Burn-Through Point Dynamic Modeling Based on Temporal Cascade Broad Echo State Learning System in Sintering Process

In an iron ore sintering process, the early prediction of burn-through point (BTP) is a pivotal and essential undertaking. The BTP stands as a cornerstone in maintaining the uninterrupted operation of the sintering process, while also providing critical assurance for both the quality and quantity of sintered ore. The inherently time-varying and dynamic characteristics of sintering process make predicting BTP a challenge. To this end, we propose an innovative temporal cascade broad echo state learning system (TCBESLS), specifically tailored for achieving accurate BTP dynamic prediction. TCBESLS ingeniously integrates the core strengths of the broad echo state network and the temporal broad learning system, enabling the real-time acquisition and efficient analysis of temporal features and dynamic characteristics in the production process. This ensures that the system can quickly respond to and adapt to complex variations within the production flow. The validation of the feasibility and effectiveness of the proposed method is rigorously conducted through comparative and ablation experiments, firmly grounded in actual production data. The experimental results conclusively demonstrate the performance and accurate predictive capabilities of this method in comparison with some advanced methods.