A Cross Domain Feature Extraction Method Based On Transfer Component Analysis For Rolling Bearing Fault Diagnosis

Feature extraction plays a significant role in the rolling bearing fault diagnosis. However, the complexity and variability of the actual working condition leads to the data unstable and fault characteristics unpredictable. Traditional machine learning methods won't work or have a poor performance under this circumstance. In this paper, we propose a cross domain feature extraction method based on the Transfer Component Analysis algorithm to solve the problem. Transfer Component Analysis, as a novel method in transfer learning field, is an efficient method for cross domain feature extraction problem. The performance of the proposed method is verified with experiments using the actual rolling bearing data.