材料科学
均质化(气候)
标准差
叠加原理
参数统计
校准
计算机科学
刚度
算法
复合材料
生物系统
结构工程
代表性基本卷
体素
反褶积
极限抗拉强度
纤维
联轴节(管道)
不确定度量化
执行机构
有限元法
航空航天
包辛格效应
方向(向量空间)
嵌入
二次方程
聚类分析
声学
分割
可扩展性
机身
拉伸试验
模数
纤维增强复合材料
灾难性故障
多尺度建模
作者
Zhiyong Wu,Fang-Ping Chen,Limin Xu,Wen-Hai Sun,Bai-Jun Yang,Weiyan Lü,Deping Lu,Hui Guo
标识
DOI:10.1177/00219983261474270
摘要
A four-order “orientation-clustering-stiffness-damage” coupling framework is proposed for 28–30 vol% carbon-fiber/epoxy short-fiber composites (7 μm nominal fiber diameter, 3–5 mm chopped length), enabling accurate cross-scale failure prediction spanning from 60 μm single-fiber interfacial debonding regions up to 300 μm component-level representative volume elements that cover the full statistical distribution of fiber orientation and spatial arrangement. Twenty-one 30 × 30 × 30 mm 3 cubic specimens were in-situ imaged under quasi-static loading via synchrotron radiation Micro-CT at 0.5 μm voxel spacing, delivering approximately 1 μm effective spatial resolution after standard modulation transfer function deconvolution and ring-artifact suppression processing. The tomographic finite-element analysis workflow was fully synchronized to a pre-calibrated six-channel acoustic-emission array with 120 dB gain and 100 kHz–1 MHz bandwidth, where 15 specimens were exclusively used for multi-parameter model calibration through a genetic algorithm optimization loop, while the remaining six completely independent specimens were adopted for rigorous blind validation under uniaxial tension, biaxial compression and in-plane shear loading conditions. This framework achieves a tensile modulus prediction error below 5% and over 92% acoustic-emission hit event agreement on the fully unseen validation dataset, which significantly outperforms conventional Mori-Tanaka mean-field homogenization and periodic representative-volume-element homogenization methods by reducing prediction deviation by more than 40%. By embedding high-order orientation tensors, statistical fiber clustering spectral descriptors and gradient-sensitive non-local damage variables into a single physically interpretable constitutive kernel, the framework delivers an experimentally calibratable, highly scalable and transferable modeling paradigm, holding great practical application potential for composite process reverse design in aerospace and automotive lightweighting industries.
科研通智能强力驱动
Strongly Powered by AbleSci AI