Parallel DEM Software for Simulation of Granular Media

  • Authors:

  • Algirdas Maknickas;Arnas Kačeniauskas;Rimantas Kačianauskas;Robertas Balevičius;Algis Džiugys

  • Affiliations:

  • Parallel Computing Laboratory, Vilnius Gediminas Technical University, Saulėtekio al. 11, 10223 Vilnius, Lithuania, e-mail: alm@sc.vtu.lt, arnka@fm.vtu.lt;Parallel Computing Laboratory, Vilnius Gediminas Technical University, Saulėtekio al. 11, 10223 Vilnius, Lithuania, e-mail: alm@sc.vtu.lt, arnka@fm.vtu.lt;Laboratory of Numerical Modelling, Vilnius Gediminas Technical University, Saulėtekio al. 11, 10223 Vilnius, Lithuania, e-mail: rkac@fm.vtu.lt, robertas.balevicius@st.vtu.lt;Laboratory of Numerical Modelling, Vilnius Gediminas Technical University, Saulėtekio al. 11, 10223 Vilnius, Lithuania, e-mail: rkac@fm.vtu.lt, robertas.balevicius@st.vtu.lt;Laboratory of Combustion Processes, Lithuanian Energy Institute, Breslaujos 3, 44403 Kaunas, Lithuania, e-mail: dziugys@isag.lei.lt

  • Venue:
  • Informatica
  • Year:
  • 2006

Quantified Score

Hi-index 0.00

Visualization

Abstract

The paper describes the development and performance of parallelalgorithms for the discrete element method (DEM) software. Spatialdomain decomposition strategy and message passing inter-processorcommunication have been implemented in the DEMMAT code forsimulation of visco-elastic frictional granular media. The novelalgorithm combining link-cells for contact detection, the staticdomain decomposition for parallelization and MPI data transfer forprocessors exchanging particles has been developed for distributedmemory PC clusters. The parallel software DEMMAT_PAR has beenapplied to model compacting of spherical particles in therectangular box. Two benchmark problems with different numbers ofparticles have been solved in order to measure parallel efficiencyof the code. The inter-processor communication has been examined inorder to improve domain decomposition topology and to achievebetter load balancing. The speed-up equal to 11 has been obtainedon 16 processors. The parallel performance study has been performedon the PC cluster VILKAS of Vilnius Gediminas Technical University,Lithuania.