Fault diagnosis of electrical machine using thermal imaging
Loading...
Date
2019-10
Authors
Lim Yong Fong
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Electrical machines are susceptible to failure at any point of time, due to numerous factors such as wear and tear, abnormal operating conditions, material fatigue and et cetera. Therefore, it is important to detect fault at early stages so that the source of problem can be identified and fixed before the condition worsens. Different methods for fault diagnosis were developed in order to assess the condition of an electrical machine, including analysing electrical parameter, vibration, and mechanical properties. In this project, emphasis is placed on diagnosing electrical machines using thermography. The advantage of utilizing thermal imaging in fault detection is that the method is noninvasive, as it does not interact directly with the machine in inspection, rather it captures the infrared radiation emitted. Abnormal conditions in general lead to unwanted hotspot in the windings, wirings, or the casing of the machine. Thus, by identifying any hotspot appearing on any parts of the machine in the thermal images, the responsible technician can be alerted to a potential failure and pre-emptive measures can be taken. To identify the hotspot in an automated fashion, MATLAB is used. The image processing toolbox provided by MATLAB has different algorithms for this purpose. Two methods were compared, which were: Otsu multithresh and K-means clustering. Dice and Jaccard indices were calculated to measure the performance of these two algorithms. The running time for each algorithm was recorded and compared too. Then, Mann-Whitney test was carried out to compare the two sets of data. In terms of accuracy, k-means clustering algorithm is a better choice in detecting hotspot in the thermal images.