| 151 | 0 | 62 |
| 下载次数 | 被引频次 | 阅读次数 |
在利用深度学习方法加速冲压成形仿真并降低计算成本的过程中，纯数据驱动方法难以准确捕捉物理规律，缺乏物理可解释性，为解决这一问题，基于图神经网络(GNNs)和物理信息神经网络(PINNs)提出了一种数据与知识混合驱动的深度学习框架，使用图神经网络表达有限元网格信息，将本构方程等物理先验知识嵌入损失函数中，构建以几何特征和材料属性为输入的深度学习模型，并将该模型用于某汽车前盖内板冲压成形性预测。结果表明：所提出神经网络模型所预测的厚度减薄率和应变场与有限元结果的最大误差仅为2.08%;采用训练好的模型预测前盖内板成形性仅需9.73 s,相比于有限元计算效率提高1000倍以上，有效地降低了汽车产品正向开发过程中的计算成本。
Abstract：In the process of accelerating stamping simulation and reducing computational costs through deep learning methods, purely data-driven approaches often fail to accurately capture physical laws and lack physical interpretability. To address this issue, a hybrid data and knowledge-driven deep learning framework was proposed, integrating graph neural networks(GNNs) with physics-informed neural networks(PINNs). In this framework, GNNs were employed to represent finite element mesh information, while constitutive equations and other physical priors were embedded into the loss function. A deep learning model was constructed with geometric features and material properties as inputs, and applied to the stamping formability prediction of an automotive front hood inner panel. The results demonstrate that the maximum error of the thickness reduction and strain fields predicted by the proposed neural network model compared with the finite element results is only 2.08%. The trained model requires only 9.73 s to predict the formability of the front hood inner panel, achieving a computational efficiency more than 1000 times higher than finite element calculation and effectively reducing the computational cost in the forward development process of automotive products.
[1]郑超.基于Autoform的汽车结构件成形数值模拟及回弹研究[D].长春:吉林大学,2015.ZHENG Chao.Research on forming numerical simulation and springback of automotive structural part based on Autoform[D].Changchun:Jilin University,2015.
[2]冯天飞,施法中.冲压成形模拟技术与SHEETFORM软件[J].机械工人(冷加工),2002,(9):48-50.FENG Tianfei,Sh I Fazhong.Stamping forming simulation technology and SHEETFORM software[J].Mechanical Worker(Cold Working),2002,(9):48-50.
[3]蔡杨,王勇超,黄庆奕,等.基于改进BP神经网络模型的2024铝合金热冲压成形工艺参数优化[J].塑性工程学报,2025,32(10):44-51.CAI Yang,WANG Yongchao,HUANG Qingyi,et al.Optimization of 2024 aluminum alloy hot stamping process parameters based on improved BP neural network model[J].Journal of Plasticity Engineering,2025,32(10):44-51.
[4]PFAFF T,FORTUNATO M,SANCHEZ-GONZALEZ A,et al.Learning mesh-based simulation with graph networks[J].2020.DOI:10.48550/ar Xiv.2010.03409.
[5]SHLOMI J,BATTAGLIA P,VLIMANT J R.Graph neural networks in particle physics[J].Machine Learning:Science and Technology,2020,2(2):021001.
[6]SANCHEZ-GONZALEZ A,GODWIN J,PFAFF T,et al.Learning to simulate complex physics with graph networks[C]//International Conference on Machine Learning.Shangri-La,2020:8459-8468.
[7]高泽,楚遵康,石稼晟,等.基于图网络的汽车零部件应力场快速预测方法研究[J].汽车工程,2024,46(1):170-178.GAO Ze,CHU Zunkang,SHI Jiasheng,et al.Research on fast prediction method of stress field of automotive parts based on graph network[J].Automotive Engineering,2024,46(1),170-178.
[8]BELBUTE-PERES F D A,ECONOMON T,KOLTER Z.Combining differentiable PDE solvers and graph neural networks for fluid flow prediction[C]//International Conference on Machine Learning.Shangri-La,2020:2402-2411.
[9]RAISSI M,PERDIKARIS P,KARNIADAKIS G E.Physics-informed neural networks:A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations[J].Journal of Computational Physics,2019,378:686-707.
[10]TAO F,LIU X,DU H,et al.Physics-informed artificial neural network approach for axial compression buckling analysis of thinwalled cylinder[J].AIAA Journal,2020,58(6):2737-2747.
[11]HAGHIGHAT E,JUANES R.Sci ANN:A Keras/Tensor Flow wrapper for scientific computations and physics-informed deep learning using artificial neural networks[J].Computer Methods in Applied Mechanics and Engineering,2021,373:113552.
[12]胡卫龙.仿真本构模型现状及适应未来智能化成形技术应具有的预测功能[J].塑性工程学报,2023,30(2):1-15.HU Weilong.Present situation of simulation constitutive models and expected predictive abilities to serve upcoming intelligent forming technology[J].Journal of Plasticity Engineering,2023,30(2),1-15.
[13]ANTONIETTI P F,DASSI F,MANUZZI E.Machine learning based refinement strategies for polyhedral grids with applications to virtual element and polyhedral discontinuous Galerkin methods[J].Journal of Computational Physics,2022,469:111531.
[14]ZHOU H,XU Q,NIE Z,et al.A study on using image-based machine learning methods to develop surrogate models of stamp forming simulations[J].Journal of Manufacturing Science and Engineering,2022,144(2):021012.
[15]ZIENKIEWICZ O C,TAYLOR R L.The finite element method for solid and structural mechanics[M].Amsterdam:Elsevier,2005.
[16]CHAKRABARTY J.Theory of plasticity[M].Amsterdam:Elsevier,2012.
基本信息:
中图分类号:TP183;TG386
引用信息:
[1]杨卓远,余海燕,贺宏伟,等.基于物理信息的图神经网络冲压成形性预测方法[J].塑性工程学报,2026,33(02):102-109.
基金信息:
国家自然科学基金资助项目(51175382)
2026-02-12
2026-02-12
2026-02-12