4D Imaging of ceramic matrix composites during polymer infiltration and pyrolysis

Abstract This study investigates microstructural evolution of pores in a woven-fabric ceramic matrix composite (CMC) laminate manufactured using multi-step polymer infiltration and pyrolysis (PIP). The evolution of pores was visualized and quantified by leveraging high-resolution X-ray micro-computed tomography (µCT) and machine-learning-based image segmentation. X-ray µCT imaging was performed after each PIP step to capture the progression of the evolving microstructure. Segmentation of tomographic images was performed semi-automatically to identify individual fiber tows and automatically to identify pores using Trainable Weka Segmentation. Analysis of segmented image data was performed to determine changes in the percent, distribution, and connectivity of pores with each manufacturing step. Results reveal that the CMC retained about 2.5% of porosity by volume, despite five consecutive PIP re-infiltrations. 3D X-ray µCT image data revealed that persistent porosity was predominantly caused by pores that formed before or during the initial C-staging of the resin. Analysis of segmented image data showed that connectivity between pores and the free surfaces increased significantly after the initial pyrolysis, and the first two re-infiltrations filled a majority of these connections. After C-staging and initial pyrolysis, pores tended to form between fiber tows and primarily in interstitial regions. Throughout PIP manufacturing, the through-thickness distribution of pores was relatively uniform, and small pores were more readily filled during consecutive re-infiltrations, likely due to their higher capillarity. It was also revealed that the shape of shrinkage cracks was similar to those that formed within unidirectional, single-tow specimens, which suggests that single-tow specimens are a good analog for studying the formation of shrinkage cracks within woven-fiber tows. Overall, this paper presents novel ways to study the 3D evolution of pores within CMCs by leveraging X-ray µCT imaging and machine-learning-based segmentation algorithms.