PHYSICAL CONSISTENCY IN UNCERTAINTY BASED DESIGN OPTIMIZATION

Resumo

A previous work demonstrated that standard probabilistic robust optimization (that takes into
account a weighted sum of the expected value and the standard deviation) may become non-consistent
from the physical point of view, when too much weight is given to the standard deviation. If this occurs,
the optimum designs may have no meaning from the physical/engineering point of view. An alternative

probabilistic robust optimization approach was then proposed, that is able to ensure physical consis-
tency of the problem a priori (i.e. before the optimization procedure is started). In this work, physical

consistency of reliability-based design optimization (RBDO) and worst-case robust optimization is in-
vestigated. We show that the RBDO and the worst-case problems are consistent under milder conditions

than standard probabilistic robust optimization. This indicates that RBDO and worst-case optimization
are inherently more consistent than probabilistic robust optimization, at least from the physical point of
view.