Construction Of Supersaturated Designs Based On Distance Criterion

Modern experiments,such as industrial and engineering experiments,involve a large num-ber of factors.Under the assumption of sparse effect,not all factors are active.So in the early stages of experiments,researchers often use screening experiments to identify active factors.Supersaturated design is often used for screening factors due to economy and practicality.Most of the current researches on supersaturated design propose corresponding design criteria and construction methods from the perspective of orthogonality.But the uniformity of design is also a very important property,and distance is a commonly used criterion for measuring design uniformity.Therefore,this article studies supersaturated design based on the distance criterion,and two methods for constructing supersaturated designs are proposed based on the distance criterion.The supersaturated design constructed in this paper not only has good uniformity,but also has good orthogonality.The specific work of this article is as follows:The first part reviews two optimal criteria used to evaluate the construction of supersatu-rated designs,that is,theE（s²）criterion and theE(f_NOD)criterion.The second part introduces a distance criterion and proposes two construction methods for two-level and three-level supersaturated designs under the given distance criterion,algorithm search and combinatorial construction.The third part is numerical simulation.The superiority of the proposed design is evaluated by comparing the distance efficiency,E（s²）value,andE(f_NOD)value.The numerical simula-tion results show that the design obtained in this paper has good uniformity and performs well under other design criteria.