Artificial intelligence applications for predicting and optimizing blast furnace slag performance

Slag, as a by-product of the blast furnace smelting process, plays a crucial role in the operating efficiency and product quality by influencing its fluidity and chemical composition. However, the use of cost-effective raw materials has altered slag performance, presenting new challenges for blast furnace operation. In this study, we focused on the CaO–SiO 2 –Al 2 O 3 –MgO quaternary slag system to analyze changes in slag viscosity, sulfur partition ratio, and melting behavior under various slag compositions. To optimize slagging conditions, we employed random forest, gradient boost regression tree, and artificial neural network to predict slag performance based on composition, evaluating accuracy through metrics like R 2 , MSE, MAPE, and MAE. The models consistently achieved an R 2 above 0.97, indicating reliable predictions. Additionally, a user-friendly software was developed by Python to enable staff to forecast slag performance under different slagging compositions. Experimental results were compared with plant data, with a margin of error of 7%. This research offers a novel approach to predicting slag behavior, providing valuable insights for practical production, mitigating the adverse effects of cost-effective materials, and ensuring the safe and steady operation of blast furnaces.