Optimization of CVC shifting mode for hot strip mill based on the proposed LightGBM prediction model of roll shifting

In the routine roll shifting mode, work rolls of the continuous variable crown (CVC) hot strip mill are always in repeated shifting positions, which affects the uniform wear of work rolls. As an available solution to the above problem, a new random shifting mode for CVC work rolls has been developed in this paper. According to the relationship between shifting position and bending force, the new CVC shifting mode shifts work rolls in a random pattern within the limits by randomly changing the bending force, so that the roll shifting is dispersed and the strip shape remains good. The Light Gradient Boosting Machine (LightGBM) algorithm is applied to build the prediction models of CVC shifting to accurately express the relationship between shifting position and bending force. Random search and Bayesian optimization are used to optimize the LightGBM models, respectively. By comparison, LightGBM with Bayesian optimization is recommended to predict roll shifting, which is more accurate and efficient than using random search. The new CVC shifting mode has been implemented by an off-line application in the 1780 mm hot rolling line. The results reveal that the proposed CVC shifting mode can well disperse roll shifting positions and accurately control strip shape.

Automated summary

A new random shifting mode for CVC work rolls has been developed to disperse roll shifting positions and maintain good strip shape by randomly changing bending force within limits. The Light Gradient Boosting Machine algorithm is applied to accurately express the relationship between shifting position and bending force, and Bayesian optimization is recommended for optimizing the prediction models. The proposed CVC shifting mode has been successfully implemented offline in a hot rolling line, demonstrating effective control of strip shape.