Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
691687 | Journal of the Taiwan Institute of Chemical Engineers | 2012 | 10 Pages |
A two-layer feed forward neural network was successfully applied to predict the adsorptive removal of textile dye direct blue 86 using microwave assisted activated carbon. The Levenberg–Marquardt back-propagation method was used to train the artificial neural network (ANN) at various experimental conditions. The pH, contact time, initial dye concentration, adsorbent dose and temperature were chosen as the input variables whereas, the dye uptake capacity was considered as the output variable. The tan sigmoid and linear transfer functions had been used to train the hidden and output layer of the network respectively. According to Levenberg–Marquardt back-propagation algorithm the optimum number of neurons was found to be five. The predicted and experimental values of the desired output variables were compared and a good correlation coefficient (0.982) was also obtained. The performance of the developed network was further improved by normalizing the experimental dataset and it was found that after normalization the MSE and validation error was reduced significantly. The sensitivity analysis was also performed to determine the most significant input parameter. The developed network was also found to be useful in predicting the adsorption capacity of an unknown material at any given experimental condition.
► A feed forward neural network was applied to predict the adsorptive removal of direct blue 86 using microwave assisted activated carbon. ► The Levenberg–Marquardt back-propagation method was used in the present study. ► The optimum number of neurons was found to be 5 as predicted by Levenberg–Marquardt method. ► The predicted and experimental values were found to be in reasonable agreement with each other. ► The developed network can be used to predict the adsorption capacity of any complex system.