This thesis investigates a critical concern in aerospace engineering, particularly about fibre- reinforced polymers (FRPs) deployed in structural laminates. The deleterious impact of microcracks within laminates, induced by thermal stresses during operational conditions, poses a formidable challenge, potentially curtailing these materials’ longevity and optimal utilization. This problem could be even more dangerous in structures used for hydrogen fuel storage. Subcritical matrix cracking can lead to the percolation of cracks through the thickness of the material and consequent fuel leakage, which can be considered a safety issue. The study aims to comprehensively comprehend the intricate mechanisms underpinning microcrack generation and propagation within non-crimp fabric (NCF) laminates manufactured through resin transfer moulding (RTM) when subjected to thermal cycling.

Utilizing X-ray computed tomography (XCT) constitutes the primary investigative modality for elucidating the progression of microcrack formation and dissemination. The NCF laminates’ low crimped textile structure, affording in-plane mechanical properties akin to traditional unidirectional angle-ply laminates, introduces a notable juxtaposition. However, stitching threads within the NCF laminates introduces a distinct challenge by inciting matrix crack formation in response to thermal fluctuations. This behaviour was not evidenced in standard woven materials.
Incorporating machine learning algorithms (ML) for crack segmentation derived from XCT imaging is a pioneering approach.

This methodology facilitates extracting pivotal geometric characteristics from segmented images, offering indispensable insights into the morphology, dimensions, spatial arrangement, and density of the incurred damage. Such meticulous analyses profoundly contribute to an enriched comprehension of damage evolution within the laminates, offering a nuanced perspective on microcrack development and its consequential impact on material integrity.
The combination of advanced imaging modalities with machine learning algorithms underscores the study’s potential to augment our discernment of damage mechanisms inherent in composite materials undergoing thermal cycling. This work promises to engender more robust and enduring composite structures for aerospace applications, thereby ensuring the development of safer and more resilient aircraft designs.