A Fault Classification Method for Rotor under Fluctuating Condition based on COTHS-CNN

Plenty of challenges hindered a proper condition monitoring of a rotor system in reality, such as rotating shaft speed fluctuating, dealing with large amount of measured data, an incomplete description of rotor state via a single sensor and etc. To tackle these issues, a sequential use of signal processing methods, Computed Order Tracking (COT) and holospectrum (HS) are utilized to extract features from raw vibrations and these features are fed into the Convolutional Neural Networks (CNN) later. The so-called Computed Order Tracking and holospectrum based Convolutional Neural Networks (COTHS-CNN) is proposed and enables the capacity of fault pattern recognition under unstable conditions for rotor. The experiment results demonstrate the effectiveness of the proposed method.