A neural network approach to evaluate density profiles from reflectometry in ASDEX Upgrade discharges with internal transport barriers

In next step devices his expected that reflectometry can be used as an alte rnative to magnetic systems in the control of plasma position and shape. Th is is particularly important in long discharges when the accumulated errors of magnetic signals may be quite significant. This is beyond the present a pplication of reflectometry and puts new requirements on the diagnostic, na mely automatic analysis of reflectometry data, real-time data processing, a nd high reliability. A key step is to demonstrate the potentialities of rea l-time analysis in present reflectometry systems. With that purpose, we pro pose a neural network approach to process simulated and experimental data m easured with reflectometry on the ASDEX Upgrade tokamak. The study shows th at the neural network approach has the potential to meet the tight timing r equirements of control applications with sufficient accuracy, provided that realistic profiles are used in the training. First tests using ASDEX Upgra de reflectometry data are promising. (C) 2000 Elsevier Science S.A. All rig hts reserved.