Integrated System Design Using Bayesian Belief Networks

Abstract

The use of exact inference discrete Bayesian Belief Networks (BBNs) can be a powerful decision-making tool when considering the conceptual design of complex systems. A design method is demonstrated which applies BBN’s to a conceptual design of a complex system. The notional example system is modeled after a direct energy weapon since it provides an excellent example of the interactions between thermal and electrical management and distribution systems. Assessing interactions and trades between subsystems in of vital importance when attempting to reduce uncertainty and risk early in the design process. It is shown how BBN’s can be used to reduce risk in the conceptual design phase by assessing the impact of uncertainty on the interaction of the various subsystems. Several scenarios are examined which demonstrate the usefulness of such a method. Emphasis is placed on how uncertainty design choices early on in the design phase may have large impacts later in the design cycle. Finally a simulated walkthrough is performed in multiple phases that show how the method may be applied in a design iteration between the system level and subsystem designer.