Abstract
Infrared thermography technology is nowadays one of the most efficient non-destructive testing techniques for diagnosing faults of electrical systems and components. Overheated components in electrical systems and equipment indicate a poor connection, overloading, load imbalance or any other defect. Employing Thermographic inspection for finding such heat-related problems before subsequent failure of the system is practised in several industries. However, an automatic diagnostic system based on artificial neural network enhances the functionality by decreasing the operating time, human efforts and also increases the reliability of the system. The present article proposes employing artificial neural network (ANN) for inspection of electrical components and classifying their thermal conditions into three classes namely normal, intermediate and critical. Two different sets of inputs were provided to the neural network classifier, firstly statistical data of the temperature profile obtained from thermal images and secondly histogram based first order statistical features along with the glcm based features and both are compared to get the performance of network created. The multilayered perceptron network (MLP) was used as the classifier and the performance of the network was compared to two different training algorithms, viz. Levenberg-Marquardt (LM) and Scaled Conjugate Gradient (SCG). The performances were determined in terms of percentage of accuracy by plotting the confusion matrix. It was found that MLP network trained using the SCG algorithm gives the highest percentage of accuracy of classification i.e., 91.5% for the Statistical data features of the temperature profile as compared to the other set of features.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have