We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Real-time robust bearing fault detection using scattergram-driven hybrid CNN-SVM.
- Authors
Mitra, Sukanya; Koley, Chiranjib
- Abstract
Industries rely on the early and efficient detection of bearing faults to enhance the service life of three-phase induction motors. Various bearing health condition monitoring techniques have been proposed for years, majorly involving sound and/or vibration sensors. This paper compares sound- and vibration-based approaches for automated and real-time bearing health diagnosis, evolving a fusion of wavelet scattering transform and a hybrid two-dimensional convolutional neural network architecture with a multi-class support vector machine classifier. In this work, data are collected from our own laboratory-based sample process control application operated through a supervisory control and data acquisition system under harsh industrial environmental conditions, including dynamicity in motor load, speed, and externally induced noise levels from a nearby induction motor. Multiple parameters, such as image resolution, color channel, and intelligent architecture, are optimized for cost-effective real-time fault detection. This study also suggests that the proposed hybrid model, driven by normalized time–frequency feature images, can work with the reduced signal sampling frequency and analog-to-digital converter bit resolution while accomplishing the performance. The robustness of the proposed model is verified on publicly available databases. Rigorously, it is observed that the individual sound and vibration sensor-based measurement systems perform equally well for bearing fault diagnosis with the help of quintessential specification tuning in the data acquisition system.
- Subjects
INDUCTION motors; SUPERVISORY control &; data acquisition systems; BEARINGS (Machinery); CONVOLUTIONAL neural networks; DATA acquisition systems; FAULT diagnosis; SUPPORT vector machines
- Publication
Electrical Engineering, 2024, Vol 106, Issue 3, p3615
- ISSN
0948-7921
- Publication type
Article
- DOI
10.1007/s00202-023-02162-1