Boruta-ISSA-KELM-Based Intelligent Diagnostic Model for Abnormal Blast Furnace

Blast furnace hanging is a frequent abnormal burden descent condition during the ironmaking process. To tackle the challenges of feature redundancy and low parameter optimization efficiency in traditional diagnostic methods under complex operating conditions, this study proposes an intelligent diagnostic model based on Boruta-ISSA-KELM. First, key diagnostic indicators for hanging are determined through blast furnace operation mechanism analysis. Subsequently, a two-stage feature selection process is implemented. Second, an improved Sparrow Search Algorithm (ISSA) is introduced. This algorithm incorporates adaptive weighting coefficients and a Lévy flight strategy to enhance the optimization efficiency of the kernel function parameter and regularization coefficient in the Kernel Extreme Learning Machine (KELM). Finally, comparative analyses with particle swarm optimization (PSO) and support vector machine (SVM) models are conducted. The proposed model achieves an accuracy of 94.78% in real-time hanging diagnosis, as validated by comprehensive evaluation metrics, including Kappa coefficient, recall rate (R), F1 score, mean absolute error (MAE), and root mean square error (RMSE). The results confirm that this swarm intelligence optimization algorithm demonstrates strong robustness, effectively meeting real-time monitoring requirements for abnormal blast furnace conditions.