Results from an International MHD Data Mining Collaboration

New data mining techniques have been successfully applied to MHD data on H-1, TJ-II and Heliotron-J, and are being implemented on LHD and W-7AS data. The motivation for automated mining of fusion databases is to distil and classify data for inclusion in fusion physics databases, and to highlight physically-interesting, previously unnoticed modes. We present results from data mining of more than 10,000 shots from H-1, TJ-II and Heliotron J, showing a range of Alfv\'{e}nic and non-Alfv\'{e}nic modes, many with well-defined poloidal mode structure and clear relation to heating configuration and plasma geometry. In the case of H-1, the dispersion relations for several of these modes have been examined in detail exploiting H-1's high resolution in rotational transform. Examples of use of this relation to provide information about rotational transform (a form of Alfven spectroscopy) are given. We also discuss possible real-time application of the cluster technique to preliminary mode identification as data is being acquired, and some initial work on application of image processing techniques to MHD spectrogram analysis.