The ability to predict acoustics of indoor spaces is highly important to for example avoid noise annoyance in open plan office, to promote speech transmission in sport halls and to optimize acoustics in concert halls. Do to the nature of sound with its small wavelengths compared to the dimensions of typical indoor spaces, no researchers have managed to solve the governing equations for room acoustics for the full audible frequency (20Hz-20kHz) and the room acoustics community has resorted to approximative methods until now. What is important for the room acoustics community, acoustic consultants and (applied) researchers, is to have access to a detailed simulation software written in a high level programming language, that can be applied to realistic environments for a larger frequency range than is possible right now. It helps to understand acoustic mechanisms of rooms as for example the influence of room shapes and material properties, as well as enables using the computed sound signals in listening experiences (via VR).
The Building Acoustics group is one of the nice recipients of the OpenSSI 2021b call of the eScience center. This open call supports researchers who want to significantly improve the run-time performance of their research software, and require additional expertise to achieve this. Our project attempts to push the frequency limits of a numerical solver of the governing equations for room acoustics, the Discontinuous Galerkin (DG) method. The TU/e Building Acoustics group has further developed this method for room acoustics applications and the solver is highly suitable for parallelization yielding an acceleration of its runtime: the method discretizes a room into non-overlapping elements where most operations are performed elementwise. As a demonstration of its merits, calculations with the accelerated DG method will be compared with detailed measurement results of an existing benchmark room at Eindhoven University of Technology.