张洪武

个人信息Personal Information

教授

博士生导师

硕士生导师

性别:男

毕业院校:德国汉诺威大学

学位:博士

所在单位:力学与航空航天学院

学科:工程力学. 计算力学. 生物与纳米力学

电子邮箱:zhanghw@dlut.edu.cn

扫描关注

论文成果

当前位置: 中文主页 >> 科学研究 >> 论文成果

A MULTISCALE COMPUTATIONAL METHOD FOR 2D ELASTOPLASTIC DYNAMIC ANALYSIS OF HETEROGENEOUS MATERIALS

点击次数:

论文类型:期刊论文

发表时间:2014-01-01

发表刊物:INTERNATIONAL JOURNAL FOR MULTISCALE COMPUTATIONAL ENGINEERING

收录刊物:SCIE、EI、Scopus

卷号:12

期号:2

页面范围:127-154

ISSN号:1543-1649

关键字:multiscale computational method; heterogeneous material; elastoplastic dynamic analysis; numerical base functions

摘要:The elastoplastic dynamic analysis of heterogeneous materials is studied based on the multiscale computational method developed in our previous work (Zhang et al., 2013). The basic principles of this method are introduced briefly. To describe the complex deformation, a 2D multinode coarse element is proposed. In addition, to improve the computational accuracy for the dynamic problems, mode base functions are introduced into the multiscale numerical base functions to consider the dynamic effect of the structure. For nonlinear elastic or elastoplastic dynamic problems, the microscopic unbalanced nodal force cannot be projected to the macroscopic level effectively only by the displacement and mode base functions when the nonlinear material deformation occurs during the computation. So a correction technique of the local displacement is applied to deal with the unprojected microscopic unbalance forces within the coarse element. Furthermore, the computational procedures of a two-scale modeling are proposed within the framework of nonlinear dynamic analysis. Extensive numerical experiments are carried out and the results are compared with the traditional finite element method (FEM) which is applied directly on the fine-scale mesh. It is shown that the proposed multiscale method can obtain excellent precision of the nonlinear dynamic response of the elastoplastic heterogeneous materials. Moreover, the computation comparisons indicate that the proposed method spends less computer memory and CPU time than the traditional FEM.