An Identification Method for Rotor Axis Orbits based on Enhanced Hierarchical Multivariate Fuzzy Entropy and Extreme Learning Machine

Abstract

The rotor system is the core equipment of industrial rotating machinery, and ensuring its safety is an essential basis for improving the productivity of the equipment. As a critical monitoring quantity reflecting the operating status of the rotor system, identification models based on axis orbits are effective means for detecting equipment faults. However, most of the existing axis orbit identification models belong to the category of image recognition, and these methods have defects such as unclear physical meaning of features and weak generalization performance. Therefore, the paper returns to the essence of axis orbits and proposes a rotor axis orbit recognition method based on multivariate swing signals, feature extraction and pattern recognition. Firstly, the mutually perpendicular swing signals of the rotor are obtained based on eddy current sensors. Secondly, we propose a feature extraction tool for extracting the multivariate signals named enhanced hierarchical multivariate fuzzy entropy (EHM_vFE), a nonlinear dynamics metric based on the enhanced hierarchical decomposition method. Next, the features of axis orbits are extracted by the EHM_vFE. Finally, some of the extracted features are input into an extreme learning machine (ELM) for model training, and the effectiveness of the method is verified with the remaining samples. We apply the proposed method to the rotor axis orbit identification case, and the results show that its recognition rate is 98.963%. In comparison experiments with recognition models based on nonlinear dynamics indicators, multivariate signal processing methods, traditional image feature extraction methods, and popular deep learning models, the proposed model shows substantial advantages, verifying the reasonableness and superiority of the proposed method. This study provides a new idea for rotor shaft fault diagnosis, which has significant reference value for promoting the development of intelligent operation and maintenance of industrial equipment.