The modelling of [Fe(CN)6]4- adsorption onto sepiolite with artificial neural network [Sepiyolitin [Fe(CN)6]4- kompleksi adsorpsiyonunun yapay sinir a?lari{dotless} i?le modellenmesi]

Artificial neural network (ANN) is the modeling method which has been succesfully used for adsorption processes in recent years. In this study, effectiveness of sepiolite in the adsorption of a strong metal-cyanide complex, [Fe(CN)6]4-, from aqueous solution was investigated. The experimental results was used as the database in forecasting model developed in ANN. Fe and CN adsorption capacities of sepiolite were forecasted by two hidden layer ANN model. Concentration, particle size, time and activation conditions were input independent variables while the capacity was forecasted as output depended variable. Total 324 data was randomly separated to two subsets as 232 training and 92 test data. Tansig-tansig-logsig functions and 8 and 7 neurons in the first and second hidden layers, respectively, resulted in the best model configuration. The highest correlation and the lowest errors achieved with this configuration were 0.99401-0.015 at training and 0.98983- 0.020 at test. Cross validation was applied to the best configuration. A nine fold cross validation resulted in 0.99442-0.015 correlation-error values for training and 0.98088-0.026 for testing. The achieved close correlation and error values after cross valudation indicated the success and confidency of the established model.