Damage Analysis on ACSR Conductors Using X-Ray with A Deep Learning Approach

Abstract

Aluminum Conductor Steel Reinforced (ACSR) conductors are widely used in electrical transmission networks due to their high tensile strength and long-term durability. Despite these advantages, ACSR conductors remain vulnerable to internal degradation such as corrosion and wire breakage, which cannot be detected through conventional visual inspection. To address this limitation, this study employs nondestructive X-ray imaging to assess internal conductor conditions and classify them into three categories: normal, internal corrosion, and broken wires. A deep learning approach based on a Convolutional Neural Network (CNN) was developed to enhance the accuracy and consistency of defect identification. The research workflow involved systematic data collection, image pre-processing, and model training focused on learning structural strand patterns and density variations associated with internal damage. Model performance evaluation showed strong learning capability, with training loss decreasing from 1.35 to 0.35 and validation loss from 0.9 to 0.25. The model achieved nearly 90% validation accuracy, surpassing training accuracy and indicating excellent generalization. Overall, the integration of X-ray imaging and deep learning demonstrates high potential for rapid, reliable, and automated detection of internal ACSR conductor defects, supporting improved maintenance planning and decision-making for transmission line asset management.