Simulation codes are everywhere. They are used to design things like automobiles, airplanes, bridges, chemical plants, and computers. Simulation codes are used in engineering to predict the behavior of systems that are being designed or to understand existing systems. With the increasing reliance on simulation codes, it is becoming critically important to determine how well they predict actual physical phenomena. Uncertainty in simulation code predictions has many sources, including the lack of knowledge of the underlying material models, the variability of the initial geometry and materials, and the degree of variability in the physical phenomenon itself.
In this presentation I will concentrate on the first issue by presenting a framework within which to handle uncertainties in the parameters associated with the material models on which a simulation code is based. The goal is to determine how these uncertainties affect our ability to predict the behavior of a physical system. This framework may accommodate the other degradations to predictability and may provide a worthwhile foundation for the verification and validation (V&V) of simulation codes.
A probabilistic methodology is presented for assessing the uncertainties in simulation predictions that arise from uncertain model parameters. The models may represent, for example, aspects of material behavior. The model parameters are often derived from uncertain measurements. A probabilistic network is proposed that helps both conceptualize and computationally implement an analysis of experiments in terms of numerous models in a logically consistent manner. This approach permits one to improve one's knowledge about the underlying models at every level of the hierarchy of validation experiments.
Keywords: simulation code validation, simulation uncertainty, uncertainty assessment, probabilistic network
Viewgraphs for this talk (pdf, 341 KB)
Return to publication list
Send e-mail to author at kmh@hansonhub.com