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.

摘要

高精度的折弯力预设是保证带材头平整度的前提，但工业现场使用的预设模型存在局限性。根据工作辊和中间辊弯曲力的板形控制特点，综合某台1420 mm六辊冷连轧机的历史生产数据，建立了基于符号回归的弯曲力预置链模型，并用年龄分层人口结构遗传规划（ALPS-GPSR）。符号回归模型的优点是能够获得显式的数学表达式，这在复杂的工业生产环境中非常重要。随着基于ALPS-GPSR的弯曲力预设模型在工业现场的应用，