A Gluing Algorithm for Network-Distributed Multibody Dynamics Simulation

Abstract

The improved performance and capacity of networks has made thecombined processing power of workstation clusters a potentiallypromising avenue for solving computationally intensive problems acrosssuch distributed environments. Moreover, networks provide an idealplatform to employ heterogeneous hardware and software to solvemultibody dynamics problems. One fundamental difficulty with distributedsimulation is the requirement to couple and synchronize the distributedsimulations. This paper focuses on the algorithms necessary to coupletogether separately developed multibody dynamics modules so that theycan perform integrated system simulation. To identify a useful couplingstrategy, candidate numerical algorithms in the literature are reviewedbriefly – namely, stiff time integration, local parameterization,waveform relaxation, stabilized constraint and perturbation. Anunobtrusive algorithm that may well serve this `gluing' role ispresented. Results from numerical experiments are presented and theperformance of the gluing algorithm is investigated.