Optimization of cyanide removal from wastewaters using a new nano adsorbent containing ZnO nanoparticles and MOF/Cu and evaluating its efficacy and prediction of experimental results with artificial neural networks

The present study examined the removal of cyanide from synthetic wastewater by a new nano-adsorbent containing synthesized ZnO nanoparticles and copper-metal organic frameworks (MOF/Cu) (the mass ratio of 1:5) (ZM15). The structure and morphology of ZM15 were examined using X-ray diffraction (XRD), Scanning Electronic Microscope (SEM) and Brunauer-Emmett-Teller (BET). To evaluate important parameters affecting cyanide removal including pH (3-9), time (30-90 min), the adsorbent weight (0.05-0.4 g), temperature (25-45 degrees C), and initial concentration of cyanide (10-100 mg/L), the RSM (response surface method) based on Design of Experiments (DOE) was utilized. The results of DOE led to a second order quadratic model with acceptable p-value and a lack-of-fit value less than and more than 0.05, respectively. According to Pareto, adsorbent weight, initial concentration, and pH are the most effective factors in cyanide removal efficiency. The optimum values of removal efficiency achieved at pH = 6, the adsorbent weight of 0.4 g, temperature 25 degrees C, contact time 60 min, and initial cyanide concentration of 10 mg/L were 65%, 76.5%, 68.5%, 67%, and 67%, respectively. Equilibrium adsorption data were examined by using Langmuir and Freundlich adsorption isotherm models, which resulted in good agreement of experimental data with Langmuir isotherm model. The results were evaluated through Perceptron Artificial Neural Networks (ANNs) and Self-Organizing Map (SOM). The input parameters were variables affecting the cyanide removal, and the output parameter or the target parameter was cyanide removal efficiency. The winning number of neurons in the middle layer was found to be 29, and the network with topology 5-29-1 and correlation coefficient of 0.91581, the Mean Square Error (MSE) of 1.7033, and the largest error value of 83915 was selected as the best network for prediction. (C) 2019 Elsevier B.V. All rights reserved.