Theoretical and experimental analysis of bispectrum of vibration signals for fault diagnosis of gears

• The bispectrum of the signal containing multiple periodic components is studied.
• Biphase verification is essential for decreasing the false alarm of bispectrum-based method.
• The bispectrum characteristics are revealed for the vibration of faulty gear.
• A diagnostic approach by biphase verification is proposed for gear faults.

Condition monitoring and fault diagnosis is an important issue for gearbox maintenance and safety. The critical process involved in such activities is to extract reliable features representative of the condition of the gears or gearbox. In this paper a framework is presented for the application of bispectrum to the analysis of gearbox vibration. The bispectrum of a composite signal consisting of multiple periodic components has peaks at the bifrequencies that correspond to closely related components which can be produced by any nonlinearity. As a result, biphase verification is necessary to decrease false-alarming for any bispectrum-based method. A model based on modulated signals is adopted to reveal the bispectrum characteristics for the vibration of a faulty gear, and the corresponding amplitude and phase of the bispectrum expression are deduced. Therefore, a diagnostic approach based on the theoretical result is derived and verified by the analysis of a set of vibration signals from a helicopter gearbox.