Orthogonal designs for computer experiments constructed from sequences with zero autocorrelation

Abstract

Designs for computer experiments constitute an important class of experimental designs. Computer experiments are used when the physical experiments are expensive or time-consuming and attracted a lot of attention in recent years. In this paper, we proposed a method for generating computer experiments with many factors and symmetric runs. These designs are suitable for computer experiments and are constructed using known sequences with zero autocorrelation function, such as T-sequences, Bases sequences, Normal sequences, and other. The results appear to be encouraging as the methodologies can transform the known sequences into designs for computer experiments without the need for a computer search. The generated designs have some favorable properties, including the symmetry in their runs which results in all the even orders effects being orthogonal to the main effects.