摘要
本文将鲸鱼优化算法引入SVR算法中,建立镁合金轧制温度、下压量、轧制线速度、轧后厚度和轧辊半径与抗拉强度和伸长率的联系。采用Person图分析了AZ31镁合金轧板轧制工艺参数与力学性能的相关性;采用SHAP解释模型结合轧制机理分析,确定了影响力学性能的主要轧制工艺参数。结果表明:轧制线速度、轧辊半径和轧后厚度相较于平均应变速率与镁合金抗拉强度和伸长率线性相关性更强;优化后的SVR模型预测抗拉强度的相关系数从0.7提升到0.86,平均绝对误差从7.4降低到5.2,平均相对误差从0.16%降低为0.11%,均方误差从8.9降低到6.2;伸长率的相关系数从0.59提升到0.85,平均绝对误差从2.3降低到1.4,平均相对误差从0.56%降低为0.39%,均方误差从3降低到1.8。由SHAP解释模型结果分析可知,抗拉强度主要受到下压量的影响,其次是轧后厚度和轧辊半径;伸长率主要受到轧制线速度的影响,其次是温度和轧辊半径。
In this study,the whale optimization algorithm(WOA)was introduced into the support vector regression(SVR)algorithm.Then,the relationships among magnesium alloy rolling parameters(temperature,reduction,rolling speed,post-rolling thickness and roll radius)and mechanical properties(tensile strength and elongation)were established.The correlations among rolling process parameters and mechanical properties were analyzed using Pearson plots.The SHAP interpretation model,combined with rolling mechanism analysis,were used to identify the key rolling parameters influencing mechanical properties.The results demonstrate that rolling line speed,roll radius,and post-rolling thickness exhibit stronger linear correlations with the tensile strength and elongation of magnesium alloys compared to the average strain rate.After optimization,the SVR model show significant improvements:for tensile strength prediction,the correlation coefficient increases from 0.70 to 0.86,MAE decreases from 7.4 to 5.2,MAPE reduces from 0.16%to 0.11%,and MSE drops from 8.9 to 6.2.For elongation prediction,the correlation coefficient improves from 0.59 to 0.85,MAE decreases from 2.3 to 1.4,MAPE reduces from 0.56%to 0.39%,and MSE decreases from 3.0 to 1.8.The SHAP interpretation model analysis reveal that tensile strength is primarily influenced by the reduction amount,followed by post-rolling thickness and roll radius,and the elongation of Mg alloy rolled sheet is mainly affected by rolling line speed,with temperature and roll radius being secondary influencing factors.
作者
刘筱
林煜昕
朱必武
刘文辉
郭鹏程
徐从昌
李落星
LIU Xiao;LIN Yuxin;ZHU Biwu;LIU Wenhui;GUO Pengcheng;XU Congchang;LI Luoxing(College of Marine Equipment and Mechanical Engineering,Jimei University,Xiamen 361021,China;Chongqing Research Institute,Hunan University,Chongqing 401135,China;College of Smart Manufacturing and Mechanical Engineering,Hunan Institute of Technology,Hengyang 421002,China)
出处
《中国有色金属学报》
北大核心
2025年第9期2925-2937,共13页
The Chinese Journal of Nonferrous Metals
基金
国家自然科学基金资助项目(52471132,52475356,52071139,U20A20275)
福建省科技计划杰出青年基金项目(2024J010031)
重庆市自然科学基金资助项目(CSTB2023NSCQ-MSX0886)。
关键词
鲸鱼优化算法
支持向量回归
AZ31镁合金
轧制工艺
Pearson相关系数
whale optimization algorithm
support vector regression algorithm
AZ31 magnesium alloy
rolling process
Pearson correlation coefficient
作者简介
通信作者:朱必武,副教授,博士,电话:18674355439,E-mail:zmbh4538@163.com。
