OA-Based Latin Hypercube Design Under Uniform Projection Criterion

The computer experimental design is an significant problem in statistics.With the development of society and science,we are able to transfer the practical problem to the mathematical model.Nevertheless,some mathematical models are extremely complicated,high-speed electronic computer and huge amounts of time and money.So,we need to find some properly efficient points to be as inputs,and then through modeling we can get a new similar model to work instead of the complex one.In this article,we introduce some popular computer experiment designs and models,and the popular methods are: uniform design,Latin hypercube design,orthogonal Latin hypercube design,OA-based Latin hypercube design.According to the hierarchical ordering principle,lower order effects are more likely to be important than higher order effects.When choosing a design,it should pay more attention to the low-dimensional projection uniformity of the designs.Hence,we propose a new design method called OA-based Latin hypercube design under uniform projection criterion.In addition,this paper introduces the commonly used modeling methods,Kriging modeling which is usually used in computer experiments.Finally,this paper uses the existing design methods and the proposed new design method to establish kriging models,and then compares their mean square error.Through the comparison,it is not difficult to find that the proposed new design method performs better.