A sampling technique enhancing accuracy and efficiency of metamodel-based RBDO: Constraint boundary sampling

Abstract

Reliability-based design optimization (RBDO) dealing with variation of output induced by uncertainty of design variables needs computationally expensive reliability analysis to calculate failure probability. Metamodel-based RBDO is one of emerging techniques used to overcome computational drawback. In this research, constraint boundary sampling is proposed to build metamodel that can predict optimum point accurately while satisfying constraints. Constraint boundary sampling is sequentially to locate sample points along constraint boundary by using kriging model and its mean squared error. Metamodel-based RBDOs with constraint boundary sampling are compared with that with conventional space-filling sampling.