Research on condition assessment method based on projection one-class classifier

• Published in: 2017 Prognostics and System Health Management Conference (PHM-Harbin) pp. 1 - 6
• Main Authors: Yongsheng Zhu, Fangzhe Wang, Xian Wang, Ke Yan, Xiaoran Zhu
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.07.2017
• Subjects: condition assessment; Data models; Degradation; Fatigue; Feature extraction; feature vectors; Indexes; kernal trick; Kernel; one-class classifiers; Rolling bearings; SVDD
• DOI: 10.1109/PHM.2017.8079129

In order to solve the problem that the anomalous samples are scarce and the model is susceptible to abnormal data, this paper introduces the idea of kernel trick in the process of constructing the projection classifier and constructs three kinds of projection one-class classifiers: Projection Support Vector Data Description (PSVDD), Projection K-means (PK-means) and Projection K-centers (PK-centers) by means of the Feature Vector Selection and Projection (FVSP). In order to further improve the performance of the condition assessment model, this paper uses the method of ensemble learning and evidence theory with PSVDD, PK-means and PK-center and puts forward a new assessment index: Health Index (HI). Finally, the fatigue life experiment of rolling bearing is carried out to verify the effectiveness of the proposed method.