NoiseModelling

An entire region of more than 500 square kilometers can be computed at once on a standard personal computer thanks to the area subdivision step (sub-domains). The separation of data between threads and the limited usage of synchronisation provide a good exploitation of the multi-core technology.

For each sub-area an extended sub-area box is defined by appending the maximum propagation distance. This extended envelope collect all sounds-sources and buildings in intersection. An union of these buildings and sources are meshed thanks to a Constrained Delaunay triangulation. This mesh provides an optimal spatial distribution of receivers points. Furthermore the connection between the receivers reduces drastically the error introduced later by the interpolation in comparison with traditional grid receivers.

At this step, the computation sub-area has all needed information: a propagation domain, a list of sound-sources and receivers coordinates. Each receiver seek for line sources in a range equal to the maximum propagation distance parameter; then each line source is subdivided into point sources with a space interval that depends on the receiver distance. When all point source locations are available, the propagation paths from the receiver to the sources are computed, by considering the direct, the reflected and the diffracted paths. For each path, the contribution of the sound source level is calculated, and then energetically summed into the corresponding receiver.