Aromatic protein-like recognition enables rapid prediction of sludge deep dewaterability

Efficient sludge dewatering remains challenging in global municipal wastewater treatment plants because water content of the dewatered sludge cake (WC_dsc) strongly determines sludge handling cost, transport demand, and downstream treatment burden. Rapid prediction of sludge dewaterability is highly desirable, yet current indicators, including capillary suction time, specific resistance to filtration, and WC_dsc or solids content, mainly reflect final dewatering outcomes and provide limited insight into water release during conditioning. Here, we identify aromatic protein-like substances, particularly tyrosine- and tryptophan-like fractions, as molecular indicators of sludge deep dewaterability. Three representative iron-based advanced oxidation process (Fe-AOP) systems, Fe²⁺ + H₂O₂, Fe²⁺ + peroxymonosulfate (PMS), and Fe²⁺ + percarbonate (SPC), were systematically evaluated through bench-scale and large-scale tests. Results showed that Fe-AOPs highly improved sludge dewatering and reduced aromatic protein-like substances in the filtrate, with stronger depletion generally corresponding to lower WC_dsc. In the 60-day large-scale test, the average WC_dsc decreased to 62.05 wt%, 58.51 wt%, and 55.28 wt% after Fe²⁺ + H₂O₂, Fe²⁺ + PMS, and Fe²⁺ + SPC treatment, respectively, while aromatic protein-like removal exceeded 55%. Based on the molecular signals, three backpropagation (BP) neural network models were developed. Among them, the Levenberg-Marquardt-Bayesian regularization-BP hybrid mode showed the best performance, with an R² of 0.837, a root mean square error of 3.774, and a mean absolute percentage error of 3.921%. These results indicate that filtrate aromatic fluorescence provides a practical molecular basis for the rapid prediction of sludge deep dewaterability.