Loading...

Intelligent Fault Diagnosis of Rolling Bearings under Diverse Operating Conditions with Limited Vibration Data Using Meta-Learning Algorithms

Shamsodini, Ali | 2025

0 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 58404 (08)
  4. University: Sharif University of Technology
  5. Department: Mechanical Engineering
  6. Advisor(s): Behzad, Mehdi; Mohammadi, Somayeh
  7. Abstract:
  8. Rotating machinery is a critical component in many industries, and its failure can lead to process interruptions and significant economic losses. Rolling bearings, as key elements of these machines, account for approximately 45–55% of failures. Therefore, timely detection of bearing faults plays a crucial role in reducing maintenance costs and improving safety. However, achieving timely fault detection is challenging due to weak signals and the presence of noise, which necessitates the use of advanced algorithms with strong diagnostic capabilities. Despite the considerable success of deep learning models in pattern recognition, their application to fault diagnosis of rotating machinery faces two major challenges: distribution shifts under real-world operating conditions and the scarcity of labeled data. In this thesis, a novel algorithm is proposed that leverages multiple datasets during training, incorporates preprocessing techniques, and adopts a meta-learning framework with a contrastive loss function and a virtual-class approach. The proposed method demonstrates desirable performance in realistic scenarios with limited data and diverse operating conditions. Evaluation on 10 benchmark laboratory datasets and one industrial dataset confirms the high diagnostic accuracy of the algorithm for timely fault detection of rolling bearings, achieving over 90% accuracy on most laboratory datasets and more than 85% on the industrial dataset
  9. Keywords:
  10. Fault Diagnosis ; Rolling Bearing ; Metalearning ; Contrastive Loss Function ; Virtual Classes ; Intelligent Fault Detection

 Digital Object List

 Bookmark

No TOC