| The noise distribution of the mechanical system is closely related to the mechanical structure design and specific operating conditions.A single microphone can obtain acoustic signals for time-frequency domain analysis,but it is challenging to obtain acoustic spatial distribution characteristics.The use of microphone arrays can invert the spatial distribution of acoustic signals according to the time difference of the signals reaching different microphones to achieve the purpose of acoustic source localization.However,in the actual scenes,due to the aperture limitation of the microphone array and other noise interference,the resolution of the lowfrequency acoustic source localization is not high enough.In rotating machinery,due to the periodic excitation of blades,low-frequency line spectra are widespread in the radiation noise of the mechanical system.Identifying low-frequency noise sources can effectively assist the fault warning and the design of noise reduction of the mechanical system.This paper studies the localization of low-frequency acoustic sources from the perspective of acoustic measurement and further applies it to centrifugal blower equipment.The main contents are as follows:(1)Based on the limitation of the array aperture for the localization of low-frequency acoustic sources under far-field conditions,the non-synchronous measurement is used to measure lowfrequency acoustic sources.For the acoustic source localization algorithm in the frequency domain,the acoustic source localization result mainly depends on two factors: the shape of the microphone array and the frequency of acoustic source localization.The lower limit of the localization frequency depends on the size of the array aperture.In order to broaden the lowfrequency working range of the acoustic source localization algorithm,this paper introduces the non-synchronous measurement method,which translates the array according to a predetermined position in the measurement plane to expand the aperture.The synthetic aperture based on the non-synchronous measurement makes the point spread function is closer to the ideal Dirac function,thereby realizing the localization of low-frequency acoustic sources.(2)A Bayesian inference algorithm based on the non-synchronous measurement is developed.The power propagation model based on the non-synchronous measurement is solved through the proposed Bayesian inference to obtain the high-resolution acoustic source localization.Aiming at the low-resolution problem of conventional beamforming,this paper proposes the concepts of the power propagation model and the convolution model.The acoustic source localization result of conventional beamforming is modeled as the linear multiplication of accurate power solution and power propagation matrix.The power propagation model is a well-posed linear equation in the real number domain.More accurate acoustic source localization results can be obtained by solving the power propagation model.In the acoustic source localization problem,the deconvolution method and the regularization method are usually used to solve the power propagation model.However,the deconvolution algorithm does not consider the influence of noise,and the regularization parameters of the regularization method are not adaptive and need to be adjusted according to the specific acoustic scene.The Bayesian inference method proposed in this paper considers the influence of noise and is sufficiently robust.Especially in the framework of nonsynchronous measurement,under the double interference conditions of cross-spectrum filling error and additive noise,the proposed Bayesian inference method shows better performance than the deconvolution and regularization algorithms.(3)The energy accuracy of the acoustic source localization algorithm is studied based on the acoustic power,and the low-frequency acoustic source localization experiments are carried out on the Bluetooth speaker and blower system in the anechoic chamber(free acoustic field)and factory workshop(non-free acoustic field)respectively.In the acoustic power experiment,the sound power level(SPL)of the Bluetooth speaker acoustic source was obtained by the hemisphere method and compared with the sound power level of acoustic imaging to obtain the energy characteristics of different acoustic source localization algorithms.In the anechoic chamber experiment,the effective combination of non-synchronous measurement and Bayesian inference can accurately obtain the high-resolution acoustic source localization results of the Bluetooth speaker under the conditions of low-frequency and low signal-to-noise ratio.In the blower system experiment on the factory floor,the test results of the centrifugal blower system showed that the main contribution of low-frequency noise came from the air outlet,the pipe flange connection,the elbow pipe,the vicinity of the volute,the gap between the volute and the box.The acoustic localization results can assist the failure warning and low-noise design of the mechanical systems.This paper studies the high-resolution algorithms of acoustic source localization,especially the low-frequency acoustic sources(500～1500Hz)that are naturally occurring in rotating machinery.As for the measurement method,the synthetic aperture method of non-synchronous measurement is adopted to broaden the operating frequency range of the acoustic source localization algorithm.As for the inversion algorithm,a Bayesian inference algorithm that considers both anti-noise and adaptability is proposed,and the power propagation model based on the non-synchronous measurement is solved.Finally,signal simulation experiments,acoustic power experiments,anechoic chamber experiments,and workshop blower system experiments verify the effectiveness of the high-resolution acoustic source localization algorithm proposed in this paper.Accurate inversion of noise sources for chemical equipment such as blowers can play an auxiliary role in the early warning of mechanical systems,and provide technical support for the design of noise reduction and condition monitoring of mechanical systems. |
PDF Full Text Request