A novel nonlinear spectrum estimation method and its application in on-line condition assessment of bearing-rotor system
Yulai Zhao,Zepeng Liu,Junzhe Lin,Qingkai Han,Yang Liu
DOI: https://doi.org/10.1016/j.measurement.2023.113497
2023-08-31
Measurement: Journal of the International Measurement Confederation
Abstract:In this paper, a novel approach for nonlinear spectrum estimation is proposed and combined with Support Vector Machine (SVM) classifier to assess the condition of bearing-rotor systems, including fault types and severity. The proposed approach involves extraction and reconstruction of target harmonics, dynamic process modeling of the reconstructed signal based on a data-driven method, and extraction of nonlinear spectrum features in the form of nonlinear output frequency response functions (NOFRFs). The focus is on the influence of model stability and accuracy evaluation on the robustness of NOFRFs. Then the robustness of the NOFRFs based on the proposed estimation approach is compared with the conventional estimation approach using simulation cases and experiments. SVM is used to classify the extracted nonlinear spectrum features. Finally, the classification accuracy of NOFRFs evaluated by the proposed approach is compared with time-domain and frequency-domain features. Neighborhood Component Analysis (NCA) is used to select sensitive features and improve classification accuracy. The proposed condition assessment strategy does not require the selection of sensitive features, ensuring the stability of assessment results, which is also a significant advantage of the proposed strategy.