Order Tracking For Fault Signal Of Rolling Element Bearings Based On Improved MCA And NCT

Rolling element bearing is not only one of the most important component in rotating machinery, but also a main resource of faults among machinery equipment whose failure often directly leads to the abnormality. According to this basis of fact, it is a great significance in fault diagnosis.The vibration signal of rolling element bearing contains components from different sources. These components which provide primary evidence of fault can be separated by their variety of morphology. Morphological component analysis (MCA) has been proven available of component-separation by sparse representation of each morphological component utilizing distinguished dictionaries. Moreover, speed fluctuation which makes the signal become non-stationary often occurs in running condition of rolling bearings. Order tracking is the most useful method that can stabilize the signal. The key to order tracking is the acquisition of instantaneous frequency (IF). Nonlinear chirplet transform (NCT) uses the concept of polynomial approximation to effectively estimate the IF. The main works of this thesis are as follows:(1) An improved MCA based on TQWT has been used for the different ratio of central frequency to bandwidth of each component in fault vibration signal. The effectiveness and disadvantages of MCA has been studied. An improved method of MCA has been proposed to overcome the difficulty of components separation in conventional filter-based method. In the proposed method, sparse dictionaries are constructed by utilizing different Q-factor wavelets to decompose the signal into rotating component, resonance component and noise component. Therefore, the fault type can be determined by envelope analysis of the resonance component. The effectiveness and accuracy of the proposed method is demonstrated by experiments.(2) Due to the time-varying feature of the signal caused by speed flutuation, a NCT-based order tracking method is proposed. By searching the peak of the time-frequency spectrum that acquired by NCT, the proposed method is capable of estimating the IF from the raw signal directly.Thus, the angular resample can be applied. The accuracy and robust of the proposed method have been proven by analysis of a simulated signal.(3) An order tracking method based on improved MCA and NCT has been proposed for rolling bearings under speed fluctuation and strong background noise. Each morphological components has been separated firstly by improved MCA. Meanwhile, the instantaneous frequancy has been extracted by NCT and peak searching. On this basis, an envelope order spectrum is utilized to identify the characteristic frequency in the order domain. The effectiveness of the proposed method is demonstrated by analyzing a real vibration signal collected from a helicopter engine gear box.