Construction Of Multi-level Space-filling Experiment Design Based On Orthogonal Arrays

Space filling design is widely used in experimental designs.For the symmetric designs with any levels,when considering all possible results of level permutation and level expansion,the average of the discrepancy defined by any reproducing kernel can be expressed as a linear combination of generalized word length pattern.Level permutation can improve the space filling property of the design,and level expansion can expand the number of levels in multiples without changing the number of trials.Based on these theories,this paper proposes a method for constructing multi-level space filling designs based on orthogonal tables.Further,this paper improves the maximum projection criterion to be applicable to the case where the factor level has repeated values.Alao this paper continue to discuss when the space filling property measurement criterion is the decreasing function of distance,or the improved maximum projection criterion,the average of these two criteria can both be expressed as a linear combination of generalized word length pattern.