Abstract
Good settleability of activated sludge is essential for the stable operation of biological wastewater treatment systems and for achieving effluent compliance. Although the Sludge Volume Index (SVI) is widely used, it is a static endpoint indicator that struggles to detect subtle changes in the early stages of sludge bulking. To address this limitation, this study developed an automated machine vision-based method for monitoring the settling process, enabling real-time generation of settling curves, extraction of dynamic features, and early warning. Compared with manual measurements, the proposed method showed a measurement error below 2%, confirming its accuracy in interface identification and measurement consistency. Based on the fitted settling curves, the maximum volumetric settling rate, vmax, was extracted as a dynamic indicator. After filamentous bulking was induced in laboratory-scale sequencing batch reactors (SBRs), vmax showed a sustained decline before the marked increase in SVI, indicating settleability deterioration approximately 2–3 days in advance. The decline in vmax was consistent with changes in floc morphology and filamentous bacterial proliferation, indicating its sensitivity to early changes in sludge structure. The method was successfully applied to full-scale secondary clarifier monitoring, demonstrating its feasibility and stability under engineering operating conditions. Overall, machine vision-based settling monitoring can provide dynamic process information to complement conventional SVI assessment and offers an effective indicator for early identification of activated sludge bulking.