Acoustic Feature Extraction Of Mechanical Impact Fault Based On Blind Signal Processing

Signal processing is the core problem of fault diagnosis.Its purpose is to extract the fault characteristics carried in the signal.Noise signal,also known as acoustic signal,is also suitable for mechanical fault diagnosis and monitoring compared with vibration signal,and has the advantages of contactless monitoring,quick and easy operation,low requirements for sensors,etc.However,due to its unique transmission mode,it has formed a special sound field environment,It is necessary to pay attention to noise and signal model;therefore,this paper will focus on the research of feature extraction of mechanical acoustic signal based on denoising algorithm and blind deconvolution algorithm in time domain.In order to further extract the buried information of acoustic signals,this paper studies and explores the acoustic signal feature extraction algorithm based on the blind signal processing convolution and aliasing model.The performance of EFICA and RobustICA in blind deconvolution in time domain is compared and verified from two aspects of theoretical model and practical application.It is found that in blind deconvolution of impulsive signals,EFICA algorithm is more steady and effectiveness.In addition,considering that the propagation process of acoustic signal is susceptible to interference,the denoising algorithm suitable for mechanical acoustic signal processing is studied.From the core theoretical basis and specific implementation steps of the algorithm,this paper explores the wavelet decomposition and reconstruction algorithm based on wavelet analysis,its advanced optimization algorithm,translation invariant wavelet denoising method,and then discusses the Kalman filter.Based on the above,an improved method is proposed.The improved algorithm has more powerful denoising effect,which can be found from the experimental results.Moreover,It can effectively denoise the impact signal.It can be applied to the fault monitoring of mechanical acoustic signal or impact signal processing in strong noise environment,and improve the signal-to-noise ratio of mechanical signal.In the process of realizing blind deconvolution in time domain,especially when the dimension of signal subspace increases greatly,clustering algorithm is necessary.In this paper,the improved KFCM algorithm for FCM and the FCM algorithm under the extended condition are introduced,Combined with the theoretical model and its application in image signal,a clustering algorithm,TF_KFCM algorithm is proposed,which can take into account the characteristics of time domain and frequency domain.Through the clustering experiment of the actual impact signal,compared with the general KFCM algorithm,the effectiveness of the new algorithm is verified.Finally,the above improved research is integrated into the mechanical fault extraction algorithm,and a new KETFE algorithm is proposed.It is applied to the simulation acoustic signal fault diagnosis experiment,from which the fault feature signal is successfully extracted,which verifies the effectiveness of the improved algorithm in the acoustic signal fault feature extraction.