Research on Sinter Quality Prediction System Based on Granger Causality Analysis and Stacking Integration Algorithm

Sinter ore quality directly affects the stability of blast furnace production. In terms both physical and chemical properties, main indicators sinter are TFe content, alkalinity, drum index. By analyzing massive historical data on production a steel company, this study proposes prediction system based Granger causality analysis stacking integration algorithm. First, real sintering in enterprises (including coal gas pressure, ignition temperature, combustion air etc.), preprocessing raw was carried out using combination feature engineering process. Second, Pearson correlation analysis, Spearman were used to screen characteristic parameters with strong influence target variable (drum Index, TFe, alkalinity). Third, model for developed algorithm pair training. Finally, program development tool realize establishment online operation system. The test results showed that predicted goodness fit alkalinity (R), index 0.942, 0.958, 0.987, respectively, calculation time met actual requirements. establishing suitable running online, real-time realized guide promptly.