The Prediction of Equilibrium Adsorption of Liquid Phase Sulfur Compounds Using Neural Network (MLP-NN) Model

This study presents the use of multilayer perceptron neural network (MLP-NN) for estimation of the amount of equilibrium sulfur adsorption in liquid phase of hydrocarbon/ sulfur compound solution. The model was precisely developed on a collection of ۱۰۷ experimental data of seven types of sulfur compounds. Developed Model account for the effect of inputs parameters, including initial sulfur concentration, adsorbent weight, solvent/ solute molecular weights, solvent/ solute densities, adsorbent particle size, temperature and Si/ Al ratio of adsorbent. Admittedly the amount of equilibrium sulfur adsorption was assumed as the only output. Graphical presentations and statistical comparisons were applied to analyze the performance and accuracy of developed model. Results indicate that MLP-NN model provide precise predictions. Developed model offers fair agreement in predictions of the adsorption experimental data points with average absolute relative deviation and overall R۲ values of ۱.۴۳% and ۰.۹۹۹۶ respectively.