Just-in-time updated DBN BOF steel-making soft sensor model based on dense connectivity of key features
Abstract: Due to the high-dimensional nonlinear nature of the BOF steelmaking production process data, although the ability of deep learning to extract abstract information is more prominent, it faces the challenge of low correlation between the extracted features and labels, and the static model cannot be applied to the forecasting requirements under changing working conditions. In order to enable deep learning to cope with these problems and maintain good prediction performance, this chapter proposes a Deep Belief Network (DBN) feature extraction model based on dense connectivity of key features. First, the key features are selected by feature importance judgment and redundancy judgment, and the selected key features are passed layer-by-layer through a densely connected structure. Second, a deep feature extraction network is formed by stacking layers to improve the feature extraction capability of the network. Finally, a Just-in-time learning (JITL) method is proposed to reduce the high-dimensional steelmaking data of the BOF while preserving the data structure by using the stream learning dimensionality reduction method to improve the accuracy of the metrics in the JITL process, so that the online fine-tuned model can be applied to the forecasting requirements under different working conditions. According to the actual BOF steel production process data, the prediction accuracy of the terminal carbon content reached 82.0% within the error range of ±0.02%, and the prediction accuracy of the temperature reached 80.0% within the error range of ±10°C.
- Location
-
Deutsche Nationalbibliothek Frankfurt am Main
- Extent
-
Online-Ressource
- Language
-
Englisch
- Bibliographic citation
-
Just-in-time updated DBN BOF steel-making soft sensor model based on dense connectivity of key features ; volume:43 ; number:1 ; year:2024 ; extent:17
High temperature materials and processes ; 43, Heft 1 (2024) (gesamt 17)
- Creator
-
Lu, Zongxu
Liu, Hui
Chen, FuGang
Li, Heng
Xue, XiaoJun
- DOI
-
10.1515/htmp-2024-0060
- URN
-
urn:nbn:de:101:1-2412141748321.301808093426
- Rights
-
Open Access; Der Zugriff auf das Objekt ist unbeschränkt möglich.
- Last update
-
15.08.2025, 7:30 AM CEST
Data provider
Deutsche Nationalbibliothek. If you have any questions about the object, please contact the data provider.
Associated
- Lu, Zongxu
- Liu, Hui
- Chen, FuGang
- Li, Heng
- Xue, XiaoJun