Performance prediction in biofilters using intelligent data driven models

Biofiltration is a versatile technique to treat hazardous airborne contaminants and Volatile Organic Compounds (VOCs) at relatively low concentrations from process industries. A compost based biofilter inoculated with mixed cultures of bacteria was examined in continuous mode for 155 days to treat toluene vapors at different loading rates (2.7 - 282.3 g/m3.hr). The performance was evaluated by the pollutant removal efficiency and elimination capacity in the biofilter. The basic motive behind the present study is to expose the perspectives of modern and sophisticated approaches such as Artificial Neural Networks (ANNs) for performance prediction in the field of biofiltration. Neural networks trained with back - propagation algorithm have already been successfully used to model air and water quality indices. An application of the Back Propagation Neural Network (BPNN) using experimental data from a biofilter is presented in this paper to predict the outlet toluene concentration, removal efficiency and elimination capacity using the easily measurable parameters in any biofiltration process. In addition to the neural model, simple regression models were also developed. The developed models were validated with a separate testing data that were not used for model formulation. The results from this study show superior performance of the ANN models compared to the regression models. These data driven models can replace conventional phenomenological - mathematical models, as they are able to generalize suitable approximations between the given set of inputs and outputs. The prediction method based on ANN can be successfully employed for long - term performance prediction in biofilters with an acceptable accuracy. © 2012 by Nova Science Publishers, Inc. All rights reserved.