Abstract: Different from the FFT-based methods, the objects of the multifractal spectrum treaded with are the signals’ geometry features. Before, the inherent limitations of the algorithm for the multifractal spectrum limited its application. Now, the wavelet local maxima modulus based method is received more attentions because it can calculate the multifractal spectrum effectively and simply. The concept of multifractal spectrum is presented systematically and the wavelet local maxima modulus based method is introduced detailedly. The way of applying the multifractal spectrum for fault diagnosis is discussed. The vibration signals of four kinds of typical faults：unbalance, oil whirl, coupling misalignment and rub-impact are then studied. The result shows that the geometry features of vibration signals can be reflected through multifractal spectrum, which provides an effective way for fault diagnosis.

Key words: Fault diagnosis, Vibration signal, Wavelet analysis, Wavelet modulus maxima Multifractal