基于随机森林算法的热轧精轧带钢宽展量预测  

Spread prediction of hot finish rolled strip steel based on random forest algorithm

作　　者：邱振波 李子正 尹宝良 邝霜 陈彤 白振华 QIU Zhen-bo;LI Zi-zheng;YIN Bao-liang;KUANG Shuang;CHEN Tong;BAI Zhen-hua(National Engineering Research Center for Equipment and Technology of Cold Strip Rolling,Yanshan University,Qinhuangdao 066004,China;Tangshan Iron and Steel Group Co.,Ltd.,Tangshan 063000,China;State Key Laboratory of Metastable Materials Science and Technology,Yanshan University,Qinhuangdao 066004,China)

机构地区：[1]燕山大学国家冷轧板带装备及工艺工程技术研究中心,河北秦皇岛066004 [2]唐山钢铁集团有限责任公司,河北唐山063000 [3]燕山大学亚稳材料制备技术与科学国家重点实验室,河北秦皇岛066004

出　　处：《塑性工程学报》2023年第8期107-114,共8页Journal of Plasticity Engineering

基　　金：河北省重大科技成果转化专项(22281001Z)。

摘　　要：基于随机森林算法对带钢的宽展量进行了预测。基于对宽展量影响因素的分析选择了多个重要的工艺因素,利用袋外数据计算了不同工艺因素的重要性。考虑模型的预测精度和训练时间,提出了一种模型综合评价指标,利用该评价指标筛选出模型具有最优综合预测性能时的重要工艺因素,并建立了宽展量预测模型。考虑现场需对带钢宽展进行在线预测,设定误差阈值并计算滑动平均绝对百分误差可实现模型预测精度的动态调整。该模型在测试数据集上具有较小的预测误差以及较快的模型调整速度,表明该优化模型具有较好的综合预测性能。Based on random forest algorithm,the spread of strip steel was predicted,based on the analysis of the influencing factors of spread,multiple important process factors were selected,and the importance of different process factors was calculated using out of bag data.Considering the prediction accuracy and training time of the model,a comprehensive evaluation index for the model was proposed,which was used to screen out important process factors when the model has the optimal comprehensive prediction performance,and the wide spread prediction model was established.Considering the need for online prediction of strip steel width on site,setting error thresholds and calculating the sliding average absolute percentage error can achieve dynamic adjustment of the prediction accuracy of the model.The model has small prediction error and fast model adjustment speed on the test dataset,indicating that the optimized model has good comprehensive prediction performance.

关 键 词：热轧精轧 宽展量 优化 随机森林 综合预测性能 

分 类 号：TG333.5[金属学及工艺—金属压力加工]