The Application Of Blind Source Separation In The Condition Of Completed And Underdetermined In Mechanical Fault Diagnosis

In the processing of condition monitoring and fault diagnosis about the mechanical vibrations, all kinds of noise and compounded signals could be obtained by sensors. Thus, it is significant to study how to separate and extract the characteristics of failure from the observed signals. The present study is involved in three technologies of blind source separation and their applications in the field of mechanical fault diagnosis. The mechanical vibration signals is taken as studying object and blind source separation (BSS) is taken as researching method. To solve the compounded fault of rolling bearing, independent component analysis (ICA) and sparse component analysis (SCA) were studied with the morphological filtering.The main contents are given as below:(1) In the background of mechanical fault diagnosis, the research status of ICA, SCA and Non-negative matrix factorization (NMF) is summarized and the application conditions of BSS in the field of mechanical fault diagnosis is also summarized.(2) A modified ICA method which is based on particle swarm optimization is applied to the fault diagnosis of rotator compound fault. Because the common gradient algorithm is adapt to get the local optimization rather than the global optimization and the iteration step plays an important role in the convergence rate to the algorithm.(3) The coupling mechanism of compounded failure about rolling bearings is studied by experiments, for various fault coupled with each other and the general BSS algorithm can't separate the failure characteristic effectively. According to the character that vibration signals of faulty rolling bearing are non-stationary and high frequency modulation, a new method which is based on morphological filtering and ICA is proposed to separate compounded signals of rolling bearing.(4) The number of observed signals must be greater than the number of original sources is a primary requirement for most of BSS methods. SCA was applied to the diagnosis of compound fault of rolling bearing. A new method which is based on morphological filtering and SCA is proposed to solve over-completed and underdetermined conditions about BSS.(5) NMF and its application in BSS are studied, and NMF were used to the fault diagnosis of rolling bearing.(6) A demo system about three aspects of BSS which is applied in the fault diagnosis of mechanical vibration signals is implemented based on theoretical investigation and experimental research. The system is tested by simulated signals and actual mechanical vibration signals.