Predictions of blast furnace permeability index based on SVM-RFE+WOA-BiLSTM model

The blast furnace permeability index is an important parameter that can quickly, intuitively, and comprehensively reflect the condition of the blast furnace. Accurate prediction of the blast furnace permeability index can help detect and prevent abnormal conditions such as pipeline issues, suspended material, collapse, and gas loss at an early stage (about 10 minutes in advance). This paper proposes a predictive model for the blast furnace permeability index that combines Support Vector Machine-Recursive Feature Elimination (SVM-RFE), Whale Optimization Algorithm (Whale Optimization Algorithm, WOA), and Bidirectional Long-Short Memory Neural Network (Bidirectional Long Short-Term Memory, BiLSTM). First, SVM-RFE is used to perform feature selection on the original high-dimensional, nonlinear, and time-series data, generating a feature importance ranking with interpretability. Then, the optimized BiLSTM using the WOA algorithm is employed to predict the blast furnace permeability index. The results show that the constructed SVM-RFE+WOA-BiLSTM model significantly reduces prediction errors compared to traditional methods and greatly improves prediction accuracy. The eight parameters selected during the feature selection process have strong interpretability, which helps blast furnace operators quickly grasp the instantaneous changes in the furnace condition and take effective measures to restore smooth operation.