Assessing the resilience of optimal solutions in multiobjective problems

Nuno Costa, João Lourenço

Research output: Contribution to journalArticlepeer-review

1 Citation (Scopus)
32 Downloads (Pure)

Abstract

Processes and products are multidimensional so researchers and practitioners have to solve problems with multiple objectives frequently. These problems have, in general, responses in conflict so they do not have a unique solution. Different approaches have been proposed in the literature to solve these problems, but many of them, including the popular desirability function approach, are not employed with the focus on the generation of Pareto frontiers. In addition, it is important to stress that some Pareto solutions may not yield the expected outcome(s) when implemented in practice. Thus, to avoid wasting resources and time in implementing a theoretical solution which does not produce the expected outcome(s), in this paper is proposed a novel metric to assess the resilience of Pareto solutions. This way, the decision-maker may identify a solution less sensitive to changes in the variables setting when their values are implemented in production process (equipments) or during its operation. Metric usefulness is illustrated using a case study, and results analysis is complemented with plots that facilitate the decision-making process.

Original languageEnglish
Article number104850
Number of pages7
JournalChemometrics And Intelligent Laboratory Systems
Volume239
DOIs
Publication statusPublished - 15 Aug 2023

Keywords

  • Desirability
  • Gradient norm
  • Implementation errors
  • Optimization
  • Pareto
  • RSM

Fingerprint

Dive into the research topics of 'Assessing the resilience of optimal solutions in multiobjective problems'. Together they form a unique fingerprint.

Cite this