Fault Diagnosis Of Rolling Bearing Based On Wavelet Package Transform And Rough Set Theory

As a fundamental component part, the rolling bearing playing a role in bearing and driving load among the mechanical devices, it is the most pivotal and important part in the transmission system. Meanwhile, due to a bad working environment, it is very easy for rolling bearing to break down. Therefore, the analysis and study of rolling bearing fault diagnosis technique has an importantly scientific and realistic meaning.The traditional method of ball bearing fault diagnosis is to analyze on vibration signal of rolling bearing with FFT-based (fast Fourier transform) frequency domain or cepstrum to identify the faults. Exactly speaking, it is only suitable for analyzing the steady signals with the Gaussian distribution (the normal probability distribution). However, the failure signal of rolling bearing is unsteady and the fault feature is generally faint, hence it is quite difficult for the traditional method to diagnose the faults. Wavelet package transform is a JTFA (Joint Time-Frequency Analysis) technique with the capability of identifying the stronger weak signals and has the trait of MRA (Multi-resolution Analysis) compared with the traditional FFT-based frequency domain analysis technique and can extract more effectively the fault feature of a signal. Wavelet package transform has higher frequency resolution and lower time resolution in the low frequency of a signal and higher time resolution and lower frequency resolution in the high frequency, called a "microscope" of signal analysis. Using wavelet package analysis technique to detect and diagnose the faults of dynamic system can achieve good results.However, there are also many shortcomings existed in the method of wavelet package transform. For instance, using wavelet package transform to diagnose the faults need have certain prior knowledge, such as the structure parameter of rolling bearing, the rotational speed of machine, the theoretical calculation formula of characteristic frequency, etc. And the rotational speed of the machine in motion can always has fluctuations, even larger variations; the characteristic frequency with relation to the rotational speed also arises certain fluctuations, even a large jump. Moreover, for many types of rolling bearings, it is very difficult to derive the theoretical formulas as to calculating their characteristic frequencies, so there are many uncertainties in diagnosing rolling bearings.Rough set theory, which is very suitable for diagnosing the mechanical faults, is a new mathematical method to handle fuzzy, incomplete and undetermined information. It needn't any prior knowledge and systematic model of mathematics, but can extract useful information only from data themselves and reveal potential rules among them and make information simplified at the same time.The thesis combined the merits of wavelet package transform and rough set theory, and then proposed a gaining method, whose most distinct trait was that it needn't calculate the fault characteristic frequency of rolling bearing but only extract the energy characteristic values of a signal, of diagnosis rules based on the combination of wavelet package transform and rough set theory. Finally, the method was used to diagnose seven types of faults of rolling bearing and achieve good results.