Artificial neural network – Imperialist competitive algorithm based optimization for removal of sunset yellow using Zn(OH)2 nanoparticles-activated carbon

Abstract

The effects of variables were modeled using multiple linear regressions (MLR) and artificial neural network (ANN) and the variables were optimized by imperialist competitive algorithm (ICA). Comparison of the results obtained using introduced models indicated the ANN model is better than the MLR model for the prediction of sunset yellow removal using zinc oxide nanoparticles-activated carbon. The coefficient of determination (R2) and mean squared error (MSE) for the optimal ANN model with 9 neurons at hidden layer were obtained to be 0.9782 and 0.0013, respectively. A nano-scale adsorbents namely as Zn(OH)2 was synthesized and subsequently loaded with AC. Then, this new material efficiently applied for sunset yellow (SY) removal, from aqueous solutions in batch process. Firstly the adsorbent were characterized and identified by XRD, FESEM and BET. Unique properties such as high surface area (>1308m2/g) and low pore size (<20Å) and average particle size lower than 45.8Å in addition to intrinsic properties of nano-scale material high surface reactive atom and the presence of various functional groups make it possible for efficient removal of (SY). The effects of adsorbent dose, pH, initial SY concentration and contact time were optimized. Fitting the experimental data of adsorption over time in the range of 30min to various models show the suitability of second-order and intraparticle diffusion models for the prediction of removal rate and their parameters (R2>0.999). The factors controlling adsorption process were also calculated and discussed. Equilibrium data fitted well with the Langmuir model at all amount of adsorbent with maximum adsorption capacity of 158.7mgg−1.