A Prestressed Concrete Cylinder Pipe Broken Wire Detection Algorithm Based on Improved YOLOv5.

Area of Science:

• Civil Engineering
• Artificial Intelligence
• Structural Health Monitoring

Background:

• Prestressed concrete cylinder pipe (PCCP) failures pose significant risks to infrastructure and economic feasibility.
• Traditional inspection methods for PCCP are labor-intensive, time-consuming, and disruptive to operations.
• Developing efficient and accurate methods for early detection of PCCP wire breaks is crucial for preventative maintenance.

Purpose of the Study:

• To develop a lightweight deep learning algorithm for identifying broken wires in PCCP.
• To improve the efficiency and accuracy of PCCP wire-break monitoring systems.
• To enable real-time monitoring without requiring high-performance hardware.

Main Methods:

• A PCCP wire-breaking test platform was established, utilizing Distributed Fiber Acoustic Sensing (DAS) for monitoring.
• A dataset of simulated broken wire signals was created using Continuous Wavelet Transform (CWT).
• A lightweight deep learning model, YOLOv5-Break, was proposed, integrating MobileNetV3, Dynamic Conv, coordinate attention, and Focal_EIoU loss.

Main Results:

• The YOLOv5-Break algorithm achieved a mean Average Precision (mAP) of 97.72% on the custom dataset.
• The model weight was reduced to 7.74 MB, significantly smaller than existing YOLO versions.
• YOLOv5-Break demonstrated superior computational efficiency compared to YOLOv8s and YOLOv9s.

Conclusions:

• The proposed YOLOv5-Break algorithm effectively identifies PCCP broken wires with high accuracy and minimal computational resources.
• Its lightweight nature and efficiency make it suitable for deployment in practical PCCP monitoring systems.
• This advancement offers a more streamlined and cost-effective solution for ensuring the integrity of PCCP infrastructure.