Diagnosing Upsets in Anaerobic Wastewater Treatment Using Bayesian Belief Networks

Abstract

Bayesian belief networks are probabilistic knowledge-based expert systems that predict the probability of an event occurring or diagnose the most probable causes of specific problems. Although Bayesian belief networks calculate the probabilities of events both before and after the introduction of evidence and are particularly useful for complicated systems with nonlinear relationships between causes and effects, they have not been widely applied to wastewater treatment systems. A Bayesian belief network for diagnosing upsets in an anaerobic wastewater treatment system was developed using the anaerobic sequencing batch reactor as a model system. A new approach for determining the conditional probabilities of the states of the variables was developed using a microbial kinetics model in conjunction with Monte Carlo simulation. The completed network suggests the most probable cause of upsets to the anaerobic sequencing batch reactor and updates its suggestions as more evidence is provided. The approach used is general and may be applied to other anaerobic treatment systems.