Abstract:Advances in Mechanical Engineering, Volume 16, Issue 10, October 2024. Bearing fault diagnosis presents challenges, such as insufficient fault samples and significant fault data distribution variation in different bearing operating conditions. These problems cause traditional deep learning models to show poor generality and accuracy during bearing fault diagnosis. To address these challenges, this paper proposed a few-shot bearing fault diagnosis method based on an Ensemble Empirical Mode Decomposition (EEMD) parallel neural network and a relation network (RN). First, the original bearing fault vibration signal was decomposed by EEMD, while the decomposed signal components were processed via Short Time Fourier Transfor (STFT) to obtain a two-dimensional time-frequency feature map. Then, a parallel neural network was used for initial fault feature extraction, after which the extracted features were fused to construct a more accurate multi-dimensional fault feature map. A more precise fault feature vector was generated via the feature embedding module of the RN, and the fault features of the support and query sets were stitched to create a fault feature vector set. Finally, the relation module of the RN was used for the nonlinear distance determination of the fault feature vector set and to generate the relation score for the few-shot variable condition bearing fault diagnosis. In this paper, EEMD module is introduced into RN to construct multi-dimensional fault characteristics of the original fault signal. Original signal decomposition, STFT transformation and splicing effectively improve the randomness and blindness of convolution operations, improve the accuracy of fault feature extraction in RN, and thus improve the overall diagnostic performance of the model. The experimental results showed that the proposed model obtained higher diagnostic accuracy than the matching network (MN) and RN meta-learning fault diagnosis (MLFD) methods. The accuracy of fault diagnosis of the model in 5Way-Nshot is >80%, and the accuracy of fault diagnosis in 5Way-10shot is the highest at 95.2%.

Residual Prototype Network-based Method for Few-Shot Fault Diagnosis of Bearing

A Meta-learning Method for Few-shot Bearing Fault Diagnosis under Variable Working Conditions

Few-Shot Rolling Bearing Fault Diagnosis with Metric-Based Meta Learning

A Novel Deep Model with Meta-learning for Rolling Bearing Few-shot Fault Diagnosis

Few-Shot Learning for Fault Diagnosis: Semi-Supervised Prototypical Network with Pseudo-Labels

Few-Shot Bearing Fault Diagnosis Based on Model-Agnostic Meta-Learning

Rolling Bearing Fault Diagnosis with Variable Load and Few Samples Based on Multifeature Fusion Meta-Learning

Few-shot bearing fault diagnosis method based on an EEMD parallel neural network and a relation network

A Curriculum-Based Meta-Learning Approach for Few-Shot Fault Diagnosis

Limited Data Rolling Bearing Fault Diagnosis with Few-Shot Learning

Few-shot bearing fault diagnosis based on meta-learning with discriminant space optimization

Meta-learning for few-shot bearing fault diagnosis under complex working conditions

A Meta-Learning Method for Electric Machine Bearing Fault Diagnosis under Varying Working Conditions with Limited Data

Meta-learning with Elastic Prototypical Network for Fault Transfer Diagnosis of Bearings under Unstable Speeds

Prior knowledge-based residuals shrinkage prototype networks for cross-domain fault diagnosis

Generic Meta-Transfer Learning Model with Special Neuronal Processing Parameters for Few-Shot Fault Bearing Diagnosis

A Novel Bearing Fault Diagnosis Method Based on Few-Shot Transfer Learning Across Different Datasets.

Brain-Inspired Meta-Learning for Few-Shot Bearing Fault Diagnosis

A Prototype Correction Network with Coordinate Attention for Few-Shot Fault Diagnosis Under Varying Working Conditions

Bearing Fault Diagnosis Based on Prototypical Network

Bearing Fault Diagnosis Based on Attention Mechanism and Deep Residual Network