Session: 06-01: Spray Modeling

Paper Number: 141698

141698 - Simulation of Droplet Stream Impinging on Ignition Assistant Device Using Gpu-Accelerated Sph Method 

Abstract: 

Drop-wall interactions are encountered in various engineering problems, including fuel spray. Specifically, drop-wall interaction and thermal analysis of the ignition assistant device are of interest for the Unmanned Aerial Vehicle (UAV) application. This device should withstand the harsh operating environment, such as high altitudes characterized by low intake pressure and temperature, and low-quality fuels with low Cetane numbers. The smoothed particle hydrodynamics (SPH) simulation is proven to be effective in handling multi-phase interactions involving heat transfer, boasting its meshless Lagrangian formulation. However, the computational cost has been an issue due to extreme spray conditions, such as high liquid drop velocity, making the simulation inevitably have a fine resolution as well as a very small timestep. This study focuses on simulating the large-scale drop-wall interaction of ignition assistant devices through the hybrid parallelization paradigm. Both the central processing unit (CPU) and graphical processing unit (GPU) are fully utilized in this programming paradigm by coupling the shared memory (OpenMP), distributed memory (MPI), and memory off-loading (OpenACC) programming. Various computing environments, ranging from desktops to clusters, are tested to show the versatility of current SPH simulation. Different levels of GPU parallelization paradigms are analyzed in the OpenACC application programming interface (API), from the most straightforward kernel approach to the CUDA-ish block execution model. The results show favorable scaling performance that gives good portability in various computing architectures. Lastly, the thermal analysis of the ignition assistant device due to fuel droplets is conducted through large-scale simulations.

Presenting Author: Yongsuk Cho Texas Tech University

Presenting Author Biography: Dr. Yongsuk Cho is a post-doctoral research associate at Texas Tech University. Dr. Cho's research focuses on (1) Drop-wall interaction, (2) Biomass pyrolysis, (3) Numerical simulations, including the Lattice Boltzmann Method and Smoothed Particle Hydrodynamics, and (4) High-performance computing. Dr. Cho received his B.S. degree from the University of New Mexico and his Ph.D. from Iowa State University - all in mechanical engineering.

Authors:

Yongsuk Cho Texas Tech University
Song-Charng Kong Texas Tech University