Research On Noise Source Localization And Recognition Based On Generalized Inverse Beamforming

In the underwater vehicle vibration and noise reduction project,it is the key technology to accurately identify the major noise sources in the mechanical noise system,to identify the special distribution and frequency characteristics of the major noise sources in both frequency domain and space domain as well as quantitively identify the contribution of each major noise source to the radiated noise of the underwater vehicle so that we can effectively take measures to reduce vibration and noise.With the development of vibration and noise reduction technology,the noise source may be located in the near field of the array.Thus it is very important to carry out the research on the method of locating and identifying the near-field noise source.Beamforming technology can effectively make up for the drawback of sound intensity and sound hologram technology.It has the advantages on measure speed,resolution in medium and high frequency range,flexibility of information processing and strong engineering practicability.It is an indispensable sound source recognition technology in the field of shipbuilding.Starting with the conventional beamforming,the principle of near-field focused beamforming based on acoustic vector array is described in this paper.A simulation comparison between near-field conventional beamforming and ac-MVDR focused beamforming is carried out and the results show that the near-field ac-MVDR focused beamforming method has a higher spatial resolution and stronger anti-aliasing ability than the conventional focused beamforming method.Focusing on the defects of traditional focused beamforming,which are poor recognition effect in low-frequency positioning and low spatial resolution or even failure when dealing with coherent and complex types of sound sources,a generalized inverse beamforming method based on the inverse problem of sound field is studied in this paper.The iterative generalized inverse beamforming(FBRM-IGIB)method based on the function beamforming normalization matrix is studied emphatically to solve the problem that traditional inverse beamforming algorithm is ill-defined.The theoretical derivation and simulation analysis show that this method can estimate the position and intensity of multiple types of sound sources regardless of its coherence.Thus it effectively solves the problem that the traditional methods have insufficient capability of processing coherent sound sources due to low spatial resolution.The parameters which may influence the positioning performance and spatial resolution of the method are analyzed and discussed.The simulation results show that the FBRM-IGIB method has higher spatial resolution and better positioning performance than the ac-MVDR,GIB and FGIB methods and the validity and applicability of the proposed method are verified.Combined with the array expansion pre-processing technique based on the beamforming normalization matrix,the aperture of the array is enlarged,the effective dynamic display range is improved,and the positioning recognition effect of the low frequency band is improved.Finally,the air dual sound source localization and the underwater vehicle radiated noise experiments are evaluated.The applicability of the proposed FBRM-IGIB method and the effectiveness of the array expansion preprocessing technique based on the beamforming regularization matrix are verified.The experimental results agree with the theoretical analysis,which provides guidance for engineering application of near-field noise source localization and recognition.