Non-contacting Acoustic Diagnosis Study For Rolling Bearing Based On Wavelet And EMD

The revolving equipments are large-scale important equipments which are widely used in the petroleum petrifies, the light industry, the transportation industry and so on. There is approximately 30 percent revolving equipment failure caused by damage of the rolling bearing. The most universal vibration examination method is judged by the high speed impact and the unusual sound based on the contact between rolling body and the fault bearing conveyer. However, the rotate speed of the low-speed revolution bearing rolling body is extremely low, so is the impact level. It is extremely difficult to recognize low speed fault rolling bearing using the vibration examination method. However, the break and the fatigue damage of the low-speed rolling bearing can be effectively distinguished using acoustic emission technique to analyze fault rolling bearing and collect acoustic emission signals from the rotational bearing. Compared with the traditional resonating demodulation diagnosis by vibration method, the acoustic emission method can be able to forecast the faults of bearing at the earliest time. The characteristic frequency and the characteristic signal are more obvious. The acoustic emission method can highlight its superiority.The traditional contacting acoustic emission inspection method maintained the sensor and the outside surface of bearing full coupling.Therefore it can slightly reduce the attenuation of AE wave. Not only the sensor received effective condition information of the bearing, but also inevitably received the disturbance information from no-examined parts. In certain specific situations such as trackside acoustic detection on train bearings, there is a relative translation between the sensor and the bearing besides revolving, it can not be directly pasted the sensor on the bearing surface or the bearing seat surface to monitor the bearing on line. Non-contacting AE method can be able to solve the above problem. In order to avoid introducing into some extra reverberations and the neighbor mechanical device noise so as to reduce the signal-to-noise ratio, the diagnosis only can be operated in the near sound field.The mechanism of causing rolling bearing AE waves and the diffusing rule were studied in the article as a fundamental research, under the laboratory condition, non-contacting AE inspection experiments to the different fault pattern of rolling bearings were carried. The AE signals obtained from rolling bearing were de-noised using wavelet noise reduction processing according to the characteristic of noise signal. Then modern signal processing technologies such as the wavelet analysis, the neural network, EMD and so on were used to analyze the characteristic of de-noise signals. Eventually fault patterns of rolling bearing were synthetically judged.Not only the formation and the expansion process of rolling bearings'attrition and superficial injury, but also the mutual friction and the collision between fault spot and the bearing component, can cause acoustic emission. The effective information received by the sensor mainly comes from the surface wave of which the AE wave attenuate and the pattern varied in the bearing and the air. The wavelet analysis can effectively pick out each kind of noise and distinguish the characteristic frequency band of AE signals from rolling bearing. The wavelet packet analysis provided a method which can make characteristic frequency spectrum of AE signal more accurately. The very good effect was get using the wavelet packet analysis to de-noise and to analyze the AE wave of rolling bearing. However, the wavelet and the wavelet packet analysis have merely offered characteristic frequency band of the bearings'acoustics signal, it is insufficient of them to carry on pattern recognition and the classifications to each kind fault of the rolling bearing. After using the wavelet packet to extract the characteristic of the AE signal, it can greatly enhance the fault diagnosis accuracy and the validity with the aid of the neural network pattern recognition technology.In addition, the EMD method auto-adapted withdrew the partial characteristic information of original signal. There are big breakthroughs of the Hilbert transformation adapting with EMD to decompose in the signal, to seek instantaneous parameter and to portray the frequency characteristic etc. Further, using EMD to decompose the wavelet de-noise signal, then making the Hilbert marginal spectrum of the effective IMF, can accurately portray the characteristic frequency of bearings'acoustic signal, and make the fault diagnosis goal well.