Graduation Semester and Year
Spring 2025
Language
English
Document Type
Dissertation
Degree Name
Doctor of Philosophy in Aerospace Engineering
Department
Mechanical and Aerospace Engineering
First Advisor
Andrew Makeev
Second Advisor
Atilla Dogan
Third Advisor
Guillaume Seon
Fourth Advisor
Brian Huff
Fifth Advisor
Catherine Kilmain
Abstract
Structural health monitoring plays a pivotal role in the maintenance of aerospace structures. A crucial component of structural health monitoring is nondestructive inspection (NDI), which encompasses various techniques such as visual inspection, ultrasonic and acoustic testing, and X-ray Computed Tomography (CT). Among these methods, X-ray CT is distinguished by its unparalleled fidelity, leading to extensive applications. Although X-ray CT provides high-fidelity defect detection, its application to large aerospace components is restricted by the size limitations of test specimens. Inclined CT (ICT) mitigates these constraints by positioning the X-ray source and detector on opposite sides of a stationary test specimen. This system geometry, however, results in limited angular data for 3D reconstructions, leading to significant artifacts that may inaccurately represent defects. This study illustrates that deep learning (DL) techniques, specifically the fine-tuned Segment Anything Model (SAM), can enhance defect recognition from ICT data. The methodology involves fine-tuning SAM with a dataset comprising 1,800 images across ten synthetic phantoms with varying defect sizes and locations. Validation of the fine-tuned model on an as-built aluminum test specimen serves as a proof-of-concept, achieving over 70% accuracy for defect detection and 98% accuracy for overall shape detection. Further validation on carbon fiber reinforced polymer specimens with Teflon inserts provided improved results compared to ICT reconstruction methods, indicating promising practical applicability. The findings suggest that DL-enhanced ICT can attain detection capabilities comparable to full CT while maintaining ICT's compatibility with large structures, making it a valuable NDI method for aerospace industry applications.
Keywords
structural health monitoring, nondestructive inspection, x-ray computed tomography, limited-view ct, inclined ct, 3d reconstruction, deep learning, segmentation, neural networks, segment anything model
Disciplines
Aviation Safety and Security | Maintenance Technology | Structures and Materials
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Recommended Citation
Metiner, Abdullah, "DEEP LEARNING-ENHANCED X-RAY COMPUTED TOMOGRAPHY FOR DEFECT DETECTION IN COMPOSITE STRUCTURES" (2025). Mechanical and Aerospace Engineering Dissertations. 425.
https://mavmatrix.uta.edu/mechaerospace_dissertations/425
Included in
Aviation Safety and Security Commons, Maintenance Technology Commons, Structures and Materials Commons