基于深度学习的2.5D陶瓷基复合材料损伤识别与评估

Deep Learning-based Damage Identification and Evaluation of 2.5D Ceramic Matrix Composites

为揭示陶瓷基复合材料的损伤演化与失效机理,开展了 2.5D-SiCf/SiC复合材料的X射线CT原位拉伸试验,得到了材料加载过程和失效断口的CT数据;采用基于深度学习的图像分割方法,利用图像处理软件建立了复合材料内部损伤的三维可视化模型;对拉伸产生的微裂纹进行了数量、体积、空间分布和空间角度分析。微裂纹分析表明:较高载荷状态下的拉伸微裂纹体积在0-5.3×10^(-4) mm^(3)之间,断口微裂纹体积在0-2.2×10^(-4 )mm^(3)之间,空间角度为0°和45°左右的微裂纹占主要部分,微裂纹与孔隙分布有很强的相关性。断口损伤分析表明:2.5D-SiCf/SiC复合材料的拉伸损伤包括纤维断裂、纤维拔出、基体开裂和界面脱粘,其中纤维拔出长度范围为16.02-250.32μm,平均长度为88.26μm;2.5D-SiCf/SiC复合材料的损伤分析表明基于深度学习的图像分割方法为揭示材料损伤演化机理与评估损伤程度提供了有效手段。

The X-ray CT in-situ tensile test of 2.5 D-SiCf/SiC composites was conducted to reveal the damage evolution and failure mechanism of the ceramic matrix composites.The CT data of the materials during the loading process and failure fracture were obtained.A deep learning-based image segmentation method was used to propose a three-dimensional visualization model of the internal damage of the composites,and the number,volume,spatial distribution and spatial angle of the microcracks generated due to the stretching were analyzed.At high loads,the tensile microcrack volumes are 0-5.3×10^(-4) mm^(3),the fracture microcrack volumes are0-2.2×10^(-4) mm^(3),and the proportion of the microcracks with spatial angles of about 0° and 45° is relatively high.Meanwhile,the microcracks are closely correlated to the pore distribution.According to the results by fracture damage analysis,the tensile damage process of 2.5 D-SiCf/SiC composites includes matrix cracking,fiber fracture,fiber pull-out and interface debonding.The fiber pull-out lengths are in the range of 16.02-250.32μm,and the average length is 88.26μm.It is indicated that the image segmentation method based on deep learning could offer an effective way to reveal the mechanism of material damages and evaluate the degree of the damages.