Computational simulation is a ubiquitous tool in engineering. Further, the explosion of computational capabilities over the last several decades has resulted in the use of computational models of unprecedented complexity to make critical design and operation decisions. One potential benefit should be to improve reliability of the engineered system while reducing margins, due to the more accurate predictions such models could produce. However, realizing this benefit requires that the models employed be carefully validated, and the simulations rigorously verified. The Center for Predictive Engineering and COmputation Sciences (PECOS) at the University of Texas at Austin is developing tools and methodologies for verification and validation in the presence of uncertainty for such simulations. Among the developments being pursued are tools and techniques for code and solution verification, calibration and validation of physics- based models using uncertain data, characterizing the uncertainties in such data, representing uncertainty due to model inadequacy and validating predictions of unobserved quantities. Central to model validation is accounting for uncertainty, and at PECOS, it is represented using Bayesian probabilities, with calibration and validation processes formulated in terms of Bayesian inference.
Researchers should cite this work as follows:
Birck Nanotechnology Building, Room 2001, Purdue University, West Lafayette, IN