摘要
以压缩试验所得数据为基础,根据BP人工神经网络算法原理,建立了Ti-50.5Ni合金高温压缩变形真应力与真应变、应变速率和变形温度关系的预测模型。结果表明,BP神经网络用于Ti-50.5Ni合金高温压缩变形本构关系建模是可行的,拟合度可达到1.3%,较好地反映了实际变形过程的特征,可弥补传统回归模型不能反应变形全过程的局限性,并消除实验过程中实际温度偏离设定温度所带来的样本误差及其对模型准确度的影响。
On the basis of the compression experiment data obtained from Gleeble-1500 Thermal Simulator, the predicting models for the relation between true stress and true strain, strain rate and temperature for Ti-50.5 Ni alloy have been developed with BP（back propagation） Artificial Neural Network method. The results show that, it is feasible to establish high temperature compression deformation constitutive relation model of Ti-50.5 Ni alloy by BPNN method. The fitting degree of the BPNN model is less than 1.3%, and can reflect the real feature of the practical deforming process. The deficiency of traditional regression model, which cannot describe the whole deforming process is filled up. The sample error caused by the deviation of actual temperature from the set temperature and its effect on the model accuracy are eliminated in the experiment.
出处
《稀有金属材料与工程》
SCIE
EI
CAS
CSCD
北大核心
2008年第1期19-23,共5页
Rare Metal Materials and Engineering
基金
国家自然科学基金面上项目(50474072)
国家自然科学基金重点项目(50634010)
长江学者和创新团队发展计划(IRT0407)
关键词
镍钛形状记忆合金
神经网络
本构关系
Ni-Ti shape memory alloy
neural network
constitutive relationship
