Recursive construction of output-context fuzzy systems for the condition monitoring of electrical hotspots based on infrared thermography
Infrared thermography technology is currently being used in various applications, including fault diagnosis in electrical equipment. Thermal abnormalities are diagnosed by identifying and classifying the hotspot conditions of electrical components. In this article, a new recursively constructed outp...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Engineering Applications of Artificial Intelligence
2015
|
| Subjects: | |
| Online Access: | http://eprints.um.edu.my/13767/1/Recursive_construction_of_output-context_fuzzy_systems_for_the.pdf http://eprints.um.edu.my/13767/ http://www.sciencedirect.com/science/article/pii/S0952197614002826 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Infrared thermography technology is currently being used in various applications, including fault diagnosis in electrical equipment. Thermal abnormalities are diagnosed by identifying and classifying the hotspot conditions of electrical components. In this article, a new recursively constructed output-context fuzzy system is proposed to characterize the condition of electrical hotspots. An infrared camera is initially used to capture the thermal images of components with hotspots, and intensity features are extracted from each hotspot. The Recursively Constructed Fuzzy System (RCFS) is then applied to automatically realize and formulate the conditions of the thermal abnormalities. On the basis of the priority level, the hotspot conditions are categorized as normal, warning, and critical. From these three categories, the conditions can be further simplified into two categories, namely, defect (warning and critical) and normal. The proposed RCFS realizes the prominent distinctions in the output domain by using a self-organizing method. The termination of the recursive algorithm finds an effective rule base to achieve an accurate representation of the datasets. The proposed system obtains less fuzzy rules with reasonable accuracy. Our survey of 253 detected regions shows that the proposed RCFS produces 92.3 and 80 testing accuracies for classifying conditions into two and three classes, respectively. The thermographic diagnostic evaluation shows that the proposed intelligent system automatically identifies the rationally acceptable limits of hotspot conditions. Therefore, the proposed system is suitable for establishing an intelligent defect analysis system. (C) 2014 Elsevier Ltd. All rights reserved. |
|---|