基于大数据挖掘技术的热轧板带钢轧后冷却多目标优化  

Multi-objective Optimization of Cooling after Hot-rolling of Strip Steel Based on Big Data Mining Technology

作　　者：孙铁军[1] 王明瑞[2] 刘斌 崔文超 李鹏威[1] 曲丽萍 SUN Tiejun;WANG Mingrui;LIU Bin;CUI Wenchao;LI Pengwei;QU Liping(College of Electrical and Information Engineering,Beihua University,Jilin 132021,China;Shougang Jingtang United Iron&Steel Co.Ltd.,Tangshan 063200,China)

机构地区：[1]北华大学电气与信息工程学院,吉林吉林132021 [2]首钢京唐钢铁联合有限责任公司,河北唐山063200

出　　处：《北华大学学报（自然科学版）》2023年第6期820-827,共8页Journal of Beihua University（Natural Science）

基　　金：吉林省教育厅科学技术研究项目(JJKH20230064KJ,JJKH20210046KJ)。

摘　　要：针对热轧板带钢轧后超快速冷却终了温度和冷却路径的二维多目标优化问题,开发了基于大数据挖掘技术的多目标优化遗传算法,用来构建冷却集管最佳开闭决策库.依据Pareto占优机制,挖掘历代种群Pareto前沿面交集所提供的特征信息建立基因库,将其随机嵌入下一代种群,以高效抑制种群进化过程中所呈现的迁徙性、漫游性和随机性;基因库采用拥挤距离策略,有利于种群在全局决策空间搜索最优解群;较优模式段的数据清洗技术维持了Pareto前沿面中的决策变量相对于优化目标的匀称性,提升了系统柔性;DNA固化技术是引导算法收敛于全局最优集群的强力引擎,提高了系统控制精度.绘制仿真曲线(基于MFC),验证算法的合理性和先进性.A two-dimensional multi-objective optimization concerning the final temperature and cooling path for hot-rolled strip steel after rolling has been addressed by developing a multi-objective optimization genetic algorithm based on big data mining techniques,which is used to construct an optimal decision database for cooling manifold operations.Following the Pareto dominance mechanism,the feature information from the intersection of Pareto frontiers of previous generations is mined to establish a gene library.This library is randomly integrated into the next generation population to efficiently suppress the migration,wandering and randomness exhibited during population evolution.The gene library employs a crowding distance strategy to facilitate the population in searching for optimal solution clusters in the global decision space.Data cleaning techniques for superior pattern segments maintain the symmetry of decision variables relative to optimization objectives in the Pareto front,enhancing system flexibility.DNA solidification technology serves as a powerful engine to guide the algorithm towards convergence to the global optimal cluster,thereby improving system control precision.Simulation curves(based on MFC) are plotted to validate the rationality and sophistication of the algorithm.

关 键 词：热轧板带钢 多目标优化 Pareto占优机制 数据清洗 DNA固化技术 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]