The MHD Documentation Database (MDDB)
An overview of the MDDB project is given in a CWGM7 presentation, available here.
Aim: provide a reference documentation database of MHD modes observed in all major stellarators.
Useful for: finding parameterisation of modes (e.g. Alfvén scaling) across many shots, also for cross-machine comparisons
How: The MHD documentation database (MDDB) will be generated using automated techniques scanning 10,000s of shots.
Data: choose Mirnov channels -- arrays on all devices, an "always-on" diagnostics, likely to have consistent configuration (sampling, gain, etc) across many shots.
Since 2007, data mining results have been obtained from H- 1, Heliotron J, TJ-II, W7AS and LHD. (also JT-60). In this time our methods have improved, however this means that results from each device are somewhat ad hoc. Different code versions, different pre-processing filters, different clustering algorithms. Code had limitations: both in how easily it could be extended, and code complexity became barrier to new users In order to make pyfusion code usable for collaborators, and to allow a systematic treatment across all devices, significant refactoring was required
will document only those modes which we can treat rigorously with our current methods
- strong, coherent modes
- no standing waves
- defined solely by Mirnov phase structure (not amplitude)
- time-independent treatment (~0.5 ms samples)
- probabilistic phase structure representation (von Mises distribution)
- single set of Mirnov coils (skip shots with missing coils)
- SVD-based mode detection, correlation threshold SV grouping
- include missing coils, and newer coils (e.g. 16 3-axis helical array in H-1)
- lower energy threshold for modes
- broadband fluctuations
- time dependence
MDDB 1.0 structure
The MDDB 1.0 schema consists of five tables:
The mode table
|Mode ID||Channel A||Channel B||Delta phase: mu||Delta phase: gamma|
The documentation table
The instance table
|Instance ID||Shot||Time (0.5ms interval)||Frequency||Signal strength|
The mapping table
|Instance ID||Mode ID||Likelihood|
The supplementary instance table (possibly sparse)
Target dataset for MDDB 1.0
|Mirnov coils||20 coil poloidal array||14 coil poloidal array||6 coil toroidal array, 15 coil helical array||15 coil poloidal array||16 coil poloidal array|
|Sample rate||1 MHz||1 MHz||1 MHz||312.5 kHz||333 kHz|
Related pages: Pyfusion code used in generating the MDDB