Generalized classification modeling of activated sludge process based on microscopic image analysis.

Area of Science:

• Environmental Engineering
• Biotechnology
• Image Analysis

Background:

• Conventional activated sludge wastewater treatment process (AS WWTP) monitoring relies on costly, time-consuming, and hazardous physico-chemical measurements.
• Existing image processing methods are plant- and state-specific, lacking generalizability across different facilities and conditions.

Purpose of the Study:

• To develop a generalized classification model for identifying the state of AS WWTPs.
• To create a model that requires no prior plant-specific information and is valid for any plant in any state.

Main Methods:

• Acquisition of AS sample images from nine plants via bright-field microscopy.
• Extraction of features based on morphological parameters of flocs.
• Application of sequential feature selection and least absolute shrinkage and selection operator (LASSO) for feature selection.
• Development of a support vector machine (SVM)-based strategy with a novel agreement solver for imbalanced data.

Main Results:

• The proposed SVM strategy achieved high accuracy (0.9423) and a kappa coefficient (κ) of 0.6681 for minority class data (bulking).
• Demonstrated superior performance compared to state-of-the-art multiclass SVMs and ensemble classifiers.
• Successfully identified different states across various AS plants.

Conclusions:

• The developed generalized classification model effectively identifies AS WWTP states using floc morphology.
• This approach offers a more accurate, generalizable, and potentially less hazardous alternative to conventional monitoring methods.
• The strategy shows promise for robust and reliable wastewater treatment process management.