Decision support system for optimizing BOF tapping cut-off in low-P steel production using machine learning and constraint programming

Controlling phosphorus in ultra-low-phosphorus steel often requires an early basic oxygen furnace (BOF) tapping cut-off to limit carryover slag, reduce tapping quantity (TQ), and increase yield variability. This study develops BOF tapping cut-off decision support system (BTDSS) that combines machine learning, similarity retrieval, and constraint programming to recommend a risk-controlled tapping cut-off. From the 9980 raw BOF and Ruhrstahl Heraeus (RH) heats, 6281 heats were retained after filtering for missing values, outliers, and metallurgically inconsistent data. RH end-point phosphorus (%P_RHend) was predicted from 26 pre-tapping features using XGBoost with recursive feature elimination and cross-validation, achieving adjusted R² of 0.9774 and 10.41% mean absolute percentage error. For a query heat, comparable heats were retrieved within each steel grade group by principal component analysis of 12 pre-tapping variables and cosine similarity to form an empirical TQ distribution. The Gaussian and Gaussian mixture models were compared using the Akaike and Bayesian information criteria. When multimodality is detected, the higher-mean mode is retained and tail trimming is applied. Trimmed quantiles determine the predicted tapping quantity (TQ_pred) and risk-controlled maximum tapping quantity (TQ_max). Based on model prediction, the phosphorus safety margin, error buffer, steel grade phosphorus limit, and similar-heat phosphorus distribution determine the upper quantile used for TQ_max. The constraint programming searches within the feasible region bounded by TQ_pred and TQ_max for tapping guidance. Across grade groups, TQ_pred was 0.34% higher than the similar-heat mean and predicted tapping yield (TY) reached 95.1%. The recommended TQ_max provided 2.2% headroom, with the maximum TY of 96.9%. Retrospective validation using 1736 new heats showed that the BTDSS-derived recommendations increased predicted TY from 94.9% to 96.2% and reduced the yield range from 13.2% to 4.5%. In the TQ_max scenario, the additional recoverable molten steel was 13107 t, indicating that the BTDSS can provide risk-aware tapping guidance with potential plant-scale benefits.