Multi-Grid and GPU-accelerated simulation of an underwater hull based on the Single-Step and Simplified Lattice Boltzmann Method

Abstract: In this work, a fully transient, three-dimensional study of the DARPA SUBOFF underwater vehicle is presented using a computationally efficient Lattice Boltzmann Method (LBM) solver, based on the Single-Step and Simplified Lattice Boltzmann Method (SSLBM). The SSLBM solver model highly turbulent flows by integrating the Vreman Large Eddy Simulation (LES) turbulence model and accelerates computations with Graphics Processing Units (GPUs). The solver is coupled with an immersed multi-grid discretization technique, expanding the fluid domain to minimize boundary condition effects, and an immersed-refinement technique to capture the boundary layer near the hull, with negligible impact on computational performance. The simulation, conducted at a Reynolds number of 1.1 million with 77.3 million grid points over 145 physical seconds, achieves a computational performance of 1,529 Million Lattice Updates per Second (MLUPS), running 50,000 time-steps in 42 minutes. Normalized velocity profiles around the trailing edge and pressure coefficients are compared with various experimental and numerical results, demonstrating that the SSLBM solver provides accurate results in a short period. This study highlights the solvers ability to efficiently handle complex flow simulations for underwater vehicles. Keywords: Lattice Boltzmann Method, GPU, Large Eddy Simulation