基于GA-BP和PSO-BP神经网络的6061铝合金板材流变应力预测模型  

Flow Stress Prediction Model of 6061 Aluminum Alloy Sheet Based on GA-BP and PSO-BP Neural Networks

作　　者：丁凤娟 贾向东[2] 洪腾蛟 徐幼林[2] Ding Fengjuan;Jia Xiangdong;Hong Tengjiao;Xu Youlin(Nanjing Forestry University,Nanjing 21003,China;Anhui Science and Technology University,Bengbu 233100,China;College of Mechanical and Electronic Engineering,Nanjing Forestry University,Nanjing 21003,China)

机构地区：[1]南京林业大学机械电子工程学院,江苏南京210037 [2]南京林业大学,江苏南京210037 [3]安徽科技学院,安徽蚌埠233100

出　　处：《稀有金属材料与工程》2020年第6期1840-1853,共14页Rare Metal Materials and Engineering

基　　金：Natural Science Foundation of Jiangsu Higher Education Institutions of China(18KJB460020);High-level(Higher Education)Science Foundation of Nanjing Forestry University(GXL2018020);Youth Science and Technology Innovation Foundation of Nanjing Forestry University(CX2018027)。

摘　　要：6061铝合金作为一种热可强化铝合金,具有良好的成形性能,但是其塑性流变应力受最终热处理工艺的加热温度、保温时间和冷却方式等参数的影响很大。因此,为了获得最终热处理工艺参数对6061铝合金板材的塑性性能及流变行为的影响,试验中以6061-T6铝合金板材为研究对象,通过单向拉伸试验、金相实验和硬度测试等方法研究不同热处理工艺参数(加热温度为500、530、560和590℃、保温时间2 h、冷却方式为空冷)对6061铝合金塑性性能和硬度的影响。通过单向拉伸试验获取不同热处理工艺参数条件下6061铝合金的真实应力-应变曲线;借助BP、GA-BP和PSO-BP神经网络构建不同热处理温度条件下6061铝合金的本构关系模型。结果表明,BP、GA-BP和PSO-BP神经网络模型均能较好的拟合不同热处理温度条件下6061铝合金的流变行为,但是PSO-BP神经网络模型对6061铝合金流变应力的预测精度更高,模型预测性能更优越,其平均绝对误差(MAE),平均相对误差(AARE)和相关系数(R^2)分别为1.89,1.56%和0.9965。Taking 6061-T6 aluminum alloy cold-rolled sheet as the research object,the plastic deformation behavior of 6061 aluminum alloy at different heat treatment temperatures(500,530,560 and 590°C)was analyzed through uniaxial tensile test,metallographic test and microhardness test.Combined with experimental data and BP,GA-BP and PSO-BP neural networks,the constitutive models of this material under different heat treatment temperature conditions were constructed.The results show that BP,GA-BP and PSO-BP neural network models can better fit the flow behavior of 6061 aluminum alloy under different heat treatment temperature conditions,but PSO-BP neural network model has higher prediction accuracy and good performance in predicting the flow stress of 6061 aluminum alloy,and its average absolute error(MAE),average relative error(AARE)and the correlation coefficient(R^2)are 1.89.1.56%and 0.9965,respectively.

关 键 词：6061铝合金 流变应力 人工神经网络 遗传算法 粒子群优化 热处理工艺 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TG146.21[自动化与计算机技术—控制科学与工程]