Abstract: A new method based on bispectral analysis for rotating machinery faults diagnosis is presented. Applying the bicoherence spectral analysis to the vibration signals, the features that signify the condition of the machine can be derived. These features quantitatively describe the nonlinear phase couple due to faults. Experiments show that bicoherence spectrum are very sensitive to the faults and can identify the fault pattern validly.

Key words: Bicoherence spectrum, Bispectrum, Faults, Nonlinear phase couple, Rotating machinery