The principle of cross-spectral beamforming with exclusion of autospectra is analyzed based on spherical wave assumption. The result shows that the interference from the self-noise in each microphone channel is suppressed and the accuracy of sound source identification is improved by excluding the auto-spectral elements from the cross-spectral matrix. Furthermore, the algorithm of sound source identification is designed and the corresponding sound source imaging software is developed. The sources in the source plane are identified accurately. Additionally, the correctness of the designed algorithm and practical applicability of the developed software are tested and verified by an example. Based on this, the noise sound source identification test of an engine on a bench is made under the rated speed condition. The result shows that the noise at the engine head is mainly from the radiation noise of the air compressor exhaust outlet and the engine body at the upper left of the crankshaft pulley, and the noises of the left side and the right side of the engine are both mainly from the radiation noise of the engine block and the oil pan.%由于传声器阵列具有自噪声的干扰,在各通道的互谱距阵中,消除对角自谱元素的波束形成,可以提高声源识别的精度.由此,建立相应的声源识别算法和平面声源的成像软件.并对某发动机在额定工况下的噪声源识别进行验证.结果表明:发动机前端噪声来源于空气压缩机排气出口和曲轴传动皮带轮的上方机体辐射；左右两侧噪声来源于发动机缸体和油底壳辐射.由此表明涉及的算法与成像软件的正确与有效性能.
展开▼
