Formulation and numerical performance of an adaptive algorithm for efficient collision detection

The 'Discrete Particle Method'(DPM) is a versatile numerical tool for improving the understanding of particle flows behavior on a meso-scopic level. A crucial point when using the DPM is the CPU-time consumption for detection of particle collisions. An adaptive algorithm for efficient particle-particle and particle-wall collision detection in a two-dimensional case is presented. The physical domain is hierarchically divided and structured as a quadtree. The algorithm ensures an efficient computation of colliding particle flows by splitting and merging the cells between each time step to keep the number of particles within a proposed range. The numerical performance of the adaptive algorithm is studied by simulating a flow particle in a 90° bend. The computational time of the adaptive algorithm is compared with the simulations performed with a uniform fixed cell structure with optimal size. The adaptive algorithm seems to be mostly advantageous inflows where the particles are not uniformly distributed, in complex geometries, and otherwise where information about the optimal cell size is not known a priori.