Deep neural network approach to estimate biaxial stress-strain curves of sheet metals
Deep neural network approach to estimate biaxial stress-strain curves of sheet metals
复制标题
DOI：
10.1016/j.matdes.2020.108970
复制
发表时间：
2020-10
影响因子：
8.4
通讯作者：
A. Yamanaka;Ryunosuke Kamijyo;Kohta Koenuma;I. Watanabe;T. Kuwabara
中科院分区：
文献类型：
--
作者：
A. Yamanaka;Ryunosuke Kamijyo;Kohta Koenuma;I. Watanabe;T. Kuwabara
为了提高板料成形模拟的精度，利用多轴材料试验结果对本构模型进行了标定。然而，多轴材料测试是耗时的，需要专门的设备。提出了两种不同的深度神经网络(DNN)方法，即二维和三维卷积神经网络(DNN-2D和DNN-3D)，用于从代表样品晶体织构的数字图像中有效地估计铝合金板材的双向应力-应变曲线。DNN-2D用于从{111}极图的数字图像中估计双向应力-应变曲线，而DNN-3D则从纹理的3D图像中估计曲线。使用合成织构数据集和基于晶体塑性的数值双轴拉伸试验获得的相应的双轴应力-应变曲线来训练这两个DNN。通过比较数值双向拉伸试验的结果，检验了两种训练的DNN的准确性。结果表明，两种DNN均能较高精度地估计双向应力-应变曲线。虽然DNN-3D提供了比DNN-2D更好的估计，但它的计算效率较低。因此，这两个DNN及其训练过程为材料建模提供了一种新的有效的虚拟数据方法。
To improve the accuracy of a sheet metal forming simulation, the constitutive model is calibrated using results from multiaxial material testing. However, multiaxial material testing is time-consuming and requires specialized equipment. This study proposes two different deep neural network (DNN) approaches, a two- and three-dimensional convolutional neural network (DNN-2D and DNN-3D), to efficiently estimate biaxial stress-strain curves of aluminum alloy sheets from a digital image representing the sample's crystallographic texture. DNN-2D is designed to estimate biaxial stress-strain curves from a digital image of {111} pole figure, while DNN-3D estimates the curves from a 3D image of the texture. The two DNNs were trained using synthetic texture datasets and the corresponding biaxial stress-strain curves obtained from crystal plasticity-based numerical biaxial tensile tests. The accuracy of the two trained DNNs was examined by comparing the results from that of the numerical biaxial tensile tests. It was observed that both the DNNs could estimate biaxial stress-strain curves with high accuracy. Though DNN-3D provides with a better estimation than DNN-2D, it displays lower computational efficiency. Thus, the two DNNs and their training procedures offer a new and efficient method to provide virtual data for material modeling.