摘要
利用光学显微镜观测和测量轴编C/C复合材料细观编织结构及其尺寸,建立轴编C/C复合材料有限元模型,通过给出组分材料有效性能的变化区间,构造组分材料性能与轴编C/C复合材料宏观有效性能的对应关系,利用径向基函数(RBF)人工神经网络(ANN)方法,对该高度非线性的对应关系进行训练,通过轴编C/C复合材料宏观实验结果,预报其组分材料的有效性能。结果表明,轴编C/C复合材料面内弹性性能基本相同,在测量时可忽略面内纤维束铺设方向的影响;人工神经网络对训练样本有一定的依赖性,但通过多次随机构造样本训练网络,可得到理想的预测结果,且人工神经网络方法具有很好的容错性,能很好地预报轴编C/C复合材料组分材料的有效性能。
The microscopic braid structure and size of the in-plane braided C/C composites were measured by optical microscope observation.The finite element model of the braided composites was established.After the change ranges of effective properties of the constituent materials were provided,the macroscopic effective behavior corresponding relation between in-plane braided C/C composites and constituent material was constructed.Then,the nonlinear corresponding relation was trained by the radial basis function（RBF） artificial neural network（ANN） method.The effective properties of the constituents were predicted by the network combining the macroscopic experimental results of the braided composites.The results show that the in-plane elastic properties of the braided composites are about the same,which can be measured along the axial direction.The network is dependent on the training samples,but reasonable predictive results can be obtained by constructing a number of random sample training network.The network has a good fault tolerance and can predict the effective behavior of constituents of the braided composites accurately.
出处
《固体火箭技术》
EI
CAS
CSCD
北大核心
2012年第5期644-649,共6页
Journal of Solid Rocket Technology
基金
国家安全重大基础研究项目(61391)
国家自然科学基金(90916027
11102051)
中国博士后科学基金(20110491069)
关键词
轴编C
C复合材料
组分材料
人工神经网络
有效性能
in-plane braided C/C composite
constituents
artificial neural network
effective behavior
