Open Access

ARTICLE

Fault Diagnosis of Industrial Motors with Extremely Similar Thermal Images Based on Deep Learning-Related Classification Approaches

Hong Zhang^1,*, Qi Wang¹, Lixing Chen¹, Jiaming Zhou¹, Haijian Shao²

1 School of Electrical Information Engineering, Jiangsu University of Technology, Changzhou, 213001, China
2 Department of Electrical and Computer Engineering, University of Nevada, Las Vegas, NV 89154, USA

* Corresponding Author: Hong Zhang. Email:

Energy Engineering 2023, 120(8), 1867-1883. https://doi.org/10.32604/ee.2023.028453

Received 19 December 2022; Accepted 13 February 2023; Issue published 05 June 2023

Abstract

Induction motors (IMs) typically fail due to the rate of stator short-circuits. Because of the similarity of the thermal images produced by various instances of short-circuit and the minor interclass distinctions between categories, non-destructive fault detection is universally perceived as a difficult issue. This paper adopts the deep learning model combined with feature fusion methods based on the image’s low-level features with higher resolution and more position and details and high-level features with more semantic information to develop a high-accuracy classification-detection approach for the fault diagnosis of IMs. Based on the publicly available thermal images (IRT) dataset related to condition monitoring of electrical equipment-IMs, the proposed approach outperforms the highest training accuracy, validation accuracy, and testing accuracy, i.e., 99%, 100%, and 94%, respectively, compared with 8 benchmark approaches based on deep learning models and 3 existing approaches in the literature for 11-class IMs faults. Even the training loss, validation loss, and testing loss of the eleven deployed deep learning models meet industry standards.

---

Graphic Abstract

---

Keywords

---