Sensitivity analysis of limit states in civil engineering: From model outputs to reliability assessment

Sensitivity analysis of limit states in civil engineering: From model outputs to reliability assessment

Paper in proceedings (conference paper)

Sensitivity analysis is a tool for measuring how the conclusion drawn from a model depends on changes in the parameters that enter it. Research on probabilistic reliability analysis in engineering disciplines shows that simply focusing on the model output is not sufficient. In stochastic models, the correlation between input and output, output variance analysis and other methods oriented only to model output are not directly related to reliability, which is one of the main objectives of structural design. The presented article provides a brief history on the development of sensitivity analysis methods, starting with methods oriented to model outputs and ending with the global sensitivity analysis of reliability oriented to the probability of failure or design quantiles. In the next part of the article, a case study examining the influence of loading, material and geometric characteristics on the probability of failure of a steel bar with a thin-walled closed cross-section subjected to torsion is presented. Sensitivity analysis is performed using two sensitivity measures that measure changes in the probability of failure due to changes in input variables. There are large differences between the proportion of first-order and higher-order sensitivity indices. Although the results of the sensitivity analysis are not the same, both sensitivity measures give similar values of total sensitivity indices and infer the same sensitivity ranking of the input variables.

Sensitivity analysis is a tool for measuring how the conclusion drawn from a model depends on changes in the parameters that enter it. Research on probabilistic reliability analysis in engineering disciplines shows that simply focusing on the model output is not sufficient. In stochastic models, the correlation between input and output, output variance analysis and other methods oriented only to model output are not directly related to reliability, which is one of the main objectives of structural design. The presented article provides a brief history on the development of sensitivity analysis methods, starting with methods oriented to model outputs and ending with the global sensitivity analysis of reliability oriented to the probability of failure or design quantiles. In the next part of the article, a case study examining the influence of loading, material and geometric characteristics on the probability of failure of a steel bar with a thin-walled closed cross-section subjected to torsion is presented. Sensitivity analysis is performed using two sensitivity measures that measure changes in the probability of failure due to changes in input variables. There are large differences between the proportion of first-order and higher-order sensitivity indices. Although the results of the sensitivity analysis are not the same, both sensitivity measures give similar values of total sensitivity indices and infer the same sensitivity ranking of the input variables.

sensitivity analysis, structures, columns, limit state

sensitivity analysis, structures, columns, limit state

OMISHORE, A.

2024

27.09.2023

AIP Publishing

World Multidisciplinary Civil Engineering-Architecture-Urban Planning Symposium - WMCAUS 2022

1551-7616

AIP conference proceedings

2928

1

United States of America

150006-1

150006-8

8

https://pubs.aip.org/aip/acp/article/2928/1/150006/2913242/Sensitivity-analysis-of-limit-states-in-civil