Research On Time-frequency Analysis Methods Of Fault Diagnosis Of Rolling Bearings Based On Acoustic Emission Techniques

Rolling bearings are the key components of rotating machineries, and are also very easy to be damaged. So it is very important to carry on the works of condition monitoring and fault diagnosis of rolling bearings, which will contribute to the development of our society and economy. Comparing with vibration methods that are used most extensively at present in the field of condition monitoring and fault diagnosis of rolling bearings, acoustic emission (AE) techniques have much superiority when applying for fault diagnosis of rolling bearings.Time-frequency analysis method can be used to process effectively non-stationary time-variation signal, and are applied extensively for fault diagnosis of rolling bearings based on vibration methods recently years. So the paper introduce time-frequency analysis method to the field of fault diagnosis of rolling bearings based on acoustic emission techniques, and carry on the research on time-frequency analysis method of fault diagnosis of rolling bearings based on acoustic emission techniques. The works include mainly four aspects. (1) Developing the acoustic emission data collecting system, together with the fault experimental platforms of rolling bearings and SWAES full-waveform acoustic emission detection instrument etc, the experimental system detecting faults of rolling bearings based on acoustic emission techniques is constituted. The acoustic emission experiments of typical faults of rolling bearings are carried on, and the characteristics of AE signals initiated by faults of rolling bearings and fault feature extraction principles are studied. (2) Three time-frequency methods of fault AE signals of rolling bearings are developed respectively: STFT (Gabor transform) analysis method, wavelet scalogram (wavelet reassigned scalogram) analysis method and WVD analysis method. The two-dimension and three-dimension time-frequency spectrums of all the methods can correctly describe AE events initiated by failures of rolling bearings, and can represent audio-visually the number, strength, frequency compositions and distribution in time-frequency surface of the pulse signals. So the effective information required by the precise diagnosis of rolling bearings can be obtained. (3) Actual measurement AE signals have three main shortcomings: containing many noise signals, having very big date size and very broad frequency scope. So it very difficult to extract feature signals from actual measurement AE signals. However, the feature extraction analysis method of wavelet packet is put forward by the paper, ant the method can solve the problems effectively. (4) A wavelet function is designed, which can be used effectively to extract fault features of AE signals of rolling bearings. The function is more effective than Daubechies wavelet function that is used widely at present, and can improve greatly the availability and accuracy of wavelet analysis of fault AE signals of rolling bearings. These research findings not only can improve greatly the efficiencies and accuracies of incipient faults prediction and diagnosis of rolling bearings, but also are favorable to promote the waveforms analysis techniques of AE signals of rolling bearings.