摘要
利用多个自组织神经网络的同时训练 ,可以一次性自动地生成计算区域内与计算区域边界上所有网格节点 ,尔后按 FGT法的基本思想将节点连接成所需的三角形或四边形单元。由于在自组织神经网络中采用了改善的目标函数 ,节点的分布可以实现自适应调节以反映网格疏密分布的要求。文末几个算例验收证明本算法具有自适应性 ,适用于凸域、凹域 。
To make use of self organized neural network, the points of FEM among the computed area and boundaries of it are produced one times. Then, this points are linked into the needed triangle or quadrilateval element by the FGT method. To apply the target function improved, the distribution of points may be adjusted according to distribution function. Some examples prove this method has the advantages to adjust autoatically, and to fit convex, non convex, mutli connected domain.
出处
《计算力学学报》
CAS
CSCD
北大核心
2001年第4期501-504,共4页
Chinese Journal of Computational Mechanics
