Research On The Analysis Method Of Cyclostationary Fault Signal Of Machinery

Due to the physical structure and running feature of rotating machinery, its fault signal is essentially cyclostationary, i.e., the statistics of the signal are periodic. It belongs to a special kind of non-stationary signal, and the extraction of fault feature is difficult by the traditional signal processing method based on the stationary theory. Aiming at the cyclcostationary property of fault signal of rotating machinery, the analysis and feature extraction methods for the cyclcostationary fault signal are researched in this dissertation.Firstly, the cyclostationary feature of the fault signal of rotating machinery is researched and analyzed home, and the similarities and differences of cyclostationary order and the least cyclostationary order are clarified; based on the exact periodicity of fault signal with angle independent variable, the affection of speed fluctuating to the cyclcostationarity of the signal is researched, it is shown that time domain average can't denoise effectively because of the non-steady rotating speed. The cyclostationary feature of rolling bearing fault signal is analyzed. Based on the broad frequency belt feature of fault signal, the author advances combining the band-pass filtering with the two order cyclic statistics to extract the fault feature of rolling bearing.Secondly, based on the cyclostationarity of fault signal, using the cyclic moment analysis to extract the fault feature is researched. The demodulation principle is deduced, and the author points out that the demodulation utility of two order cyclic moment is the result of the spectrum difference. Aimed at the complicated cyclic moment spectrum, Hilbert transform combined with two order cyclic moment is advanced. The hunting domain of cyclic frequency is researched, and the author points out that the time lag...