02.08.2013 Views

Research Needs for Magnetic Fusion Energy Sciences - US Burning ...

Research Needs for Magnetic Fusion Energy Sciences - US Burning ...

Research Needs for Magnetic Fusion Energy Sciences - US Burning ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

maintaining multiple codes, which take different approaches to solving the same problem. it is<br />

crucial to verify a code be<strong>for</strong>e beginning tests against experimental results — however, it is not<br />

necessary or desirable to wait until each model is “complete” be<strong>for</strong>e proceeding. For fusion problems,<br />

no model will ever really be complete and testing of well-verified, though imperfect, models<br />

can help guide code development by focusing on areas of disagreement.<br />

The next element in this Thrust is the design and execution of validation experiments to assess<br />

the degree to which a calculation describes the real world. The essential activity is close and careful<br />

comparison between model output and experimental measurements, with an emphasis on<br />

quantitative measures and attention to errors and uncertainties in both code and experiment<br />

(see, <strong>for</strong> example, P.W. Terry et al, PoP 15, 062503, 2008). it is a physical problem, meant to build<br />

confidence in the models and one without a clearly defined endpoint. That is, validation is not a<br />

one-time test where a code is “approved” <strong>for</strong> all time if it passes or discarded if it fails, but is instead<br />

part of an iterative process <strong>for</strong> improving the fidelity of the models. identifying the cause<br />

of discrepancies will be among the most difficult of the challenges posed by this Thrust, but will<br />

also be among the most scientifically rewarding. For some fusion science problems, which require<br />

extrapolation past currently accessible regimes, we will need to infer the correctness of the<br />

underlying physical model to a higher degree than if we were only interpolating between accessible<br />

data points. The relation between the various processes can be illustrated in Figure 1. The<br />

first step in the validation ef<strong>for</strong>t, <strong>for</strong> each case study, will be the identification of critical physics<br />

issues requiring testing with due consideration to the uniqueness and sensitivity of particular<br />

measurements and model predictions. defining measurement needs is critical and will lead<br />

to requirements <strong>for</strong> innovative diagnostics to be developed and deployed as part of this Thrust.<br />

it is highly desirable to compare predictions at various levels of integration, sometimes called a<br />

“primacy hierarch.” For example, to validate a calculation of turbulent transport, one will want<br />

to compare fluctuation levels and spectra, correlations, fluxes and profiles. Thus, the diagnostic<br />

challenge will be significant and require substantial resources and attention.<br />

Increasing<br />

Realism in<br />

Physics and<br />

Geometry<br />

Coupling of<br />

Physics<br />

Complexity<br />

Complete<br />

Systems<br />

Subsystem<br />

Cases<br />

Unit<br />

Problems<br />

Figure 2. Validation hierarchy. Model testing is most effectively carried out by employing a set of experiments<br />

which probe the model at various levels of physical and geometric complexity.<br />

281<br />

Decreasing<br />

Number of<br />

Code Runs<br />

Quantity and<br />

Quality of Data<br />

Info on Initial<br />

and Boundary<br />

Conditions

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!