MODELING OF MALACHITE GREEN ADSORPTION ONTO AMBERLITE IRC-748 AND DIAION CR-11 COMMERCIAL RESINS BY ARTIFICIAL NEURAL NETWORK

Abstract

In this study, the malachite green adsorption process using Amberlite IRC-748 and Diaion CR-11 resins was modelled by artificial neural network method. In the model created for this study, adsorbent dosage, initial malachite green concentration and contact time parameters, which are the independent variables of the adsorption process, were used as input. Adsorption percentage values, which are the dependent variables of the adsorption process, were obtained as output. Mean squared error (MSE) and determination coefficient (R2) values were obtained from the models created using thirty-one experimental data for adsorption of malachite green with Amberlite IRC-748 and thirty-eight experimental data for adsorption with Diaion CR-11. By evaluating these values together, the most appropriate training algorithm, transfer function in the hidden layer and the number of neurons in the hidden layer were defined. Accordingly, for both Amberlite IRC-748 and Diaion CR-11 resins, the optimum training algorithm was determined as Levenberg-Marquardt back-propagation and the optimum hidden layer transfer function as tan sigmoid. The optimum number of neurons in the hidden layer was identified as 13 for Amberlite IRC-748 and 12 for Diaion CR11. The MSE, R2all and R2test values of the models produced with the optimum parameters were obtained as 0.000261, 0.9972, 0.9903 for Amberlite IRC-748 and 0.000482, 0.9932, 0.9931 for Diaion CR11, respectively.