Scholarlink

<p>To address the computational demands of penetration-induced damage simulations, this paper presents a graphics processing unit (GPU)-accelerated framework for explicit finite element (FE) analysis with erosion contact. Leveraging stream concurrency and a thread-to-element mapping strategy, the framework enables efficient GPU-parallel computation for multiple element types. Key developments include a GPU-based surface reconstruction procedure for dynamically updating contact topology after element erosion, a lock-free bucket sorting strategy combined with a parallel local search for contact detection, and the parallelization of the defense node algorithm for contact force computation. The integrated framework has been applied to several engineering case studies, demonstrating substantial performance gains. Numerical experiments show that the approach achieves a maximum speedup of 85.37 relative to a serial CPU baseline and 12.09 compared to a 16-core parallel implementation, providing a practical and efficient tool for high-performance explicit FE simulations of penetration-induced damage.</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

A GPU-accelerated framework for explicit finite element analysis with erosion contact

  • 期刊论文

  • 发表日期：

    2026年5月26日

查看全文
  • Xinggang Cao,
  • Xiang Zhao,
  • Yongjie Pei,
  • Zhenhui Liu,
  • Yong Cai,
  • Xiangyang Cui

摘要

To address the computational demands of penetration-induced damage simulations, this paper presents a graphics processing unit (GPU)-accelerated framework for explicit finite element (FE) analysis with erosion contact. Leveraging stream concurrency and a thread-to-element mapping strategy, the framework enables efficient GPU-parallel computation for multiple element types. Key developments include a GPU-based surface reconstruction procedure for dynamically updating contact topology after element erosion, a lock-free bucket sorting strategy combined with a parallel local search for contact detection, and the parallelization of the defense node algorithm for contact force computation. The integrated framework has been applied to several engineering case studies, demonstrating substantial performance gains. Numerical experiments show that the approach achieves a maximum speedup of 85.37 relative to a serial CPU baseline and 12.09 compared to a 16-core parallel implementation, providing a practical and efficient tool for high-performance explicit FE simulations of penetration-induced damage.