The Research Of Rolling Bearing Fault Diagnosis Under Time-varying Rotational Speed And Gear Noise

Performing fault diagnosis of rolling element bearings plays a significant role in the steady operation of the mechanical system. For one thing, the operation of most mechanical systems contains angular motions which makes the rolling bearing one of the most commonly used component, for another, as its name suggests, the rolling bearing bears much of the load. These together with the large life dispersion and harsh operating environment make the bearing one of the most failure-prone machinery components in the rotating machine. In real engineering, reliable bearing condition monitoring is crucial to avoid unexpected machine breakdowns.Owing to the key roles the rolling bearing plays in the general mechanical systems, many researchers in the field of fault diagnosis have done lots of works on it and an effective Fourier transform (FT)-based method called resonance demodulation have been proposed. However, it will success only with the hypothesis of steady rotational speed and the analysis object is also restricted to the vibration signal of rolling bearing itself without other vibration sources. With the trend of complication and integration of mechanical machine, the more changeable working mode and complicate environment will increase the difficulty of the fault diagnosis of the rolling bearing. In real engineer, large range of speed variation will change the characteristic of rolling bearing vibration and the gear noise, one of the most commonly interruptions, will block the feature identification of faulty rolling bearing in both time-domain and frequency-domain. As such, this thesis plans to solve the problem of rolling bearing fault diagnosis under varying rotational speed and the interruption of gear noise in the basis of the detail analysis of the characteristic of faulty rolling bearing and gear vibration under the time-varying rotational speed with the framework of envelope analysis and order tracking.Meshing vibration of gearbox is one of the most common interruptions towards the fault diagnosis of the rolling bearing. It’s necessary to eliminate this kind of noise before performing the bearing fault diagnosis. However, two problems, the time-varying rotational speed working mode and the difficulty of locating ideal position of reference transducer, will disable the traditional gear noise elimination algorithms. So, a concept of instantaneous dominant meshing multiple (IDMM) trend is put forward at first based on the analysis of gear noise characteristic under time-varying rotational speed; and, at second, an ideal reference signal which only relates to the gear noise and the determination rules of key parameters (adaptive filter length and step-size parameter) for adaptive noise cancellation algorithm are constructed based on the IDMM trend; and the final gear noise elimination algorithm based on the newly constructed ideal reference signal and the key filtering parameters for the time-varying rotational speed mode is proposed without additional auxiliary device and hurt of rolling bearing part.Acquiring the changing rule of bearing rotational frequency is the precondition of eliminating the influence caused by time-varying rotational speed. However, for a rolling element bearing, the signal components related to rotational speed usually cannot be directly extracted from the vibration signal itself or the other representations if no speed sensor, such as the tachometer, is available. To solve this problem, this thesis plans to use the strategy of constructing a candidate trend. In specific, the comparison between the two traditional bearing fault diagnosis algorithms separately based on the envelope spectrum of the time domain and order envelope spectrum of the angular domain are performed at first and the two basic characteristics of the candidate trend are summarized as follows:it should be synchronous with the bearing rotational speed trend and the ratio between them should be a constant value. A qualified candidate trend, the instantaneous fault characteristic frequency (IFCF), is then constructed in terms of the vibration characteristic of faulty rolling bearing. At last, signal filtering via kurtogram based spectral kurtosis (SK) analysis and the short time Fourier transform (STFT) are used together to generate the time-frequency representation (TFR) of the filtered signal with clear IFCF trend line which can then be extracted used the amplitude-sum based spectral peak search algorithm.The order tracking method is the most commonly used and efficient algorithm to eliminate interruption stemming from the time-varying rotational speed. However, if this algorithm is employed to analysis the vibration signal of faulty rolling bearing under time-varying rotational speed, there will exist deviation between the real resampled outcome and the ideal one which will then affect the bearing fault diagnosis in the next step. To fulfill this gap, the unchangeable characteristic of the peak time of the impulse response caused by the bearing fault under time-varying rotational speed is deduced using the calculation model of the impulse response at first; and the reason of envelope deformation is then explained by analyzing how the unchangeable peak time affects the traditional order tracking algorithm; and the IFCF trend based fault angular resampling is finally accomplished using the algorithm segments fitting and envelope deformation elimination.There exist obvious differences between the IFCF trend based resampling outcome and the traditional one in the angular domain, and the demodulate gear rotational order will act as the interruption when the envelope analysis is used for the final bearing fault diagnosis. To bridge these gaps, the thesis explains the fault characteristic order (FCO) spectrum based fault diagnosis principle at first, then discusses the way how the demodulated gear rotational order affects the identifying of the fault type, and finally constructs the fault characteristic order template to realize the fault diagnosis of the rolling bearing by directly locating the bearing rotational order.To testify the effectiveness of the proposed algorithms, the simulation signal is firstly constructed based on the vibrational characteristic of the faulty bearing and gear noised under the time-varying rotational speed. And the real data of the mixed signal under the synchronous and asynchronous operation situations are then acquired by using and reforming the SpectraQuest machinery fault simulator at the University of Ottawa lab. Furthermore, the constructed simulated signal and measured real data are used for verifying the effectiveness of the algorithms proposed in every chapters and the flowchart of the whole algorithm is summarized vividly using a mixed vibration signal of rolling bearing with outer race fault and gear noised under the synchronous operating mode. Finally, conclusions of the whole thesis are draw and further research directions are prospected.