Computational intelligence techniques for modeling of dynamic adsorption of organic pollutants on activated carbon.

The objective of this work is to compare the efficiency of three computational intelligence techniques: Artificial Neural Networks (ANNs), Support Vector Machines (SVMs) and Adaptive Neuro-Fuzzy Inference System (ANFIS) to model the dynamic adsorption of organic pollutants on activated carbon. A comparison study was enhanced using five models: ANN with conventional transfer functions, ANN with new transfer function called "SPOCU", SVM, SVM hybrid with Dragonfly optimisation algorithm (SVM-DA) and ANFIS. A set of data points, collected from scientific papers containing the dynamic adsorption kinetics of adsorption on activated carbon, was used in the modelling process. The studied parameters were molar mass, initial concentration, flow rate, bed height, BET surface area, time and concentration of non-dimensional effluents. Overall, the developed model was able to accurately estimate 11,763 experimental data points gathered from the literature. The performance of the optimised models has been evaluated using different metrics between the experimental and the predicted data. Results show that SVM-DA model can estimate accurately the dynamic adsorption of organic pollutants on activated carbon against the other tested models. Also a graphical user interface is developed in this paper in order to keep the traceability of the estimated results. [ABSTRACT FROM AUTHOR]