• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2022, Vol. 58 ›› Issue (16): 2-20.doi: 10.3901/JME.2022.16.002

• 特邀专栏: 高性能塑性成形制造(上) • 上一篇    下一篇

扫码分享

塑性成形快速数值仿真方法的研究进展

詹梅1,2, 董赟达1,2, 翟卓蕾1,2, 樊晓光2, 石志鹏1,2, 安强1,2   

  1. 1. 西北工业大学深圳研究院 深圳 518057;
    2. 西北工业大学材料学院 西安 710072
  • 收稿日期:2021-11-30 修回日期:2022-04-10 出版日期:2022-08-20 发布日期:2022-11-03
  • 作者简介:詹梅,女,1972年出生,博士,教授,博士研究生导师。主要研究方向为高性能轻量化薄壁复杂构件精确塑性成形制造理论、技术及装备。E-mail:zhanmei@nwpu.edu.cn
  • 基金资助:
    广东省基础与应用基础研究基金联合基金(2019B1515120047)和国家自然科学基金(52130507)资助项目

Review on Fast Numerical Simulation Method for Plastic Forming

ZHAN Mei1,2, DONG Yunda1,2, ZHAI Zhuolei1,2, FAN Xiaoguang2, SHI Zhipeng1,2, AN Qiang1,2   

  1. 1. Research & Development Institute of Northwestern Polytechnical University in Shenzhen, Shenzhen 518057;
    2. School of Materials Science and Engineering, Northwestern Polytechnical University, Xi'an 710072
  • Received:2021-11-30 Revised:2022-04-10 Online:2022-08-20 Published:2022-11-03

摘要: 精确高效的数值仿真预测模型是塑性成形技术向数字化、智能化发展的关键技术之一。为实现大型复杂构件先进塑性成形工艺研究的实时化,形成了大规模离散模型精简/降维、求解器算法改进与高性能并行计算三种高效仿真方法,从降低模型规模、加速/省略仿真中的耗时流程与提升算法的设备使用效率三方面提速。围绕这三方面,综述网格密度动态控制方法与多种实用壳单元对模型规模缩减的策略,介绍机器学习、物质点法与虚拟元法等创新算法对仿真流程的优化,并讨论并行算法在同构与异构平台上的研究进展。通过对高效仿真方法研究现状的整理,评估了其在塑性成形数值仿真领域中应用前景,并展望了该领域未来可能的发展趋势。

关键词: 网格密度控制, 壳单元, 机器学习, 物质点法, 虚拟元法, 并行计算, GPU

Abstract: Accurate and efficient numerical prediction model is the kernel about the digitization and intellectualization of plastic forming. To study advanced plastic forming of large-scale complex components in real time, there exists three fast simulation method: massive model simplifying or dimensionality reducing, solver algorithm modifying and high-performance parallel computing. And these methods are introduced in terms of reducing model scale, accelerating or avoiding time-consuming steps and improving device efficiency. The two major approaches about model scale reducing are introduced: mesh density dynamic controlling technique and shell element model. Subsequently, a review of some innovative numerical algorithms for improving calculation process is provided, such as machine learning, material point method and virtual element method. Then, the research progress of parallel techniques on homogeneous and heterogeneous computing platforms is discussed. Finally, the application prospect and development of these numerical simulation methods for plastic forming are also presented.

Key words: mesh density controlling, shell element, machine learning, material point method, virtual element method, parallel computing, GPU

中图分类号: