Projects (V: Vacation, H: Honours, MPh: M.Phil, PhD: PhD)

Application of Data-Mining Techniques to Plasma Phenomena and Data Quality Assessment for H-1: (V,H,MPh,PhD)
Recent advances in databases and intelligent algorithms have enabled machine-based "trawling" of very large databases, to find both prescribed or already known phenomena, and by looking for data that does not fit known patterns, discover new behaviour. Association rules or clustering algorithms even allow machines to discern patterns without human guidance. H-1 provides a rich variety of instability phenomena, from the drift frequency range (low kHz) to higher magnetohydrodynamic (MHD) frequencies up to ~ 1MHz, from several types of sensor array sets. These include two 20 channel magnetic probe arrays ("Mirnov"), two arrays of visible light detectors (miniature photomultiplier arrays) two "soft" X-ray arrays and spatially scanning scanning plasma density interferometers.

The "Datamining" project is a combination of development of algorithms as described above, and physical understanding of the phenomena, with a weighting to best fit the student's interest and ability.

A related, but smaller project is the Data Quality Assessment Project (V,H,MPh) . This is more of a software-oriented project, creating a set of rules and a procedure for checking the quality of the data and plasma parameters recorded each time we pulse H-1.  Our data are numerous and diverse, and expensive to acquire, so if a data set is flawed, we would like to know quickly, to save wasting resources, and avoid the need to repeat the data taking.
The set of rules or classifications that define “good” data should be developed from two directions:

    a/ experts write some simple rules, and then fine tune them on old data, and new data is at arrives.
    b/ experts classify good and bad data, pulse by pulse and machine learning techniques are used to discover new rules that define bad data.

 For both projects, the most efficient way to get started would be to use the IDL data visualization language to evaluate the required quantities - we have much existing code, and it allows quick debugging, and has great interactive graphics. This could be an alternative with a relatively short learning curve. 
Alternatively you could use Java, which is more widespread, but requires a bit more development time.  Also if you are not familiar with Java, you would have to be prepared to spend a substantial part your time learning the language. 

Python, which is in between (but  free), would be particularly useful for the data mining, as most of the original work by Dave Pretty is in Python.

 We have an extensive mysql database of shots, and David McLenaghan's Java “eScope” data viewer has simple data mining capabilities ready and working if you want a head start.  

Effect of different discharge conditions on the quality of the H-1 vacuum system(V,H,MPh):
You will use the new computer controlled residual gas analyser to compare different regimes of glow discharge conditioning, baking, and pulsed discharge cleaning in their efficiency at removing carbon and oxygen from the internal walls of the H-1 vacuum vessel.  This could be extended to analysing the visible spectral lines of full scale H-1 plasma to relate the discharge conditioning to the final improvement in plasma purity.

Analysis of Alfvén eigenmode data from H-1 plasma: (adaptable to all levels: V,H,MPh,PhD)
One of the major concerns in predictions of the plasma parameters required for next step fusion experiments (e.g. ITER)  is the possible deleterious effect of the excitation of large amplitude Alfvén eigenmodes, manifesting as magnetic field fluctuations which can cause energy transport across the confining magnetic field. Similar fluctuations are observed in high temperature H-1 plasmas.  You would analyse some of the substantial data set of frequency vs. time data ("voiceprint") looking for agreement or otherwise with a few simple theories, or for different characteristic phenomena.  We have data viewing and data mining software already, which you are encouraged to use, or you could improve this, or develop your own approach.  For Honours and Vacation students, a senior research student or Postdoc would be able to provide close support with software and analysis. There are many directions in which this work could be extended to a PhD, including sophisticated data analysis and data mining, theoretical analysis/basic plasma stability studies, or extension of the probe/detector/signal processing hardware.  

Real Time Signal Processing and Analysis for Plasma Experiments: (V,H,MPh)
A LabView-based control system for fast acquisition of vacuum magnetic surfaces (Poincare  plots) will be extended to include data acquisition and initial analysis of a 720 view x 64 channel tomographic system.  This system will display data in various raw and analysed forms during acquisition to allow the quality of data to be improved, and to feed back into the experiment to allow the scan parameters to be chosen strategically.  Using “Labview” and Research Systems Inc’s “IDL”, we expect that a measurement campaign of days can be performed in hours.

Mapping of Magnetic Field Configurations in H-1:(H,V,MPh)
The unique heliac plasma shape is generated by a carefully designed, fully three dimensional magnetic field.  The magnetic field lines can be directly mapped by launching electrons, and collecting them on a 64 wire grid array.  The data is tomographically inverted by existing algorithms.  This apparatus has been recently upgraded, and the project is to map the magnetic field for various configurations, looking at sensitivity to small changes in current, formation of magnetic islands, and the chaotic region at the plasma edge.

Soft XRay emissions from H-1: (adaptable to all levels, V,H,MPh,PhD)
Use a new 16 channel absolute extreme ultraviolet array (AXUV http://www.ird-inc.com/axuvope.html)  to obtain soft xray data from the H-1 heliac plasma.  This can be used as a “radiation bolometer” or with a thin foil, as a “soft” (200eV-1000eV) Xray detector, and should allow an estimate of temperature and impurity distributions in the plasma. At the honours level, the project would include at least a theoretical component or some detailed calculation and analysis of profiles. At the PhD level, the project would include both of the above, and some original research into the observed phenomena.

Boyd.Blackwell@anu.edu.au
x 52482  

wait several rings past the tone change to be forwarded to my mobile.