摘要
为了解决热轧带钢凸度预测精度低、泛化能力差的问题,提出了高斯核和多项式核混合的支持向量机(SVM)预测模型。针对混合核支持向量机参数难以确定的问题,提出了使用佳点集、不完全伽玛函数自适应权重和可选择反向学习策略改进的河马算法(IHO)对混合核参数进行寻优。仿真实验结果表明,改进的河马算法的寻优速度快、收敛精度高。在凸度预测实验中,与随机森林、核极限学习机、单一高斯核支持向量机、多项式核支持向量机预测模型相比,混合核支持向量机预测模型的精度分别提高了18.49%、15.75%、28.76%和10.27%,对于实现轧制参数精准优化、有效改善板形边浪、楔形等缺陷具有重要意义。
In order to solve the problems of low prediction accuracy and poor generalization ability of hot-rolled strip convexity,a support vector machine(SVM)prediction model with a mixture of Gaussian kernel and polynomial kernel was proposed.For the problem that the parameters of the hybrid kernel support vector machine were difficult to determine,an improved hippopotamus optimization algorithm(IHO)was proposed to optimize the hybrid kernel parameters by using good point sets,incomplete gamma function adaptive weights and op-tional reverse learning strategy.Simulation experiment results show that the IHO algorithm has a fast optimization speed and high conver-gence accuracy.In the convexity prediction experiment,compared with the random forest,kernel extreme learning machine,single Gaussian kernel support vector machine and polynomial kernel support vector machine prediction models,the accuracy of the hybrid kernel support vector machine prediction model is improved by 18.49%,15.75%,28.76%and 10.27%,respectively,which is of great significance for achieving the accurate optimization of rolling parameters and effectively improving the defects such as plate edge waves and wedges.
作者
刘文广
李子轩
谢天伟
周亚罗
张瑞成
Liu Wenguang;Li Zixuan;Xie Tianwei;Zhou Yaluo;Zhang Ruicheng(Shougang Jingtang Iron and Steel United Co.,Ltd.,Tangshan 063200,China;College of Electrical Engineering,North China University of Science and Technology,Tangshan 063210,China;Beijing Shougang Co.,Ltd.,Beijing 064404,China)
出处
《锻压技术》
北大核心
2025年第7期132-142,共11页
Forging & Stamping Technology
基金
河北省自然科学基金资助项目(F2018209201)
唐山市科技局科技计划资助项目(22130213G)。
关键词
混合核支持向量机
带钢凸度
河马算法
热轧
预测精度
hybrid kernel support vector machine
strip convexity
hippopotamus algorithm
hot rolling
prediction accuracy
