基于改进ESE算法的多目标优化试验设计方法

Multi-objective experimentation design optimization based on modified ESE algorithms

最优化拉丁超立方试验设计是试验设计领域的热点问题。传统的拉丁超立方试验设计优化算法以试验点间正交属性度量准则或均匀性度量准则为单一优化目标,以单一优化准则求得的试验设计并不能确保其他的优化准则最优。以正交度量准则和均匀性度量准则为优化目标,提出基于改进ESE算法的多目标优化试验设计方法。算例测试证明,与已有算法相比,本文提出的新算法能得到更优的试验设计。

The optimization of Latin hypercube design（LHD） is an important area of the space filling experimentation.There are procedures to find good LHD by minimizing the pairwise correlations or maximizing the inter-site distance.we have shown that these two criteria need not agree with each other,that is,minimizing of pairwise correlations can result in LHDs where the inter-site distances is high and vice versa.Therefore,a multi-objective optimization approach to find good LHD is proposed by combining correlation and distance performance measures with modified ESE algorithm.Several examples are presented to show that the new algorithm is fast and the optimal designs are good in terms of both the correlation and distance criteria.