Recent innovations and current performance of conventional boundary element computations.

Abstract

As long as boundary element methods have existed, research has been ongoing to make the calculations faster and applicable for larger problems. The rapid rise in computing power, both in terms of processor speed and storage capacity, has increased the problem size that can be addressed and has led to many novel strategies for reducing the computational burden. For example, many of the matrices can be stored at coarse frequency spacing over the range of interest and interpolated at much finer resolution, reducing the computation times enormously. Other strategies for speeding the computations that will be discussed include condensed acoustic meshes, optimized math kernel libraries, and multipole expansions. Results will be presented to indicate typical computation times and problem size limitations.