Study of acid orange 7 removal from aqueous solutions by powdered activated carbon and modeling of experimental results by artificial neural network

In this work, removal of acid orange 7 (AO7) by powdered activated carbon, from aqueous solutions with initial concentrations of 150 ppm to 350 ppm and initial pH values of 2.8, 5.8, 8.0 and 10.5 at 25°C was studied. Experiments were done in batch mode and the experimental solutions were agitated periodically. All concentrations were measured spectrophotometrically at 483 nm and three times replicated. In most cases, after 75 min contact time, the most of AO7 removal is performed. The maximum equilibrium removal of acid orange 7 (AO7) was 96.24% for its initial concentration of 150 ppm at pHi = 2.8, and minimum equilibrium removal was 48.05% for initial concentration of 350 ppm at pHi = 5.8. At the similar experimental conditions, application of different initial pH values altered the AO7 removal percent no more than 9.06%. It is found that the adsorption system follows the second-order adsorption rate expression and the constants of the rate expression at different conditions were calculated which are comparable and often higher than other adsorbents in adsorption of other dyes. The constants of Langmuir equation, Q and b, and constants of Freundlich equation, Kf and 1/n, were calculated and results show that the adsorption process is favorable. Comparison of R2 values shows that fitting of Freundlich equation to experimental data is better than Langmuir equation. The experimental results were also modeled by artificial neural network with mean relative error of 5.81%. This model was developed in Matlab 6.5 environment using a 3-layer feed forward backpropagation network with 3, 2 and 1 neurons in first, second and third layers, respectively.

Results of this study show that removal percent of acid orange 7, decreases with increasing its initial concentration. It can be seen that maximum removal percent of AO7 is 96.24% at pHi = 2.8 and Ci = 150 ppm and the minimum one is 48.05% at pHi = 5.8 and Ci = 350 ppm. The results indicate that at similar experimental conditions, application of different initial pH values altered the acid orange 7 removal percent no more than 9.06%. The plots of t/q vs. t show good linear relationships with R 2 values greater than 0.9988. This indicates that the adsorption system belongs to the second–order kinetic model. Linear regression coefficients of 1/qe = f (1/Ce) and log qe = f (log Ce) plots are greater than 0.9 and so, the equations of Langmuir and Freundlich are well fitted to the experimental data at equilibrium conditions. Of course, according to R2 values, fitting of Freundlich equation to experimental data is better than Langmuir equation. The mean relative error of modeling of the experiments by artificial neural network is 5.81%, which is good