基于热连轧实测数据的模型钢族层别优化

Optimization for steel grade family of model based on measured data during hot continuous rolling

钢族层别是热连轧过程控制最重要的基础配置数据之一,对轧制模型设定精度具有重要影响。现有的钢族层别划分以人工经验为主,主要依靠碳当量来对钢族进行划分,存在划分不合理、难以满足实际生产需要等不足。鉴于此,提出一种新的线聚类算法来对钢族层别分类进行优化。首先,大量收集某钢族层别的轧制历史数据;其次,绘制各带钢变形速率与变形抗力数据的散点图,并观察数据散点的带状分布特征;最后,采用线聚类算法来优化钢族层别的分类,挑选分离度较大的钢种到新的钢族。目前,该技术已在宝钢1 880等多个热连轧模型调试中使用,对热连轧工程的达标达产和产品验证起到重要支撑作用。

The steel grade family is one of the most important basic configuration data for the process control during hot continuous rolling. It has an important influence on the setting precision of rolling models. The existing divisions of the steel grade family are mainly based on the artificial experience, mainly relying on the carbon equivalent, which is unreasonable and difficult to meet the actual production needs. In view of this, a new line clustering algorithm to optimize the classification of steel grade family is proposed. First, collect a large number of rolling historical data from a certain steel grade family. Then, draw a scatter diagram of the strip steel about the deformation rate and the deformation resistance, and observe the distribution characteristics of the data. Finally, use the linear clustering algorithm to optimize the classification of the steel grade family, and select the steel with larger separation degree to an new steel group. At present, the technology has been used in the commissioning of several hot rolling models such as Baosteel 1 880, which has played an important role in the standard production and the product verification of the hot rolling engineering.