Abstract
This research focuses on the development of a machine learning-based approach for the early diagnosis of blade rubbing in rotary machinery. In this paper, machine learning-based diagnostic methods are used for blade rubbing early diagnosis, and the faults are simulated using experimental models. The experimental conditions were simulated as follows: Excessive rotor vibration is generated by an unbalance mass, and blade rubbing occurs through excessive rotor vibration. Additionally, the severity of blade rubbing was also simulated while increasing the unbalance mass. And then, machine learning-based diagnostic methods were applied and the trends according to the severity of blade rubbing were compared. This paper provides a signal processing method through feature analysis to diagnose blade rubbing conditions in machine learning. It was confirmed that the results of the unbalance and blade rubbing represent different trends, and it is possible to distinguish unbalance from blade rubbing before blade rubbing occurs. The diagnosis using machine learning methods will be applicable to rotating machinery faults like blade rubbing; furthermore, the early diagnosis of blade rubbing will be possible.