1. Application of the gradient boosting decision tree in the online prediction of rolling force in hot rolling.
- Author
-
Wang, Qiuna, Song, Lebao, Zhao, Jianwei, Wang, Haiyu, Dong, Lijie, Wang, Xiaochen, and Yang, Quan
- Subjects
- *
HOT rolling , *DECISION trees , *ARTIFICIAL neural networks , *RANDOM forest algorithms , *PREDICTION models , *FORECASTING - Abstract
The prediction accuracy of the rolling force is crucial for the strip hot rolling process, which will significantly affect the dimensional control accuracy of the strip product. The rolling force prediction models used in actual strip hot rolling production before are mainly analytical models and simple neural network models, but with the improvement of product quality, these models can no longer meet the requirements of high-end product production. A new online model based on the gradient boosting decision tree (GBDT) method is proposed to improve the accuracy of the online prediction of rolling force, in which the random forest method based on feature importance is adopted to select feature parameters. A model self-training function was developed in the control system to ensure the accuracy of the model used online. By comparing various machine learning methods, the results show that the rolling force prediction model proposed by the GBDT is better than that based on other regression methods. The established model has been successfully applied to predict the rolling force for the finishing rolling in a 2250 mm strip hot rolling production line. Compared with the traditional model, the rolling force prediction accuracy and thickness control accuracy are significantly improved. [ABSTRACT FROM AUTHOR]
- Published
- 2023
- Full Text
- View/download PDF