Reliability based design optimization (RBDO) is necessary for the design of complex engineering systems with quantified levels of risk. To achieve failure probabilities in the order of 10-3 in large systems such as space vehicles and aerospace structures typical component level failure probabilities have to be of the order of 10-4 to 10-7. Quantification of risk using reliability based approaches relies on availability of statistics of the random variables that affect the response. In engineering applications where extensive tests cannot be performed, we introduce uncertainty into characterization of the random variables. The uncertainty is introduced from the use of small sample sizes to estimate the distribution parameters (mean and variance â?" location and scale) of a chosen distribution function; uncertainty could also arise from the selection of an incorrect probability distribution function (PDF). Including data uncertainty

introduces yet another level of complexity to the already computationally expensive problem of RBDO. Inverse reliability measures have been proposed and used in RBDO to minimize computational effort. This poster presents methods to obtain confidence bounds for the inverse reliability measure, namely the probabilistic sufficiency factor, resulting from data uncertainty. The estimated confidence bounds are them approximated using a response surface for use in optimization. A bi-objective optimization is performed to achieve a design with the greatest reliability and lowest sensitivity to data uncertainty.