Use of BDST and an ANN model for prediction of dye adsorption efficiency of Eucalyptus camaldulensis barks in fixed-bed system Behzat Balci a , Olcayto Keskinkan a, * , Mutlu Avci b a Department of Environmental Engineering, Faculty of Engineering and Architecture, Cukurova University, 01330 Balcalı/Adana, Turkey b Department of Computer Engineering, Faculty of Engineering and Architecture, Cukurova University, 01330 Balcalı/Adana, Turkey article info Keywords: Dyes Adsorption columns Tree barks Eucalyptus camaldulensis BDST model, neural networks abstract In this study, the Bohart and Adams’ model taking into account bed depth, and influent dye concentration was studied to exhibit adsorption process of textile dyes (Basic Blue 41 – BB41 and Reactive Black 5 – RB5) in glass columns using tree barks (Eucalyptus camaldulensis). Adsorption capacity coefficient values are determined using the Bohart and Adams’ bed depth service model. The model indicated that adsorp- tion properties of E. camaldulensis barks conform for tertiary treatment for textile BB41 and RB5 contain- ing wastewaters. An artificial neural network (ANN) based model for determining dye adsorption capability of bed system is also developed. The breakthrough curves of adsorption are also exhibited by this model. Results showed that ANN model could describe present system. Results showed that with the increases of bed height, and the decreases of influent dye concentrations, the breakthrough time was delayed. Ó 2010 Elsevier Ltd. All rights reserved. 1. Introduction Artificial neural networks (ANNs) have become widely used in various research areas where the available information is experi- mental. ANNs introduce an easy mathematical function approxi- mation for any linear and nonlinear systems. Topology of the neural networks consists of input layer, hidden layers and output layer. The neural network training method develops the input– output relation for the modeled system utilizing data sets (Sato, Sha, & Palosaari, 1999). Various researchers used the ANN for exhibit the performance of adsorption systems successfully (Brasquet & Le Cloirec, 1999; Du, Yuan, Zhao, & Li, 2007; Kumar & Porkodi, 2009; Robinson, Chandran, & Nigam, 2002; Yetilmezsoy & Demirel, 2008). On the other hand, dye contamination in aqueous wastewater from industries is a serious problem because dyes are not biode- gradable and tend to suppress photosynthetic activity in aquatic habitats by preventing the sunlight penetration. Dyes have also toxicological characteristics which are the main issues for environ- mentalists and have been the subject of growing attention for some years. Removal of textile dyes from wastewaters is one of the major problems in wastewater treatment technology. Traditional treatment methods such as ion exchange, chemical pre- cipitation, and membrane separation are often ineffective and very expensive when they are used for the removal of dyes. Currently, the most widely used and effective physical method for the treatment of colored wastewater is adsorption. The most convenient method for designing adsorption systems is the use of adsorption isotherms. The theoretical adsorption capacity of the adsorbent for a particular contaminant can be determined by calculating its adsorption isotherm (Tchobanoglous, 2003). The performance of a given adsorption system can be demonstrated through the use of adsorption isotherms. The degree to which adsorption will occur and the resulting equilibrium relationships are correlated according to the empirical relationship of Freundlich and the theoretically derived Langmuir relationship (Eckenfelder, 1989). In most wastewater flowing systems, since the contact time is not sufficiently long for the attainment of equilibrium, the data obtained under batch conditions are generally not adequate. Hence, it is required to perform equilibrium studies by using col- umns (Zhou, Zhang, Zhou, & Guo, 2004). Activated carbon is the most popular and widely used adsor- bent. In most industries, activated carbon columns are employed for the treatment of toxic, non-biodegradable wastewaters and as a tertiary treatment following biological oxidation (Eckenfelder, 1989). However, it is expensive because of the chemicals required for its regeneration after pollutant removal; the higher the quality, the greater the cost. Some natural materials not only have excellent adsorbability of dyes, but also have biocompatibility, 0957-4174/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.eswa.2010.07.084 Abbreviations: BB41, C.I. Basic Blue 41; RB5, C.I. Reactive Black 5; ANN, artificial neural network; BDST, bed depth service time model; MLP, multilayer perceptron. * Corresponding author. Fax: +90 3223386126. E-mail address: olcayto@cu.edu.tr (O. Keskinkan). Expert Systems with Applications 38 (2011) 949–956 Contents lists available at ScienceDirect Expert Systems with Applications journal homepage: www.elsevier.com/locate/eswa