Multifractal dissolved oxygen diagnostics for early warning of aerobic granular sludge destabilization

While aerobic granular sludge (AGS) stability is fundamentally governed by the size-structure heterogeneity of granules, conventional bulk indicators fail to capture these distributional shifts until macroscopic settling deteriorates. To bridge this diagnostic gap, this study introduces a non-invasive monitoring paradigm that extracts latent structural information from dissolved oxygen (DO) fluctuations via multifractal analysis. Over a 150-day operation of two sequencing batch reactors, we demonstrated that structural divergences directly dictate DO fluctuation signatures. In the conventional reactor, granule overgrowth and fragmentation drove the particle size distribution (PSD) toward a heavy-tailed, Cauchy-like regime. Conversely, a size-managed reactor employing periodic floc-granule replacement maintained a narrow, near-symmetric PSD. Mechanistically, the emergence of heavy-tailed PSDs generated intermittent, extreme DO fluctuations. Consequently, the correlation dimension (D₂) declined significantly, establishing a three-tier temporal hierarchy of destabilization: D₂ captured incipient structural drift, followed sequentially by dynamic settling indicators (t_d and t_s) and ultimately the conventional sludge volume index (SVI). Within this framework, D₂ and the left spectral width (Δα_left) serve complementary roles: D₂ provides a threshold-based early alarm, while Δα_left specifically diagnoses the accumulation of oversized granules, as confirmed by controlled mixing experiments. Furthermore, intervention experiments revealed that these multifractal descriptors respond reversibly to size management, though recovery efficiency drops sharply if intervention is delayed. Ultimately, this study transforms routine DO sensor data into a zero-CAPEX, two-step alarm-diagnosis protocol, enabling timely intervention before AGS destabilization becomes irreversible.