Performance optimization of uncertain and dynamic high-dimensional wind-excited systems

Abstract

This paper focuses on the development of an efficient design optimization framework for wind-excited systems that is capable of handling not only high-dimensional and complex probability spaces, but also high-dimensional spaces of design parameters. Data-driven simulation models are utilized in assessing the system-level probabilistic measures. To efficiently solve the performance-based design optimization problem, a framework is proposed that is based on approximately decoupling the stochastic simulation from the optimization process. Local approximation models, constructed from results of a single stochastic simulation, are used to define a deterministic composite function that relates the design parameters to the system-level performance metrics. The explicit nature of this relationship is then exploited to define a sequence of deterministic optimization sub-problems that yield solutions to the original stochastic optimization problem. To illustrate the applicability of the proposed approach, a large-scale building system is optimized under stochastic wind tunnel-informed excitations and subject to system-level loss constraints.