You are here

BISON Validation

January 29, 2013 - 11:54am

Addthis

BISON Validation

Predictive Maturity

Work continued on the previously developed discovery, accumulation, and assessment (DAA) process to plan, track, assess, and communicate VU activities and results. DAA was applied to the BISON sensitivity analysis described above, and the results were exported to Synopsis, the DAA management tool. [SNL, LANL, INL]

Building on previous sensitivity studies of the LIFE-IV nuclear fuels code, a recently completed VU study focused on a methodology by which experimental campaigns may be devised to improve code calibration. Specifically, a principal component analysis is performed on the input parameters of the experiments, and the experiments that offer the least residual error when reconstructed from the principal components are proposed as those that should be conducted first. In this way, the experiments are ranked according to their ability to "cover" the input design space, or in this case, a low-dimensional representation of the input design space.

A comparison of two sequences examined in this study is provided in Fig. 9, which shows the predictive maturity index (PMI) approaching an asymptotic value more quickly with the early incorporation of experiments that offer more coverage of the input design space (vis-à-vis lower reconstruction errors) and also yield better-calibrated models (insofar as stability of the model predictions is concerned). [LANL]

Figure 10 shows the BISON DAA hierarchy with color-coding to provide a snapshot of the completeness of the BISON validation effort. The placeholders used to determine completeness are specific questions or items that must be identified, such as the physical quantities measured in the experiment. For BISON, about 230 placeholders need to be filled to complete the description of the validation effort. One can see that the initial steps in validation have been taken (e.g. acquiring a description of the experiment and data), and that more planning, verification, simulation, and validation comparison work remains to be done. If this data is collected periodically, one can plot project progress over time with Synopsis.

Addthis