鄰域型粒子群演算法於結構設計之應用

Neighborhood Particle Swarm Optimization for Structural Designs

本文主要在探討一種具有群體智慧的概念方法－－粒子群演算法，應用於全域最佳化問題，利用粒子族群具有探測與開發的特色，在問題空間中搜尋全域的最佳解。本文利用函數測試例模擬粒子群演算法運動軌跡後，為了加強粒子群演算法在區域尋搜的能力，嘗試利用網路拓撲的概念來改良粒子族群相互連繫的行為模式，稱之為鄰域型粒子群演算法（Neighborhood Particle Swarm Optimization，NPSO）的搜尋策略，將其應用於幾個有限制條件的工程結構設計問題，並與其他全域型演算方法比較。測試結果顯示，加入鄰域型搜尋策略的NPSO較原本PSO來得穩定快速，且NPSO與其他方法比較，NSPO不僅能更精確地搜尋到問題最佳解，且具有相當不錯的計算效率。 This paper is to study swarm's intelligence based on the particle swarm algorithm for global optimization problems. This algorithm having the characteristics of exploration and exploitation of particles in the problem space searches for the optimal solution. Firstly, in this study the simulation of searching trajectory of swarm with PSO algorithm is investigated through a testing function. To improve the performance of PSO algorithm, the concept of network topology of particles in swarm, neighborhood-called particle swarm algorithm (Neighborhood Particle Swarm Optimization, NPSO) search strategy, is introduced to solve several constrained structural design problems. The results show that the searching performance, such as reliability and convergent speed are better than those of the standard PSO. Comparison of NPSO and several other methods, the NPSO can not only find more accurately the final solution to the problem, but also having rather computation efficiency.