This study is focused on optimizing electromagnetic acoustic transducer (EMAT) sensors for enhanced ultrasonic guided wave signal generation in steel cables using CAD and modern manufacturing to enable contactless ultrasonic signal transmission and reception. A lab test rig with advanced measurement and data processing was set up to test the sensors’ ability to detect cable damage, like wire breaks and abrasion, while also examining the effect of potential disruptors such as rope soiling. Machine learning algorithms were applied to improve the damage detection accuracy, leading to significant advancements in magnetostrictive measurement methods and providing a new standard for future development in this area. The use of the Vision Transformer Masked Autoencoder Architecture (ViTMAE) and generative pre-training has shown that reliable damage detection is possible despite the considerable signal fluctuations caused by rope movement.
Read full abstract