Research On Timbre Feature Of Structural Sound Based On HSP And LPC

In the process of solving the NVH problem,it is often necessary to determine the sound radiation source to suppress the noise.Some existing methods of sound source identification and location have some limitations,and the operation process is complicated,Therefore,in order to identify the sound source directly according to the timbre features of noise,the extraction and analysis of structural timbre features are studied as follows:Firstly,analyzed the mechanical structure noise system and the voice signal system,it is found that both system have high similarity from incentive source to transfer path,then the feasibility of applying the speaker recognition algorithm to the field of mechanical fault diagnosis is demonstrated;Combined with the content of speech timbre recognition,the timbre of structural sound is defined;after the excitation signal in the noise radiated by a structure is stripped,the remaining part of the fixed information is taken as the timbre feature of the noise radiated by the structure,to a certain extent,this fixed component can be regarded as the transfer function from excitation to response.Secondly,in order to extract and analyze the timbre features of structural noise,The basic algorithm principle of Homomorphic Signal Processing(HSP)and Linear Prediction Coding(LPC)is analyzed,and the possibility of structural timbre feature extraction is demonstrated theoretically;The simulation results show that the two algorithms can extract the transfer function curve contained in the response signal under different excitation input,the feasibility of timbre extraction is demonstrated.Thirdly,in order to further analyze the two feature extraction algorithms,based on MATLAB,the simulation of signal is carried out to explore the influence of two algorithms on the accuracy of simultaneous interpreting of transfer function curves under different excitation types,damping ratios of different transmission systems,and parameter settings of the algorithm itself.The simulation results show that the two algorithms can separate the path information and incentive information from the response signal,and the two algorithms are highly complementary.In view of this,the results of the two algorithms are fused to improve the accuracy,the simulation results show that this strategy is more accurate in the estimation of system transfer function.Fourthly,the noise timbre characteristic curves of the engine under five working conditions are extracted,through the correlation verification,the timbre characteristic curves under five working conditions are highly similar,and the correlation coefficients are higher than 0.97.The modal analysis of the right cover of the gearbox is carried out.The maximum error is less than 5% by comparing the peak frequency in the timbre characteristic curve of the right cover with the first ten natural frequencies of the right lid,Moreover,the voice characteristic curve of the right lid is more consistent with the frequency response curve of the right lid.The feasibility of timbre feature of structural sound extraction and the effectiveness of the algorithm are proved.Finally,the noise problem of air conditioning in an SUV is analyzed based on the characteristic of timbre analysis strategy.By extracting and analyzing the timbre features of the noise,it is determined that the HVAC air conditioning box is one of the main radiators of the interior noise,and the 760 Hz vibration generated by the compressor is transferred to the air conditioning box through the low pressure pipe and converted into nois.Through this case,the practicability of the method is proved.