基于数据驱动的连续退火生产过程建模与操作优化  

Data-driven modeling and operation optimization for continuous annealing process

作　　者：杨大雷 周永[2] 刘珧[2] 王显鹏[3] 

机构地区：[1]宝钢工业技术服务有限公司,上海201900 [2]宝山钢铁股份有限公司,上海201900 [3]东北大学,辽宁沈阳110819

出　　处：《宝钢技术》2015年第6期70-75,80,共7页Baosteel Technology

摘　　要：连退生产过程包含众多相互耦合的控制变量,并且带钢质量的检测采用离线方式,从而导致对于调质度为T5的带钢,实际生产中经常发生带钢硬度波动较大,难以实现对带钢产品质量的精确控制问题。针对这一问题,提出了一种基于数据驱动的连续退火生产过程建模方法,能够根据带钢退火前的生产信息及当前的生产过程信息实现对带钢产品质量的在线预报。在此基础上,提出了连续退火生产过程的多目标操作优化模型与算法,并开发了优化软件系统。基于实际生产过程数据的仿真试验结果表明,该软件系统能够为生产现场提供一套最优的控制变量设定值,从而实现产品质量最优化、能源消耗最小化和机组生产效率最大化的目标。Continuous annealing process consists of a number of control variables that are correlating with each other,and strip quality testing is carried out using the offline mode. This causes frequent occurrence of hardness fluctuations in practical production for T5 strips and makes it very difficult to achieve a precise control on strip quality. To solve the problem,a data-driven modeling method is proposed for the continuous annealing production process,which is able to provide online prediction of strip quality based on strip information before continuous annealing and current information of production process. Based on this modeling method, a multi-objective operation optimization model and the corresponding algorithm are proposed,and subsequently an optimization software system for the continuous annealing production process is developed. The simulation results based on practical production process data indicate that the software system is able to provide a set of optimal settings of control variable for practical production,based on which the optimization of product quality,minimization of energy consumption,and maximization of production efficiency can be achieved.

关 键 词：连续退火 数据驱动建模 多目标操作优化 

分 类 号：TG156.2[金属学及工艺—热处理]