利用遗传算法和神经网络响应面来实现复合材料结构优化设计英文.docxVIP

Composite Structural Optimization by Genetic Algorithm andNeural Net work Response Surface Model ingXU Yuan2ming , L I Shuo , RON G Xiao2min( S chool of A e ron a u t ical S cience a n d En gi nee ri n g Tech nology , B ei j i n g U ni ve rsi t y of A e ron a u t ics10 008 3 , Chi n a)a n d A st ron a u t ics , B ei j i n gAbstract : Neural2Net wor k Respo nse Surfaces ( NN RS) is applied to replace t he act ual expensive finite element analysis during t he co mposite st ruct ural op timizatio n p rocess. The Ort hot ropic Experiment Met hod (O EM) is used to select t he most app rop riate design samples for net wor k t raining. The t rained respo nse surfaces can eit her be objective f unctio n or co nst raint co nditio ns. Toget her wit h ot her co nven2 tio nal co nst raint s , an op timizatio n model is t hen set up and can be solved by Genetic Algorit hm ( GA) . This allows t he separatio n bet ween design analysis modeling and op timizatio n searching. Through an ex2 ample of a hat2stiffened co mposite plate design , t he weight respo nse surface is co nst ructed to be objective f unctio n , and st rengt h and buckling respo nse surfaces as co nst raint s ; and all of t hem are t rained t hrough NAS TRAN finite element analysis. The result s of op timizatio n st udy illust rate t hat t he cycles of st ruc2 t ural analysis can be remar kably reduced or even eliminated during t he op timizatio n , t hus greatly raising t he efficiency of op timizatio n p rocess. It also observed t hat NN RS app ro ximatio n can achieve equal or even bet ter accuracy t han co nventio nal f unctio nal respo nse surfaces.Key words : neural net wor k ; genetic algorit hm ; respo nse surface ; co mposite st ruct ural op timizatio n 利用遗传算法和神经网络响应面来实现复合材料结构优化设计 , 徐元铭 ,李烁 ,荣晓敏. 中国航空 学报 (英文版) , 2005 , 18 (4) : 310 - 316 .摘 要 :运用正交试验设计选择设计样本 ,建立神经网络响应面 ,以代替复合材料结构优化中的大 量的有限元分析 ;将神经网络响应面作为目标函数或者约束条件 ,加上其他常规约束条件进行优 化模型的建立 ,再应用遗传算法 ( GA) 进行优化 ,这可以实现设计分析与设计优化的分离