Data-Driven Approach to Reliability-Based Design Optimization by Using Order Statistics

Abstract

This paper deals with reliability-based design optimization of structures considering uncertainty in input distribution, where the structure is guaranteed, with the specified confidence level, to have the target reliability. A sufficient condition for this confidence-level constraint is given by using the principal component analysis and the order statistics applied to a given random samples of uncertain parameters. Concrete formulations of this approach are presented for the compliance optimization problem of trusses. The obtained optimization problem that conservatively approximates the original reliability-based design optimization problem with the specified confidence level is solved with a robust optimization approach.