We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
A hybrid method based on deep learning and ensemble learning for induction motor fault detection using sound signals.
- Authors
Shirdel, Shahryar; Teimoortashloo, Mazdak; Mohammadiun, Mohammad; Gharahbagh, Abdorreza Alavi
- Abstract
Fault detection in induction motors has an important role in saving costs in various industries. Due to the wide range of faults in induction motors, using different electrical and mechanical properties such as motor current, vibration, heat, and flux to detect faults in different conditions is necessary. In this research, A fault detection scheme using sound signals has been proposed. Sound signals are not sensitive to faults, or their signal-to-noise ratio is lower than other properties but acquiring sound signals does not need direct contact with the motor or to stop the motor in many cases, and sound sensors (microphones) have a lower cost than other types of sensors. The proposed method uses a hybrid combination of Mel-frequency cepstral coefficients (MFCCs), Wavelet scattering, ensemble learning, convolutional neural network (CNN) classification methods. The results of ensemble learning and CNN are combined in a novel FUSION step, and the final result is determined. The proposed method was implemented on a standard database, and its efficiency was proved compared to the state-of-the-art methods.
- Subjects
INDUCTION motors; CONVOLUTIONAL neural networks; DEEP learning; MOTOR learning; DATABASES; SIGNAL-to-noise ratio
- Publication
Multimedia Tools & Applications, 2024, Vol 83, Issue 18, p54311
- ISSN
1380-7501
- Publication type
Article
- DOI
10.1007/s11042-023-15996-5