MODELING THE TREATMENT PERFORMANCE OF A SUBMERGED MEMBRANE BIOREACTOR USING ARTIFICIAL NEURAL NETWORK

摘要

Membrane bioreactor (MBR) technology has become increasingly popular in wastewater and water treatment. MBR is a suspended growth activated sludge treatment system that relies upon membrane equipment for separation of liquids/solids prior to discharge of the treated effluent. For many industrial processes as well as for conventional wastewater treatment, mathematical models are used to simulate process performance and derive optimization and control approaches. Although MBRs have been shown to be very effective in treating various types of wastewaters and producing high quality water for further reuse, there is limited information in the literature regarding modeling process performance or plant operation of MBRs. Furthermore, no study was found in the available literature which investigated the applicability of artificial neural network (ANN) methodology to predict treatment performance of a MBR system treating domestic wastewater for further reuse. However, Cinar et al. investigated the modeling of a submer ged MBR treating cheese whey wastewater by the ANN. Modeling by ANN methods has been increasingly researched and employed in environmental applications in recent years. Better control of wastewater treatment plants may be achieved by the use of robust models to predict certain key parameters based on past observations. Models based on ANN may be very effective at capturing the non-linear relationships existing between variables in complex systems like MBRs. Thus, the main objective of this work was to investigate the applicability of ANN modeling method to estimate effluent chemical oxygen demand (COD), ammonia nitrogen (NH3-N) and total suspended solids’ (TSS) concentrations in a pilot-scale MBR treating domestic wastewater. The data used in modeling efforts was obtained from this pilot-scale MBR, which was continuously operated aerobically about 8 months. A single ANN model structure was developed for all three effluent parameters using the correlations among the past information of influent and effluent data.