基于组合优化算法的结构稳健优化设计方法研究与应用

【作者】 项立银；

【导师】 顾克秋；

【作者基本信息】 南京理工大学 ， 机械设计及理论， 2011， 硕士

【摘要】 在传统的结构优化设计研究中,设计变量通常被视为确定的量,并没有考虑它们可能产生的的不确定性,这也最终会导致结构性能的不稳定。因此有必要考虑设计变量实际中存在的变差,本文即以火炮中的典型部件为研究对象,研究和探索了结构稳健优化设计方法在典型机械结构中的应用。首先对结构稳健优化设计方法进行了深入的研究,在iSIGHT中进行了优化算法的二次开发,将多岛遗传算法与改进Hooke-Jeeves算法相结合,提出一种基于组合优化算法的6σ结构稳健优化设计方法。其次建立了闭气结构和后大架的三维实体参数化模型,并在ABAQUS里进行静力学刚强度分析,评估其结构性能。在iSIGHT中,搭建结构稳健优化设计平台,集成PRO/E和ABAQUS,实现机械结构的“设计—分析—稳健优化—再设计”。最后提出一种中心加密试验设计方法,对初始设计空间进行更加有效的探索；为了得到设计响应和设计变量之间的关系,需要进行大量的实验,这必将消耗大量的资源和时间,因而本文分析研究了响应面与Kriging两种近似模型,将Kriging近似模型应用到实例中,极大地降低了计算时间；多岛遗传算法中的迁移操作保持了优化解的多样性,提高了包含全局最优解的机会,但收敛速度较慢,本文开发的改进Hooke-Jeeves算法,可以进行有效的局部加速搜索,提高了得到全局最优解的机会,应用本文开发的这种组合优化算法,可以很好的提高结构稳健优化设计的效率。闭气结构与后大架稳健优化设计的成功,说明了本文研究与探索的结构稳健优化设计方法是具有一定的实用价值,可以解决机械结构设计中的一些实际问题。更多还原

【Abstract】 In the conventional structural optimization design, design variables are generally considered to be certain, not taking into account their possible uncertainties.That may eventually lead to instability of strcucture performance. So it’s necessary to take into account the actural existing variations. Taking some typical components of artillery as the research object, some key technologies of structural robust design method applied in the typical mechanical structures are researched and explored.Firstly, the structural robust design method is studied deeply. A secondary development of optimization algorithm is made based on iSIGHT. The multi-island genetic algorithm and improved Hooke-Jeeves algorithm are combined to propose a six sigma structural robust optimization design method.Secondly,3-D geometry parameterized models of an obturator and a back trail are accomplished.Their statics stiffiness and strength analysis is conducted in ABAQUS to evaluate their structural performances.A structural robust optimization design platform is established in iSIGHT, integrating PRO/E and ABAQUS. The mechanical structure’s design -analysis - robust optimization-design is realized.Lastly, a center denseer design of experiment method is proposed to take some more effective explorations to initial design space. In order to get the relationship between responses and design variables, need to do a lot of experiments. That will cost large amounts of resources and time. Therefore, two approximate models, response surface model and Kriging model are studied in thesis. Kriging approximation model is applied to the examples and the computational time is greatly reduced. The transfer operations of multi-island genetic algorithm can keep the diversity of the optimal solution, increase the chance of containing global optimal solution, but its convergence speed is slow. The improved Hooke-Jeeves algorithm can speed up the search in local effectively. The chance of getting global optimal solution is increased the optimization efficiency is improved.The success of robust optimization design to an obturator and a back trail shows that the research and exploration of structural robust design method has a certain applicable value, which can solve some actual problems of the mechanical structure design.更多还原