New machine learning model-based fault diagnosis of induction motors using thermal images

Ibrahim, Alasmer, Anayi, Fatih ORCID: https://orcid.org/0000-0001-8408-7673 and Packianather, Michael ORCID: https://orcid.org/0000-0002-9436-8206 2022. New machine learning model-based fault diagnosis of induction motors using thermal images. Presented at: 2nd International Conference on Advance Computing and Innovative Technologies in Engineering (ICACITE), Greater Noida, India, 28-29 April 2022. Proceedings of 2nd International Conference on Advance Computing and Innovative Technologies in Engineering. IEEE, pp. 58-62. 10.1109/ICACITE53722.2022.9823832

Full text not available from this repository.

Abstract

Identification of the induction motor health condition is a significant task in the industry, which can ensure a safe operation. A viable machine learning model-based fault diagnosis is proposed in this paper using thermal images data. A hybrid model is developed which ensures the robust performance of the proposed application. It is investigated under three faulty motor conditions that include bearing faults, broken rotor bar faults, as well as stator faults with three different motor running conditions namely 1480 rpm, 1450 rpm, and 1380 rpm. Several experiments have been conducted in the lab including different faulty cases with reference cases (healthy motor). Grey Level Co-occurrence Matrices (GLCM) are adopted for feature extraction purposes. Then, an optimization algorithm-based feature selection is utilized to select the most appropriate features for building the classification system. Some intelligence algorithms are applied to classify motor faults. GLCM is implemented which uses second-order statistics of features to infer the degree of correlation between pairs of pixels. It defines the distance and angle between the pixels. The extracted parameters from GLCM such as contrast, dissimilarity, homogeneity, ASM, energy, and correlation. In order to select the efficient features and reduce the dimensionality of data, invasive weed optimization (IWO) is adopted. Two classification algorithms namely Extreme Gradient Boosting (XGBoost), and Random Forest (RF) are employed to characterize the motor conditions. The proposed classification algorithms are trained using a cross-validation strategy to secure a high classification performance. Some evaluation metrics are calculated to assess the model performance such as the accuracy, precision, sensitivity, and F1_ score. The achieved results showed that the hybrid model was capable of appropriately predicting and identifying motor faults. Furthermore, the presented model can be used for additional thermal image processing applications.

Item Type:	Conference or Workshop Item (Paper)
Date Type:	Published Online
Status:	Published
Schools:	Engineering
Publisher:	IEEE
ISBN:	9781665437905
Last Modified:	28 Dec 2024 15:47
URI:	https://orca.cardiff.ac.uk/id/eprint/152042

Citation Data

Cited 5 times in Scopus. View in Scopus. Powered By Scopus® Data

Actions (repository staff only)

Edit Item