.png){kind=icon}
.png){kind=icon}
Return
A hybrid approach for feature reconstruction and its applications in liquid steel end-point temperature prediction during vacuum tank degassing
.png){kind=icon}
PRE
AI
W
W
DOI：10.7717/peerj-cs.3696.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This study puts forward a novel approach for the feature reconstruction of the liquid steel end-point temperature prediction model in the vacuum tank degassing (VTD) process. A new feature-reconstruction method is developed to tackle the issue of limited variable characteristics in the discrete VTD steelmaking process. Basic mathematical functions are employed to generate new features from the initial dataset, aiming to adequately expand the feature search space. Subsequently, the generated features are extracted through the score of the maximal information coefficient. This integrated method that combines feature generation and feature selection is termed feature reconstruction, which enables the effective selection of an optimal set of superior features. These features are then utilized as inputs for a neural regressor to predict the liquid steel end-point temperature. Comprehensive experimental comparisons are conducted between the proposed feature-reconstruction method and four typical machine learning algorithms (artificial neural networks (ANN), support vector regression (SVR), extreme learning machine (ELM), and self-adaptive Adaboost.RT (SAE-ELM)) using real-world steelmaking plant data. The results verify that the proposed method shows promising enhancement in temperature prediction precision.
Keywords:
Feature reconstruction
Machine learning
Vacuum tank degassing
Temperature prediction
Journal
IF:
2.5
Papers:
3.3K
Citations:
6.9K
Organization
Cited Papers