基于人工神经网络的短纤维增强复合材料设计  

Short fiber reinforced composite material design based on artificial neural network

作　　者：李骁 夏佳佳 张向奎[2] LI Xiao;XIA Jiajia;ZHANG Xiangkui(School of Automotive Engineering,Dalian University of Technology,Dalian 116024,Liaoning,China;School of artificial Intelligence,Dalian University of Technology,Dalian 116024,Liaoning,China)

机构地区：[1]大连理工大学汽车工程学院,辽宁大连116024 [2]大连理工大学人工智能学院,辽宁大连116024

出　　处：《计算机辅助工程》2024年第2期17-23,60,共8页Computer Aided Engineering

摘　　要：为指导短纤维增强复合材料设计,提出一种基于人工神经网络的方法。首先,结合取向平均法和自洽假设建立数学模型,计算出不同体积分数下SFRCs的弹性性能;然后,建立2个不同的人工神经网络,将数学模型计算得到的数据集投入训练,由此可以由SFRCs的体积分数快速预测出弹性性能,也可以由已知的SFRCs的弹性性能反向推导出纤维的体积分数。与传统方法相比,该方法可以有效减少重复实验次数,降低成本与周期,对复合材料实验的设定有一定的参考意义。In order to guide the design of short fiber reinforced composites,a method based on artificial neural network is proposed.First,a mathematical model is established by combining orientation averaging method and self-consistent hypothesis,to calculate the elastic properties of SFRCs under different volume fractions.Then,two different artificial neural networks are built,and the data sets obtained from the mathematical model are trained.The elastic properties of SFRCs can be quickly predicted from the volume fraction of SFRCS.The fiber volume fraction can also be inversely derived from the known elastic properties of SFRCs.Compared with the traditional method,it can effectively reduce the number of repeated experiments,reduce the cost and cycle,and has certain reference significance for the setting of composite experiments.

关 键 词：短纤维增强复合材料 细观力学 人工神经网络 弹性性能 训练 体积分数 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TB334[自动化与计算机技术—控制科学与工程]