Efficient space-filling and near-orthogonality sequential Latin hypercube for computer experiments

In this paper, a general and efficient method for constructing LHDs possessing favorable characteristics of orthogonality and better computational efficiency is first proposed. Next, a sequentially enlarging strategy is derived based on the initial design. During the enlarging process, optimization of space-filling property is carried out, and a near-orthogonal space-filling augmented LHD is subsequently obtained. The enlarging process is repeated iteratively until the desired accuracy is achieved. Numerical examples and two mechanical applications are utilized in order to confirm and emphasize the effectiveness of this proposed method.