Molten Steel Temperature Prediction in Ladle Furnace Using a Dynamic Ensemble for Regression

Abstract

The accurate prediction of molten steel temperature is of great significance to the control of tapping temperature in ladle furnace. The more accurate the prediction is the better performance the controller will attain. In order to further improve the accuracy of existing data-driven predictive models, we propose a dynamic ensemble for regression to predict molten steel temperature. In contrast to existing ensemble models, we only select one base model with the highest competence from the pool for each test pattern, rather than fusing all base models. We can thus alleviate the effect of weak base models on the ensemble. Specifically, the operation of dynamic selection can be implemented by estimating the competence of all base models in the region of competence. In addition, the proposed ensemble generation method has taken into account the diversity and accuracy, which are crucial factors of deriving better ensembles. In order to investigate the effectiveness of our predictor, we compare it with several competitors proposed in literature on a real-world dataset. The experimental result has approved the superiority of our predictor.

A Proposed Methodology to Evaluate Machine Learning Models at Near-Upper-Bound Predictive Performance—Some Practical Cases from the Steel Industry

“…Additional developments include the comparison of the predictive performances from three ML trained on data from five different EAF [17] and the evaluation of multiple ML models predicting the EE consumption with the aim to retrieve the optimal melting time for a given heat [18]. The published papers on ML models predicting the LRF end-point temperature are numerous [19][20][21][22][23][24][25][26][27]. However, the research primarily concerns the comparison of the predictive performance of models created using existing or newly developed ML modeling frameworks.…”

Dynamic Regressor/Ensemble Selection for a Multi-Frequency and Multi-Environment Path Loss Prediction

“…However, it failed on some public datasets [54]. Other applications in which dynamic ensemble regression was applied include prediction of mass flow rate and volume fraction [55], water quality prediction, and in predicting the temperature of molten steel in a ladle furnace [56]. We therefore investigate on DRS and DES to find out which of them is suitable for a path loss dataset.…”