Microstructurele variabiliteit en het effect ervan op mechanische eigenschappen in filamentgewikkelde composieten

Filament-wound composite pressure vessels (CPVs) are the industry standard for storage of hydrogen fuel in fuel cell-based electric vehicles. The filament winding winds continuous fibres over a mandrel or liner to produce axisymmetric parts. Thanks to continuous fibres and optimised layup sequences tailored to loading directions, filament-wound composite structures can withstand high load levels. Therefore, this process is ideal for manufacturing CPVs which operate under nominal working pressures of up to 700 bar. Due to the limited fibre tension and environmental control during filament winding, CPVs often suffer from microstructural defects. As a result, the design safety factors for composite pressure vessels are large to compensate for the performance degradation and variability caused by these defects. This increases the material usage and, consequently, the cost of the CPVs. The manufacturing process can significantly influence the extent and characteristics of microstructural defects. Limited studies describe the effect of manufacturing parameters on defects in filament-wound composites, as most of the research in that field is performed on flat plates. Such an analysis is crucial to optimise the microstructure of CPVs and minimise the microstructural variability and defects. The overarching goal of this thesis was to understand the effect of manufacturing parameters on the microstructural variability in CPVs. Voids, fibre misalignment and fibre volume fraction are the most crucial defects with respect to the burst strength. The present work aimed to: (1) characterise the CPV microstructure in an accurate and representative manner using X-ray computed tomography (XCT), (2) determine the effect of manufacturing parameters on microstructural defects, and (3) numerically analyse the effect of crucial defects on the stress redistribution around broken fibres in hoop layers. For this analysis, eleven CPVs manufactured with different manufacturing parameters were analysed. Various factors affect the accuracy of the characterisation of microstructural features through XCT. The scanning parameters and quantification techniques must be tailored to the feature of interest, and enough samples should be analysed for a representative analysis. In this thesis, the scanning parameters were obtained through a sensitivity analysis depending on the microstructural feature. The quantification techniques were selected such as to ensure the accuracy of the results, minimum bias, and analysis efficiency. The traditional greyscale thresholding was found insufficient for the complex voids, and a deep-learning-based segmentation tool was utilised. Separate quantification techniques were selected for the average and local variations in fibre-related defects. Statistical analysis was used to determine the number and location of the analysed samples in both circumferential and axial directions. The analysis shows that at least four well-separated samples along the circumferential direction extracted from one axial location sufficiently represent the microstructure in the CPV's cylindrical section. The developed scanning and analysis protocols for the CPVs can be extended to other large composite components. The trends in void characteristics, fibre misalignment, and fibre volume fraction were studied as a function of manufacturing parameters. The manufacturing parameters considered in this study are: three variations of winding tension and two variations of winding speed, liner pressure during winding, and resin take-up. The increase in winding tension causes an increase in the average void fraction. The change in winding speed and resin take-up also affects the void fraction. The trends in void size are weak, and void shape and location did not show specific trends. Both in-plane and out-of-plane misalignments show a slight increase or decrease with increasing winding tension and winding speed, respectively. The trends observed for fibre misalignment were weak in all cases. The fibre volume fraction varies radially for all CPVs, with an average increase of 14% in the innermost hoop layer compared to the vessel's nominal fibre volume fraction. Only resin take-up had some influence on the average fibre volume fraction and thickness of the innermost hoop layer. The manufacturing parameters influence the microstructural defects to various degrees. The microstructural defects also show an interdependence. Fibre misalignment is higher in the vicinity of complex voids. The little influence of manufacturing parameters on void shape indicates that they do not affect the high misalignment near complex voids either. The fibre volume fraction is lower in high void fraction regions. The voids, depending on their size and shape, also affect the fibre packing density. The analysis of voids reveals that they show the highest sensitivity to manufacturing parameters and can affect the local fibre volume fraction and packing. This necessitated an analysis of their effect on the stress redistribution in the innermost hoop layer, which is loaded in a circumferential direction under burst conditions (analogous to longitudinal fibre direction). A 3D finite element model simulating the stress redistribution near a broken fibre is utilised. The interaction of the void with the broken fibre and its intact neighbours increases the stress recovery lengths and affects the stress overload in the intact fibres. This effect is more pronounced if the void interacts with both broken fibres and immediate neighbours. If the void is not touching the broken fibre, the primary effect comes from the change in fibre packing. The results show that the voids may increase the possibility of fibre break cluster development, depending on their location with respect to a broken fibre. The results obtained from this study established a correlation between the CPV microstructure and the manufacturing parameters. The study also created a vast data repository on various defect characteristics. The correlation between these defects was established. Based on the observed trends, recommendations for optimising manufacturing parameters are provided to minimise defects in the innermost hoop layer of CPV, which can help reduce the burst performance variability and consequently the design safety factor.