Design Decisions Under Conditions Of Uncertainty And Risk

Principal Investigators
Margaret Wiecek, James Reneke, Georges Fadel (Clemson U.)

Industry
Peter Fenyes (General Motors)

Government
Mark Brudnak (TARDEC)

The problem of rational selection from among competing alternatives has been identified as central to application areas as diverse as engineering design, management, and finance. The decisions are recognized as difficult because of multiple selection criteria which may depend on several disciplines and a decision environment containing elements of uncertainty and risk. Our long-term objective is to develop a methodology for making design decisions under uncertainty and risk and introduce them into the ARC context as aids to the decision makers which can be implemented quickly and require limited computational resources. Our short-term objective is to perform a preliminary study justifying the need for that methodology and revealing special features of the proposed approach which are not offered by other methodologies for decision-based design.

This preliminary study will clarify fundamental issues encountered when making engineering design decisions under conditions of uncertainty and risk, and will establish the need for a general methodology to evaluate alternatives with respect to multiple, conflicting and interacting performance criteria. Such situations arise in many areas of human activity such as business, management, and engineering, and across government and corporate sectors. For example, in the engineering design context, evaluation of alternatives encompasses selection of design solutions, proposals, suppliers, materials, etc., and requires decision-making under uncertainty and risk. Further, assessing system performance may require testing the physical artifact or simulation of system models. These efforts, if undertaken in support of decision making, should take place within a framework provided by decision models in order to assure a comprehensive and least cost review. The proposed research is the first step in developing a family of models that provide such a framework.