Related Experiment Video
Updated: Jul 16, 2026
08:27
Image Recognition and Parameter Analysis of Concrete Vibration State Based on Support Vector Machine
Published on: January 5, 2024
Measurement-Efficient Few-Shot Vibration Fault Diagnosis via Physics-Informed Self-Supervised Learning and Adaptive
Zongzhe Ni1, Xiancheng Ji1, Jianjun Yi1
1School of Mechanical and Power Engineering, East China University of Science and Technology, Shanghai 200237, China.
Sensors (Basel, Switzerland)
|July 15, 2026
Summary
This study introduces a new few-shot fault diagnosis method for rotating machinery. It efficiently uses vibration data, improving accuracy and reducing costs by adaptively stopping measurements.
Area of Science:
- Mechanical Engineering
- Artificial Intelligence
- Signal Processing
Background:
- Vibration-based fault diagnosis is crucial for rotating machinery.
- Limited fault labels and variable measurement lengths hinder practical applications.
- Longer data improves reliability but increases costs; shorter data risks unreliability.
Purpose of the Study:
- To develop a measurement-efficient framework for few-shot fault diagnosis under data constraints.
- To enable models to identify fault classes and determine optimal measurement duration.
- To improve the accuracy-cost trade-off in vibration-based diagnostics.
Main Methods:
- Proposed a framework combining unlabeled data knowledge, physically constrained augmentation, and adaptive early stopping.
- Utilized a shared one-dimensional feature extractor.
- Evaluated on UORED-VAFCLS and Paderborn University bearing datasets.
Main Results:
- Achieved robust diagnostic performance with fewer acquired vibration windows compared to fixed-length inference.
- In the PU-Hard 8-shot setting, attained 80.26% accuracy using an average of 1.2432 windows.
- Reduced evaluation cost J from 0.3929 to 0.2596.
Conclusions:
- Adaptive measurement significantly improves the accuracy-cost trade-off in few-shot vibration diagnosis.
- The proposed framework offers a practical solution for real-world machinery health monitoring.
- Demonstrated the effectiveness of integrating prior knowledge and adaptive strategies.