(Georgia Institute of Technology, 2006-01)
Duncan, Scott Joseph; Paredis, Christiaan J. J.; Bras, Berdinus A.
Often in the early stages of the engineering design process, a decision maker lacks the information needed to represent uncertainty in the input parameters of a performance model. In one particular form of severely deficient information, a nominal estimate is available for an input parameter, but the amount of discrepancy between that estimate and the parameter’s true value, as well as the implications of that discrepancy on system performance, are not known. In this paper, the concepts and techniques of information-gap decision theory (IGDT), an established method for making decisions robust to severely deficient information, are examined more closely through application to a design problem with continuous design variables. The uncertain variables in the chosen example problem are parameters of a probability distribution, so the relationship between IGDT and design approaches considering precise and/or imprecise probabilities is explained. Insight gained from a walkthrough of the design example is used to suggest the types of problems an IGDT approach will or will not effectively solve as well as potential limitations that could be encountered when solving more complex problems.
(Georgia Institute of Technology, 2006)
Bras, Berdinus A.; Paredis, Christiaan J. J.
When making design decisions in environmentally benign design and manufacture, the decision maker is often faced with extreme uncertainty. Due to a lack of understanding of the complex dynamics of environmental and societal systems, it is very difficult to judge the impact different design alternatives have on the environment, the economy and the society, especially in the distant future.
In this paper, two formalisms are illustrated for making design decisions under extreme uncertainty. The formalisms are probability bounds analysis and info-gap decision theory. We introduce the basic concepts for both formalisms, discuss the advantages and limitations, and identify under which circumstances they are useful in the context of design decision making. One can think of both decision methods as having a built-in sensitivity analysis allowing the decision maker to judge whether a decision can be made confidently based on the current information, or whether additional information needs to be gathered.