基于改进蜣螂优化自学习算法的轧辊偏心在线辨识  

Roll Eccentricity Online Identification Based on Improved Dung Beetle Optimization Self-learning Algorithm

作　　者：韩啸[1] 张飞[1] 郭强[1] 王丽君[2] 黄学忠 Han Xiao;Zhang Fei;Guo Qiang;Wang Lijun;Huang Xuezhong(National Engineering Research Center for Advanced Rolling and Intelligent Manufacturing,University of Science and Technology Beijing,Beijing 100083;School of Automation and Electrical Engineering,University of Science and Technology Beijing,Beijing 100083;Technical Research Institute of Guangxi Beigang New Material Co.,Ltd.,Beihai Guangxi 536017)

机构地区：[1]北京科技大学高效轧制与智能制造国家工程研究中心,北京100083 [2]北京科技大学自动化学院,北京100083 [3]广西北港新材料有限公司技术研究院,广西北海536017

出　　处：《中国仪器仪表》2025年第2期29-35,共7页China Instrumentation

基　　金：国家重点研发计划重点专项(2022YFB3304002-02);广西重点研发计划(桂科AB21196025)。

摘　　要：轧辊偏心直接影响轧制力的分布,进而影响轧制精度。偏心离线辨识方法在生产中效果不明显甚至引起反作用。针对这一问题,本文提出了一种融合鱼鹰算法与自适应t分布的蜣螂算法优化的自学习算法的在线辨识方法。首先,该算法使用Logistic混沌映射序列对种群进行初始化;同时,融合鱼鹰算法改进了蜣螂滚球行为;此外,引入自适应t分布扰动策略。其次,使用改进蜣螂算法寻优,得到最佳的支撑辊划分数值。最后,采用自适应系数的自学习方法对偏心进行在线辨识,并增加幅值限定和周期自动校正方法,实现对偏心的在线调整。结果表明,通过对轧制过程中数据的不断更新和学习,该方法在轧制过程中能够有效地跟踪轧辊偏心的变化,提高了偏心在线辨识的准确性。Roll eccentricity directly affects the distribution of rolling force,thus influencing rolling precision.Offline eccentricity identification methods have shown unclear effectiveness or even caused adverse effects in production.To address this issue,this paper proposes an online identification method based on a self-learning algorithm optimized with a combination of the Osprey algorithm and adaptive t-distribution within the Dung Beetle Optimizer Algorithm(OTDBO).Firstly,the algorithm initializes the population using a Logistic chaotic mapping sequence.Concurrently,it integrates the fish-eagle algorithm to enhance the rolling behavior of the dung beetles.In addition,an adaptive t-distribution disturbance strategy is introduced.Secondly,the improved dung beetle algorithm is employed for optimization to obtain the optimal division values for the backup roll.Finally,an adaptive coefficient self-learning approach is adopted for online eccentricity identification.Additionally,amplitude limitation and automatic periodic correction methods are incorporated to achieve online adjustment of eccentricity.The results indicate that by continuously updating and learning from data during the rolling process,this method effectively tracks the variations in roll eccentricity,thereby enhancing the accuracy of online eccentricity identification.

关 键 词：蜣螂优化算法 自学习算法 轧辊偏心 在线辨识 

分 类 号：TG3[金属学及工艺—金属压力加工]