Application of artificial intelligence in metallurgy: increasing the efficiency of the walz kiln

The article discusses the application of neural network methods and machine learning algorithms to optimize the operation of the Waelz kiln at a metallurgical plant. Based on real production data, a comparison of the efficiency of various algorithms (LGBM Regression, CatBoost, Decision Tree, Random Forest, etc.) for predicting the zinc content in clinker was carried out. The best results were shown by LightAutoML (R2 = 0.9339), CatBoost (R2 = 0.9218) and LGBM Regression (R2 = 0.9195). The key parameters influencing the process were determined: zinc content in the charge, compressed air consumption, dust chamber pressure and coke loading. Recommendations for process control were developed, including raw material quality control, optimization of gas-dynamic parameters and monitoring of reducing conditions. The results of the study demonstrate the potential of AI solutions for improving the efficiency of metallurgical production in the context of digital transformation of industry.