed intelligent models exhibit good accuracy for prediction of ILs viscosity, CMIS method shows better performance compared to other approaches. Finally, an outlier analysis (using leverage statistical algorithm) has been performed on the whole data set to detect the probable doubtful ILs viscosity data. Outcomes of this study show that intelligent based modeling algorithms are robust tools that can be replaced by difficult and time-consuming processes of measuring the physicochemical properties (here viscosity) of ILs.