K₂CO₃-Fe₂O₃ catalyzes sludge ceramsite formation: ML-elucidated nucleation-growth kinetics

Urban sewage sludge generation continues to rise annually, making its resource recovery into lightweight ceramsite a prominent research focus. This study systematically performed 30 distinct thermogravimetric (TG) experimental conditions-each replicated three times independently (90 total runs)-spanning three sludge-to-shale mass ratios (3:7, 5:5, and 7:3) and two heating rates (10 and 20 °C/min). Using the Satava-Sesták method, 15 solid-state kinetic mechanism functions were evaluated; the random nucleation and growth model (No. 10-12) was identified as optimal, with an average coefficient of determination (R²) > 0.81. A machine learning dataset comprising 240 valid samples (multi-condition × multi-temperature-point) was constructed, with input features including component mass fractions, heating rate, and temperature, and the output being the fitted apparent activation energy. Analysis using four machine learning models-including Random Forest and Gradient Boosting Decision Tree (GBDT)-revealed that mass loss in Stage II exhibited the strongest statistical association with porosity (GBDT feature importance = 0.36), while mass loss in Stage III showed the strongest statistical association with compressive strength (GBDT importance = 0.48); these associations reflect predictive correlations-not causal relationships. Based on this, optimized ceramsite achieved a porosity of 58% and a compressive strength of 19 MPa. Prediction error was quantified using mean absolute percentage error (MAPE): the best-performing LightGBM model yielded MAPE = 8.7% for mechanism function No. 12, with all models exhibiting MAPE values strictly within the range of 5.2%-9.2%. Mechanistic interpretation indicates a strong synergistic effect between K₂CO₃ and Fe₂O₃ during pyrolysis-K₂CO₃ initiates nucleation, whereas Fe₂O₃ promotes crystal grain growth, jointly establishing a self-regulating "gas evolution-melt formation-pore stabilization" tripartite mechanism.