A robust-design framework that seeks parameter choices satisfying system requirements across a set of scenarios drawn directly from data (experimental or synthetic realizations of uncertain parameters or operating conditions), rather than from an assumed probability model. Robustness to measurement error and to overfitting the finite dataset is obtained by also requiring satisfaction over perturbed versions of each scenario, and a tunable relaxation trades a lower objective value against less robustness by optimally discarding a chosen number of outlier scenarios and permitting perturbed scenarios to violate the requirements with a small acceptable probability, spanning risk-averse and risk-agnostic formulations. The design of an aeroelastic wing serves as the illustration.
This paper proposes risk-averse and risk-agnostic formulations to robust design in which solutions that satisfy the system requirements for a set of scenarios are pursued. These scenarios, which corr…