GPU-accelerated model for fast, three-dimensional fluid-structure interaction computations.

Area of Science:

• Computational fluid dynamics
• Fluid-Structure interaction (FSI) modeling
• Numerical methods for physics simulations

Background:

• Fluid-Structure interaction (FSI) is crucial in many scientific and engineering fields.
• Simulating FSI, especially with moving boundaries, presents significant computational challenges.
• Existing methods often require substantial computational resources and time.

Purpose of the Study:

• To introduce an efficient methodology for one-way Fluid-Structure interaction (FSI).
• To develop a workflow for embedding moving geometries within Lattice-Boltzmann Method (LBM) computations.
• To validate and evaluate the performance of the proposed FSI methodology.

Main Methods:

• Utilized a Graphics Processing Unit (GPU) accelerated Lattice-Boltzmann Method (LBM) implementation.
• Developed an efficient workflow for integrating polygonal mesh-based moving geometries into LBM.
• Validated the method with a synthetic experiment (expanding/contracting vessel) and the 3D peristaltic flow problem.

Main Results:

• The methodology was successfully validated against known solutions for peristaltic flow.
• Achieved an average execution time of approximately one hour per computation.
• Demonstrated that the Reynolds number significantly influences the flow regime and streamlines.

Conclusions:

• The proposed GPU-accelerated LBM method provides an efficient approach for one-way FSI simulations.
• The workflow effectively handles moving geometries in complex fluid dynamics problems.
• The study highlights the impact of Reynolds number on flow characteristics in deforming tubes.