A NEURAL-NETWORK MODEL FOR PREDICTION OF BINARY ADSORPTION USING SINGLE SOLUTE AND LIMITED BINARY SOLUTE ADSORPTION DATA

A simple neural network model was used to predict binary solute adsorp tion onto granular activated carbon (GAG). While some data on binary a dsorption were required, the neural network could be effectively train ed using predominately single solute adsorption data, and only a limit ed number of data sets (<10) were necessary for effective performance. Once trained, the network was capable of predicting binary solute ads orptions even for systems showing nonideality.