A Method to Assess Shield Corrosion Sizes in Cable Defects Detection Based on Time-frequency Domain Reflectometry

Kun Zhao,Tiantian Wang,Junbai Chen,Zixing Xu,Lei Jin,Hongjie Li
DOI: https://doi.org/10.1109/aceee62329.2024.10651927
2024-01-01
Abstract:Shield corrosion defects threaten the integrity of power cables. Therefore, the detection and accurate sizing as-sessment of these defects is crucial. However, existing methods generally fail to precisely evaluate the size of corrosion defects. To address this issue, we propose an approach that integrates Time-Frequency Domain Reflectometry (TFDR) with neural network algorithms to detect and size shield corrosion defects. Models of both intact and corroded cable shields are established, with reflected TFDR signals computed subsequently. A four-layer neural network is employed and trained by the data from these signals and achieves sizes assessment of shield corrosion defects. A simulation verifies the performance of the network whose correlation coefficient is over 0.999. An experiment is carried out and the results show that the maximum errors for estimating the angles and lengths of corrosion defects are limited to 5.88% and 13.43%, respectively. This hybrid TFDR-neural network methodology demonstrates substantial potential in enhancing power cable maintenance strategies, marking a significant advancement in the operational management of power systems.
What problem does this paper attempt to address?