Modeling the azo-dye removal from aqueous solution by adsorption on expanded modified vermiculite using statistical design and neural network

In this work, the adsorption characteristics of anionic azo-dye of Direct Orange 26 (DO26) from aqueous solution on expanded vermiculite (EV) was investigated. The optimum adsorption condition for effective parameters such as pH, contact time, adsorbent dosage and initial dye concentration in 3 levels was obtained using response surface methodology based on central composite design. Results showed the maximum adsorption capacity of 43.76 and 66.32 mg/g for direct orange 26 (natural) and direct orange 26 (modified), respectively, in pH of 2, adsorbent dosage of 0.1 g, initial dye concentration of 75 ppm, contact time of 50 min and temperature of 25 °C. On the other hand, Group method of data handling (GMDH) as a neural network was used to forecast the removal efficiency of DO26. The results demonstrated that response surface methodology (RSM) as a statistical approach had a better performance than GMDH to predict the removal percentage. The investigated adsorption process was fitted well with Freundlich isotherm. Also, adsorption kinetic data was in good agreement with pseudo-second order model. Thermodynamic data indicated that the adsorption process was spontaneous and endothermic with ∆G = −3.12 and − 3.41 (kJ/mol at T = 25 °C), ∆H = 93.515 and 24.586 kJ/mol and ∆S = 325.88 and 94.164 kJ/mol.K for DO26 (natural) and DO26 (modified), respectively.