桁架结构截面优化设计的改进模拟退火算法  

Improved simulated annealing algorithm for cross-section design optimization of truss structures

作　　者：顾元宪[1] 项宝卫[2] 赵国忠[1] 

机构地区：[1]大连理工大学工业装备结构分析国家重点实验室工程力学系,大连116024 [2]浙江台州学院计算机系,临海317020

出　　处：《计算力学学报》2006年第5期546-552,共7页Chinese Journal of Computational Mechanics

基　　金：国家自然科学基金(10032030;10228206;10302006);国家自然科学基金创新(2005)研究群体基金(10421202)资助项目

摘　　要：将模拟退火算法应用于桁架结构截面尺寸优化设计,提出若干方法改进了算法的鲁棒性、计算效率和求解精度。通过一批经典问题,同时与传统结构优化算法和遗传算法进行了比较。数值结果表明,本文的改进模拟退火算法具有很高的优化求解精度,计算效率有显著提高且优于遗传算法,有望在结构优化设计问题中发挥其特点。The simulated annealing （SA） algorithm is applied to the design optimization of cross-section sizes of bars of truss structures. Some approaches have been proposed to improve the basic procedures of SA algorithm such as the determination of initial temperature, the generation and acceptance of solutions, and the convergence criterion based on a newly defined relative precision. The important algorithm control parameters such as the number of random seeds and the length of Markovian chain are studied carefully by numerous tests. These improvements and studies have enhanced the robustness, efficiency and accuracy of the SA algorithm in the solution of structural design optimization problems. A certain number of traditional example problems and a relatively complicated 200-bar problem of cross- section design optimization for truss structures are solved with the present improved SA algorithm and compared with related examples in literatures. The solution accuracy is compared with the traditional structural optimization algorithms, and convergent efficiency is compared with the Genetic Algorithm （GA）. The numerical results and comparisons of test algorithm of present paper has high solution preci creased to be much better than GA. It is hopeful to to make use of its featured advantages. Sl examples have demonstrated that the improved SA on, and its solution efficiency has noticeably inapply SA algorithm in structural design optimization to make use of its featured advantages.

关 键 词：模拟退火 桁架结构 截面优化 遗传算法 

分 类 号：O224[理学—运筹学与控制论]