A Cross-Machine Intelligent Fault Diagnosis Method with Small and Imbalanced Data Based on the ResFCN Deep Transfer...

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Machine Learning

Background:

• Intelligent fault diagnosis (IFD) for mechanical equipment often relies on transfer learning, typically assuming source and target domains share identical equipment.
• Real-world applications face limitations due to the difficulty of obtaining data from identical equipment for source domain training.
• Existing methods inadequately address the temporal characteristics inherent in time-series monitoring data.

Purpose of the Study:

• To propose a novel cross-machine IFD method that overcomes the limitations of traditional transfer learning approaches.
• To effectively leverage time-series features from monitoring data for improved fault diagnosis.
• To enable accurate fault diagnosis for target equipment using data from dissimilar source equipment.

Main Methods:

• A residual full convolutional neural network (ResFCN) transfer learning model is employed.
• Sliding window (SW)-based data segmentation is utilized to process time-series data.
• Network pretraining on a source domain, followed by model fine-tuning on a target domain, is implemented.

Main Results:

• The proposed cross-machine IFD method demonstrates superior performance compared to existing approaches.
• The method achieves high accuracy and robustness in fault diagnosis, even with small and imbalanced datasets.
• Experimental validation using two public datasets confirms the effectiveness of the approach.

Conclusions:

• The developed ResFCN-based transfer learning model successfully addresses the cross-machine fault diagnosis challenge.
• Incorporating time-series features and domain adaptation techniques enhances diagnostic accuracy and applicability.
• This method offers a practical solution for real-world industrial fault diagnosis scenarios.