Sound source identification based on beamforming is widely used today as a spatial sound field visualization technology in wind tunnel experiments for vehicle development. However, the conventional beamforming technique has its inherent limitation, such as bad spatial resolution at the low frequency range, and limited system dynamic range. To improve the performance, three deconvolution methods CLEAN, CLEAN-SC and DAMAS were investigated and applied to identify wind noise sources on a production car in this paper. After analysis of vehicle exterior wind noise sources distribution, correlation analysis between identified exterior noise sources and interior noise were conducted to study their energy contribution to vehicle interior. The results show that the algorithm CLEAN-SC based on spatial source coherence shows the best capability to remove the sidelobes for the uncorrelated wind noise sources, while CLEAN and DAMAS, which are based on point spread functions have definite limitations.
Considering the testing car, the main noise source of exterior is from the wheelhouse region, then follows the rearview mirror with much lower sound energy. However, noise from the mirror contributes most to the vehicle interior, while the contribution from wheelhouse region ranks the second place. In addition, windshield wipers and door handle can do perceptible contributions to vehicle interior noise at some characteristic frequency bands.