Bearing remain life prediction based on weighted complex SVM models

作者：Shaojiang Dong, Jinlu Sheng, Zhu Liu, Li Zhong, Hanbing Wei · 发表于：Journal of Vibroengineering · 年份：2016 · DOI：10.21595/jve.2016.16910 · 被引用次数：11 · 研究领域：Machine Fault Diagnosis Techniques、Gear and Bearing Dynamics Analysis、Lubricants and Their Additives

Aiming to achieve the bearing remaining life prediction, this research proposed a method based on the weighted complex support vector machine (SVM) model. Firstly, the features are extracted by time domain, time-frequency domain method, so as the extract the original features. However, the extracted original features still with high dimensional and include superfluous information, the multi-features fusion technique principal component analysis (PCA) is used to merge the features and reduce the dimension. And the bearing degradation indicator is constructed based on the first principal component, which can indicate the bearing early failure state precisely. Then, based on the life condition indicator, the weighted complex SVM model is used to achieve the bearing remain life prediction, in this model, the particle swarm algorithm (PSO) method is used to select the SVM internal parameters, the phase space reconstruction algorithm is used to determine the structure of the SVM. Cases of actual were analyzed, the results proved the effectiveness of the methodology.