Bi-Objective Optimization through Pareto Frontier Analysis and Artificial Neural Network for Adsorptive Removal of Fluoride by a Novel Al/Olivine Composite

Abstract

A new model for the bi-objective optimization of the removal efficiency and adsorption capacity was developed through an artificial neural network (ANN) and Pareto frontier analysis for the adsorptive removal of fluoride. A novel low-cost natural material, olivine, was impregnated with aluminum and was used for defluoridation. The preparation process of the adsorbent was optimized using the response surface method with a 32 factorial design. The significant variation of fluoride removal efficiency from 39.33% to 95.4% showed the noticeable effect that the preparation process parameters have on the defluoridation potential of the adsorbent. The ANN was applied to model the significant parameters of the adsorption process with the removal efficiency and adsorption capacity. It showed a correlation coefficient of more than .99 and a mean square of error of 2.35 and .18, respectively. A decision tool with Pareto frontier analysis was applied to the simulated data set of the ANN model to generate the overall optimal solutions. The removal efficiency and the adsorption capacity varied from 94.68% to 98.96% and 4.88 to 1.92 mg/g, respectively, in the optimum points.