Image Registration-Based Bolt Loosening Detection of Steel Joints

Area of Science:

• Civil Engineering
• Structural Health Monitoring
• Computer Vision

Background:

• Self-loosening of bolts due to repetitive loads and vibrations compromises structural integrity in civil infrastructure.
• Existing bolt loosening detection methods often rely on extensive and costly sensor deployment.
• Computer vision offers a promising, low-cost, contactless alternative for structural health monitoring.

Purpose of the Study:

• To propose and validate a non-contact bolt loosening detection method using computer vision.
• To demonstrate the effectiveness of analyzing image registration errors for identifying bolt rotation.
• To provide a cost-efficient approach for monitoring numerous bolted steel joints.

Main Methods:

• Utilized a consumer-grade digital camera to capture images of bolted steel joints over time.
• Implemented image registration techniques to align sequential images of the monitored structures.
• Analyzed differential features in registration errors to detect bolt rotation indicative of loosening.

Main Results:

• The proposed vision-based method successfully detected bolt loosening in laboratory experiments.
• The approach demonstrated robustness across various structural configurations, including gusset plates, column flanges, and girder webs.
• Differential features in registration errors effectively served as indicators for bolt loosening.

Conclusions:

• Vision-based monitoring presents a viable and cost-effective solution for detecting bolt loosening in civil structures.
• The developed method offers a practical alternative to traditional sensor-based monitoring systems.
• Accurate detection of bolt loosening enables timely structural maintenance and enhances safety.