GPU-based collision analysis between a multi-body system and numerous particles
The use of a graphics processing unit (GPU) is an ideal solution for problems on data-parallel computations. The serial CPU-based program for collision analysis between a multi-body system and numerous particles is rebuilt as a parallel program that uses the advantages of a GPU. In this study, a GPU is used to effectively perform multi-body dynamic simulation with particle dynamics. The multibody system has 20 circular objects, 19 spring-damper force elements, and 2 revolute joints. The motion equations are formulated using the Cartesian coordinate system, and the implicit Hilber-Hughes-Taylor integration algorithm is used for the integral equation. To detect collisions between a multi-body system and particles or between particles, a spatial subdivision algorithm and a discrete element modeling are used. The developed program is verified by comparing the results with ADAMS. The numerical efficiencies of the serial program using CPU and the parallel program using GPU are compared according to the number of particles. The results show that the greater the number of particles, the more computing time can be saved. For example, when the number of particles is 900, the computing speed of the parallel analysis program is about five times faster than that of the serial analysis program.
Keywords: Graphics processor unit, Parallel programming, Multi-body dynamics, Spatial subdivision, Discrete element method