首页|基于机器学习的Ti-5Al-1.5Mo-1.8Fe低成本钛合金热加工图预测

基于机器学习的Ti-5Al-1.5Mo-1.8Fe低成本钛合金热加工图预测

扫码查看
原文链接
  • 国家科技期刊平台
  • NETL
  • NSTL
  • 万方数据
为了研究Ti-5Al-1.5Mo-1.8Fe低成本钛合金的热变形行为,运用Instron 5869 压缩实验机进行热压缩实验.构建了以变形温度、应变速率、应变为输入变量和流变应力为输出变量的 6 种机器学习模型,预测不同条件下该合金的流变应力值并评估检验模型的预测性能.根据预测表现最好的LSTM神经网络模型的预测数据绘制预测加工图,对照实验加工图评估检验其预测能力.结果表明:预测加工图能够较为准确地反映出Ti-5Al-1.5Mo-1.8Fe合金在应变为 0.499 时的可加工区域,与实验加工图的吻合程度较高,该方法能较好地预测 Ti-5Al-1.5Mo-1.8Fe 合金的热变形行为.
Prediction of processing map based on machine learning for low-cost Ti-5Al-1.5Mo-1.8Fe titanium alloy
In order to study the thermal deformation behavior of low-cost titanium alloy,a Ti-5Al-1.5Mo-1.8Fe alloy was compressed for hot compression experiment by using Instron 5869 thermal compressor.Six machine learning models taking deformation temperature,strain rate and the degree of strain as input variables and adopting flow stresses as output variables were established.The flow stress values of alloy under different conditions were predicted and the prediction performance of these models were evaluated.The predicted processing map was drawn according to the prediction data of LSTM neural network model with the best prediction performance,and the predictive ability of the model was evaluated and verified by its comparison with experimental processing map.The results show that the processable regions of Ti-5Al-1.5Mo-1.8Fe alloy with the strain of 0.499 can be accurately reflected by the predicted processing map,in good accordance with the experimental map.The as-proposed method has better prediction for the thermal deformation behavior of Ti-5Al-1.5Mo-1.8Fe alloy.

low-cost titanium alloyTi-5Al-1.5Mo-1.8Fe alloythermal deformation behaviorhot compression experimentflow stressmachine learning modelLSTM neural network modelprocessing map

牟义强、张洺川、乔泽、王枫、覃美玲、徐勤思

展开 >

沈阳航空航天大学 民用航空学院,辽宁 沈阳 110136

攀枝花市午跃科技有限公司,四川 攀枝花 617000

低成本钛合金 Ti-5Al-1.5Mo-1.8Fe合金 热变形行为 热压缩实验 流变应力 机器学习模型 LSTM神经网络模型 加工图

国家自然科学基金辽宁省自然科学基金

51871220LACT-007

2024

10.7688/j.issn.1000-1646.2024.03.08

沈阳工业大学学报
沈阳工业大学

沈阳工业大学学报

CSTPCD北大核心
影响因子：0.62
ISSN：1000-1646
年,卷(期)：2024.46(3)
  • 24