Artificial neural networks for modeling and optimization of phenol and nitrophenols adsorption onto natural activated carbon

An artificial neural network (ANN) approach was developed to predict the adsorption efficiency (W%) of phenol and nitrophenols onto activated carbon. We have studied the backpropagation of a three- layer feedforward network with Levenberg Marquardt, which describes the relationship between the adsorption efficiency as output and the operation conditions as contaminants (Phenol, Nitrophenols), initial contaminant concentration (C i ), pH and contact time. This model has been validated comparing it with both experimental measurement and simulated analysis and showed high agreement with very low percentage of error (0.5%) and high Pearson correlation (R 2 = 0.9868). The sensitivity analysis has also shown that the contact time was the most important influential parameter in this process. Based on the sensitivity analysis and neural networks model, we have developed an optimization algorithm (ANNi) for the calculation of the contact time into adsorption process when the initial conditions are well known and adsorption efficiency is required. ANNi could perform assessment with a minimal error. This technique is a very promising tool for modeling and optimization of the adsorption onto activated carbon process minimizing time and operation cost.