Bispectral analysis for diagnosis of gear pitting fault

A mechanical system shows always unstable, non-Gaussian and nonlinear characteristics to some extent, especially when the system suffer from defect, therefore it is difficult for conventional vibration analysis based on Fourier transform to find weak fault information. High-order spectral analysis (HOSA) provides a more effective tool. In this paper bispectrum and 1(12)-dimension spectrum were used to analyze vibration signals from gearbox with various pitting faults and to extract the fault feature by slice. Through comparison between normal and fault conditions it can be concluded that bispectral analysis has a stable performance and is very sensitive to faults.