摘要
针对传统轧制力预测模型存在的假设多、计算误差大、泛化性能差的问题,提出一种利用蜣螂优化算法(DBO)优化深度极限学习机(DELM)的冷连轧轧制力预测模型(DBO-DELM)。基于Bland-Ford-Hill轧制力模型分析,针对冷连轧各机架分别选取DELM轧制力预测模型的特征参数;采用国内某四机架冷连轧机组的实际生产数据,在用蜣螂优化算法优化确定DELM的偏置和权重参数的基础上,训练生成各机架的DBO-DELM轧制力预测模型。测试结果表明,DBO-DELM轧制力预测模型预测相对误差在±5%以内均可达77%以上,比现有MA-SVM、ELM-AE等模型在第二、第三机架上提高6%以上,第一、第四机架上提高10%以上。DBO-DELM轧制力预测模型在多个机架上相比于现有神经网络模型在预测精度上有明显的提高,并展现出良好的泛化能力,为冷连轧轧制力预测提供了一种有效的方法。
Aiming at the problems of many assumptions,large computational errors and poor generalisation performance of the traditional rolling force prediction model,a cold rolling force prediction model(DBO-DELM)using the dung beetle optimizer algorithm(DBO)to optimise the deep extreme learning machine(DELM)is proposed.Based on the Bland-Ford-Hill rolling force model,the characteristic parameters of the DELM rolling force prediction model are selected for each frame of cold continuous rolling.Using the actual production data of a four-frame cold continuous rolling unit in China,the DBO-DELM rolling force prediction model for each frame is generated based on the optimisation of the bias and weight parameters of DELM by the dung beetle optimizer algorithm.The test results show that the relative error of the DBO-DELM rolling force prediction model can reach more than 77%within±5%,which is more than 6%higher than that of the existing MA-SVM,DBN,ELM-AE models in the second and third frame,and more than 10%higher than that of the first and fourth frame.In summary,the DBO-DELM rolling force prediction model has obvious improvement in prediction accuracy compared with the existing neural network model in multiple frames and shows good generalisation ability,which provides an effective method for the rolling force prediction in cold continuous rolling.
作者
李晓阳
朴春慧
王雪雷
张明志
LI Xiaoyang;PIAO Chunhui;WANG Xuelei;ZHANG Mingzhi(School of Information Science and Technology,Shijiazhuang Tiedao University,Shijiazhuang 050043,China;Shijiazhuang Yangwang Electromechanical Technology Co.,Ltd.,Shijiazhuang 051432,China;Beijing National Railway Research&Design Institute of Signal&Communication Co.,Ltd.,Beijing 100070,China;Hebei Key Laboratory for Electromagnetic Environmental Effects and Information Processing,Shijiazhuang 050043,China)
出处
《哈尔滨理工大学学报》
北大核心
2024年第6期112-123,共12页
Journal of Harbin University of Science and Technology
基金
河北省重点研发计划项目(21355902D).
关键词
冷轧
数据驱动模型
蜣螂优化算法
深度极限学习机
轧制力预测
cold rolling
data-driven modelling
dung beetle optimizer
deep extreme learning machine
rolling force prediction
