Using AMNS data within an Integrated Tokamak Modelling Environment

Transcription

1 Using AMNS data within an Integrated Tokamak Modelling Environment Presented by: David Coster (AMNS Coordinator, IMP3 Leader, Deputy Task Force Leader) TF Leader : G. Falchetto Deputies: R. Coelho, D. P. Coster EFDA CSU Contact Person: D. Kalupin

2 Background The EFDA Task Force on Integrated Tokamak Modelling coordinates the development of a coherent set of validated simulation tools benchmarks these tools on existing tokamak experiments provides a comprehensive simulation package for ITER and DEMO plasmas 2

3 ITM use of AMNS data AMNS data should be centralized and managed Version control of data imported to the ITM-TF data base is mandatory. The provenance of the data must be accurate and stored in the ITM database For production runs with ITM-TF codes using AMNS data it is important that the data have been given a stamp of approval by an expert. 3

4 ITM use of AMNS data The data should be comprehensive, ubiquitous and easily used This means identifying what data is needed not always easy! The AMNS data must be communicated to ITM- TF codes via a standardised interface this should also ensure coherence between different ITM-TF codes needing the same type of data All AMNS data used by codes should be available through the AMNS data interface no back doors 4

5 Approach in the ITM Atomic Data Data selection Data import Data storage Data transport User interface design Supporting multiple languages Physics Code Molecular Data Nuclear Data AMNS Database AMNS Library Physics Code Physics Code Surface Data Physics Code 5

6 Driven by user codes requirements Data selection fully differential cross sections for all relevant processes and species don t really want to resolve all excited states so some sort of effective cross-section might be useful» bundling data availability imported all of the ADAS ADF11 rate coefficients 6

7 Data storage Need to have a data storage mechanism that allows us to store all of the AMNS data that are needed allows us to store all of the additional meta-data associated with the AMNS data allows us to effectively and efficiently use the data 7

8 Data import Need to store the data from outside sources in the AMNS database external data is available in a wide variety of formats is not always of the same quality need to capture the meta-data about the imported data source of the data import procedure quality estimate 8

9 Data transport Need to make the AMNS data available to ITM codes that might be running in a distributed environment remote data access is supported by the ITM s UAL (Universal Access Layer) for CPOs (Consistent Physical Objects) could also create local copies of the database which can pose a problem of data consistency also means that we want to limit the number of files SOLPS uses a subset from > 1400 files, > 180 MB 9

10 Making data available Need to make the AMNS data available through a standardized interface allows the data storage mechanism to change without affecting user codes (hopefully) ensures that the data are correctly used in the ITM codes allows users codes access to newer versions of the data when such data becomes available without the user having to re-code the data access routines 10

11 Making data available, II Challenges: within the ITM we have codes written in Fortran C C++ Matlab Python many codes will also be used outside of the ITM environment and code developers are not that keen on having to support one version within the ITM and one for use outside of the ITM 11

12 ITM approach AMNS task under the ITM leadership Developed a data format for data storage actively evolving as we learn more about what data is needed and discover that it doesn t fit Developed a data access paradigm implemented this paradigm in a data access library in Fortran have C bindings for this library are working on a Python interface 12

13 Details: data storage ITM uses CPOs (Consistent Physical Objects) to store data experimental data (from present tokamaks) simulation data machine descriptions AMNS data Each CPO is described by an XML Schema automatic tools then provide data access routines in Fortran, C++, Matlab, Python, Java automatic tools provide for the storage, retrieval and transport of CPO data backends are (currently) MDSplus and HDF5; with add-ons ASCII data is also supported 13

14 AMNS CPO ITM CPOs are identified by USERNAME / DEVICE / DATA_VERSION / SHOT_NUMBER / RUN_NUMBER For the AMNS system we have chosen to use USERNAME is the actual user, or if the user has no data, then amnsdata DEVICE is amns DATA_VERSION is managed centrally and is changed when a new version of the CPO data structure is released (currently 4.10a ) SHOT_NUMBER is derived from the heaviest species involved in a process and equals ZN + ZM*1000 (ZN is the nuclear charge, ZM is the number of nucleons if the reaction depends on the isotopen and 0 otherwise) RUN_NUMBER represents the AMNS version number of the data Doing things this way allows us to leverage the already existing ITM infrastructure and to benefit from future developments 14

15 AMNS CPO Design concept is still evolving Want to support data provided as table R(x 1. x 2,.., x n ) linear interpolation at the moment (1d, 2d, 3d, ) for R or log(r) for x i or log(x i ) data stored in the AMNS database formula R(x 1. x 2,.., x n ) hopefully the formula has the right asymptotic behaviour! index to a predefined function stored in the database, together with any coefficients 15

16 AMNS CPO, II 16

17 Implementation Physics code Access to AMNS data only via interface initialization (2) finalization (2) querying parameters (2) setting parameters (2) getting data (1) Separation between use of the data and the implementation of the data Code author doesn t need to become an expert in AMNS Ensures compatibility between codes AMNS implementation Only accessed by a set of defined calls Implementation by AMNS experts Different versions can be supported Different implementations possible Analytic formulae Table lookup Old versions should always be recoverable (even if wrong) Should become easier to implement new data 17

18 A prototype has been implemented As a F90 module using derived types Interface will handle error estimates in the AMNS data! call ITM_AMNS_SETUP(amns) query%string='version' call ITM_AMNS_QUERY(amns,query,answer)... call ITM_AMNS_SETUP_TABLE(amns, lr_rx, species_lr, amns_lr) query%string='source' call ITM_AMNS_QUERY_TABLE(amns_lr,query,answer)... set%string='nowarn' call ITM_AMNS_SET_TABLE(amns_lr,set)... call ITM_AMNS_RX(amns_lr,rate(:,:),ne(:,:),te(:,:))... call ITM_AMNS_FINISH_TABLE(amns_lr)) call ITM_AMNS_FINISH(amns) An example 18

19 Some examples Electron impact ionization (ADAS ADF11 SCD) 19

20 Some examples D-D fusion cross section (Bosch-Hale formula) 20

21 Some examples Differential elastic cross section for W (at 90 o ) (Tokesi, preliminary data) 21

22 Coronal example: W 22

23 Coronal example: W radiation No guarantees about the correctness of the data used, or for the program that calculated these results It was done to demonstrate how the ITM AMNS system is used. 23

24 Documentation Entries in the ITM web site derived from the AMNS database. 24

25 A sample entry Documentation 25

26 Data needs are still evolving ITM has Some thoughts data consumers» data needs, not always clearly formulated data providers data packagers» format of the AMNS CPO» data import» AMNS access library Not always properly synchronized 26

27 Some thoughts Having an IAEA recommended data set will make the ITM-AMNS job of selecting data easier If the IAEA data is in a standard format, then importing the data will also be easier If the IAEA data set has the provenance information standardized, then this will also make it easier when importing the data Decisions on how to categorize and store the data in the IAEA database might affect the organization within the ITM The ITM is likely, though, to have requirements that the IAEA database doesn t 27