Inference Uncertainty Quantification Instead of Full-scale Testing

We give simple examples illustrating the concept and importance of inference uncertainty, which can be defined as the difference between what is measured (the observable quantity) and what is desired (the unobserved quantity). Quantification of uncertainty arising from inference has an important role to play in lieu of full-scale testing, because system-level uncertainties may not be observable by observing separate effects tests. Yet, little attention has been paid to this type of uncertainty, which is prevalent in numerous scientific and engineering applications. We propose that inference uncertainty can be mathematically characterized using different theories of uncertainty, including probability theory. A metric for the quantification of margins and uncertainties relating to factor of safety is discussed, and an example of information integration is illustrated. (Manuscript approved for unlimited, public release, LA-UR-08-1669, Unclassified.)