The coupled structural-acoustic Boundary Element Method (BEM) has been widely used in vibro-acoustic analysis for decades now, due to its ease of model building and high accuracy. Recently, evolutions of BEM to solve large Aero-Vibro-Acoustic (AVA), space and underwater problems have been introduced. However, the large-frequency range of excitations, sometimes compounded by the large-size nature of the models at hand implies that such problems have a large number of degrees of freedom. As a consequence, while BEM can provide an accurate solution, the time and memory requirements related to model size and frequency range tend to create practical limitations. The advent of affordable large-scale distributed memory computing power enables a new generation of BEM solvers to push forward the capabilities of BEM in addressing large problems. It enables coupled indirect BEM problems to be solved on large clusters and distribute all CPU-intensive steps on a large number of CPUs. In this paper, the theory behind DMP (Distributed Memory Processing) for BEM will be introduced
and illustrated on an automotive wind noise case.
2015 – ICSV22, Florence, Italy
Denis Blanchet (ESI Group, Germany)
Anton Golota (ESI Group, Germany)
Nicolas Zerbib (ESI Group, France)
Lassen Mebarek (ESI Group, France)