Identification of process faults in an urban waste water treatment plant

Wastewater treatment is a complex process, the heart of which is the biological reduction of harmful impurities. When the process deviates from normal operations, a fault occurs with the release of pollutants and pathogens into public waters. The objective of this research is to explore the ability to identify the operational state of the wastewater treatment plant from state variables that are multivariate, non-linear, and time dependent. We contrast the capabilities of classical linear statistical models, nonparametric statistical methods, and neural network models.