Semnan University Press Journal of Modeling in Engineering 2008-4854 17 56 2019 04 21 Experimental Study and neural network modeling for prediction of refractive index of pure and binary alcohol mixtures Experimental Study and neural network modeling for prediction of refractive index of pure and binary alcohol mixtures 375 387 3825 10.22075/jme.2018.14568.1442 FA Fatemeh Kord Faculty of Chemical Engineering, Babol Noshiravani University of Technology Kamyar Movagharnejad Faculty of Chemical Eng., Babol Noshiravani University of Technology Journal Article 2018 05 05 In this study, the refractive index of pure alcohols and binary alcohol mixtures were investigated experimentally and theoretically. In the experimental approach, a refractometer was used to measure the refractive index of the samples. Aa artificial neural network in form of a multilayer perceptron was also used to model and predict the measured refractive data. The input parameters of the network included temperature, molecular weight, the group contributions of CH3, CH2 and OH for the pure materials and for the binary mixtures the additional parameters of mole fractions, molecular weight and group contributions of both components have to be considered. The refractive index of the pure or binary alcohol mixture consist the only output parameter of the network. 70% of the experimental data were considered for train, 15% for test and 15% for the validation of the neural network. The optimum neural architecture for the pure compounds consisted of 10 neurons in the hidden layer with 0.08457 mean absolute relative error and the optimum network for binary mixtures consisted of 12 neurons in the hidden layer with 0.07121 absolute relative error. Comparison of the results showed a good agreement between the experimental data and the neural network outputs and the high accuracy of the model. In this study, the refractive index of pure alcohols and binary alcohol mixtures were investigated experimentally and theoretically. In the experimental approach, a refractometer was used to measure the refractive index of the samples. Aa artificial neural network in form of a multilayer perceptron was also used to model and predict the measured refractive data. The input parameters of the network included temperature, molecular weight, the group contributions of CH3, CH2 and OH for the pure materials and for the binary mixtures the additional parameters of mole fractions, molecular weight and group contributions of both components have to be considered. The refractive index of the pure or binary alcohol mixture consist the only output parameter of the network. 70% of the experimental data were considered for train, 15% for test and 15% for the validation of the neural network. The optimum neural architecture for the pure compounds consisted of 10 neurons in the hidden layer with 0.08457 mean absolute relative error and the optimum network for binary mixtures consisted of 12 neurons in the hidden layer with 0.07121 absolute relative error. Comparison of the results showed a good agreement between the experimental data and the neural network outputs and the high accuracy of the model. https://modelling.semnan.ac.ir/article_3825_ef0a025b4d696734123c1d2c47426cfa.pdf