Using Deep Learning to Find Damage in Composite Structures
Author Information
Author(s): Azad Muhammad Muzammil, Munyaneza Olivier, Jung Jaehyun, Sohn Jung Woo, Han Jang-Woo, Kim Heung Soo
Primary Institution: Dongguk University-Seoul
Hypothesis
Can deep learning models effectively detect and assess damage severity in composite structures using Lamb waves?
Conclusion
The study demonstrates that a deep learning framework can accurately localize and assess damage severity in composite structures using Lamb waves.
Supporting Evidence
- The CNN model achieved a test accuracy of 92.19% for damage severity assessment.
- The CNN model showed a mean absolute error of 12.20 mm for damage localization.
- ANN and CNN models achieved 100% accuracy in identifying healthy states.
- The study utilized a dataset of 3240 damage signals and 360 baseline signals.
Takeaway
Researchers used smart computer programs to find and measure damage in special materials called composites, helping keep them safe and strong.
Methodology
The study used three deep learning models (ANN, CNN, GRU) to analyze Lamb wave signals from composite structures for damage detection and localization.
Potential Biases
The reliance on specific data augmentation techniques may introduce bias in model training.
Limitations
The study primarily focused on laminated composite plates and may not generalize to more complex structures.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website