ORIGINAL PAPER Stochastic modeling applications for the prediction of COD removal efﬁciency of UASB reactors treating diluted real cotton textile wastewater Kaan Yetilmezsoy Zehra Sapci-Zengin Published online: 13 October 2007 Ó Springer-Verlag 2007 Abstract A three-layer Artiﬁcial Neural Network (ANN) model (9:12:1) for the prediction of Chemical Oxygen Demand Removal Efﬁciency (CODRE) of Upﬂow Anaer- obic Sludge Blanket (UASB) reactors treating real cotton textile wastewater diluted with domestic wastewater was presented. To validate the proposed method, an experi- mental study was carried out in three lab-scale UASB reactors to investigate the treatment efﬁciency on total COD reduction. The reactors were operated for 80 days at mesophilic conditions (36–37.5°C) in a temperature-con- trolled water bath with two hydraulic retention times (HRT) of 4.5 and 9.0 days and with organic loading rates (OLR) between 0.072 and 0.602 kg COD/m 3 /day. Five different dilution ratios of 15, 30, 40, 45 and 60% with domestic wastewater were employed to represent seasonal ﬂuctuations, respectively. The study was undertaken in a pH range of 6.20–8.06 and an alkalinity range of 1,350– 1,855 mg/l CaCO 3 . The concentrations of volatile fatty acids (VFA) and total suspended solids (TSS) were observed between 420 and 720 mg/l CH 3 COOH and 68–338 mg/l, respectively. In the study, a wide range of inﬂuent COD concentrations (COD i ) between 651 and 4,044 mg/l in feeding was carried out. CODRE of UASB reactors being output parameter of the conducted anaerobic treatment was estimated by nine input parame- ters such as HRT, pH, COD i concentration, operating temperature, alkalinity, VFA concentration, dilution ratio (DR), OLR, and TSS concentration. After backpropagation (BP) training combined with principal component analysis (PCA), the ANN model predicted CODRE values based on experimental data and all the predictions were proven to be satisfactory with a correlation coefﬁcient of about 0.8245. In the ANN study, the Levenberg-Marquardt Algorithm (LMA) was found as the best of 11 BP algorithms. In addition to determination of the optimal ANN structure, a linear-nonlinear study was also employed to investigate the effects of input variables on CODRE values in this study. Both ANN outputs and linear-nonlinear study results were compared and advantages and further developments were evaluated. Keywords Textile wastewater  Anaerobic treatment  COD removal efﬁciency  Neural network  Backpropagation algorithm  Modeling 1 Introduction Wastewater efﬂuents from textile industries are character- ized by high volumes and extremely variable composition, which can include biodegradable and non-biodegradable dyes, organic matter, salts and toxic substances (Isik and Sponza 2004). The variety of raw materials, chemicals, processes and also technologic variations applied to the processes cause complex and dynamic structure of envi- ronmental impact of textile industry (Sapci and Ustun 2003). These industries have shown a signiﬁcant increase in the use of synthetic complex organic dyes as the col- ouring material. Therefore, the discharge of dye house wastewater into the environment is aesthetically displeas- ing, impedes light penetration, damages the quality of the receiving streams and may be toxic to treatment processes, to food chain organisms and to aquatic life (Talarposhti et al. 2001). K. Yetilmezsoy (&)  Z. Sapci-Zengin Department of Environmental Engineering, Faculty of Civil Engineering, Yildiz Technical University, 34349 Yildiz, Besiktas, Istanbul, Turkey e-mail: yetilmez@yildiz.edu.tr 123 Stoch Environ Res Risk Assess (2009) 23:13–26 DOI 10.1007/s00477-007-0191-5