本文主要是針對連續變數、離散變數、混合變數之最佳化設計問題，提出一種結合人工蜂群演算法(Artificial Bee Algorithm, ABC)與隨機鄰點演算法(Stochastic Neighborhood Search Algorithm, SN)的混合搜尋演算法，即 ABC_SN。 ABC 為一隨機搜尋法，與粒子群演算法模擬鳥群智能行為相似，藉由蜜蜂的智能行為發展出來的演算法，具有全域搜尋的能?，其概?簡單且?需調整過多??。SN 是由差分進化演算法概念衍伸出來的一個隨機搜尋法。過去研究結果顯示，ABC 過度廣域的搜索造成了搜索精準度不高，且容易陷入局部最佳解，為了改善 ABC 的搜索精準度不高與容易陷入局部最佳解的這兩個問題，因此本文使用 SN 針對 ABC 的問題加以改善，將 ABC 與 SN 整合後期望增加其搜索精度與脫離局部最佳解。 在將 ABC 與 SN 整合方式中，本文共使用了兩種形式，藉由數個結構輕量化設計問題、含束制條件數學式問題與無束制條件數學式問題將分別用來探討其適用性和影響求解品質與效率的相關參數。 比較結果發現其中一種方式是將 ABC 的工蜂與觀察蜂階段中，隨機方式產生食物源更改為隨機鄰點方式來產生，其整體來看求解品質較佳。而只將 ABC 的觀察蜂階段中，隨機方式產生食物源更改為隨機鄰點方式來產生的話，則是強健性略優。 相較於兩種方式設計結果之比較，本文再將求解品質較佳的方式，稍作修改，因而在增加兩種分析形式並觀察是否能更進一步增加其強健性。 Previous studies showed that ABC have low accuracy and poor search in local space. In order to improve the drawback of ABC, this article raised a hybrid heuristic searching algorithms that combine with ABC (Artificial Bee Colony Algorithm) with SN (Stochastic Neighborhood Search Algorithm). These hybrid heuristic searching algorithms are devoted to problem solve for optimization problems with discrete, continuous and mixed variables. ABC is a random search method, and was developed by simulating the intelligent behavior of the honey bee. SN is a random neighborhood search method, it was modified from the differential evolution algorithm. For integration of ABC algorithm and SN algorithm, this article were used in two forms. This two forms will used to solve optimum design problem and discuss the applicability and the solution quality and efficiency, those problem include non-constrains mathematical problem 、constrains mathematical problem and structural problem. We can found that one of the model which is modify in employed bee phase and onlooker phase has better solution quality. And the other one which is modify in onlooker phase is slightly stronger. For increase stronger of the model which have better solution quality, this article will addition the other two analytical forms.
