| Summary: | Infrared (IR) imaging technology has been widely used in various applications mainly in temperature analysis as many objects emit electromagnetic radiation, especially in the electrical equipment that flows with strong current. The equipment could simply lead to overheating and may cause electrical equipment and devices to malfunction. Recently, infrared thermography (IRT) has appeared as an important tool in preventive maintenance for power distribution monitoring and detect defections in electrical equipment based on absolute and relative temperature data as a non-contact temperature distribution measuring technique. However, the conventional method is inefficient and inaccurate since requiring tedious work to manually record the temperature of thermal images. This study is proposed to automatize the power faults detection method using a deep learning method of convolution neural network (CNN) based on thermal imaging. The study involves used of data acquisition, pre-processing to resize and color pre-processing to make it compatible with CNN. Technically, data augmentation is used to double the data in the study and the proposed CNN model frameworks is designed for extracting features. The results of an accuracy score under normal and fault conditions have shown better method's average accuracy of 83.3% for normal conditions and 82.8% for fault conditions. This study suggested that the detection method using CNN based on thermal imaging has ability to identify between normal and defect condition of power distribution equipment. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023.
|
|---|