Three-dimensional settling map for intelligent detection of activated sludge (AS) properties and health evaluation of AS based wastewater treatment

Rapid and reliable detection of activated sludge (AS) is essential for smart wastewater treatment but remains challenged by the interference susceptibility and poor generalization of current methods. Here, a Three-Dimensional Settling Map (3D-SM) was proposed to dynamically encode the complete sludge settling process as a unified spatiotemporal–optical fingerprint. Using a self-developed platform and adaptive image processing, this map was extracted and decoded via a deep learning model to directly quantify key AS parameters—MLSS, SVI₃₀, and SV₃₀—with high accuracy (MLSS: R²=0.957, SV₃₀: R²=0.953, SVI₃₀: R²=0.962). The 3D-SM showed low sensitivity to tested environmental factors (e.g., pH, conductivity, and color) within the evaluated range and enabled short-term state prediction (R²>0.60) under investigated conditions. Furthermore, it supported threshold-based sludge-state assessment and exploratory identification of filamentous bacteria enrichment, achieving>92.7% accuracy in identifying settling dysfunctions and 99.3% accuracy in classifying filamentous levels on the test set. A 71-day cross-site validation at one industrial AO treatment plant confirmed consistent performance (R²>0.801, accuracy>95.2%) within the tested operational range. This work establishes 3D-SM as a foundational, interference-resistant tool that transforms settling dynamics into an AI-parsable digital signature, providing a robust at-line approach for supportive AS intelligent monitoring.