Cepstrum editing for enhancing bearing fault diagnosis of a gearbox
Abstract
Bearings are a critical component in rotational machinery and especially in gearbox. The vibration signals generated by bearings are impulsive, non-periodic, and low amplitude. The signals are often buried in the high-amplitude components like gearmesh, imbalance, and misalignment signals. These high amplitude effects make it difficult to identify bearing fault frequencies even using enveloping techniques. To improve the diagnostics, it is important to increase the bearing fault signal-to-noise ratio. In this paper, a cepstrum technique was used to systematically remove high amplitude signals associated with gearmesh and shaft speeds. Residual was signal was then obtained to increase the bearing fault signal-to-noise ratio. Spectra kurtosis was performed to obtain optimum filter bands to extract the fault frequency components of the bearing. The envelope analysis was successfully applied to determine bearing fault frequencies. Vibration signals with different types of seeded faults were collected on a machinery fault simulator (MFS) and a drive train diagnostics simulator (DDS). The test results have proven the effectiveness of the cepstrum editing methodology. The Cepstrum analysis is computational very efficient and powerful tool for bearing and gearbox fault diagnosis.
