GPU-Accelerated Multi-Material FVM-SPH with Immersed Boundary Method for High-Fidelity Modeling of Shock-Structure Interactions
Please login to view abstract download link
Shock wave propagation from explosive detonations and its damaging effects on adjacent structures involve complex compressible fluid flows and intense fluid-structure interactions (FSI), posing significant computational challenges. This paper proposes a novel GPU-accelerated hybrid framework that couples multi-material finite volume method (FVM) with smoothed particle hydrodynamics (SPH) to simulate the entire process of shock wave dynamics and structural response, including fracture propagation and fragmentation. The FVM efficiently resolves explosive detonation and shock propagation in fluids using a six-equation model with adaptive Riemann solvers, while SPH captures large structural deformations and material failure in a meshless Lagrangian framework. The immersed boundary method (IBM) enables robust two-way coupling between FVM and SPH domains, ensuring accurate momentum and energy transfer across fluid-structure interfaces. Leveraging GPU parallelization, the framework achieves high computational efficiency, enabling simulations with millions of nodes/particles. Five benchmark cases-shock-bubble interaction, 2D/3D underwater explosions, reinforced concrete damage under blast loads, and dynamic fracture of steel tubes-are used to validate the method. Simulation results show strong agreement with experimental data. The GPU acceleration achieves a 350 times speedup over CPU-based SPH, making the framework practical for large-scale FSI problems. This work demonstrates the unified FVM-SPH-IBM approach is capable of handling multi-material compressible flows, extreme structural deformations, and fragmentation, offering a powerful tool for defense and engineering applications.
