Development of a Bayesian belief network for anaerobic wastewater treatment

grantor: University of TorontoA Bayesian belief network (BBN) was developed to determine the most probable causes of failure for an anaerobic sequencing batch reactor (ANSBR) attempting to achieve a COD removal efficiency and/or a methane production rate greater than 85% and 12 L/d, respectively. The BBN could also be used in prognosis mode to increase operational efficiency and to design experiments. This expert system would minimize the technical expertise required during the operation of an anaerobic treatment system with its extremely complex microbiological community. The Bayesian belief network consisted of nodes representing variables, and arrows representing cause-effect relationships between the variables. The conditional probabilities were calculated using an ANSBR model developed previously at the University of Toronto. By comparing the probabilities of the states before and after the introduction of evidence (using Bayes' Theorem), the most likely causes and/or effects could be determined, with the best suggestions for system adjustments.M.A.Sc