Scholarlink

<p>A new method is developed to improve the computational efficiency of MPM calculations in MPM-FEM hybrid analyses with the help of a dynamic load balancing (DLB) technique. This method achieves proper communication between control points associated with the B-spline basis functions assigned to the cores corresponding to not only the adaptively located subdomains for the MPM, which change with the deformation of the solid phase, but also the separately located subdomains for the MPM and stabilized FEM. Since higher-order B-spline functions are used in the MPM-FEM to reduce the cell-crossing error that often arises in the MPM with linear basis functions, material points and control points that need to communicate variable values are not always located at domain boundaries, further increasing the complexity of interdomain communication. A novel contribution of this study is the high computational efficiency of the MPM calculations, which allow for proper communication under this complexity. The first example is an MPM run for the solid phase only aimed at investigating the scalability and parallel efficiency of the developed DLB. The second and third examples show that the DLB technique for MPM calculations achieves excellent computational efficiency even when the MPM-FEM hybrid method results in inconsistent domain decomposition patterns for the fluid and solid phases.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Speeding up MPM calculations in MPM-FEM hybrid method incorporating inconsistent domain decomposition patterns for solid and fluid phases

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年2月11日

查看全文
  • Shaoyuan Pan,
  • Soma Hidano,
  • Reika Nomura,
  • Lu Jing,
  • Yohei Miki,
  • Masatoshi Kawai,
  • Shuji Moriguchi,
  • Kengo Nakajima,
  • Kenjiro Terada

摘要

A new method is developed to improve the computational efficiency of MPM calculations in MPM-FEM hybrid analyses with the help of a dynamic load balancing (DLB) technique. This method achieves proper communication between control points associated with the B-spline basis functions assigned to the cores corresponding to not only the adaptively located subdomains for the MPM, which change with the deformation of the solid phase, but also the separately located subdomains for the MPM and stabilized FEM. Since higher-order B-spline functions are used in the MPM-FEM to reduce the cell-crossing error that often arises in the MPM with linear basis functions, material points and control points that need to communicate variable values are not always located at domain boundaries, further increasing the complexity of interdomain communication. A novel contribution of this study is the high computational efficiency of the MPM calculations, which allow for proper communication under this complexity. The first example is an MPM run for the solid phase only aimed at investigating the scalability and parallel efficiency of the developed DLB. The second and third examples show that the DLB technique for MPM calculations achieves excellent computational efficiency even when the MPM-FEM hybrid method results in inconsistent domain decomposition patterns for the fluid and solid phases.