Sensitivity estimation of first excursion probabilities of linear structures subject to stochastic Gaussian loading. (May 2021)

• Record Type: Journal Article
• Authors: Valdebenito, Marcos A.; Misraji, Mauricio A.; Jensen, Héctor A.; Franco Mayorga, C.
• Is Part Of: Computers & structures. Volume 248(2021)
• Publication Date: 2021-05
• Subjects: First excursion probability -- Sensitivity analysis -- Stochastic Gaussian loading -- Linear structure -- Directional Importance Sampling; Structural engineering -- Data processing -- Periodicals; Electronic data processing -- Structures, Theory of -- Periodicals; 624.171
• DOI: 10.1016/j.compstruc.2021.106482
• Languages: English
• ISSNs: 0045-7949

Abstract

Highlights: Sensitivity of probability is estimated applying Directional Importance Sampling. Sensitivity is calculated in a post-processing step of probability estimation. Sensitivity analysis of unit impulse response functions is required. Precision of sensitivity estimates is quantified in terms of standard deviation. Abstract: This contribution focuses on evaluating the sensitivity associated with first excursion probabilities of linear structural systems subject to stochastic Gaussian loading. The sensitivity measure considered is the partial derivative of the probability with respect to parameters that affect the structural response, such as dimensions of structural elements. The actual calculation of the sensitivity demands solving high dimensional integrals over hypersurfaces, which can be challenging from a numerical viewpoint. Hence, sensitivity evaluation is cast within the context of a reliability analysis that is conducted with Directional Importance Sampling. In this way, the sought sensitivity is obtained as a byproduct of the calculation of the failure probability, where the post-processing step demands performing a sensitivity analysis of the unit impulse response functions of the structure. Thus, the sensitivity is calculated using sampling by means of an estimator, whose precision can be quantified in terms of its standard deviation. Numerical examples involving both small- and large-scale structural models illustrate the procedure for probability sensitivity estimation.