< Back to previous page

Project

Adjoint-based divertor shape optimization for the ITER reactor using plasma edge simulations with realistic vessel geometry

To be able to interpret experimental results from nuclear fusion experiments and to design the second generation divertor for ITER, numerical codes are indispensable tools. The numerical domain of typical plasma edge simulations is bounded by an outer most flux surface. However, to find the optimal design for the ITER heat exhaust device or divertor, which spreads the heat load as much as possible, a numerical shape optimization algorithm needs to be employed. To extract physically relevant optimal design, it is necessary to expand the numerical domain upto the vessel walls. Therefore, the first task entails the development of a robust grid generation tool for grids upto the vessel wall.

The so-called wide grids need to comply with different requirements. Firstly, the grid cells should align properly with the magnetic field to avoid numerically induced transport. Secondly, local refinement towards the vessel wall is also highly recommended as typically strong gradients are found in these regions. These two requirement are already conflicting as the vessel wall is in most cases not aligned nor perpendicular to the magnetic field. Therefore, comprimises will need to be made. On top of that, the wide grids will need to be able to cope with automatic grid adaptions as for the shape optimization algorithm, the shape of the vessel wall will be altered during the iteration process.

Next, a second major challenge consists of the necessary enhancements of the plasma model. Typically, a fluid approach based on the Braginskii equations is employed to model the plasma inside the reactor. This approach assumes high density plasma or a high number of collisions between the plasma particles per volume. However, in the regions closer to the vessel wall and further away from the core and the divertor, the plasma is less dens and shows more kinetic behavior. Therefore, changes in the plasma model will need to be made to capture this expected kinetic behavior in the correct way. A second model enhancement is the addition of plasma drifts which can highly influence the plasma distribution in the extended regions. Drifts are transport phenomena caused by the interaction between  the magnetic field and f.e. the electric field, the pressure, etc. These transport phenomena are purely convective and thereby rather instabile. 
For this reason, stabilization schemes need to be implemented to be able to add the plasma drifts into the plasma edge simulations.

Finally, the optimization algorithm itself will be developed. To avoid computationally expensive finite difference optimization, an adjoint based approach will be used which does not become more expensive with increasing number of design variable. The implementation of this adjoint approach involves differentation of the used models. As this would be quite cumbersome to do manually in a state of the art plasma edge code, algorithmic differentation will be employed. After that, every tool is in place to recieve an optimal divertor design for the ITER reactor.

Date:19 Oct 2021 →  Today
Keywords:Nuclear fusion modelling
Disciplines:Modelling and simulation, Physics of (fusion) plasmas and electric discharges, Numerical computation
Project type:PhD project