Robustness-Based Design Optimization of Multidisciplinary System Under Epistemic Uncertainty

Abstract

This paper proposes formulations and algorithms for design optimization of multidisciplinary systems under both aleatory uncertainty (i.e., natural or physical variability) and epistemic uncertainty (due to sparse or imprecise information) from the perspective of system robustness. The availability of sparse and interval data regarding input or design random variables introduces uncertainty about their probability distribution type and distribution parameters. A single-loop approach is developed for the design optimization, which does not require any coupled multidisciplinary uncertainty propagation analysis. Thus, the computational complexity and cost involved in estimating the mean and variation of the objective and constraints are greatly reduced. A decoupled approach is used to unnest the robustness-based design from the analysis of nondesign epistemic variables to achieve further computational efficiency. The proposed methods are illustrated for a mathematical problem and a practical engineering problem (fire-detection satellite), where the information on the random inputs is only available as sparse point data and/or interval data.