Design optimal experiments for parameter identification of a dynamic model with perturbed inputs

Research output: Chapter in Book/Report/Conference proceedingChapter

1 Scopus citations

Abstract

An optimization approach for sensor placement and input sequences to improve parameter identifiability in spatiotemporally dependent experiments, modelled by partial differential equations, is developed. Robust design criteria using bilevel optimization are applied in the approach. The determinant (D-optimality), the smallest eigenvalue (E-optimality) or the inverse of condition value (modified E-optimality) of the Fisher information matrix are applied as criteria. A greedy algorithm is used in the optimization procedure to avoid extremely long calculation time. The developed approach is illustrated towards the identification of kinetic parameters in a transient axial dispersion reactor model with perturbed inlets. Improvement of the parameter identifiability for the model is investigated fitting synthetic data generated by high-fidelity simulations with preset kinetic parameters. The performance of the three objective functions is also compared.

Original languageEnglish (US)
Title of host publicationComputer Aided Chemical Engineering
PublisherElsevier B.V.
Pages3481-3486
Number of pages6
DOIs
StatePublished - Jan 2024

Publication series

NameComputer Aided Chemical Engineering
Volume53
ISSN (Print)1570-7946

All Science Journal Classification (ASJC) codes

  • General Chemical Engineering
  • Computer Science Applications

Fingerprint

Dive into the research topics of 'Design optimal experiments for parameter identification of a dynamic model with perturbed inputs'. Together they form a unique fingerprint.

Cite this