施文鹏,孙岑花,李佳俊,等.基于人工神经网络智能算法的9310钢本构模型优化[J].精密成形工程,2024,16(3):171-180.
SHI Wenpeng,SUN Cenhua,LI Jiajun,et al.9310 Steel Constitutive Model Optimization Based on Artificial Neural Network Intelligent Algorithm[J].Journal of Netshape Forming Engineering,2024,16(3):171-180. |
| 基于人工神经网络智能算法的9310钢本构模型优化 |
| 9310 Steel Constitutive Model Optimization Based on Artificial Neural Network Intelligent Algorithm |
| 投稿时间：2024-01-05 |
| DOI：10.3969/j.issn.1674-6457.2024.03.019 |
| 中文关键词: 9310钢 本构模型 Arrhenius型本构模型 人工神经网络(ANN) 智能算法优化 |
| 英文关键词: 9310 steel constitutive model Arrhenius constitutive model artificial neural network (ANN) intelligent algorithm optimization |
| 基金项目:江西省自然科学基金面上项目(20232BAB204050) |
|
| 摘要点击次数: 1020 |
| 全文下载次数: 400 |
| 中文摘要: |
| 目的 研究9310钢在变形温度为800~1 200 ℃、应变速率为0.01~50 s−1和高度压下量为70%条件下的热变形行为，建立预测效果相对较好的9310钢本构模型。方法 使用Gleeble-3800热模拟机对9310钢进行等温恒应变速率热压缩实验，基于热压缩实验数据，分析了应变速率对9310钢流动软化效应的影响，建立了考虑应变补偿的Arrhenius本构模型与支持向量回归(SVR)本构模型，并进行了模型精度分析，之后引入人工神经网络(ANN)智能算法优化了Arrhenius本构模型。结果 与变形温度相比，应变速率对9310钢流动软化效应的影响更为显著。相较于支持向量回归(SVR)本构模型，考虑应变补偿的Arrhenius本构模型精度更高，其相关系数R为0.993 4，平均相对误差(AARE)和均方误差(MSE)分别为0.055 6和89.362，它在预测高应变速率(1、10、50 s−1)流动应力时出现了较大偏差，经ANN智能算法优化后，相关系数R提高至0.999 1，AARE和MSE分别降至0.019 9和9.998，且绝对误差在±10 MPa以内的预测流动应力占比为98.34%。结论 在低应变速率(0.01 s−1)下软化效应更强，在高应变速率(10 s−1)下再结晶程度较低，软化效应较弱。ANN智能算法优化后的Arrhenius本构模型具有较高的精度，能较准确地预测9310钢的流动行为。 |
| 英文摘要: |
| The work aims to study the thermal deformation behavior of 9310 steel under the conditions of deformation temperature of 800-1 200 ℃, strain rate of 0.01-50 s−1 and high depression of 70%, and to establish a constitutive model of 9310 steel with a relatively good prediction effect. An isothermal constant strain rate thermal compression test was carried out on 9310 steel using Gleeble-3800 thermal simulator, and the influence of strain rate on the flow softening effect of 9310 steel was analyzed based on the thermal compression experimental data, and an Arrhenius constitutive model and a support vector regression (SVR) constitutive model considering strain compensation were established, and the model accuracy was analyzed. Compared with the deformation temperature, the strain rate had a more significant effect on the flow softening effect of 9310 steel. Compared with the Support Vector Regression (SVR) constitutive model, the Arrhenius constitutive model considering strain compensation had higher accuracy, with a correlation coefficient R of 0.993 4, an average relative error (AARE) and a mean square error (MSE) of 0.055 6 and 89.362, respectively, and a high strain rate (1, 10, 50 s−1). After the optimization of the ANN intelligent algorithm, the correlation coefficient R was increased to 0.999 1, the AARE and MSE were reduced to 0.019 9 and 9.998, respectively. The proportion of the predicted flow stress with an absolute error of ±10 MPa was 98.34%. The softening effect is stronger at low strain rate (0.01 s−1). The degree of recrystallization is lower at high strain rate (10 s−1), and the softening effect is weaker. The Arrhenius constitutive model optimized by ANN intelligent algorithm has high accuracy and can accurately predict the flow behavior of 9310 steel. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
渝公网安备 50010702501719号