Corrosion Defects Detection in the High Voltage Cable Aluminum Sheath Using Ultrasonic Guided Wave Based on Auto-encoder Network

The corrosion of the high voltage cable aluminum sheath is one of the main problems for faults in cable accessory grounding systems. This paper proposes an ultrasonic guided wave corrosion defects detection method for the high voltage cable aluminum sheath based on auto-encoder network. Firstly, this paper analyzes the identification principle of auto-encoder networks in the corrosion detection of the aluminum sheath. Then, an experimental platform is established to obtain the guided wave signals at different defect positions. Finally, the auto-encoder network model extracts the characteristics of the guided wave signals and achieves the identification of corrosion defect locations by training the model. The training results match the training label values, and the average training result error of the model is 5.4596mm. The results show that the auto-encoder network model can eliminate the interference of weld on corrosion defects detection, and the model can accurately identify the location of corrosion defects in the high voltage cable aluminum sheath.