Robust dual-response optimization

Research output: Contribution to journalArticleScientificpeer-review

Abstract

This article presents a robust optimization reformulation of the dual-response problem developed in response surface methodology. The dual-response approach fits separate models for the mean and the variance and analyzes these two models in a mathematical optimization setting. We use metamodels estimated from experiments with both controllable and environmental inputs. These experiments may be performed with either real or simulated systems; we focus on simulation experiments. For the environmental inputs, classic approaches assume known means, variances, or covariances and sometimes even a known distribution. We, however, develop a method that uses only experimental data, so it does not need a known probability distribution. Moreover, our approach yields a solution that is robust against the ambiguity in the probability distribution. We also propose an adjustable robust optimization method that enables adjusting the values of the controllable factors after observing the values of the environmental factors. We illustrate our novel methods through several numerical examples, which demonstrate their effectiveness.
Original languageEnglish
Pages (from-to)298-312
JournalIIE Transactions: Industrial Engineering Research and Development
Volume48
Issue number3
DOIs
Publication statusPublished - 2016

Fingerprint

Probability distributions
Experiments

Keywords

  • robust optimization
  • dual-response optimization
  • simulation optimization

Cite this

@article{cdbd17f661c94c3da9baf7921c4f68d8,
title = "Robust dual-response optimization",
abstract = "This article presents a robust optimization reformulation of the dual-response problem developed in response surface methodology. The dual-response approach fits separate models for the mean and the variance and analyzes these two models in a mathematical optimization setting. We use metamodels estimated from experiments with both controllable and environmental inputs. These experiments may be performed with either real or simulated systems; we focus on simulation experiments. For the environmental inputs, classic approaches assume known means, variances, or covariances and sometimes even a known distribution. We, however, develop a method that uses only experimental data, so it does not need a known probability distribution. Moreover, our approach yields a solution that is robust against the ambiguity in the probability distribution. We also propose an adjustable robust optimization method that enables adjusting the values of the controllable factors after observing the values of the environmental factors. We illustrate our novel methods through several numerical examples, which demonstrate their effectiveness.",
keywords = "robust optimization, dual-response optimization, simulation optimization",
author = "Ihsan Yanikoglu and {den Hertog}, Dick and J.P.C. Kleijnen",
year = "2016",
doi = "10.1080/0740817X.2015.1067737",
language = "English",
volume = "48",
pages = "298--312",
journal = "I.I.E. Transactions: Industrial Engineering Research and Development",
issn = "0740-817X",
publisher = "Taylor and Francis Ltd.",
number = "3",

}

Robust dual-response optimization. / Yanikoglu, Ihsan; den Hertog, Dick; Kleijnen, J.P.C.

In: IIE Transactions: Industrial Engineering Research and Development, Vol. 48, No. 3, 2016, p. 298-312.

Research output: Contribution to journalArticleScientificpeer-review

TY - JOUR

T1 - Robust dual-response optimization

AU - Yanikoglu, Ihsan

AU - den Hertog, Dick

AU - Kleijnen, J.P.C.

PY - 2016

Y1 - 2016

N2 - This article presents a robust optimization reformulation of the dual-response problem developed in response surface methodology. The dual-response approach fits separate models for the mean and the variance and analyzes these two models in a mathematical optimization setting. We use metamodels estimated from experiments with both controllable and environmental inputs. These experiments may be performed with either real or simulated systems; we focus on simulation experiments. For the environmental inputs, classic approaches assume known means, variances, or covariances and sometimes even a known distribution. We, however, develop a method that uses only experimental data, so it does not need a known probability distribution. Moreover, our approach yields a solution that is robust against the ambiguity in the probability distribution. We also propose an adjustable robust optimization method that enables adjusting the values of the controllable factors after observing the values of the environmental factors. We illustrate our novel methods through several numerical examples, which demonstrate their effectiveness.

AB - This article presents a robust optimization reformulation of the dual-response problem developed in response surface methodology. The dual-response approach fits separate models for the mean and the variance and analyzes these two models in a mathematical optimization setting. We use metamodels estimated from experiments with both controllable and environmental inputs. These experiments may be performed with either real or simulated systems; we focus on simulation experiments. For the environmental inputs, classic approaches assume known means, variances, or covariances and sometimes even a known distribution. We, however, develop a method that uses only experimental data, so it does not need a known probability distribution. Moreover, our approach yields a solution that is robust against the ambiguity in the probability distribution. We also propose an adjustable robust optimization method that enables adjusting the values of the controllable factors after observing the values of the environmental factors. We illustrate our novel methods through several numerical examples, which demonstrate their effectiveness.

KW - robust optimization

KW - dual-response optimization

KW - simulation optimization

U2 - 10.1080/0740817X.2015.1067737

DO - 10.1080/0740817X.2015.1067737

M3 - Article

VL - 48

SP - 298

EP - 312

JO - I.I.E. Transactions: Industrial Engineering Research and Development

JF - I.I.E. Transactions: Industrial Engineering Research and Development

SN - 0740-817X

IS - 3

ER -