Roller bearing fault diagnosis based on locality preserving projection

A novel method for roller bearing fault diagnosis was presented based on locality preserving projection (LPP) and adaptive boosting algorithm (Adaboost). The original dataset for vibration signals was constructed, including time domain parameters, frequency domain parameters, and time-frequency domain parameters. Successively, dimension reduced features from the original dataset were extracted by using LPP. And finally, the adaptive boosting algorithm was applied for training and classification. The situations of normal condition, inner race defect, outer race defect, and ball defect of roller bearings were analysed. To verify its advantages, some comparative trials and simulation results show its effectiveness and superiority.