Interpretable symbolic machine learning and optimization framework for reducing arcing time in energy‐intensive steelmaking.
Saved in:
| Title: | Interpretable symbolic machine learning and optimization framework for reducing arcing time in energy‐intensive steelmaking. |
|---|---|
| Authors: | Saha, Somashish1 (AUTHOR), Behera, Narottam1,2 (AUTHOR), Lahiri, Sandip Kumar1 (AUTHOR) sklahiri.che@nitdgp.ac.in |
| Source: | Canadian Journal of Chemical Engineering. Jul2026, Vol. 104 Issue 7, p3574-3592. 19p. |
| Subjects: | Genetic programming, Arc furnaces, Feature selection, Machine learning, Energy consumption, Process optimization, Process control systems |
| Abstract: | This study presents an interpretable, data‐driven framework for predicting and optimizing arcing time in direct reduced iron (DRI)‐based electric arc furnace (EAF) steelmaking using genetic programming (GP) and multi‐gene genetic programming (MGGP). Unlike traditional machine learning models, GP and MGGP generate explicit symbolic equations that not only deliver high prediction accuracy but also reveal the mathematical relationships between process inputs and arcing time. This transparency enhances operator trust and facilitates practical deployment, as the derived equations can be directly used for manual verification and process control. Real industrial data comprising 22 process variables from over 8692 heats were used for model training and validation. To ensure interpretability and industrial applicability, a novel feature‐selection algorithm using iterative MGGP relevance analysis was implemented yielding 12 key inputs. The resulting symbolic models were integrated with genetic algorithms (GA) to identify optimal input conditions for minimizing arcing time, leading to reduced energy consumption. MGGP models trained and validated on an 80/20 split, achieved R2 = 0.89 and RMSE = 0.906 min on training data, and R2 = 0.889 and RMSE = 0.932 min on test data. GA optimization reduced mean arcing time by 1.45 min and energy consumption by 4%, boosting productivity by 3.54% from 211.95 to 219.45 t/h. Sensitivity trends matched established metallurgical principles. The hybrid MGGP–GA framework generates transparent symbolic equations and prescriptive control settings, striking an effective balance between predictive fidelity and interpretability. This approach is well suited for real‐time EAF energy optimization and operator acceptance. [ABSTRACT FROM AUTHOR] |
| Copyright of Canadian Journal of Chemical Engineering is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
Be the first to leave a comment!
