Research On The Method Of Sound Source Identification In Mechanical Noise Fault Diagnosis

In the actual production environment,the captured signals from microphones are the mixed equipment noise signals.Therefore,before the noise diagnosis,it is necessary to accurately identify the sound source corresponding to the equipment from the mixed signals.In view of this,the characteristics of mechanical noise were analyzed at first,and then the algorithms of sound source localization and separation in sound source identification were studied in this paper.Considering the actual situation,the relative algorithms were improved,fused and simulated.Finally,an experiment platform of sound source identification was preliminary designed to verify it.The contents are as follows:(1)The development statues,advantages and disadvantages of the fault source identification in noise diagnosis at local and abroad were summarized.According to this,the source identification method based on blind source separation was selected.Moreover,the MUSIC algorithm,TCT algorithm and SOBI algorithm were studied,and their shortcomings in practical application were analyzed.(2)An improved 2D-MUSIC algorithm based on spatial rectangular array was proposed.In this algorithm,the two-dimensional spectrum search was divided into two parts: the one dimensional search of the elevation angle and one dimensional matching of the azimuth angle,which reduced the computational complexity and avoided the problem of elevation location ambiguity.Moreover,combined with step-varied search and spectral peak recognition technology,the one-dimensional search process was optimized,and the angle of sound source was extracted.The superiority of the improved algorithm was verified by the MATLAB simulation,and the design parameters of the sound source identification system were determined through the research of the performance parameters.(3)The TCT algorithm was improved.By the covariance matrix of the focusing frequency points,the construction process of the focusing matrix was avoided,and the focusing frequency selection efficiency of the different bandwidth signals was improved by the fusion with the SOBI algorithm.The performance of the improved algorithm was verified by simulation experiments.(4)Aiming at the unknown frequency mechanical sound source with broadband and narrowband signal,the solution of the sound source identification was discussed,and the realization process of the fusion algorithm was described in detail.This algorithm firstly completed the separation of multiple source signals by SOBI algorithm,then focused the known sound source in the frequency domain,and finally realized the rapid sound source localization through the improved 2D-MUSIC algorithm on the basis of focused spatial spectrum function.The feasibility of the sound source identification method was verified by the simulation experiment of the actual hydraulic motor noise signals.(5)Combined with the actual project requirements and the results of simulation experiments,a set of sound source identification system based on STM32 development board and IPC was designed,whose software platform was MATLAB and C++ Builder.After the test of the system function,the practicability of this sound source identification method was verified by the experiment of two simulated sound source in the open environment.