Optimization of different vibration acceleration and velocity features for faults diagnosis in rotating machines

An earlier initiative to develop a multiple speeds vibration-based faults detection method that could eliminate or significantly reduce the need for per-forming rotating machines faults diagnosis at individual speeds was based on the unification of features computed from acceleration-based vibration data. While this unified multi-speed analysis (UMA) provided useful insights to the possibili-ties of simplifying rotating machine faults classification, there are still concerns that the sole dependence on acceleration-based features alone for all faults classes (e.g. bearing, gears, blades and rotor-related faults classes) may or may not be to-tally realistic. Therefore, the current study combines acceleration-based time do-main features (i.e. kurtosis, crest factor, root-mean-square) and velocity-based frequency domain features (i.e. spectrum energy, 1x-5x) so as to improve its applicability for a wider range of rotating machine faults. The findings from the current study do not only validate the potentials of multi-speed faults classification for rotor-related conditions, but also indicates that significant enhancements can be achieved through the incorporation of velocity features.