Abstract—In this work, A Neural Network (NN) was applied for the comparison of equilibrium study of binary system Co-Cu ions using adsorption isotherm model. The Absolute Average Deviation (ADD) was used to represent the sorption equilibrium values of Co and Cu binary system and a simplex optimization method was also applied for parameter identification of binary isotherms. The configuration of the Backpropagation neural network (BPNN) giving the smallest mean square error was three-layer and Levenberg- Marquardt backpropagation training algorithm (4-11-2). The obtained modeling results have shown that the use of NN has better adjusted the equilibrium data of the binary system when compared with the conventional sorption isotherm models. Keywords- Adsorption, Isotherm models, binary system, Neural Network. I. INTRODUCTION DSORPTION process involves separation of a substance or adsorbate from one phase, followed by its accumulation onto the substance of the adsorbing phase or adsorbent [1], [2]. Equilibrium condition is attained when the concentration of the solute remains constant, as a result of zero net transfer of solute adsorbed and desorbed from adsorbent surface. The equilibrium adsorption isotherms describe these relationships between the equilibrium concentration of the adsorbate in the solid and liquid phase at constant temperature [1]-[3], also propose the involved interactive force in the process. As the isotherm indicates the adsorption capacity of the sorbent, it enables the evaluation of adsorption performance, the involved mechanisms, and parameters to be improved, which are of critical importance in optimizing the use of the adsorbents [2]-[5]. J. Kabuba* is with the Minerals Processing and Technology Research Centre, Department of Metallurgy, School of Mining, Metallurgy and Chemical Engineering, Faculty of Engineering and the Built Environment, University of Johannesburg, PO BOX 526, Wits 20150, South Africa Tel. +27 11 559 6215; Fax: +27 11 559 6491. (e-mail: johnk@uj.ac.za). A. Mulaba-Bafubiandi is with the Minerals Processing and Technology Research Centre, Department of Metallurgy, School of Mining, Metallurgy and Chemical Engineering, Faculty of Engineering and the Built Environment, University of Johannesburg, PO BOX 526, Wits 20150, South Africa, Tel. +27 11 559 6215; Fax: +27 11 559 6491. (e-mail: amulaba@uj.ac.za). The analysis of experimental equilibrium data by fitting into different isotherm models is an important step to propose suitable model for process design [1]. The most widely applied isotherms for data modeling are the Langmuir and Freundlich, which are linear regression and developed based on thermodynamic equilibrium [6],[7]. Although several isotherms have been proposed to describe the equilibrium adsorption of binary system, it was found that the mathematical description of sorption isotherms of the analyzed ions in the presence of one or two additional ions in the solution is complicated from the theoretical point of view [8]. In this case a non-linear technique, NN was proposed. The NN is an information processing tool that is capable of establishing an input–output relationship by extracting controlling features from a database presented to the network and NN based predictive models are powerful in terms of learning the nonlinear relationships to understand and solve, and thereby achieving ability to predict accurately. The objective of this study is to compare the equilibrium study of binary system using adsorption isotherm models and NN. II. NEURAL NETWORK A. Basic concepts Neural network (NN) is an information processing system that is inspired by the way such as biological nervous systems e.g. brain. The objective of a NN is to compute output values from input values by some internal calculations [9]. The basic component of a NN is the neuron, also called “node”. Fig. 1 illustrates a single node of a NN. Inputs are represented by a 1 , a 2 ... and a n , and the output by O j . There can be many input signals to a node. The node manipulates these inputs to give a single output signal [10]. The values w ij , w 2j ... and w nj , are weight factors associated with the inputs to the node. Weights are adaptive coefficients within the network that determine the intensity of the input signal. Every input (a 1 , a 2 , .., a n ) is multiplied by its corresponding weight factor and the node uses summation of these weighted inputs (w 1j a 1 , w 2j a 2 , …, w nan ) to estimate an output signal using a transfer function. The other input to the node, b j is the node’s internal threshold, also called bias. This is a randomly chosen value that governs the node’s net input through (1): Comparison of equilibrium study of binary system Co-Cu ions using adsorption isotherm models and Neural Network John Kabuba * and AF Mulaba-Bafubiandi A International Conference on Mining, Mineral Processing and Metallurgical Engineering (ICMMME'2013) April 15-16, 2013 Johannesburg (South Africa) 126