Analysis of diagnostic data in fusion experiments is usually dealt with separately for each diagnostic, in spite of the existence of a large number of interdependencies between global physics parameters and measurements from different diagnostics. In this article, we demonstrate an integrated data analysis model, applied to the W7-AS stellarator, where diagnostic interdependencies have been modeled in a novel way by using so called Bayesian graphical models. A Thomson scattering system, interferometer, diamagnetic loop, and neutral particle analyzer are combined with an equilibrium reconstruction, forming together one single model for the determination of quantities such as density and temperature profiles, directly in magnetic coordinates. The magnetic coordinate transformation is itself inferred from the measurements. Influence of both statistical and systematic uncertainties on quantities from equilibrium calculations, such as position of flux surfaces, can therefore be readily estimated together with uncertainties of profile estimates. The model allows for modular addition of further diagnostics. A software architecture for such integrated analysis where possibly large number of diagnostic and theoretical codes need to be combined, will also be discussed.
Read full abstract