Adaptive Least Squares Support Vector Machine Predictor for Blast Furnace Ironmaking Process

The blast furnace system is one of the most complex industrial systems, with ongoing challenges in predicting silicon content. This study aims to develop an adaptive least squares support vector machine (LS-SVM) predictor specifically for silicon prediction in blast furnace hot metal. Employing a recursive updating algorithm, the predictor enhances the LS-SVM model's precision while avoiding the lengthy computational processes typically associated with model updates. The theoretical implications include a significant reduction in computational complexity from O(n^3m + m^4) to O(n^3 + m^3), allowing for more efficient predictions. Experiments conducted on two different blast furnace datasets demonstrate that the proposed adaptive LS-SVM predictor is effective, achieving a high hitting percentage while saving time. This approach not only addresses theoretical and experimental challenges in the ironmaking process but also contributes to the operational efficiency of blast furnace systems.