Increased computing power has enabled designers to efficiently perform robust design analyses of engine systems. Traditional, filtered Monte Carlo methods involve creating surrogate model representations of a physics-based model in order to rapidly generate tens of thousands of model responses as design and technology input parameters are randomly varied within user-defined distributions. The downside to this approach is that the designer is often faced with a large design space, requiring significant postprocessing to arrive at probabilities of meeting design requirements. This research enhances the traditional, filtered Monte Carlo robust design approach by regressing surrogate responses of joint confidence intervals for metric responses of interest. Fitting surrogate responses of probabilistic confidence intervals rather than the raw response data changes the problem the engineer is able to answer. Using the new approach, the question can be better phrased in terms of the probability of meeting certain requirements. A more traditional approach does not have the ability to include confidence in the process without significant postprocessing. The process is demonstrated using a turboshaft engine modeled using the numerical propulsion system simulation (NPSS) program. The new robust design process enables the designer to account for probabilistic impacts of both technology and design variables, resulting in the selection of an engine cycle that is robust to requirements and technology uncertainty.