Semi-Supervised Prototype Network with CBAM and Data Selector for Few-Shot Bearing Fault Diagnosis

In this paper, an improved semi-supervised prototypical network method is proposed to improve the performance of the bearing fault diagnosis model in the context of data scarcity. Firstly, a metric-based meta-learning method, the prototype network model, is introduced to train a general fault featur...

Full description

Saved in:
Bibliographic Details
Published in2022 4th International Conference on Data-driven Optimization of Complex Systems (DOCS) pp. 1 - 6
Main Authors Zhou, Wenkang, Li, Ning
Format Conference Proceeding
LanguageEnglish
Published IEEE 28.10.2022
Subjects
Fault diagnosis
Feature extraction
Few shot
Industries
Learning systems
Machine learning
Measurement
Meta learning
Prototypes
Semi-supervised learning
Online AccessGet full text

Cover

More Information
Summary:In this paper, an improved semi-supervised prototypical network method is proposed to improve the performance of the bearing fault diagnosis model in the context of data scarcity. Firstly, a metric-based meta-learning method, the prototype network model, is introduced to train a general fault feature learner. And an attention mechanism module is introduced to the model to better extract the fault features of bearings. Secondly, unlabeled fault data are utilized to tune the original prototype in a semi-supervised manner to improve the performance of the model. Thirdly, to reduce the disturbing influence of new classes of data in unlabeled samples on prototypes, a data selector is designed in the semi-supervised model. The proposed method is verified on the public bearing fault dataset, and outperforms the common machine learning methods and meta learning methods.
DOI:10.1109/DOCS55193.2022.9967716