Research On Signal Noise Reduction And Underdetermined Blind Extraction In Bearing Fault Diagnosis

Bearings are one of the most important parts in industrial production,and their operating state greatly affects the overall safety and the process of production equipment.Therefore,the research on bearing condition monitoring and fault diagnosis is particularly important.The method is to analyze and process the signals collected by the sensors.However,the signals collected under actual working conditions usually contain strong noise with complex sources,multiple faults are coupled to each other,and the number is more than the number of collected signals.The underdetermined Blind Source Separation algorithm is the insufficient to solve the above problems.Therefore,the research on bearing faults,especially the noise reduction of composite fault signals and the separation of underdetermined blind sources are of great significance.This paper mainly studies the noise reduction and blind separation of the composite fault signal of the bearing,and proposes an improved method to reduce the noise of complex noise signals.At the same time,an underdetermined blind extraction method based on improved noise reduction and Compressed Sensing is proposed to solve the problem of underdetermined blind extraction.The main work of this paper includes the following contents:First of all,from the practical application of the project,the source of the topic and the research background of this article are briefly explained.The current status of domestic and foreign researches on Blind Signal Processing technology,underdetermined Blind Source Separation technology and the noise reduction are summarized and existing problems are summarized.Secondly,in view of the problems of complex noise sources,strong background noise,and non-linearity of the signals collected in the actual scene,a noise reduction method combining improved Morphological Filtering and improved Wavelet Threshold noise reduction is proposed.The proposed method uses an improved noise reduction method to filter,and at the same time separates the bearing composite faults by the Sparse Component Analysis,and the Fast Fourier Transform and envelope analysis of the separation results provide frequency domain verification of the separated signals.The simulation results and the analysis results of the acoustic signals collected at the bearing site verify the effectiveness of the improved noise reduction method.Finally,aiming at the problem that multiple noise and multiple fault signals overlap with each other under actual working conditions,which leads to insufficient underdetermined blind extraction,an underdetermined blind extraction algorithm based on improved Morphology-Wavelet Threshold noise reduction and Compressed Sensing is proposed.First,the improved Morphology-Wavelet Threshold noise reduction method is used to reduce noise and the mixed matrix is estimated by Fuzzy C-means clustering,then the K-means Singular Value Decomposition is used to train the sparse dictionary,and finally the Orthogonal Matching Pursuit algorithm is used to recover the source signal.Underdetermined blind extraction experiments of bearing composite faults prove the effectiveness of the method.