Artificial neural network as an applicable tool to predict the binary heat capacity of mixtures containing ionic liquids

Abstract

Ionic liquids (ILs) are amazing solvents gain an increasingly attention in the different areas of chemistry and chemical engineering industries during the past decade. Similar to every promising solvent, knowing the physiochemical properties of the ILs seems to be crucial to develop new designed ILs based industries. In this direction, the present study extends an artificial neural network (ANN) to correlate the binary heat capacity of ILs. To verify the proposed network, 1571 binary heat capacity data points were collected from the previously published literatures and divided in to two subsets namely training and testing subsets. The optimum parameters of the network including the number of hidden layer, numbers of neurons and transfer functions in hidden and output layers were obtained using these subsets data points. In addition, the predictive capability of the optimized network was validated using the testing data points (not considered in the training stage). The optimized network configuration consisted of one hidden layer with sixteen neurons and tansig and purelin transfer functions for the hidden and output layers. The obtained results from the training and test stages show that the proposed network was able to accurately predict the binary heat capacity of ILs binary mixtures with total absolute average relative deviation (AARD%) of 1.60% and relation coefficient (R2) value of 0.9975.