TY - GEN
T1 - An automatic defect detection system based on deep learning for fasteners in the power system
AU - Yang, Tao
AU - Ma, Zhongjing
AU - Wang, Tianyu
AU - Fu, Jiaxin
AU - Zou, Suli
N1 - Publisher Copyright:
© 2022 Technical Committee on Control Theory, Chinese Association of Automation.
PY - 2022
Y1 - 2022
N2 - Fasteners on transmission lines are widely used in various power connection components, and the loosening and missing of fastener cotter pins can trigger problems such as falling off of components, which has a huge potential safety hazard. It is necessary to propose an automatic fastener defect detection method to ensure the stable operation of the transmission line. In this paper, a three-stage cascade system for automatic detection of fastener defects is constructed in a coarse-to-fine mode, including fastener location network (FLN), feature refinement network (FRN), and defect diagnosis network (DDN). First, the proposed FLN improves the YOLOV4 model based on a blurred module to locate fasteners. Then, the proposed FRN is applied to extract the semantic information of the fasteners and refine the features. Finally, the proposed DDN is used to classify the cotter pin defects. To verify the adaptability and accuracy of the method, a considerable amount of experiments and analyses have been performed, and the results show that the proposed detection method reaches an accuracy of 98.4% and testing speed of 0.64s, achieving state-of-the-art (SOTA) performance.
AB - Fasteners on transmission lines are widely used in various power connection components, and the loosening and missing of fastener cotter pins can trigger problems such as falling off of components, which has a huge potential safety hazard. It is necessary to propose an automatic fastener defect detection method to ensure the stable operation of the transmission line. In this paper, a three-stage cascade system for automatic detection of fastener defects is constructed in a coarse-to-fine mode, including fastener location network (FLN), feature refinement network (FRN), and defect diagnosis network (DDN). First, the proposed FLN improves the YOLOV4 model based on a blurred module to locate fasteners. Then, the proposed FRN is applied to extract the semantic information of the fasteners and refine the features. Finally, the proposed DDN is used to classify the cotter pin defects. To verify the adaptability and accuracy of the method, a considerable amount of experiments and analyses have been performed, and the results show that the proposed detection method reaches an accuracy of 98.4% and testing speed of 0.64s, achieving state-of-the-art (SOTA) performance.
KW - Cascade System
KW - Deep Learning
KW - Fastener Detection
KW - Power Transmission Line
UR - http://www.scopus.com/inward/record.url?scp=85140434012&partnerID=8YFLogxK
U2 - 10.23919/CCC55666.2022.9902337
DO - 10.23919/CCC55666.2022.9902337
M3 - Conference contribution
AN - SCOPUS:85140434012
T3 - Chinese Control Conference, CCC
SP - 6599
EP - 6604
BT - Proceedings of the 41st Chinese Control Conference, CCC 2022
A2 - Li, Zhijun
A2 - Sun, Jian
PB - IEEE Computer Society
T2 - 41st Chinese Control Conference, CCC 2022
Y2 - 25 July 2022 through 27 July 2022
ER -