Limit-State Function Sensitivity under Epistemic Uncertainty: A Convex Model Approach

Abstract

This work proposes a limit-state sensitivity index to identify the input variables of a structure or system which possess a significant impact on its state for the case where the input variables are subject to epistemic uncertainty. By introducing the concept of a nonprobabilistic limit-state measure, the proposed sensitivity index can represent the individual or joint influence of the input parameters. The proposed sensitivity index is applicable in conjunction with different convex set models, such as the hyperrectangular or hyperellipsoidal models, as well as hybrid models. The basic properties of the sensitivity index are discussed in detail and its numerical estimation form is carried out. Two test examples are presented to prove efficiency, and a comparison with two existing sensitivity indices is also performed. Finally, the proposed sensitivity index is applied to the sensitivity analysis of a composite radome structure to quantify the influence of interval variables on the maximum displacement and total strain energy.