New results for space-filling orthogonal designs

Computer experiments are popularly used to study complicated systems in many fields. Orthogonality and space-filling properties are two important design properties for such experiments. Space-filling properties guarantee effective factor screening while the orthogonality allows the estimates of the main effects in linear models to be uncorrelated with each other. In this paper, we develop a general rotation method for constructing a new class of orthogonal designs. The proposed method is easy to operate and flexible in both run size and number of levels. The resulting designs are not only orthogonal but also entertain attractive space-filling properties. These designs can be either symmetric or asymmetric. Related theoretical supports are well established. The numerical results validate the advantages of the proposed designs.