Prediction of ITER H-mode power threshold by means of various statistical techniques
The prediction of the additional heating power required to reach the H-mode regime is a key issue in the design of the International Thermonuclear Experimental Reactor (ITER). Power law scalings, which fit the multi-machine ITER threshold database, were calculated using global parameters. The inclusion of different variables showed that even the best log-linear fit still exhibits considerable data scattering. The resulting uncertainty for ITER is large. Reduction of the scattering of the experimental data has been attempted using more powerful statistical tools. Results obtained by several statistical techniques, such as neural networks, discriminant analysis, system identification are presented. Using these techniques, the information content in the data set is characterized and a model structure is elaborated which gives an indication of the underlying physical process, by choosing, semi-automatically, only the �useful� parameters. As an exploratory step, a rather flexible model containing different variables and the inner gap was fitted. This model is used to estimate the probability of being close to the transition as a function of the additional power. Applied to a randomly selected validation data set, this model could correctly discriminate between pure L-mode and LH timeslices in 85% of the cases. Ohmic LH transitions are also well predicted by this model.
Bibliographic Reference: Article: 1998 International Congress on Plasma Physics, 25th EPS Conference on Controlled Fusion and Plasma Physics, Prague (CZ), June 29 � July 3, 1998
Record Number: 199811271 / Last updated on: 1998-10-27
Original language: en
Available languages: en