A hybrid approach using multiple linear regression and random forest regression to predict molten steel temperature in a continuous casting tundish

ABSTRACT

The temperature control of molten steel is essential to ensure operational stability in a steelmaking plant. The calculation of thermal losses in the steelmaking plant’s operations depends on highly dynamic variables, which motivates the construction of predictive models for the steel temperature. This paper proposed a hybrid ensemble method using multiple linear and random forest regression to predict the end molten steel temperature at the secondary refining required to achieve a target tundish temperature. Combining these two methods makes it possible to account for the linear and non-linear relationships in the data. The implemented models were trained on industrial data, and their performance was assessed using root mean squared error (RMSE) and a custom accuracy metric. The results showed that the proposed hybrid method achieves up to 5% better accuracy compared to linear regression or random forest regression methods alone, thus can enhance molten steel prediction in steelmaking plants.

KEYWORDS:

• Steel temperature prediction
• Continuous casting superheat
• Regression analysis
• Machine learning
• Boosting algorithm
• Hybrid ensemble regression
• Multiple linear regression
• Random forest regression

Acknowledgements

The authors are grateful to Gerdau Ouro Branco for kindly providing the industrial data for this study. They are also thankful to Redemat, Universidade Federal de Ouro Preto, CNPq and FAPEMIG for all their support.

Disclosure statement

No potential conflict of interest was reported by the authors.