Applied Research On The Theory Of Single Channel Blind Source Separation In Fault Diagnosis Of Rolling Bearings

Along with the improvement of automation, the mechanical systems are more and more complex and precise, and condition monitoring and fault diagnosis of mechanical systems is increasingly important. Rotating machinery is one of the most widely used in mechanical equipment. About 30% of rotating machinery fault was caused by rolling bearings. Single Channel Blind Source Separation algorithm in the application of the fault diagnosis of rolling bearings.Firstly, the background and significance of the dissertation are elucidated, and the current situation of Blind Source Separation and its development in the field of mechanical fault diagnosis are introduced.The common algorithms of Independent Component Analysis are studied, then the number of Gauss-noise and the non-gaussian of source signal influenced on separation effect is studied. Simulation results show that the Gauss-noise number of the source signal is less than one; and the noise component of source signal is less, the separation effect is better.In order to meet the positive definite constraint conditions of Single Channel Blind Source Separation, the Single Channel Blind Source Separation method based on Preprocessing and Principal Component Analysis is studied. Its principle is that the single channel signal is preprocessed firstly, the main components is extracted by means of PCA to estimate the number of source signals. Simulation results show that the different methods based on preprocessing can affect the separation efficiency. To seek an effective preprocessing method is the key to Single Channel Blind Source Separation.Frequency Slice Wavelet Transform can reconstruct any frequency band without relying on wavelet function. Interesting frequency bands are extracted and reconstructed, then the fault feature information of bearing damage is extracted effectively. On the basis of studying characteristics of FSWT, a method of Single Channel Blind Source Separation based on FSWT is proposed, and the method of Single Channel Blind Source Separation based on FSWT can realize the fault feature extractions of rolling bearings.Based on the problems of ambiguities and threshold setting of Blind Source Separation method, Constrained Independent Component Analysis is studied, Combining with the characteristics of rolling bearing damage failure, the Constrained Independent Component Analysis algorithm is adopted based on pulse algorithm, and the effect of the cycle, phase, width of reference signal on the extraction results is studied. On this basis, a blind source separation method of single channel bearing vibration signal based on pulse CICA is proposed. Its principle is that the single channel signal is preprocessed by FSWT firstly, the envelope signals of the reconstruct signals can be obtained, the pulse reference signals are established based on the prior knowledge of fault characteristic frequency, envelope signals of the reconstructed signals and the reference signals are selected as the input matrix of CICA algorithm. This algorithm can realize the fault diagnosis of rolling bearings.