Dynamic modeling of filamentous bulking in lab-scale activated sludge processes

For years now, biological wastewater treatment plants rely on activated sludge systems in which a complex ecosystem, constituted mainly of bacteria and protozoa, (bio)degrade the incoming pollutants. Filamentous bulking, a phenomenon in which the filamentous organisms dominate the activated sludge is still a widespread problem in the operation of activated sludge processes with often severe economic and environmental consequences. Image analysis offers promising perspectives for early detection of filamentous bulking because the morphology parameters of the activated sludge respond rather fast to changing process conditions. This paper aims at exploiting this information in black box models to predict the evolution of the sludge volume index (SVI), a laboratory measurement currently exploited to quantify the sludge settleability. More specifically, dynamic ARX models are investigated as a function of organic loading and digital image analysis information (such as the total filament length per image and some representative mean floc shape parameters). The model’s performances are compared on the basis of a squared errors like quality criterion. While the identification results are very promising, the validation of the models on other independently generated data sets, depends on which data set is used for identification. The best performing models have (a combination of) the total filament length, one of the floc elongation parameters and the fractal dimension as inputs.