Defect Detection of Gas Insulation Switch by Infrared Thermography Technology With an Improved Yolo Algorithm

Gas-insulated switch-gear (GIS) defects, such as internal bubbles, surface cracks, and high pressure breakdowns, critically threaten power transmission reliability. However, conventional nondestructive detection methods like ultrasound and X-ray face the challenges in anti-interference capability, cost, and structural complexity. To address those issues, this study has proposed an infrared thermography-based defect detection method by combing with an enhanced YOLOv7 algorithm for accurate and efficient defect localization identification and depth estimation. The proposed method modifies the YOLOv7 framework by replacing standard CBS modules with Fused-MBConv layers in the backbone to improve feature extraction efficiency while integrates a coordinate attention (CA) mechanism in the neck layer to enhance spatial defect recognition. The experiments have been carried out with the infrared thermal images of insulation rod defects, including flashover, block bubbles, scratches, with the impurities at depths within 0.4–1 mm. The experimental results have demonstrated significant performance gains. The proposed method achieves a mean Average Precision (mAP) of 10.6% and 13% higher than baseline YOLOv7 and YOLOv5, respectively. Moreover, improved computational speed and detection accuracy has been seen via comparison with the fellow algorithms. These results validate the effectiveness of the method, which provide a reference solution for overcoming existing technological barriers for GIS defect diagnosis.