一種用于單元制造系統(tǒng)設(shè)計(jì)的多目標(biāo)遺傳算法(外文翻譯)

MAGHSUD SOLIMANPURy, PREM VRATz and RAVI SHANKAR}*

包含中文翻譯和英文原文，內(nèi)容詳細(xì)完整，建議下載參考！

中文：2400 字
英文：8100 字符



多目標(biāo)遺傳所算法的最新發(fā)展相當(dāng)廣泛而迅速。有很多的多目標(biāo)進(jìn)化算法，如那些歸功于Schaffer (1985), Hajela and Lin (1992), Horn and Nafpliotis (1993), Srinivas and Deb (1994)等的算法。不同多目標(biāo)進(jìn)化算法的回顧和分類可在Fonseca and Fleming (1995) and Coello (1999)中看到。這些顯著點(diǎn)說明不同方法之間的差異是由于每個(gè)染色體的適應(yīng)度函數(shù)所采用的不同策略造成的。Zitzler et al. (2000) 把不同的多目標(biāo)進(jìn)化算法分為三類，包括標(biāo)準(zhǔn)選擇，聚類選擇和Pareto選擇。該算法采用標(biāo)準(zhǔn)選擇策略，如Schaffer (1985)提出的向量評估遺傳算法，在選擇階段目標(biāo)之間的轉(zhuǎn)化。在向量評估遺傳算法中，出現(xiàn)在交配池中的種群被每個(gè)目標(biāo)選擇一部分。選擇執(zhí)行聚合的方法使用傳統(tǒng)的多目標(biāo)優(yōu)化技術(shù)，而多目標(biāo)被合并成一個(gè)數(shù)量的目標(biāo)方程。Pareto選擇為基礎(chǔ)的算法，利用Pareto的系統(tǒng)定義來排列當(dāng)前種群的解決方案?，F(xiàn)有的文獻(xiàn)中有不同的Pareto排列規(guī)則（如Goldberg 1989, Srinivas and Deb 1994）。本文中的做法屬于第一類，其中選擇階段基于適應(yīng)度函數(shù)。向量評估遺傳算法受到批評，因?yàn)閷τ诿總€(gè)目標(biāo)，它最終結(jié)果會聚集在最有方案附近(Fonseca and Fleming 1995)，這是因?yàn)?，在向量評估遺傳算法中，下一代交配池的每個(gè)部分的選擇每次都是基于一個(gè)目標(biāo)，而其他的目標(biāo)則被忽略掉(Srinivas and Deb 1994)。本文中所提出的方法是基于結(jié)合所有目標(biāo)的適應(yīng)度方程，因此，預(yù)計(jì)該方法會克服前面所提到的局限 ......



Recent developments in multi-objective evolutionary algorithms are quite extensive and rapidly growing. There are many multi-objective evolutionary algorithms such as those due to Schaffer (1985), Hajela and Lin (1992), Horn and Nafpliotis (1993), Srinivas and Deb (1994), etc. in the literature. A review and classification of different multi-objective evolutionary algorithms can be found in Fonseca and Fleming (1995) and Coello (1999). The salient point indicating the differences between these approaches is due to the strategy by which the fitness of each chromosome is assigned. Zitzler et al. (2000) classified different multi-objective evolutionary algorithms into three categories including criterion selection, aggregation selection and Pareto selection. The algorithms employing a criterion selection strategy, e.g. the 1420 M. Solimanpur et al. vector evaluated genetic algorithm (VEGA) proposed by Schaffer (1985), switch between the objectives during the selection phase. In VEGA a certain fraction of the population appearing in the mating pool is selected with regard to each objective.The methods performing aggregation selection use conventional multi-objective optimization techniques where multiple objectives are combined into a scalar objective function. Pareto selection-based algorithms use the systematic definition of Pareto solutions to rank the solutions in the current population. There are different rules for ranking Pareto solutions in the literature (e.g. Goldberg 1989, Srinivas andDeb 1994). Our approach in this paper falls into the first category in which the selection phase is based on fitness functions. The VEGA approach is criticized because it clusters final solutions around the best solution with respect to each objective (Fonseca and Fleming 1995). This is mainly due to the fact that inVEGA each fraction of the ......