Metal-organic framework MIL-100(Fe) for dye removal in aqueous solutions: Prediction by artificial neural network and response surface methodology modeling

Abstract

In this study, a metal organic framework MIL-100(Fe) was synthesized for rhodamine B (RB) removal from aqueous solutions. An experimental design was conducted using a central composite design (CCD) method to obtain the RB adsorption data (n = 30) from batch experiments. In the CCD approach, solution pH, adsorbent dose, and initial RB concentration were included as input variables, whereas RB removal rate was employed as an output variable. Response surface methodology (RSM) and artificial neural network (ANN) modeling were performed using the adsorption data. In RSM modeling, the cubic regression model was developed, which was adequate to describe the RB adsorption according to analysis of variance. Meanwhile, the ANN model with the topology of 3:8:1 (three input variables, eight neurons in one hidden layer, and one output variable) was developed. In order to further compare the performance between the RSM and ANN models, additional adsorption data (n = 8) were produced under experimental conditions, which were randomly selected in the range of the input variables employed in the CCD matrix. The analysis showed that the ANN model (R² = 0.821) had better predictability than the RSM model (R² = 0.733) for the RB removal rate. Based on the ANN model, the optimum RB removal rate (>99.9%) was predicted at pH 5.3, adsorbent dose 2.0 g L^-1, and initial RB concentration 73 mg L^-1. In addition, pH was determined to be the most important input variable affecting the RB removal rate. This study demonstrated that the ANN model could be successfully employed to model and optimize RB adsorption to the MIL-100(Fe).

Keywords: Artificial neural network; Central composite design; Metal organic framework; Response surface methodology; Rhodamine B.