Outlier treatment for improving parameter estimation of group contribution based models for upper flammability limit

Jerome Frutiger, Jens Abildskov, Gürkan Sin

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

Abstract

Flammability data is needed to assess the risk of fire and explosions. This study presents a new group contribution (GC) model to predict the upper flammability limit UFL oforganic chemicals. Furthermore, it provides a systematic method for outlier treatment inorder to improve the parameter estimation of the GC model. The new method identifies and removes outliers based on the empirical cumulative distribution plot. It is compared to outlier detection based on Cook’s distance and normal cumulative distribution.
Original languageEnglish
Title of host publicationProceedings of the 25th European Symposium on Computer Aided Process Engineering
EditorsKrist V. Gernaey, Jakob K. Huusom, Rafiqul Gani
PublisherElsevier
Publication date2015
Pages503-508
ISBN (Print)978-0-444-63429-0
DOIs
Publication statusPublished - 2015
Event25th European Symposium on Computer Aided Process Engineering : 12th International Symposium on Process Systems Engineering - Copenhagen, Denmark
Duration: 31 May 20154 Jun 2015
http://www.pse2015escape25.dk/

Conference

Conference25th European Symposium on Computer Aided Process Engineering
CountryDenmark
CityCopenhagen
Period31/05/201504/06/2015
Internet address
SeriesComputer Aided Chemical Engineering
Volume37
ISSN1570-7946

Keywords

  • Outlier treatment
  • Flammability limit
  • Group contribution

Fingerprint Dive into the research topics of 'Outlier treatment for improving parameter estimation of group contribution based models for upper flammability limit'. Together they form a unique fingerprint.

Cite this