Risk-based treatment of uncertainty in trade space exploration : application via Monte Carlo simulation on a manned, mini-submersible model

Abstract

In design, modeling and simulation are commonly used to answer questions of interest as it is both inefficient and expensive to physically build and evaluate numerous possibilities. Any modeling effort aims to build the simplest model while capturing the real-world trends appropriately. When modeling highly complex systems or pushing technological bounds, variables in the model will possess elements of uncertainty. In a trade space approach, different design combinations may exhibit different uncertainty profiles. Omitting uncertainties in the modeling effort can bias design combinations in the overall trade space in terms of capability and cost as well as misrepresent the value of tradeoffs between designs. Therefore, if the uncertainties are not represented, the decision-maker is accepting an unknown level of risk when selecting a design. This thesis proposes that uncertainty in early stage design is not well represented, despite its playing a major role in a system's ultimate success. This research explicitly accounts for uncertainty in model inputs via probability distributions instead of simply applying "best estimate" deterministic values. These distributions are sampled via Monte Carlo simulation to generate uncertainty profiles for different design combinations, thereby increasing the validity of the model outputs. This approach for capturing the implications of uncertainty in early stage design allows for a more accurate representation of design risk. Ultimately, the deterministic design points in the trade space are quantitatively and qualitatively evaluated against the design points incorporating uncertainty. Understanding that model outputs can only ever be as good as model inputs, the exploration of the effect of uncertainty on the design trade space is important. An example of Trade Space Exploration for the conceptual design of a manned, mini-submersible is used to demonstrate an approach for quantifying and visualizing uncertainty to inform decision-making. This case study suggests that visualizing risk at the system level in a typical performance versus cost context is valuable.