用人工神经网络模拟三维编织复合材料的力学性能  

Simulating Mechanical Behavior of 3D Braided Composites Using an Artificial Neural Network

作　　者：李水乡[1] 孙慧玉[2] 黄传奇[3] 

机构地区：[1]北京大学力学与工程科学系 [2]中国科技大学力学与机械工程系 [3]南京航空航天大学飞行器系

出　　处：《南京航空航天大学学报》1997年第4期397-401,共5页Journal of Nanjing University of Aeronautics & Astronautics

基　　金：航空科学基金

摘　　要：三维编织复合材料由于其材料结构及编织工艺的复杂性和众多工艺参数的影响，目前尚未建立成熟的力学模型。本文采用人工神经网络ＢＰ（ｂａｃｋｐｒｏｐａｇａｔｉｏｎ）算法，将编织工艺参数作为人工神经网络的输入，将弹性模量及强度性能作为输出，建立了编织工艺参数与力学性能的人工神经网络关系模型，并讨论了ＢＰ算法及网络构造。这种人工神经网络关系模型对于三维编织复合材料的实验、生产和应用，工艺参数的选取以及理论模型的研究都有重要的参考价值。本文最后给出了三维编织复合材料拉伸和压缩性能的实验结果与人工神经网络预测结果，通过对比显示，其模拟效果令人满意。As the complexity of braided technique and effects of various technological parameters, a reasonable mechanical model of 3D braided composites has not been developed yet. This paper presents an artificial neural network relation model with technological parameters and the mechanical behavior of 3D braided composites using BP algorithm. These parameters are used as inputs, the elastic module and the strength are used as outputs. The BP(back propagation) algorithm and the network construction are also discussed. The neural network relation model is an important reference to the experiment, manufacture, application of 3D braided composites and selection of technological parameters for studying theoretical model. Additionally, the paper provides a contrast between the tension/compression test results and the simulating results by the artificial neural network. It is shown that the simulation is successful.

关 键 词：神经网络 纤维增强 复合材料 材料力学性质 

分 类 号：TB330.1[一般工业技术—材料科学与工程]