An efficient and green carbon recovery from fine slag: Machine learning-assisted evaluation of enhanced gravity separation

Coal gasification fine slag (CGFS) contains considerable residual carbon, but its separation is challenging due to fine particle size and complex phase associations. This study compares conventional flotation and enhanced gravity separation for CGFS, integrating response surface methodology and machine learning to predict combustible recovery and analyze parameter effects. Results show that flotation achieved only ∼54% combustible recovery even at high reagent dosages, while enhanced gravity separation reached up to 96.8% under optimized conditions (20 Hz frequency, 0.020 MPa pressure, 30 g/L solid concentration). Fuerstenau curves confirm higher separation efficiency (1.587 vs. 1.288) for gravity separation. The machine learning model achieved superior prediction accuracy (R² = 0.95) compared to RSM (R² = 0.85), revealing parameter influence order: frequency > pressure > concentration. This study demonstrates that enhanced gravity separation offers a reagent-free, highly effective alternative for CGFS upgrading, while data-driven modeling provides significant advantages over conventional regression methods for complex separation processes.