Bearing Early Fault Detection Using Local Tangent Space Alignment and Hypothesis Testing

Bearing is an essential component in the rotatory machine. Its failure causes the sudden failure of industrial machines, which increases downtime and productivity loss. Early fault detection involves finding the fault initiation point where bearing enters from healthy to a failure state. Detecting the fault initiation in the bearing is essential to avoid deteriorating conditions and catastrophic failure of the machines. Extracting significant health indicators from bearing condition monitoring data is vital for detecting the accurate incipient fault. Initially, various features are extracted using signal processing techniques. The wrapper-based feature selection method is used to select the significant feature subset. Then, the Local tangent space alignment method is used to extract health indicators by fusing selected features. Finally, Hypothesis testing is used to detect the early fault in bearing, and the Support vector machine method is used to distinguish the healthy and fault states. The proposed methodology is verified using bearing accelerated life test data.