.png){kind=icon}
.png){kind=icon}
Return
Neural computing-based multi-view vision system for precise bolt looseness detection
.png){kind=icon}
PRE
AI
DOI：10.1080/13287982.2025.2588821.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Maintaining the integrity of bolted joints is vital across engineering domains. Bolt loosening under dynamic loads can compromise structural safety. We present a robust neural approach using YOLOv11 for detecting loosening through pose-based estimation of nutbolt relative rotation. A dataset of 419 bolt images and 385 nut images supports training and evaluation. The method achieves an average relative angular error of 0.94 with prediction confidence up to 98%, and remains reliable under challenging illumination, with a maximum error of 3. This innovative approach offers a highly reliable and accurate tool for monitoring bolt loosening, even detecting minute rotational movements, thereby enhancing the safety and reliability of bolted connections in diverse engineering structures.
Keywords:
Bolt looseness detection
YOLOv11
pose estimation
structural reliability
Journal
A
IF:
1.3
Papers:
25
Citations:
0