Generalized modeling of the sludge volume index and suspended solids to monitor activated sludge bioflocculation using image analysis

Abstract

The performance of an activated sludge wastewater treatment plant depends on bioflocculation that is monitored by physical measurements such as the sludge volume index (SVI) and mixed liquor suspended solids (MLSS). The estimation of SVI and MLSS has been proposed using image analysis based modeling which is time-efficient and valid for multiple plants operating in different states. The methodology includes the sequence of image acquisition using bright-field microscopy, a robust segmentation of flocs, partitioning of flocs based on different ranges of their equivalent diameters, extraction of morphological features, and modeling of SVI and MLSS using the features. It is proposed that bright-field microscopy at lower magnification to capture the flocs is sufficient to model SVI and MLSS. A robust approach for image segmentation is adopted by integrating state-of-the-art image segmentation algorithms. It is hypothesized that flocs in different ranges of equivalent diameter respond differently to the variation in the operating state. Hence, flocs and their respective image analysis features are categorized based on the range of equivalent diameter. Finally, stepwise regression is used for feature selection and model identification to explore the feasibility of generalization of models to multiple plants in different states regarding SVI and MLSS.