The meta-model approach for simulation-based design optimization

E. Stinstra

Research output: ThesisDoctoral ThesisScientific

Abstract

The design of products and processes makes increasing use of computer simulations for the prediction of its performance. These computer simulations are considerably cheaper than their physical equivalent. Finding the optimal design has therefore become a possibility. One approach for finding the optimal design using computer simulations is the meta-model approach, which approximates the behaviour of the computer simulation outcome using a limited number of time-consuming computer simulations. This thesis contains four main contributions, which are illustrated by industrial cases. First, a method is presented for the construction of an experimental design for computer simulations when the design space is restricted by many (nonlinear) constraints. The second contribution is a new approach for the approximation of the simulation outcome. This approximation method is particularly useful when the simulation model outcome reacts highly nonlinear to its inputs. Third, the meta-model based approach is extended to a robust optimization framework. Using this framework, many uncertainties can be taken into account, including uncertainty on the simulation model outcome. The fourth main contribution is the extension of the approach for use in integral design of many parts of complex systems.
LanguageEnglish
QualificationDoctor of Philosophy
Awarding Institution
  • Tilburg University
Supervisors/Advisors
  • den Hertog, Dick, Promotor
Award date22 Nov 2006
Place of PublicationTilburg
Publisher
Print ISBNs905668180X
Publication statusPublished - 2006

Fingerprint

Computer simulation
Design of experiments
Design optimization
Large scale systems
Uncertainty
Optimal design

Cite this

Stinstra, E. (2006). The meta-model approach for simulation-based design optimization. Tilburg: CentER, Center for Economic Research.
Stinstra, E.. / The meta-model approach for simulation-based design optimization. Tilburg : CentER, Center for Economic Research, 2006. 143 p.
@phdthesis{713f828a47164a19af00e6f5e668a535,
title = "The meta-model approach for simulation-based design optimization",
abstract = "The design of products and processes makes increasing use of computer simulations for the prediction of its performance. These computer simulations are considerably cheaper than their physical equivalent. Finding the optimal design has therefore become a possibility. One approach for finding the optimal design using computer simulations is the meta-model approach, which approximates the behaviour of the computer simulation outcome using a limited number of time-consuming computer simulations. This thesis contains four main contributions, which are illustrated by industrial cases. First, a method is presented for the construction of an experimental design for computer simulations when the design space is restricted by many (nonlinear) constraints. The second contribution is a new approach for the approximation of the simulation outcome. This approximation method is particularly useful when the simulation model outcome reacts highly nonlinear to its inputs. Third, the meta-model based approach is extended to a robust optimization framework. Using this framework, many uncertainties can be taken into account, including uncertainty on the simulation model outcome. The fourth main contribution is the extension of the approach for use in integral design of many parts of complex systems.",
author = "E. Stinstra",
year = "2006",
language = "English",
isbn = "905668180X",
series = "CentER Dissertation Series",
publisher = "CentER, Center for Economic Research",
school = "Tilburg University",

}

Stinstra, E 2006, 'The meta-model approach for simulation-based design optimization', Doctor of Philosophy, Tilburg University, Tilburg.

The meta-model approach for simulation-based design optimization. / Stinstra, E.

Tilburg : CentER, Center for Economic Research, 2006. 143 p.

Research output: ThesisDoctoral ThesisScientific

TY - THES

T1 - The meta-model approach for simulation-based design optimization

AU - Stinstra, E.

PY - 2006

Y1 - 2006

N2 - The design of products and processes makes increasing use of computer simulations for the prediction of its performance. These computer simulations are considerably cheaper than their physical equivalent. Finding the optimal design has therefore become a possibility. One approach for finding the optimal design using computer simulations is the meta-model approach, which approximates the behaviour of the computer simulation outcome using a limited number of time-consuming computer simulations. This thesis contains four main contributions, which are illustrated by industrial cases. First, a method is presented for the construction of an experimental design for computer simulations when the design space is restricted by many (nonlinear) constraints. The second contribution is a new approach for the approximation of the simulation outcome. This approximation method is particularly useful when the simulation model outcome reacts highly nonlinear to its inputs. Third, the meta-model based approach is extended to a robust optimization framework. Using this framework, many uncertainties can be taken into account, including uncertainty on the simulation model outcome. The fourth main contribution is the extension of the approach for use in integral design of many parts of complex systems.

AB - The design of products and processes makes increasing use of computer simulations for the prediction of its performance. These computer simulations are considerably cheaper than their physical equivalent. Finding the optimal design has therefore become a possibility. One approach for finding the optimal design using computer simulations is the meta-model approach, which approximates the behaviour of the computer simulation outcome using a limited number of time-consuming computer simulations. This thesis contains four main contributions, which are illustrated by industrial cases. First, a method is presented for the construction of an experimental design for computer simulations when the design space is restricted by many (nonlinear) constraints. The second contribution is a new approach for the approximation of the simulation outcome. This approximation method is particularly useful when the simulation model outcome reacts highly nonlinear to its inputs. Third, the meta-model based approach is extended to a robust optimization framework. Using this framework, many uncertainties can be taken into account, including uncertainty on the simulation model outcome. The fourth main contribution is the extension of the approach for use in integral design of many parts of complex systems.

M3 - Doctoral Thesis

SN - 905668180X

T3 - CentER Dissertation Series

PB - CentER, Center for Economic Research

CY - Tilburg

ER -

Stinstra E. The meta-model approach for simulation-based design optimization. Tilburg: CentER, Center for Economic Research, 2006. 143 p. (CentER Dissertation Series).