基于遗传算法优化的含氢Ti65合金人工神经网络本构模型的构建  

Constitutive modelling of artificial neural network for hydrogenated Ti65 alloy based on genetic algorithm optimization

作　　者：朱铭[1] 夏敏 田壵 邓磊[1] 金俊松[1] 王新云[1] ZHU Ming;XIA Min;TIAN Zhuang;DENG Lei;JIN Junsong;WANG Xinyun(State Key Laboratory of Materials Processing and Die&Model Technology,Huazhong University of Science and Technology,Wuhan 430074,China;Jiangsu Pacific Precision Forging Co.,Ltd.,Taizhou 225500,China)

机构地区：[1]华中科技大学材料成形与模具技术全国重点实验室,武汉430074 [2]江苏太平洋精锻科技股份有限公司,江苏泰州225500

出　　处：《兵器装备工程学报》2024年第8期232-239,共8页Journal of Ordnance Equipment Engineering

基　　金：国家重点研发计划课题(2022YFB3706903)。

摘　　要：本研究对不同氢含量(未置氢、0.13 wt.%、0.25 wt.%、0.34 wt.%和0.43 wt.%氢)的Ti65合金试样在其α+β两相区和β单相区、0.001 s^(-1)应变速率范围内进行等温压缩,研究了含氢Ti65合金的高温流变行为,建立了综合考虑氢含量、变形温度、应变、应变速率的含氢Ti65合金GA-BP神经网络本构模型,并将所建模型通过二次开发集成入有限元软件中,对含氢Ti65合金等温热压缩过程进行模拟。结果表明:4-12-12-1结构的GA-BP神经网络本构模型的相关系数和平均绝对误差分别为0.9982和0.46%,模型具有较高的预测精度和泛化能力,能够用于局部置氢Ti65合金热塑成形过程的分析。The study conducted isothermal compression tests of Ti65 alloy samples at different hydrogen contents(unhydrogenated,0.13 wt.%,0.25 wt.%,0.34 wt.%,and 0.43 wt.%hydrogen)in theα+βtwo-phase andβsingle-phase regions at a strain rate range of 0.001 s^(-1) to investigate the high-temperature deformation behavior of hydrogenated Ti65 alloys and construct a GA-BP constitutive model for Ti65 alloys that considers hydrogen content,deformation temperature,strain,and strain rate.The model was integrated into the finite element software to simulate the isothermal compression process of hydrogenated Ti65 alloy samples.The results showed that the correlation coefficient and the average relative absolute errors value of the 4-12-12-1 structure GA-BP constitutive model were 0.9982 and 0.46%,respectively,with high prediction accuracy and generalization ability.The simulation results of isothermal compression indicated that the established GA-BP constitutive model had high simulation accuracy and could be used for analyzing the thermoplastic forming process for locally hydrogenated Ti65 alloy.

关 键 词：置氢处理 Ti65合金 人工神经网络 遗传算法 本构模型 

分 类 号：TJ05[兵器科学与技术—兵器发射理论与技术] TG316[金属学及工艺—金属压力加工]