| Summary: | 碩士 === 國立臺灣大學 === 機械工程學研究所 === 88 === This thesis studies the effects of penalty factors of exterior penalty method for structural optimization using genetic algorithms and develops a fuzzy reasoning system to determine the penalty level automatically at each design generation. First, three fuzzy variables are defined for describing the number of violated constraint in the population, the level of constraint violation, and the penalty level of the exterior penalty method. Fuzzy reasoning rules are established to deal with the relationship among these three fuzzy variables. Then, a fuzzy algorithm is developed to determine the level of penalty factors at each design generation. Finally, a structural optimization program integrating genetic algorithms, ANSYS finite element software, exterior penalty method for numerical search ,and a fuzzy reasoning system for determining the penalty level. With the developed program, design processes of some test examples and practical structural design problems are given. From these results, it shows that the adjustable penalty level determined by the fuzzy reasoning system gives quite satisfactory convergent result for most cases.
|
|---|
