东北大学学报（自然科学版）

东北大学学报(自然科学版) ›› 2006, Vol. 27 ›› Issue (3): 312-315.DOI: -

• 论著 • 上一篇    下一篇

结构优化设计中的组合遗传算法

范鹤;刘斌;张延年;韩桂武;   

  1. 东北大学资源与土木工程学院;东北大学资源与土木工程学院;沈阳建筑大学土木工程学院;东北大学资源与土木工程学院 辽宁沈阳110004;辽宁沈阳110004;辽宁沈阳110015;辽宁沈阳110004
  • 收稿日期:2013-06-23 修回日期:2013-06-23 出版日期:2006-03-15 发布日期:2013-06-23
  • 通讯作者: Fan, H.
  • 作者简介:-
  • 基金资助:
    国家自然科学基金资助项目(50508008);;

Combinatorial genetic algorithm in optimum design of structure

Fan, He (1); Liu, Bin (1); Zhang, Yan-Nian (2); Han, Gui-Wu (1)   

  1. (1) School of Resources and Civil Engineering, Northeastern University, Shenyang 110004, China; (2) School of Civil Engineering, Shenyang Architecture University, Shenyang 110015, China
  • Received:2013-06-23 Revised:2013-06-23 Online:2006-03-15 Published:2013-06-23
  • Contact: Fan, H.
  • About author:-
  • Supported by:
    -

RichHTML

0

PDF (PC)

729

摘要： 针对标准遗传算法存在的早熟收敛、随机振荡和收敛速度慢等缺陷,采取改进措施.利用混沌序列的随机遍历性生成初始种群,并把相对差商算法的优化解加入到初始种群中,改善初始种群的性能.采用适应度的指数尺度变换改进传统的适应度评价函数.相对差商算法局部搜索能力强,而遗传算法具有较强的全局搜索性,发挥两者的优势,提出组合遗传算法.把相对差商算法作为一个与选择、交叉、变异平行的遗传算子嵌入到改进遗传算法中,提高局部寻优能力,防止早熟收敛.通过十杆平面桁架的数值算例来验证组合遗传算法应用的可行性和有效性,组合遗传算法的优化结果也远好于标准遗传算法和改进遗传算法.

关键词: 结构优化设计, 相对差商算法, 标准遗传算法, 改进遗传算法, 组合遗传算法

Abstract: In view of the shortcomings of simple genetic algorithm, such as premature convergence, random oscillation and slow convergence, some measures are taken to improve them. Generating original population via the ergodicity of chaotic sequence, the optimized results generated from relative difference quotient algorithm are added to the original population to improve its performance. Introduces the fitness scaling to improve traditional fitness function. Relative difference quotient algorithm is strong in local search, while genetic algorithm with a parallel operation is highly effective in global search. To play the role of both advantages, a combinatorial genetic algorithm (CGA) is proposed, taking the relative difference quotient as a genetic operator which is parallel to selection, crossover and mutation and embedded into the improved genetic algorithm so as to upgrade the local optimization ability and avoid premature convergence. A numerical example of a ten-bar truss was given to demonstrate the validity and feasibility of CGA in its application, and it is found that the optimized results of CGA are superior to SGA and IGA.

中图分类号: