Hybrid semiparametric modeling: A modular process systems engineering approach for the integration of available knowledge sources

Cristiana Rodrigues De Azevedo, Victor Grisales Díaz, Oscar Andrés Prado-Rubio, Mark J. Willis, Véronique Préat, Rui Oliveira, Moritz Von Stosch

Research output: Chapter in Book/Report/Conference proceedingChapterpeer-review

4 Citations (Scopus)

Abstract

This chapter reviews different methods of inferring system descriptions - models - from data, and discusses their merits in light of data requirements and consideration of available a priori knowledge, which can describe either the structure of interactions and/or the way/form of the occurring interactions. It focuses on hybrid semiparametric modeling. Hybrid models combine fundamental models with data-driven or heuristic-driven models. The attractive traits of hybrid models are their good extrapolation capabilities, low data requirements, and transparent structure, which helps to develop process understanding. All of these properties stem from the incorporated fundamental knowledge. The chapter describes two applications to demonstrate how hybrid models can be used to solve problems that are not easily solved using fundamental understanding alone, but where the fundamental backbone can be exploited by combining it with neural networks.

Original languageEnglish
Title of host publicationSystems Engineering in the Fourth Industrial Revolution: Big Data, Novel Technologies, and Modern Systems Engineering
PublisherWiley
Pages345-373
Number of pages29
ISBN (Electronic)9781119513957
ISBN (Print)9781119513896
DOIs
Publication statusPublished - 24 Jan 2020

Keywords

  • Data-driven model
  • Heuristic-driven models
  • Hybrid semiparametric modeling
  • Modular process systems engineering approach
  • Neural networks
  • Parallel hybrid models
  • Serial hybrid models

Fingerprint

Dive into the research topics of 'Hybrid semiparametric modeling: A modular process systems engineering approach for the integration of available knowledge sources'. Together they form a unique fingerprint.

Cite this