A Novel Detection Method of Rail Insulation Defects based on FFC-Swin-Transformer
ID：134
Submission ID：139 View Protection：ATTENDEE
Updated Time：2025-11-03 11:45:38
Hits：94
Oral Presentation
Start Time：2025-11-09 11:20 （Asia/Shanghai）
Duration：15min
Session：[S5] 5.AI-driven technology » [S5] 5.AI-driven technology
No files
Abstract
As urban rail transit grows, stray current seriously threatens the safety of power systems and oil/gas pipelines. Rail-to-earth insulation the key to control stray current. Aiming at the problem that it is difficult to detect and locate the rail insulation defects of DC traction power supply system, a novel rail insulation defect detection method based on the FFC-Swin-Transformer is proposed. By establishing a four-layer "catenary-rail-SCCN-earth" equivalent circuit model and integrating multi-train operation conditions, a rail potential dataset is efficiently generated via parallel computing. Time-domain signals of rail potential are converted to frequency-domain features using Fourier transform, and the fused time-frequency information is fed into an improved FFC-Swin-Transformer network to achieve accurate detection of rail insulation states. Experimental results show that after training on 22496 sample groups, the model achieves a test accuracy of 82.95%, effectively identifying section insulation defects and exhibiting promising engineering application potential.
Keywords
DC metro system,stary current,defect detection,rail potential,rail insulation
Speaker
Comment submit