YOLOv8-Based Real-Time Detection of Screw in High-Voltage Equipment for Smart Grid

In high-voltage equipment and large transmission grid frameworks, one of the most dangerous defects is the loosening of screw fasteners. Due to the diversity of screw loosening and the complexity of failures, this paper proposes an object detection approach based on the YOLOv8 algorithm for identifying screw looseness. The method leverages advanced deep learning techniques to effectively detect and classify screw defects in various operating conditions. By training the YOLOv8 model on a custom dataset of screw images, the proposed system demonstrates high accuracy and efficiency in detecting loosened screws, even under challenging conditions. The results show that the YOLOv8-based method achieves an accuracy of 94%, presenting the potential of this approach in enhancing the reliability of maintenance operations and preventing accidents caused by equipment failures.