Modeling the effectiveness of ozone as a water disinfectant using an artificial neural network

Data from an experimental study on the use of ozone to inactivate a parvovi rus in a synthetic and an actual industrial water source was analyzed using an artificial neural network (ANN). The goal of this analysis was to predi ct the necessary ozone dose to disinfect the water as a function of specifi c environmental conditions. The network consisted of six inputs (time, alka linity, organic carbon concentration, initial virus concentration, sonicati on, and ozone residual) and one output (virus concentration). The network w as effective in predicting the outcome of ozone disinfection under conditio ns not previously encountered in training. A sensitivity analysis revealed that the network learned relationships among the variables similar to accep ted trends in the disinfection process. A comparison with current EPA proce dures also showed the effectiveness of the ANN approach.