Inner Wall Abnormality Detection of Rotating Shaft Based on Deep Learning.

To address the challenges of inconsistent detection standards, high labor intensity, and low efficiency in existing manual detection methods, a method based on an improved YOLOv5 algorithm is proposed for detecting abnormalities in the inner wall of rotating shafts. An industrial endoscope is used to illuminate and capture images of the inner wall of the rotating shaft, and the YOLOv5 6.x version model is trained using different optimizers to detect abnormalities in the inner wall. The model is trained on a dataset of images captured from an industrial site, and experimental results show that the YOLOv5 model optimized with AdamW achieves an average mAP improvement of 4.4% compared to the model optimized with SGD, enabling effective detection of inner wall abnormalities in rotating shafts.