多学科优化中的近似模型及其在艇体结构优化中的应用  

Approximation methods of multidisciplinary design optimization and their application in shell hull of submarine

作　　者：许辉[1] 周奇[1] 

机构地区：[1]中国舰船研究设计中心,湖北武汉430064

出　　处：《舰船科学技术》2014年第12期6-10,共5页Ship Science and Technology

基　　金：国防基础科研重点资助项目

摘　　要：近似技术是多学科设计优化关键技术之一。系统介绍多学科设计优化中常用的3种近似模型:响应面模型(Response Surface Methodology,RSM)、Kriging模型、径向基函数(Radical Basis Function,RBF)模型。对比分析3类近似模型在典型数学函数不同维度下的拟合性能。结果表明,Kriging模型的综合拟合性能最优,RBF模型次之,RSM拟合性能不佳。在此基础上,通过拉丁超立方试验设计选取样本点,基于Kriging模型构造潜艇平行舯体某舱段耐压壳体结构响应的近似模型,应用粒子群优化算法完成了耐压壳体结构轻量化设计。最优化设计方案质量较初始方案减少了11.99%。Approximation method is one of the key technologies for multidisciplinary design optimization of vessels. In this paper,three commonly approximation methods,including response surface method,Kriging,radical basis function are introduced concise. A comparison of three approximation for their fitting performance is conducted through a typical mathematical functions with different dimensions. The results shows that Kriging,with its optimum comprehensive performance,is best suitable for replacing finite element model,the radical basis function is in the middle position,while the approximation performance of response surface is poor. Then,Sample points selected by latin hypercube experimental design method,and a Kriging medols are constructed for shell struct of submarine. Error analysis showed that the models can meet the engineering application requirements.Finally,The structural lightweight design results are proposed base on the particle swarm optimization. Optimized design program quality decrease 11. 99% compared to the weight of the initial design scheme.

关 键 词：近似模型 拟合性能 粒子群优化 潜艇壳体 拉丁超立方试验 

分 类 号：U661[交通运输工程—船舶及航道工程]