Modelling the strip thickness in hot steel rolling mills using least-squares support vector machines

Yuri A. W. Shardt, Siamak Mehrkanoon, Kai Zhang, Xu Yang*, Johan Suykens, Steven X. Ding, Kaixiang Peng

*Corresponding author for this work

Research output: Contribution to journalArticleAcademicpeer-review

20 Citations (Web of Science)

Abstract

The development and implementation of better control strategies to improve the overall performance of a plant is often hampered by the lack of available measurements of key quality variables. One way to resolve this problem is to develop a soft sensor that is capable of providing process information as often as necessary for control. One potential area for implementation is in a hot steel rolling mill, where the final strip thickness is the most important variable to consider. Difficulties with this approach include the fact that the data may not be available when needed or that different conditions (operating points) will produce different process conditions. In this paper, a soft sensor is developed for the hot steel rolling mill process using least-squares support vector machines and a properly designed bias update term. It is shown that the system can handle multiple different operating conditions (different strip thickness setpoints, and input conditions).

Original languageEnglish
Pages (from-to)171-178
Number of pages8
JournalCanadian Journal of Chemical Engineering
Volume96
Issue number1
DOIs
Publication statusPublished - Jan 2018
Externally publishedYes

Keywords

  • soft sensors
  • steel mill
  • support vector machines
  • process systems engineering
  • SOFT SENSORS
  • REGRESSION
  • SELECTION

Cite this