A fault diagnosis method of the two-dimension image fractal theory based on time-frequency image

The feature extracting is an important, difficult step in the fault diagnosis processing through the analysis of plant vibrations, essentially because of the strong nonlinearity of the vibration signals collected from machine. Fourier analysis is unable to describe the time-frequency(T-F) localized characteristic because it is global transform. Here, we adopt a new time-frequency methods--Local wave method, to analysis the fault vibration signals. Because the Local wave T-F spectrum can be showed in the gray image, so the fault features were extracted using the fractal dimension and moment invariants for the two-dimension local wave T-F images. The experiments show that this method is feasible.