Periodic Audze-Eglājs Criterion for Orthogonal and Regular Triangular Grids.

Periodic Audze-Eglājs Criterion for Orthogonal and Regular Triangular Grids.

Paper in proceedings (conference paper)

This paper studies a criterion that can be used for optimization of the designs of experiments (DoE) in cases when the uniform filling of the sampling space is demanded. The criterion studied in this paper is based on the Audze-Eglājs criterion and it has recently been modified in order to suppress its previously discovered defects. The principle remains but it is enriched by taking into account the periodicity of the final design, therefore it is called periodic Audze-Eglājs (PAE) criterion. To increase the efficiency of the optimization process of designs the knowledge of the minimum value of the optimization criterion is often required. Hence, two ideally space-filling deterministic designs are examined to find the lower bound on the PAE criterion. Numerical studies support the presumption that it is possible to interpolate the minimum value of PAE for sample sizes where the perfect deterministic designs with regular patterns cannot be achieved.

This paper studies a criterion that can be used for optimization of the designs of experiments (DoE) in cases when the uniform filling of the sampling space is demanded. The criterion studied in this paper is based on the Audze-Eglājs criterion and it has recently been modified in order to suppress its previously discovered defects. The principle remains but it is enriched by taking into account the periodicity of the final design, therefore it is called periodic Audze-Eglājs (PAE) criterion. To increase the efficiency of the optimization process of designs the knowledge of the minimum value of the optimization criterion is often required. Hence, two ideally space-filling deterministic designs are examined to find the lower bound on the PAE criterion. Numerical studies support the presumption that it is possible to interpolate the minimum value of PAE for sample sizes where the perfect deterministic designs with regular patterns cannot be achieved.

Optimization, design of experiments, periodic Audze-Eglais criterion, lower bound, uniform space-filling, orthogonal grid, full factorial design, Latin Hypercube Sampling

Optimization, design of experiments, periodic Audze-Eglais criterion, lower bound, uniform space-filling, orthogonal grid, full factorial design, Latin Hypercube Sampling

ŠMÍDOVÁ, M.; SADÍLEK, V.; ELIÁŠ, J.; VOŘECHOVSKÝ, M.

2016

01.09.2015

Praha, Česká republika

978-1-905088-63-8

International Conference on Civil, Structural and Environmental Engineering Computing

1

15

15