Incipient bearing fault diagnosis based on improved Hilbert-Huang transform and Support Vector Machine

The detection and diagnosis of equipment failures are of great practical significance and paramount importance in the sense that an early detection of these faults may help to avoid performance degradation and major damage. In this work, a novel methodology based on improved Hilbert-Huang transform (HHT) and support vector machine (SVM) was proposed for incipient bearing fault diagnosis with insufficient training data. Singular value decomposition (SVD) was employed to detect periodic features, and then extending of the original signal was carried out based on support vector regression (SVR). A screening process was conducted to select the vital intrinsic mode functions (IMFs). Finally, features extracted from the obtained IMFs were applied to identify different bearing faults based on SVM. To investigate the property of proposed method, an experimental test rig was designed such that varying sizes defects of a test bearing could be seeded, and it's concluded that the effectiveness of the proposed algorithm in early bearing fault diagnosis even with insufficient training data.