Technology Development And Application Study On Compressor Noise Source Localization Base On Beamforming

Many researches have been executed on the sound source localization base on frequency domain beam forming technology,however,they were started at an array with too many microphones,and proven by experiment in an extreme acoustic environment,such as semi-anechoic room or anechoic room.Such environment is not possible to be realized in a general manufacturer.Therefore,it is very interesting to have the study on sound source localization by minimized number of microphones.During the new product development of compressors,there is always noise level requirement.When it is very high or with high frequency noise,project team has to take much time to improve.On the other hand,project timing is always tough,the time for the problem solving of high noise level is limited.If sound source can be identified quickly,it will support project team to find root cause quickly.This article firstly simulated the performance(width of main lobe,cut-off frequency,max level of side-lobe)of different microphone arrays by normal frequency domain beam forming algorithm and frequency domain cross-power spectrum algorithm in MATLAB,with 24 microphones available in company,worked out optimized microphone arrays under the above two algorithms.Then,a performance comparison was performed between the optimized arrays and the other different array was simulated on different frequencies,measurement distances and algorithms in MATLAB.It proved the conclusion that the optimized arrays have better performance than the others.Afterwards,a moveable support was designed and made according to the optimized microphone arrays.In the noise engineering laboratory of company,a beam forming experiment was conducted on a surface with higher noise of a compressor,which has been identified by sound intensity scanning method.The experiment has proved that the frequency domain cross-power spectrum algorithm is better than another one on accurate localization.It also demonstrated that the optimized microphone arrays with only 24 sensors are able to identify the locations of typical sound frequencies of a compressor.