Transfer fault diagnosis of bearing installed in different machines using enhanced deep auto-encoder

Abstract

The collected vibration data with labeled information from bearing is far insufficient in engineering practice, which is challenging for training an intelligent diagnosis model. For this purpose, enhanced deep transfer auto-encoder is proposed for fault diagnosis of bearing installed in different machines. First, scaled exponential linear unit is used to improve the quality of the mapped vibration data collected from bearing. Second, nonnegative constraint is adopted for modifying the loss function to improve reconstruction effect. Then, the parameter knowledge of the well-trained source model is transferred to the target model. Finally, target training samples with limited labeled information are adopted for fine-tuning the target model to match the characteristics of the target testing samples. The proposed approach is applied for analyzing the measured vibration signals of bearings installed in different machines. The analysis results show that the proposed approach holds better transfer diagnosis performance compared with the existing approaches.