Summary

In an industrial context, the noise produced by machinery can harm workers’ health and result in deafness or hearing loss. To alleviate these problems, effective acoustic solutions must be put in place upstream. The first step in an acoustic study of noise reduction consists in locating the position of noise sources and ranking their contributions. In an industrial setting, workers are distributed around the machinery and the number of machines can be large. This makes the sound field complex because of the many reflections off walls and objects, so it is often hard to identify the main sound sources. The positions of sources can be located using a network of microphones called an acoustic antenna. The performance of source location techniques depends on the number of microphones, their distribution and the processing of the associated signal.

The purpose of this activity was to develop a spherical acoustic antenna associated with a time-based technique relying on intercorrelations of microphone signals, called Generalized Cross-Correlation (GCC) in the literature, to identify the position of the noisiest sound sources in an industrial setting. The GCC technique requires a time segment of microphone signals to generate a source map; it is known to be robust even in complex acoustic environments where the reverberant sound field makes a significant contribution (usually the case in industrial premises). The spherical antenna allows for the location of sound sources throughout a space.

In this study, the criteria applied to assess the quality of the sound map were presented. Then an optimization criterion for the antenna’s geometry was suggested. The optimized antenna was then manufactured and a panoramic camera (able to take photos or videos) added so that photos of the premises could be superimposed on the sound map. Finally, the spherical antenna that was developed, associated with the GCC technique, was tested in controlled conditions (hemi-anechoic and reverberant chambers), then in a workshop where several operators were using machines. The results show that it is possible to locate the positions of different machines despite their different sound contents and numerous reflections.