中高应变速率轧制AZ31镁合金板的抗拉强度预测  

Prediction of tensile strength of AZ31 magnesium alloy sheet rolled at medium-high strain rate

作　　者：朱必武[1,4] 蒋昊 刘筱 郭鹏程[3] 魏福安 徐从昌 李落星 ZHU Biwu;JIANG Hao;LIU Xiao;GUO Pengcheng;WEI Fuan;XU Congchang;LI Luoxing(Hunan Engineering Research Center of Evaluation of Forming Technology and Damage Resistance of High Efficiency Light Alloy Components,School of Mechanical and Electrical Engineering,Hunan University of Science and Technology,Xiangtan 411201,China;Salt Lake Chemical Engineering Research Complex,Qinghai University,Xining 810016,China;Chongqing Research Institute,Hunan University,Chongqing 401135,China;College of Marine Equipment and Mechanical Engineering,Jimei University,Xiamen 361021,China)

机构地区：[1]湖南科技大学机电工程学院,高功效轻合金构件成形技术及耐损伤性能评价湖南省工程研究中心,湘潭411201 [2]青海大学部省合建盐湖化工大型系列研究设施,西宁810016 [3]湖南大学重庆研究院,重庆401135 [4]集美大学海洋装备与机械工程学院,厦门361021

出　　处：《中国有色金属学报》2024年第6期1998-2007,共10页The Chinese Journal of Nonferrous Metals

基　　金：国家自然科学基金资助项目(52071139);湖南省自然科学基金资助项目(2023JJ30252,2023JJ30262);重庆市自然科学基金资助项目(CSTB2023NSCQ-MSX0886);青海大学盐湖化工大型系列研究设施开放研究项目(2023-DXSSKF-04)。

摘　　要：本文采集了不同轧制温度、应变速率以及压下量等三个工艺参数下中高应变速率轧制的AZ31镁合金抗拉强度的27组样本,通过惯性权重、学习因子的改进和引入变异操作对PSO-BP神经网络进行改进,并与BP、PSO-BP神经网络对比,进行抗拉强度的预测。结果表明:BP神经网络不能预测AZ31镁合金抗拉强度的非线性变化,PSO-BP神经网络和改进的PSO-BP(IPSO-BP)神经网络均能较好地预测AZ31镁合金抗拉强度的非线性变化;这三个模型中IPSO-BP神经网络预测最为准确,相较于PSO-BP神经网络,其平均绝对误差从15.3764降低至3.4288,平均相对误差从5.94%降低至1.32%,均方误差从251.3662降低至20.7199,相关系数从0.7753提高至0.8937;通过Pearson相关性计算判断出应变速率、压下量对抗拉强度的影响均大于轧制温度,而应变速率与抗拉强度呈负相关关系,压下量与抗拉强度呈正相关关系。The 27 groups of the tensile strength of AZ31 magnesium alloy rolled at medium and high strain rate under different rolling temperatures,strain rates and reduction were collected.The PSO-BP neural network was improved by modifying inertia weight,learning factor and introducing variation operation.The tensile strength predicted by improved PSO-BP(IPSO-BP) was compared with BP and PSO-BP neural network.The results show that BP neural network cannot predict the nonlinear change of tensile strength of AZ31 magnesium alloy,PSO-BP and IPSO-BP neural network can predict the nonlinear change of tensile strength of AZ31 magnesium alloy.Among the three models,IPSO-BP neural network has the most accurate prediction.In contrast to PSO-BP neural network,the absolute error decreases from 15.3764 to 3.4288,the average relative error decreases from 5.94% to 1.32%,and the mean square error decreases from 251.3662 to 20.7199.The correlation coefficient increases from 0.7753 to 0.8937.The effect of strain rate and reduction on tensile strength is greater than that of rolling temperature by analyzing Pearson correlation calculation.Strain rate is negatively correlated with tensile strength and reduction is positively correlated with tensile strength.

关 键 词：AZ31镁合金 神经网络 轧制工艺 抗拉强度 Pearson相关系数 

分 类 号：TG146.22[一般工业技术—材料科学与工程]