Prediction of activated sludge filamentous bulking using ATP data and neural networks

Abstract

AbstractNeural networks (NNs) were used to predict the onset of filamentous bulking, as described by trends in stirred sludge volume index, in a pulp and paper activated sludge treatment system. Variables related to activated sludge biomass viability, namely specific oxygen uptake rate (SOUR) and adenosine triphosphate (ATP), were used as inputs to the NN. ATP data were shown to improve NN performance in providing an early warning signal for bulking, both in terms of accuracy and prediction delay. A warning signal system was developed to provide operators with enough time to react and further investigate the causes of bulking.