Detection of three-phase induction motor faults using deep network-based transfer learning techniques

Ibrahim, Alasmer, Anayi, Fatih ORCID: https://orcid.org/0000-0001-8408-7673 and Packianather, Michael ORCID: https://orcid.org/0000-0002-9436-8206 2022. Detection of three-phase induction motor faults using deep network-based transfer learning techniques. Presented at: IM IEEE 2nd International Maghreb Meeting of the Conference on Sciences and Techniques of Automatic Control and Computer Engineering (MI-STA), 23-25 May 2022. 2022 IEEE 2nd International Maghreb Meeting of the Conference on Sciences and Techniques of Automatic Control and Computer Engineering (MI-STA). IEEE, pp. 133-138. 10.1109/MI-STA54861.2022.9837619

Full text not available from this repository.

Abstract

As the induction motors in the industry operate in difficult and confined environments, condition monitoring the condition of the motor is crucial to avoid production loss. A new hybrid application that uses a pre-trained model-based ImageNet weight is proposed for fault diagnosis using the induction motor thermal images. A deep transfer learning network based on the Efficient Net-B0 model is developed as a feature extractor. As the use of pre-trained model architecture does not need a manual attribute extraction, the classification performance and decrease in computational complexity can be improved without a noticeable assignment. The methodology of this work includes two stages: the first one consists of the use of deep learning algorithms for feature extraction. Secondly, the classification assessment of the induction motor faults is considered to validate the system's performance. The model performance has been examined by applying thermal images of the induction motor. The achieved results have precisely proved the superiority and robustness of the combination of these techniques based on pre-trained weight and learning algorithms to extract the most appropriate features to build an effective classification system for induction motor fault diagnosis topic. The proposed model has achieved the highest accuracy of 91.9% by applying an Efficient Net-B0 network.

Item Type:	Conference or Workshop Item (Paper)
Date Type:	Published Online
Status:	Published
Schools:	Engineering
Publisher:	IEEE
ISBN:	978-1-6654-7918-9
Last Modified:	30 Nov 2022 08:55
URI:	https://orca.cardiff.ac.uk/id/eprint/153171

Actions (repository staff only)

Edit Item