Design Space of Regional-Jet Wing
Kazuhisa Chiba, Shinkyu Jeong, Shigeru Obayashi, Hiroyuki Morino · 2005
TheDataMining technique isanimportant facet ofsolving multi-objective optimization problem. Because itisoneoftheeffective mannertodiscover thedesign knowledge inthemulti-objective optimiza- tionproblem whichobtains large data. Inthepresent study, twoData Miningtechniques havebeenperformed fora large-scale, real-world Multidisciplinary Design Optimization (MDO)toprovide knowledge regarding thedesign space. TheMDO amongaerodynamics, structures, andaeroelasticity oftheregional-jet wingwascarried outusing high- fidelity evaluation models onAdaptive RangeMulti-Objective Genetic Algorithm. Asaresult, ninenon-dominated solutions weregenerated andusedfortradeoff analysis amongthreeobjectives. Allsolutions evaluated during theevolution wereanalyzed fortheinfluence ofdesign variables using aSelf-Organizing Map(SOM)andafunctional Analysis ofVariance (ANOVA) toextract keyfeatures ofthedesign space. SOM andANOVAcompensated withtherespective disadvantages, thenthe design knowledge could beobtained moreclearly bythecombination between them.Although theMDO results showedtheinverted gull- wingsasnon-dominated solutions, oneofthekeyfeatures foundby DataMining wasthenon-gull winggeometry. Whenthis knowledge was applied tooneoptimum solution, theresulting design wasfound tohave better performance compared withtheoriginal geometry designed inthe conventional manner. x X~~~~~~~~~X