A scalable GPU-parallelized framework for accelerating force calculation in discrete dislocation dynamics

doi:10.1016/j.enganabound.2025.106440

A scalable GPU-parallelized framework for accelerating force calculation in discrete dislocation dynamics

2025-09-09

0

PRE

AI

DOI：10.1016/j.enganabound.2025.106440

Original

Original request for help

Save

Abstract

En 中文

In multiple engineering and scientific fields, Discrete Dislocation Dynamics (DDD) has emerged as a crucial tool for simulating and investigating the plastic behavior of materials. As dislocation line density continuously increases with plastic deformation magnitude, the computation of elastic interactions between dislocations has become a critical bottleneck affecting the efficiency of large-scale DDD simulations. This paper proposes a GPU-accelerated framework to enhance the computational efficiency of inter-dislocation interaction forces. First, a spatially dynamic hash table is established to enable rapid searching and updating of dislocation segments, thereby adapting to the evolutionary characteristics of dislocation networks. The dislocation interaction range is partitioned into far-field and near-field regions based on the fast multipole method. GPU Parallel strategies and optimization approaches are developed, demonstrating the synergistic acceleration effect of the proposed methodology. Benchmark tests and large-scale simulation examples are subsequently presented for evaluation.

A scalable GPU-parallelized framework for accelerating force calculation in discrete dislocation dynamics

Abstract

Journal

Organization