Elsevier

Applied Soft Computing

Volume 131, December 2022, 109724
Applied Soft Computing

A framework based on heterogeneous ensemble models for liquid steel temperature prediction in LF refining process

https://doi.org/10.1016/j.asoc.2022.109724Get rights and content

Highlights

  • Utilizing heterogeneous ensemble models to predict the liquid steel temperature in the LF refining process.

  • Proposing Recursive Feature Increase to facilitate the construction of heterogeneous ensemble models.

  • Proposing Recursive Search Optimization to optimize the hyper-parameters of heterogeneous ensemble models.

Abstract

The precise control of liquid steel temperature in the ladle furnace (LF) refining process is vital for stabilizing and improving the quality of liquid steel, necessitating a capable prediction system. To achieve better predictive performance, an effective prediction framework based on heterogeneous ensemble models is proposed in this paper, which mainly consists of three parts: (1) utilizing single models and tree-based ensemble models to constitute the heterogeneous ensemble model; (2) proposing Recursive Feature Increase (RFI) to facilitate the construction of the heterogeneous ensemble model, including Stacking and Majority Voting; (3) proposing a new optimization algorithm, namely Recursive Search Optimization (RSO), to optimize the hyper-parameters of the heterogeneous ensemble model. Through the verification of the collected industrial production data, it is found that the proposed framework in this paper possesses higher fitting and generalization ability, which is of great significance for engineering applications such as liquid steel temperature prediction in the LF refining process.

Section snippets

Code metadata

Permanent link to reproducible Capsule: https://doi.org/10.24433/CO.3289734.v1.

Smelting process description and data collection

Before continuous casting, the compact steelmaking process mainly includes Consteel electric arc furnacedeoxidation and alloyingLF refining. The Consteel electric arc furnace is in the early stage of producing liquid steel, while the LF refining is to adjust the composition and temperature of liquid steel to improve product quality and continuous casting effect. The deoxidation and alloying is an important stage linking the Consteel electric arc furnace and the LF refining, in which the

Review of tree-based ensemble models

Typical representatives of tree-based ensemble models, including Random Forests [16], Extremely Randomized Trees (Extra Trees) [17], eXtreme Gradient Boosting (XGBoost) [18], Gradient Boosting Decision Tree (GBDT) [19], and AdaBoost-CART [20], are adopted as main candidate models for the heterogeneous ensemble model. Tree-based ensemble models include two main frameworks, Bagging and Boosting [21], [22], [23], [24], which are better strategies to trade off the fitting and generalization ability

Construction of heterogeneous ensemble models

The heterogeneous ensemble model mainly includes two model structures, Stacking and Majority Voting (MV), as shown in Fig. 3. The main difference between Stacking and MV is that in layer 2, Stacking requires a suitable meta-learner to integrate the results of layer 1 to form the final result, while MV only votes for the final result based on the results of layer 1.

To facilitate the construction of the heterogeneous ensemble model, Recursive Feature Increase (RFI) is proposed in this paper. The

Hyper-parameter optimization of heterogeneous ensemble models

In the above section, the RFI is employed to construct the heterogeneous ensemble model. In this section, we will optimize the hyper-parameters of the heterogeneous ensemble model to further improve the predictive performance.

Conclusion

In this study, a novel framework based on heterogeneous ensemble models is proposed for the liquid steel temperature prediction in the LF refining process. In this process, various machine learning models, including single models and tree-based ensemble models, are selected as candidate models for the heterogeneous ensemble model, and meanwhile, it is found that tree-based ensemble models possess better predictive performance for the liquid steel temperature prediction in the LF refining

CRediT authorship contribution statement

Chao Chen: Conceptualization, Methodology, Data curation, Project administration, Investigation, Validation, Writing – review & editing, Software. Nan Wang: Writing – review & editing, Project administration, Funding acquisition, Supervision. Min Chen: Writing – review & editing, Project administration, Funding acquisition, Supervision. Xumei Yan: Writing – review & editing, Data curation.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

The authors gratefully acknowledge the National Natural Science Foundation of China. [Grant numbers: 52074077, 52174301, and 51974080] and the Fundamental Research Funds for the Central Universities supported by the Chinese Education Ministry [Grant number N2125018].

References (49)

  • SunJ.L. et al.

    Based on multi-algorithm hybrid method to predict the slope safety factor-stacking ensemble learning with Bayesian optimization

    J. Comput. Sci-Neth.

    (2022)
  • CuiS.Z. et al.

    A stacking-based ensemble learning method for earthquake casualty prediction

    Appl. Soft Comput.

    (2021)
  • FreundY. et al.

    A decision-theoretic generalization of on-line learning and an application to boosting

    J. Comput. System Sci.

    (1997)
  • GonzálezS. et al.

    A practical tutorial on bagging and boosting based ensembles for machine learning: Algorithms, software tools, performance study, practical perspectives and opportunities

    Inform. Fusion

    (2020)
  • AcostaS.M. et al.

    Relevance vector machine with tuning based on self-adaptive differential evolution approach for predictive modelling of a chemical process

    Appl. Math. Model.

    (2021)
  • CherkasskyV. et al.

    Practical selection of SVM parameters and noise estimation for SVM regression

    Neural Netw.

    (2004)
  • LinG.C. et al.

    Using support vector regression and k-nearest neighbors for short-term traffic flow prediction based on maximal information coefficient

    Inform. Sciences

    (2022)
  • LiuQ.S. et al.

    Prediction model of rock mass class using classification and regression tree integrated AdaBoost algorithm based on TBM driving data

    Tunn. Undergr. Space Technol.

    (2020)
  • HuangX.L. et al.

    Fair-AdaBoost: Extending AdaBoost method to achieve fair classification

    Expert Syst. Appl.

    (2022)
  • YaoJ.R. et al.

    Novel hybrid ensemble credit scoring model with stacking-based noise detection and weight assignment

    Expert Syst. Appl.

    (2022)
  • Żogała SiudemB. et al.

    Fast stepwise regression based on multidimensional indexes

    Inform. Sci.

    (2021)
  • LarroqueA. et al.

    Cylinder drag minimization through wall actuation: A Bayesian optimization approach

    Comput. Fluids

    (2022)
  • JoyT.T. et al.

    Fast hyperparameter tuning using Bayesian optimization with directional derivatives

    Knowl.-Based Syst.

    (2020)
  • KarabogaD. et al.

    On the performance of artificial bee colony (ABC) algorithm

    Appl. Soft Comput.

    (2008)
  • Cited by (3)

    The code (and data) in this article has been certified as Reproducible by Code Ocean: (https://codeocean.com/). More information on the Reproducibility Badge Initiative is available at https://www.elsevier.com/physical-sciences-and-engineering/computer-science/journals.

    View full text