Over the past several years, an acknowledged need to manage uncertainty has grown in several fields for reasons including price volatility in commodities, trends in major economies around the world, regulatory actions, and climate change. The academic study of uncertainty spans several centuries and many disciplines. The authors survey this body of literature from the perspective of engineering design and find that despite the mathematical richness and breadth of research activity, the opportunity exists to develop a design method to aid engineering decision makers as they strive to incorporate uncertainty into thir decision-making processes. Three fields of study stand out as a basis for this work: Design For Manufacturability for its structured design methods, Decision Analysis for its transparency of application, and Information Visualization for its potential to improve the intuitiveness of probabilistic quantities. This oportunity is discussed by revisiting the sources of uncertainty and building on the well-accepted distinction between inherent irreduceable random system behavior (aleatory uncertainty) and imperfect human understanding and characterization (epistemic uncertainty). Together with these two sources of uncertainty, the authors consider two activities intimately associated with uncertainty: measuring and predicting. All together, these two sources of uncertainty and these two activities constitute the basic building blocks of the scientific method. In studying uncertainty, if it has been useful to distinguish between the sources of uncertainty, the authors argue it is also useful to distinguish between the activities. From this perspective, it becomes aparent that while distinguishing aleatory from epistemic uncertainty is not always possible, distinguishing prediction from measurement can be quite simple, especially when the events of concern occur in the future. In addition, while measurement uncertainty has received the majority of research attention to date, it is prediction uncertainty that remains the larger challenge for engineering decision makers to characterize, communicate, and manage. For these reasons, the authors suggest the need to develop a structured method to incorporate prediction uncertainty into a product development decision process.
展开▼