OPTIMIZATION OF MULTI-FIDELITY DATA USING CO-KRIGING FOR HIGH DIMENSIONAL PROBLEMS

Document Type : Original Article

Authors

1 Postdoc researcher, Department of Mechanical Engineering, Vrije Universiteit Brussel, Pleinlaan 2 -1050 Brussels- Belgium.+

2 Assistant Professor, Helwan University, Faculty of Engineering - Mattaria, Department of Mechanical Power Engineering, Masaken El-Helmia, 11718 Cairo, Egypt.

3 Professor, Department of Mechanical Engineering, Vrije Universiteit Brussel, Pleinlaan 2 -1050 Brussels- Belgium.

Abstract

ABSTRACT
This paper deals with an efficient and multi-fidelity design strategy for high dimensional industrial problems. The most significant factors have been determined based on the Muschelknautz method of modeling (MM) using the screening approach. For cyclone separator, only four (from seven) geometrical parameters are significant. An optimized sampling plan based on random Latin hypercube (LHS) has been used to fit Co-Kriging based on CFD data and an analytical model for estimation of pressure drop. Co-Kriging exhibits better accuracy than ordinary Kriging and blind Kriging if only the high fidelity data is used. For global optimization, the Co-Kriging surrogate in conjunction with genetic algorithms (GA) is used. CFD simulations based on the Reynolds stress turbulence model are also used in this study. A new set of geometrical ratios (design) has been obtained (optimized) to achieve minimum pressure drop. A comparison of numerical simulation of the new design and the Stairmand design confirms the superior performance of the new design compared to the Stairmand design.

Keywords

Files

Share

How to cite

Statistics