期刊
INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN ENGINEERING
卷 65, 期 1, 页码 1-17出版社
WILEY
DOI: 10.1002/nme.1430
关键词
shape optimization; unstructured remeshing; truss analogy; analytical sensitivity analysis; consistent tangent; local minima
A novel unstructured remeshing environment for gradient-based shape optimization using triangular finite elements is presented. The remeshing algorithm is based on a truss structure analogy; in solving for the equilibrium position of the truss system, the quadratically convergent Newton's method is used. Exact analytical sensitivity information is therefore made available to the shape optimization algorithm. The overall computational efficiency in gradient-based shape optimization is very high. In solving the truss Structure analogy, we compare Our quadratically convergent Newton solver with a previously proposed forward Euler solver; this includes notes regarding mesh uniformity, element quality, convergence rates and efficiency. We present three numerical examples; it is then shown that remeshing may introduce discontinuities and local minima. We demonstrate that the effects of these on gradient-based algorithms are alleviated to some extent through mesh refinement, and may largely be overcome with a simple multi-start strategy. Copyright (c) 2005 John Wiley & Sons, Ltd.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据