Dimensional reduction applied to the reliability-based robust design optimization of composite structures

Abstract

The need to quantify the uncertainty associated with the main design features and properties of composite laminates has been recognized by the scientific and industrial communities. Recently, robustness and reliability assessment have been combined in a structural design framework called Reliability-based Robust Design Optimization (RBRDO). However, reliability assessment is known to aggravate significantly the efficiency of evolutionary algorithms, in structural design optimization. The problem becomes particularly difficult in the design optimization of composite laminate structures. In this paper, it is proposed the application of an analytical dimensional reduction technique of the uncertainty space associated with reliability assessment, based on the approximate local solution of Sobol’ indices. This approach has a negligible computational cost. It is demonstrated that several random mechanical properties of the optimal design solutions are not important for reliability assessment and can be frozen. The proposed methodology allows the RBRDO problem to be solved much more efficiently.