Challenges in the Development of Uncertainty Aware, Multi-Phase Equations of State

The rise of simulation as a development tool for high-consequence

systems makes it imperative to understand the uncertainties inherent

in simulation results, to provide the highest levels of confidence to

decision makers. The material equation of state (EOS) is a key factor

in determining simulation accuracy. Various techniques have been

developed for propagating the uncertainty of data used to calibrate

EOS models into simulation results. All these share commonalities in

that they typically evaluate some analytic EOS model at numerous

locations in the parameter space. Single phase EOS models are often

simple enough that this is a straightforward process. However, as EOS

models and data become more complicated, particularly as multiple

material phases are introduced, a number of difficult challenges

arise. Here we examine two of these challenges, ensuring the stability

of a desired phase diagram and fully incorporating uncertainty from

experimental sources. Using the framework of Bayesian inference, a

method for constraining multi-phase models to an arbitrary phase

diagram via priors will be presented. Additionally, the full

uncertainty in data, such as from shock measurements, will be

incorporated through the error-in-variables technique. SNL is managed

and operated by NTESS under DOE NNSA contract DE-NA0003525.