Wavelet-based multi-fractal spectrum for machine defect identification

This paper presents a technique based on the wavelet-based multi-fractal singularity spectrum for rotary machine defect identification. Specifically, vibration signals measured by accelerometers are decomposed into a series of scales, with each scale corresponding to a sub-frequency band, by means of the continuous wavelet transform (CWT). The multi-fractal spectrum is then calculated from the wavelet coefficient modulus-maxima lines. Comparing to other signal processing techniques, the inherently flexible time-frequency resolution property of the wavelet transform characterizes the scaling properties of the multi-fractal spectrum, thus is more effective in singularity identification. Experimental studies on rolling bearings and a gearbox have shown that the presented technique provides an effective tool for defect identification.