A reinforced combinatorial particle swarm optimization based multimodel identification of nonlinear systems

Ahmed A. Adeniran, Sami El Ferik*

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

5 Scopus citations


Several industrial systems are characterized by high nonlinearities with wide operating ranges and large set point changes. Identification and representation of these systems represent a challenge, especially for control engineers. Multimodel technique is one effective approach that can be used to describe nonlinear systems through the combination of several submodels, where each is contributing to the output with a certain degree of validity. One major concern in this technique, especially for systems with unknown operating conditions, is the partitioning of the system's operating space and thus the identification of different submodels. This paper proposes a three-stage approach to obtain a multimodel representation of a nonlinear system. A reinforced combinatorial particle swarm optimization and hybrid K-means are used to determine the number of submodels and their respective parameters. The proposed method automatically optimizes the number of submodels with respect to the submodel complexity. This allows operating space partition and generation of a parsimonious number of submodels without prior knowledge. The application of this approach on several examples, including a continuous stirred tank reactor, demonstrates its effectiveness.

Original languageEnglish
Pages (from-to)327-358
Number of pages32
JournalArtificial Intelligence for Engineering Design, Analysis and Manufacturing: AIEDAM
Issue number3
StatePublished - 1 Aug 2017

Bibliographical note

Publisher Copyright:
© 2016 Cambridge University Press.


  • K-Means
  • Multimodel
  • Nonlinear Systems
  • Particle Swarm Optimization
  • Systems Identification

ASJC Scopus subject areas

  • Industrial and Manufacturing Engineering
  • Artificial Intelligence


Dive into the research topics of 'A reinforced combinatorial particle swarm optimization based multimodel identification of nonlinear systems'. Together they form a unique fingerprint.

Cite this