摘要
通过不同温度(760、810、860和910℃)和不同应变速率(0.01、0.1和0.5 s^(-1))下的等温拉伸实验,研究了TA15钛合金板材的拉伸变形和软化行为。为了对TA15钛合金的高温拉伸变形行为实现精确预测,建立了基于深度神经网络(DNN)的TA15钛合金高温拉伸本构模型。结果表明,DNN模型能够准确地预测TA15钛合金在不同拉伸变形条件下的流动应力,预测结果与实际结果的平均绝对误差为1.3%,相关系数可达0.999。并且,相比于单个隐含层的神经网络,具有多个隐含层的DNN模型具有更高的预测精度和更好的泛化能力。此外,构建了TA15钛合金热加工图,并且通过理论分析验证了此热加工图,结果表明了此热加工图的准确性和有效性。
The tensile deformation and softening behavior of TA15 titanium alloy sheet were studied by isothermal tensile experiments at different temperatures(760,810,860,and 910℃)and different strain rates(0.01,0.1 and 0.5 s^(-1)),and a high temperature tensile constitutive model of TA15 titanium alloy based on deep neural network(DNN)was established to accurately predict the high temperature tensile deformation behavior of TA15 titanium alloy.The results show that DNN model accurately predicts the flow stress of TA15 titanium alloy under different tensile deformation conditions,the average absolute error between the predicted results and the actual results is 1.3%,and the correlation coefficient reaches 0.999.Moreover,compared with a neural network with a single hidden layer,DNN model with multiple hidden layers has higher prediction accuracy and better generalization ability.In addition,the hot processing map of TA15 titanium alloy is constructed,and the theoretical analysis is carried out to verify the hot processing map,which shows the accuracy and effectiveness of the hot processing map.
作者
唐学峰
黄振
温红宁
王欣
刘鹏
刘钊
王新云
Tang Xuefeng;Huang Zhen;Wen Hongning;Wang Xin;Liu Peng;Liu Zhao;Wang Xinyun(State Key Laboratory of Material Processing and Die and Mould Technology,Huazhong University of Science and Technology,Wuhan 430074,China;Daijiehanjin(Cangzhou)Precision Mould Co.,Ltd.,Huanghua 061100,China;School of Mechanical Engineering,Shanghai Dianji University,Shanghai 201306,China)
出处
《锻压技术》
CAS
CSCD
北大核心
2021年第9期67-76,共10页
Forging & Stamping Technology
基金
国家自然科学基金重大项目(52090043)
华中科技大学自主创新基金(2020kfyXJJS053)。
关键词
深度神经网络
TA15钛合金
高温拉伸
本构模型
热加工图
deep neural network
TA15 titanium alloy
high temperature tensile
constitutive model
hot processing map
