Purification and feature extraction of shaft orbits for diagnosing large rotating machinery

Vibration-based diagnosis has been employed as a powerful tool in maintaining the operating efficiency and safety for large rotating machinery. However, due to some inherent shortages, it is not accurate enough to extract the features of malfunctions by using traditional vibration signal processing techniques. In this paper, a high-resolution spectrum is firstly proposed to calculate the amplitude, frequency and phase details of sinusoidal harmonic and sub-harmonic vibration in large rotating machinery. Secondly, on the basis of a high-resolution spectrum, a purified shaft orbit is reconstructed to remove the interference terms. The moment and curve features, which are invariant to translation, scaling and rotation of the shaft orbit, are introduced to extract the features from the purified vibration orbit. This novel scheme is shown to be very effective and reliable in diagnosing several types of malfunctions in gas turbines and compressors under operating conditions as well as in the run-up stages.