A parallel acceleration GPU algorithm for large deformation of thin shell structures based on peridynamics

Loaded shell structures may deform, rotate, and crack, leading to fracture. The traditional finite element method describes material internal forces through differential equations, posing challenges in handling discontinuities and complicating fracture problem resolution. Peridynamics (PD), employing integral equations, presents advantages for fracture analysis. However, as a non-local theory, PD requires discretizing materials into nodes and establishing interactions through bonds, leading to reduce computational efficiency. This study introduces a GPU-based parallel PD algorithm for large deformation problems in shell structures within the compute unified device architecture (CUDA) framework. The algorithm incorporates element mapping and bond mapping for high parallelism. The algorithm optimizes data structures and GPU memory usage for efficient parallel computing. The parallel computing capabilities of GPU expedite crack analysis simulations, greatly reducing the time required to address large deformation problems. Experimental tests confirm the algorithm's accuracy, efficiency, and value for engineering applications, demonstrating its potential to advance fracture analysis in shell structures.