Preset model of bending force in 6-high universal crown tandem cold rolling mill based on symbolic regression

ABSTRACT

Highly accurate bending force presetting is a prerequisite to ensure the well flatness of the strip head, however, there are limitations in the preset model used in industrial sites. According to the flatness control characteristics of work roll and intermediate roll bending forces, this paper integrates the historical production data of a 1420 mm 6-high tandem cold rolling mill and establishes a chain model of bending force preset based on symbolic regression, which implements with age-layered population structure genetic programming (ALPS-GPSR). The advantage of symbolic regression models is the ability to obtain explicit mathematical expressions, which is very important in complex industrial production environments. With the application of ALPS-GPSR based bending force preset model in the industrial site, the hit rate of flatness within 6IU in the strip head for three specifications of the strip has increased by 2.86%, 1.61%, and 2.19% respectively on average, achieving a satisfactory result.

KEYWORDS:

• Preset of bending force
• 6-high universal crown mill
• tandem cold rolling
• flatness control
• data-driven model
• symbolic regression
• genetic programming
• age-layered population structure

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The authors confirm that the data and material supporting the findings of this work are available within the article.

Authors’ contributions

Zedong Wu performed the modelling work and completed the first draft; Quan Yang guided the revision of the first draft; Xiaochen Wang provided constructive suggestions on modelling; Dong Xu made additions to the structure and completeness of the first draft; Jianwei Zhao helped write and debug part of the algorithm; Jingdong Li collected the data and performed the validation.

Consent for publication

This work is approved by all authors for publication.

Additional information

Funding

This work was supported by the National Natural Science Foundation of China [grant number 51975043], China Postdoctoral Science Foundation [grant number 2021M69035].