Multi-strategy System and Program Design for Multimodal Optimization

• Main Authors: Chen Wen-Tsai, 陳文財
• Other Authors: Lin Chyi-Yeu
• Format: Others
• Language: zh-TW
• Published: 1996
• Online Access: http://ndltd.ncl.edu.tw/handle/74058416100678084946

碩士 === 國立臺灣科技大學 === 機械工程研究所 === 84 === Simulated annealing belongs to a category of combinatorial global optimization methods. The use of multiple-state simulated annealing provides the ability of locating the global optimum as well as other local optima existing in the design space. The primary drawback of simulated annealing is the slow convergence rate to an accurate optimum. Crystallization strategy can be used to predict near-optimum designs from a number of designs already evaluated and start subsequent local searches from those identified near-optimum designs. Traditional gradient-based nonlinear programming methods can efficiently lead to an optimum with no guarantee that the optimum is a global optimum. This thesis developed a multiple- strategy optimization system and program to combine strengths of all three optimization methods and provide a number approaches to search the global and other local optima in a multimodal optimization problem with high efficiency.