Related Experiment Video
Updated: Jun 16, 2025
08:27
Author Spotlight: Efficient Image Recognition Using Directional Gradient Histogram Technique and Support Vector Machines
Published on: January 5, 2024
1.0K
Defect Recognition in Composite Materials Using Terahertz Spectral Imaging with ResNet18-SVM Approach
Zhongmin Wang1, Jiaojie Chen1, Yilong Xin1,2
1Institute of Automation, Qilu University of Technology (Shandong Academy of Sciences), Jinan 250100, China.
Materials (Basel, Switzerland)
|June 13, 2025
Summary
This study introduces a novel terahertz time-domain spectroscopy (THz-TDS) method for detecting internal defects in multilayer composites. The ResNet18-SVM approach achieves high accuracy, enabling precise defect localization and depth resolution for nondestructive evaluation.
Area of Science:
- Materials Science
- Non-destructive Testing
- Signal Processing
Background:
- Multilayer composite materials are prone to internal defects affecting performance.
- Conventional defect detection methods struggle with depth localization.
- Accurate defect identification is crucial for quality control and structural integrity.
Purpose of the Study:
- To develop an advanced defect detection method for multilayer composites.
- To improve the depth resolution and accuracy of defect localization.
- To leverage terahertz time-domain spectroscopy (THz-TDS) and deep learning for defect analysis.
Main Methods:
- Utilized terahertz time-domain spectroscopy (THz-TDS) for data acquisition.
- Applied continuous wavelet transform (CWT) to generate time-frequency images from spectral data.
- Employed a ResNet18 model combined with a support vector machine (SVM) classifier for defect identification.
Main Results:
- The ResNet18-SVM method achieved a high accuracy of 98.56% in defect detection.
- Successfully identified three distinct types of internal defects in multilayer composites.
- Demonstrated superior performance compared to four other deep learning models and three classifiers.
Conclusions:
- The proposed THz-TDS and ResNet18-SVM method offers effective feature extraction and depth resolution.
- This technique shows significant potential for the nondestructive evaluation of multilayer structures.
- The study highlights a robust approach for advanced material inspection and quality assurance.
