Integrated Interpretative Transport Modelling

Integrated analysis of fusion data Usually, diagnostic data and modelling results are dealt with separately in the analysis of fusion data, i.e., experimental results are passed to modelling codes for addressing physical issues. But interdependencies of modelled and measured parameters are numerous. In addition to the increase of the reliability of physics results, these entanglements allow for ac-cessing barely measurable quantities more accurately, e.g., particle and heat fluxes. The goal of this paper is to introduce a framework for inclusion of modelling results to data analysis. Application to W7-AS data and transport modelling will be discussed. The notion of integrated modelling and integrated data analysis [1] is used here for the concise combination of different sources of information such as different diagnostics results. The methods of Bayesian probability theory will be employed in order to arrive at a comprehensive modelling and data analysis. Technically, the approach uses raw data and their respective uncertainties and yields estimates for all parameters entering the model, in particular those quantities, which are of physical interest. The motivation for a framework allowing for integrated analyses arises from physically motivated requirements in large, long-pulse fusion devices. E.g., the operation of W7-X plasmas on time scales large with respect to the typical configuration relaxation time (τ L/R > 10 s) requires formalized on-line analyses of validated physics quantities. Moreover, huge amounts of data require automated approaches. For detailed off-line investigations, the inclusion of interdependencies enhances both the reliability of analyses and evidence for quantities which are not directly measurable. A simple picture of the integrated approach consists of modules representing individual di-agnostics measurements or modelling results, all of which are linked through physical relationships. An ubiquitous case for such links in fusion experiments is mapping of different, spatially resolved measurements onto a common grid given by magnetic equilibria. The example discussed in this paper focuses on transport models. Probabilistic methods are briefly discussed as a prerequisite of integrated modelling. In addition to several benefits to be discussed in this paper, Bayesian probability theory also allows one to deal with so-called systematic errors, which is essential for a framework for joint treatment of different diagnostics and modelling. Aspects of software architecture, such as WebServices or GRID computing, allowing the communication of codes are beneficial for the realization of the integrated approach ; more details can be found in Ref. [2].