Acoustic Imaging Using Cyclostationarity

Cyclostationary source is the main noise source of a rotating machinery when it is in the working condition.In rotating key components such as gears and bearings,the generated vibration and noise signals are known as cyclostationary signals.Cyclostationarity provides a powerful tool for signal processing and analysis,but it is currently mostly used for vibration signal analysis instead of acoustic signal processing.The near-field acoustic holography,which is an effective noise source identification,location and visualization reconstruction technology,is widely used in the field of noise detection and control.However,due to the complexity of the sound field of rotating machinery,the traditional near-field acoustic holography is difficult to apply.In this context,the present work is aimed to develop a blind signal extraction method based on cyclostationarity,systematically evaluate the effectiveness of the developed algorithm through simulation,propose an acoustic imaging optimization approach based on the developed blind signal extraction and Bayesian near-field acoustic holography methods,and finally verify the feasibility of the proposed approach through experiments.First,the Bayesian near-field acoustic holography method is researched and constructed.The forward propagation model of the sound field is introduced,and the theory of estimating the optimal parameters based on the Bayesian theory is summarized.Therefore,an accurate reconstruction of the sound source is achieved.Numerical simulation of the localization of the point sound source is carried out,and the effectiveness of the method is verified by comparing the theoretical value with the reconstructed value.Secondly,three kinds of blind extraction methods based on cyclostationary characteristics,known as multivariate cyclic regression,subspace blind extraction and reduced-rank cyclic regression,are studied.The algorithm programming for each of these methods is realized one by one with MATLAB.The constructed algorithms are then evaluated through simulation experiments.The influence of algorithm parameters on the extraction accuracy is systematically explored,and the reduced-rank cyclic regression is found to have the best effect.Finally,a joint blind signal extraction and Bayesian sound source reconstruction method is proposed.Based on the measurement of the sound pressure of the signal radiated by the sound source obtained by microphone array,the sound source reconstruction experimental research is carried out.In the reverberation room,the sound source localization and reconstruction experiments are performed,and the Bayesian near-field acoustic holography is used with the proposed acoustic imaging approach for sound source reconstruction.It is shown that this method is able to eliminate the interference of other noise sources,and reconstruct only the cyclostationary source of interest.The acoustic imaging approach proposed in the present work not only has the advantages of non-contact measurement and visual reconstruction,but also overcomes the limitations of near-field acoustic holography for complex sound fields.As the proposed acoustic imaging approach can and further broaden the application range of near-field acoustic holography,it can thereby lay the foundation for the research on the noise control and structural optimization of mechanical systems.