Related Experiment Video
Updated: May 14, 2025
11:34
Subsurface Defect Localization by Structured Heating Using Laser Projected Photothermal Thermography
Published on: May 15, 2017
11.0K
Shearography-Based Near-Surface Defect Detection in Composite Materials: A Spatiotemporal Object Detection Neural
Guanlin Li1,2, Yao Hu1,2, Hao Wang1,2
1School of Optics and Photonics, Beijing Institute of Technology, Beijing 100081, China.
Nanomaterials (Basel, Switzerland)
|April 11, 2025
Summary
This study introduces YOWO_SS3D, a neural network for shearography defect detection using simulated data. It achieves high accuracy (96.99%) without costly experimental data, significantly outperforming existing methods.
Area of Science:
- Materials Science
- Non-Destructive Testing
- Artificial Intelligence
Background:
- Shearography is a key non-destructive testing (NDT) method for composite materials.
- Neural networks offer efficient defect detection but require extensive datasets, which are costly to acquire experimentally.
- Existing simulation methods for shearography data lack accuracy, hindering neural network performance.
Purpose of the Study:
- To develop a cost-effective method for training neural networks for shearography defect detection using only simulated data.
- To design a spatiotemporal object detection network capable of learning from simulated shearography phase map sequences.
- To improve the accuracy and applicability of neural networks in shearography NDT.
Main Methods:
- Utilized phase map sequences from shearography as the data medium.
- Designed and implemented the YOWO_SS3D spatiotemporal object detection network.
- Trained the YOWO_SS3D network using 4000 frames of simulated phase map data.
Main Results:
- The YOWO_SS3D network achieved 96.99% detection accuracy on experimental phase maps.
- This significantly surpasses the 65.37% accuracy of YOLOv4 trained on the same simulated data.
- The network effectively learns spatial and temporal features from simulated data.
Conclusions:
- The YOWO_SS3D network enables accurate defect detection in shearography using solely simulated data.
- This approach eliminates the need for expensive experimental datasets, promoting wider adoption of AI in NDT.
- The YOWO_SS3D network is deployable for practical defect detection tasks, reducing costs and improving efficiency.
