Automated magnetic divertor design for optimal power exhaust
Blommaert, Maarten; Gauger, Nicolas Ralph (Thesis advisor); May, Georg (Thesis advisor); Frank, Martin (Thesis advisor); Reiter, Detlev (Thesis advisor)
Jülich / Forschungszentrum Jülich GmbH, Zentralbibliothek, Verlag (2017) [Book, Dissertation / PhD Thesis]
Page(s): xxiv, 219 Seiten : Illustrationen, Diagramme
The so-called divertor is the standard particle and power exhaust system of nuclear fusion tokamaks. In essence, the magnetic configuration hereby `diverts' the plasma to a specific divertor structure. The design of this divertor is still a key issue to be resolved to evolve from experimental fusion tokamaks to commercial power plants. The focus of this dissertation is on one particular design requirement: avoiding excessive heat loads on the divertor structure. The divertor design process is assisted by plasma edge transport codes that simulate the plasma and neutral particle transport in the edge of the reactor. These codes are computationally extremely demanding, not in the least due to the complex collisional processes between plasma and neutrals that lead to strong radiation sinks and macroscopic heat convection near the vessel walls. One way of improving the heat exhaust is by modifying the magnetic confinement that governs the plasma flow. In this dissertation, automated design of the magnetic configuration is pursued using adjoint based optimization methods. A simple and fast perturbation model is used to compute the magnetic field in the vacuum vessel. A stable optimal design method of the nested type is then elaborated that strictly accounts for several nonlinear design constraints and code limitations. Using appropriate cost function definitions, the heat is spread more uniformly over the high-heat load plasma-facing components in a practical design example. Furthermore, practical in-parts adjoint sensitivity calculations are presented that provide a way to an efficient optimization procedure. Results are elaborated for a fictituous JET (Joint European Torus) case. The heat load is strongly reduced by exploiting an expansion of the magnetic flux towards the solid divertor structure. Subsequently, shortcomings of the perturbation model for magnetic field calculations are discussed in comparison to a free boundary equilibrium (FBE) simulation. These flaws in the magnetic model are then overcome by elaborating a strategy to include the full FBE code into the optimal design approach. Using the full model, results are then presented in application to the novel WEST (tungsten (W) Environment in Steady-state Tokamak) divertor. Finally, one-shot optimization methods are considered for further acceleration of the optimal design procedure. Instead of fully solving state and adjoint equations in each optimization iteration, one-shot methods perform only a single iteration of state and adjoint solver in each optimization iteration. To reduce the cost of design updates, a grid deformation method is derived for strictly flux-aligned grids. Starting from a literature review, a novel one-shot strategy is then elaborated that features the globalization approach of state-of-the-art one-shot methods while yielding increased efficiency and practical usability. On an unconstrained test case, the novel method shows stable convergence.