Modeling and optimization of chromium adsorption onto clay using response surface methodology, artificial neural network, and equilibrium isotherm models

Response surface methodology (RSM) was used for optimization of the adsorbent dosage, initial solution pH, initial ion concentration, and contact time in removal of Cr (III) with local nanoclay. The adsorption process was modeled by RSM and artificial neural network (ANN). The process was done in batch mode by central composite design (CCD) and the same design was applied for training ANN. The optimum condition was determined to be 500mg/L for adsorbent dosage, initial pH of 5, initial chromium concentration of 180mg/L and 20 min of contact time. In these conditions q(RSM) = 238.7780mg/g and q(ANN) = 237.5152mg/g which indicate removal percentage of 72.45% and 72.07%, respectively. RRSM2=0.9784 and RANN2=0.9834 indicate that the two models can predict the adsorption of Cr3+ properly. The two parameter Langmuir, Freundlich, Dubinin-Radushkevich (D-R), Temkin and three parameter Redlich-Peterson (R-PT), Sips and Toth isotherm models were applied to equilibrium data by minimizing the sum of squared errors (SSE), sum of the absolute errors (SAE), average relative errors (ARE), Hybrid fractional error function (HYBRID), Marquardt's percent standard deviation (MPSD), and nonlinear chi-square test error functions. The results showed that the HYBRID error function gives the lowest value and the R-PT model fits the data better than other isotherm models.