Intelligent Thermal Condition Monitoring Of Electrical Equipment Using Infrared Thermography
Infrared thermographic inspection system is widely being utilized for defect detection in electrical equipment. Conventional inspection based on the temperature data interpretation and evaluation the condition of the equipment is subjective and depends on the human experts. Implementation of an auto...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2013
|
| Subjects: | |
| Online Access: | http://eprints.usm.my/46367/1/A.%20S.%20Nazmul%20Huda24.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Infrared thermographic inspection system is widely being utilized for defect detection in electrical equipment. Conventional inspection based on the temperature data interpretation and evaluation the condition of the equipment is subjective and depends on the human experts. Implementation of an automatic diagnostic system based on artificial neural network reduces operating time, human efforts and also increases the reliability of system. In this thesis, an automatic features extraction system from thermal image of defects and the intelligent classification of thermal condition based on neural network are proposed. The proposed system extracts first order histogram based features and grey level co-occurrence matrix features from the segmented regions and evaluates the effectiveness of these features for defect characterization. Three feature selection techniques namely principal component analysis, the discriminant analysis and individual feature performance analysis are employed to find out the useful and important statistical features. In this study, multilayered perceptron network is proposed for classifying thermal condition into two classes namely normal and defective. The multilayered perceptron neural networks are trained using various training algorithms. Additionally, the present research introduces a computer aided defect diagnosis system where the defected region is found by manual thresholding and intensity features are extracted from each segmented region. The results prove that the statistical features are capable to classify thermal condition and the neural networks achieve the accuracy around 73~78% |
|---|