B2–B2.5 Code Benchmarking
B2–B2.5 Code Benchmarking
ITER-IO currently (and since about 15 years) employs the SOLPS4.xxx code for its divertor design, currently version SOLPS4.3. SOLPS.xxx is a special variant of the B2-EIRENE code, which was originally developed by an European consortium (FZ Jülich, AEA Culham, ERM Belgium/KU Leuven) in the late eigh...
Saved in:
Personal Name(s): | Dekeyser, W. |
---|---|
Baelmans, M. / Voskoboynikov, S. / Rozhansky, V. / Reiter, D. / Wiesen, S. / Kotov, V. / Börner, P. | |
Contributing Institute: |
Plasmaphysik; IEK-4 |
Imprint: |
Jülich
Forschungszentrum Jülich GmbH Zentralbibliothek, Verlag
2011
|
Physical Description: |
III, 52 S. |
Document Type: |
Report |
Series Title: |
Berichte des Forschungszentrums Jülich
4337 |
Link: |
OpenAccess |
Publikationsportal JuSER |
ITER-IO currently (and since about 15 years) employs the SOLPS4.xxx code for its divertor design, currently version SOLPS4.3. SOLPS.xxx is a special variant of the B2-EIRENE code, which was originally developed by an European consortium (FZ Jülich, AEA Culham, ERM Belgium/KU Leuven) in the late eighties and early nineties of the last century under NET contracts. The particular code version SOLPS4.3 is at present jointly maintained and upgraded at ITER-IO and FZ Jülich. Other versions of B2-EIRENE (and SOLPS.xxx) have been advanced by other research groups, notably at IPP Garching, in various different directions. Most important amongst those for our present work is the SOLPS5.2 version with the probably most advanced model available today for classical drifts, currents and electric fields in the plasma fluid module, developed by V. Rozhansky and co-workers (St. Petersburg State Polytechnical University). Surprisingly, merging the various versions of B2-EIRENE into one single, all-including code package requires significant effort. Until today even the very similar edge plasma codes within the SOLPS family, if run on a seemingly identical choice of physical parameters, still sometimes disagree significantly with each other. It is obvious that in computational engineering applications, as they are carried out for the various ITER divertor aspects with SOLPS4.3 for more than a decade now, any transition from one to another code must be fully backward compatible, or, at least, the origin of differences in the results must be identified and fully understood quantitatively. No computational engineer can allow his ability to solve an urgent design problem be limited by somebody else’s decision to change a code. Already in the past significant effort went into code-code benchmarking, even within the SOLPS family, but sometimes with limited success. We believe that one of the shortcomings of all previous code-code benchmarks was the application of the entire SOLPS model to highly integrated, multi-physics cases, i.e., to much too complex test problems, which has prevented the community from identifying all the details of code version differences on a fully quantitative level. And there are a number of ways by which different numerical results on the same problem can result. The most important of which probably are: • unintentional choice of different code input options • numerical effects, such as grid size, or convergence level • not fully documented internal code changes, which affect the underlying physical model The first and last point is a matter of code documentation and code operation. The second issue is unavoidable, but the effects can be quantified, e.g. by running physically identical cases on different grid sizes, or by measures of convergence levels, such as residuals (see Section 3.3 of the present report). In this report we document efforts undertaken in 2010 to ultimately eliminate the third issue. For the kinetic EIRENE part within SOLPS this backward compatibility (back until 1996) was basically achieved (V. Kotov, 2004-2006) and SOLPS4.3 is now essentially up to date with the current EIRENE master maintained at FZ Juelich [1]. In order to achieve a similar level of reproducibility for the plasma fluid (B2, B2.5) part, we follow a similar strategy, which is quite distinct from the previous SOLPS benchmark attempts: the codes are “disintegrated” and pieces of it are run on smallest (i.e. simplest) problems. Only after full quantitative understanding is achieved, the code model is enlarged, integrated, piece by piece again, until, hopefully, a fully backward compatible B2 / B2.5 ITER edge plasma simulation will be achieved. The status of this code dis-integration effort and its findings until now (Nov. 2010) are documented in the present technical note. This work was initiated in a small workshop by the three partner teams of KU Leuven, St. Petersburg State PU and FZ Jülich, held at FZ Jülich in February 2010, and it has been carried out as a preparatory step for a wider SOLPS re-unification activity foreseen in the coming year under a F4E (Fusion for Energy, Barcelona) ITER service contract. The material of this report (with exception of appendix D) was made available to ITER IO Cadarache in November 2010, and, slightly later, presented at IPP Garching at a wider “SOLPS meeting" by one of the authors: S. Voskoboynikov. |