Topological Optimization of Two Dimensional Structure Subjected Static Loads:Application of Genetic Approach

• Main Authors: Huang, Yung-Chin, 黃永欽
• Other Authors: Wu Chun-Yin
• Format: Others
• Language: zh-TW
• Published: 1997
• Online Access: http://ndltd.ncl.edu.tw/handle/32745160194844959620

碩士 === 大同工學院 === 機械工程學系 === 85 === ABSTRACT In this study, a new approach is developed for multi-model domain system in structural design. This approach will provide designers with multiple choices of the optimum structural topology. It is very different from the detail design, we do not have to guess the optimal shape of the structure in advance. We just need to specify a feasible region within which the component has to fit, the support locations, and the applied loads in the beginning of the program. The topology generation can thus be automated as part of the design optimization process. This system combines with niche-based genetic algorithms, adaptive resonance theory and finite element method to optimize two-dimensional continuous structure with static loads. The aim of this study is to find the optimal shape of 2-D structure with the minimum weight and best distribution of stress. Genetic algorithms are used to be the main searching method in this study. The binary one-dimensional arrays are used to represent the structural configurations and they are utilized as the chromosomes in genetic algorithms. The structures are grouped into discontinuous, overload and non-overload structures and their fitness are evaluated with different fitness functions. We degenerate the infeasible solution, discontinuous, overload structures, by lower their fitness. In order to overcome the drawback falling into a local optimum by simple genetic algorithm ( SGA ), we propose three nechie-based GA models, ART-based SGA, crowding and sharing. Each model will be tested with different loading conditions and result are quite reasonable. We can understand the niche-based GAs is helpful to find different types of structural shapes and it is more efficient than traditional SGA.