Scalable Metric Meta-learning for Cross-domain Fault Diagnosis of Planetary Gearbox Using Few Samples

Abstract

Planetary gearbox is one of the key components of rotating machinery, and it has important significance to accurately identify different faults types during its operation. Recently, the deep learning techniques have been broadly used in the field of fault diagnosis. However, most of the existing intelligent fault diagnosis methods of planetary gearbox usually require a large number of labeled samples, and the training and testing samples should be the same distribution, which is difficult to achieve in actual industrial scenarios. To address the above challenges, a new scalable metric meta-learning method is proposed for cross-domain fault diagnosis of planetary gearbox with few samples. In the method, a scalable distance metric function is designed which can efficiently evaluate the similarity of fault samples and improve the generalization performance. The proposed method is used to analyze the vibration signals collected from planetary gearbox, and a number of cross-domain experiments with different few samples are used to fully verify the effectiveness of our method. The results indicate that the performance of our method is superior to other comparison methods.