A quantitative study of off-axis bending damage in 3D braided ceramic matrix composites

Due to their complex interwoven microstructure, 3D braided composites are tougher and offer a higher fatigue resistance than traditional laminated counterparts. However, their failure mechanisms under complex loading conditions remain insufficiently understood, requiring advanced characterization techniques. This paper investigates the failure mechanisms in the braided ceramic composites under combined bending, torsion, and shear loading conditions. By combining micro-CT scan images and acoustic emission (AE) data, neural networks are utilized to process the CT images, while unsupervised machine learning is applied to classify the AE signals, facilitating the identification and characterization of damage mechanisms. The development of internal damage is then reconstructed to provide a better understanding of the damage mechanisms in these composites under the complex loading conditions. The damage takes the form of matrix cracking, yarn breakage, and yarn stripping, based on the clustering analysis of the AE signals. A quantitative evolution of the three damage modes under different flexural conditions is also presented. The AE signals elucidate the underlying mechanisms responsible for the volume of damage. An increase in damage volume is shown to be associated with an increase in the cumulative energy of the AE signals and the unsupervised machine learning method highlights the correlation between AE signal features and damage mechanisms. The test data indicates a high degree of yarn and matrix failure in the composites, with yarn breakage being the predominant damage mechanism, occurring along the braid path of the material. By integrating micro-CT analysis and AE-based unsupervised learning, this study provides a novel and automated framework for understanding damage progression in 3D braided CMCs.