Back-propagation neural network for performance prediction in trickling bed air biofilter

Experimental studies were carried out with a laboratory-scale biotrickling filter to treat a gaseous stream contaminated with benzene, toluene and xylene (BTX) operated in a continuous mode. The biotrickling filter initially acclimatised with toluene was used to treat BTX compound individually at loading rates ranging from 7.2 g/mhr to 62.2 g/mhr, operated in a sequential mode. The results showed removal efficiencies as high as 100% when operated with toluene as the sole carbon source. An application of the back-propagation neural network to this experimental data is presented in this paper. The performance parameters namely, elimination capacity and removal efficiency were predicted from the experimental observation by selecting the appropriate network topology. The sensitive internal parameters of the network were selected using the 2 fractional factorial design. The neural-network-based model was found to be an efficient data-driven tool to predict the performance of a biotrickling filter.