Loading...

Seif, Mohammad Reza | 2024

0 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: English
  3. Document No: 57721 (58)
  4. University: Sharif University of Technology, International Campus, Kish Island
  5. Department: Science and Engineering
  6. Advisor(s): Behzad, Mehdi; Mohammadi, Somayeh
  7. Abstract:
  8. Bearings play a critical role in the functionality of rotating equipment across various industries, accounting for approximately 50% of equipment failures due to bearing malfunctions. Accurate life prediction of bearings is essential not only for preventing unexpected breakdowns and subsequent damage but also for minimizing unnecessary replacements of functional bearings, which can lead to increased operational costs. With the rise of artificial intelligence, numerous predictive models have been developed; however, many of these require extensive datasets, which are often unavailable in industrial settings. Data collection is typically irregular, based on the sensitivity of the equipment, and occurs on a weekly or monthly basis. This results in a significant challenge of insufficient data for effective model training, rendering deep learning and transfer learning approaches suboptimal. In this thesis, the application of support vector regression (SVR) and relevance vector regression (RVR) combined with bootstrapping method are explored and both machine learning methods perform well with limited datasets. The parameters including kurtosis, crest factor, peak, peak-to-peak, RMS, and bearing condition according to the ISO 13373-3 standard are analyzed to identify the most effective indicators for early failure detection. To enhance predictive accuracy, an adaptive algorithm that enables the model to adjust to evolving data patterns is incorporated. After thorough evaluations, the most effective models are used to validate them with real industrial data. Results demonstrated that the RVR model significantly outperformed the SVR model, achieving an impressive prediction accuracy of 95.7% when utilizing average values from the dataset. This research underscores the potential of tailored machine learning approaches to address the challenges of data scarcity in predictive maintenance of bearing systems
  9. Keywords:
  10. Condition Monitoring ; Vibrational Analysis ; Support Vector Machine (SVM) ; Relevance Vector Machine ; Rolling Element Bearings (REBs) ; Bootstrapping ; Degradation Trend Prediction

 Digital Object List

 Bookmark

No TOC