Robust design and optimization of stochastic wind-excited systems: an adaptive kriging-based approach

Abstract

This research proposes a robust design framework for wind-excited systems in which performance is estimated at a system level in terms of state-of-the-art performance-based design metrics. In particular, the robust design problem is formulated as a stochastic optimization with objective the minimization of the variance of the performance metric. Constraints are also imposed on the initial cost of the system and expected value of the performance metric. To effectively treat the performance metrics within the optimization problem, adaptive kriging models of the deagreggated loss metrics are defined in terms of the second order statistics of the demands. By then relating the demand statistics to the design variables through the concept of the Auxiliary Variable Vector, a deterministic optimization sub-problem is defined that can handle high-dimensional design variable vectors and general stochastic excitation. By solving a sequence of sub-problems, each formulated in the solution of the previous, solutions to the original robust design problem are found. A case study consisting in a large-scale system subject to stochastic wind excitation is used to illustrate the applicability of the proposed framework.