Load capacity probabilistic sensitivity analysis of thin-walled beams

Maria Kotelko*, Pawel Lis, Martin Macdonald

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

12 Citations (Scopus)
114 Downloads (Pure)


The paper deals with the load carrying capacity probabilistic variance-based sensitivity analysis of thin-walled box-section girders subjected to pure bending. The lower- and upper-bound load capacity estimation is performed using two different analytical methods. The sensitivity analysis performed is based on the methodology of the Monte-Carlo method. The analysis is carried out using the polynomial decomposition and multi-dimensional linear regression. The sample results obtained are presented in diagrams and pie charts showing the sensitivity of load capacity to different random input variables (material properties and geometrical parameters). The variance-based analysis (Anova) of lower-bound and upper-bound load capacity estimation is carried out, from which some conclusions are derived, if (and how) assumed changes in standard deviations of input variables influence the magnitude of the load capacity and differences in upper-bound and lower-bound load capacity estimations. The results of Anova tests are shown in sample histograms. Some final conclusions concerning the efficiency of the applied models and the statistically significant influence of input random variables (yield stress, wall thickness, height and length of the beam) upon the upper-bound and lower bound estimation of the load capacity as well as the difference of these two estimations, are presented.
Original languageEnglish
Pages (from-to)142-153
Number of pages12
JournalThin-Walled Structures
Early online date27 Feb 2017
Publication statusPublished - Jun 2017


  • thin-walled girders
  • bending
  • load-capacity
  • sensitivity analysis


Dive into the research topics of 'Load capacity probabilistic sensitivity analysis of thin-walled beams'. Together they form a unique fingerprint.

Cite this