This paper examines the effect of grid resolution on the solution of a B2-EIRENE simulation of a partially detached ITER divertor plasma. Due to the large amount of required computational time in coupled finite-volume/Monte-Carlo codes, simulations of the plasma edge are typically run on coarse grids. However, new averaging techniques make simulations with finer grids feasible. Starting from the original numerical grid, simulations are performed on two successively finer grids. Results of the numerical error analysis reveal that the discretization errors are large. Peak values are particularly sensitive to grid resolution and can increase more than 40% for the same model input parameters. However, this effect is partially offset by a modification of the operational space of the ITER divertor in this case. By choosing the numerical parameters more adequately, a total numerical error of only 15% has been achieved within a feasible computational time.
Key words Plasma edge modelling, accuracy, statistical noise.Present computational techniques for coupled finite-volume/Monte-Carlo codes for plasma edge modeling under ITER or DEMO conditions face serious challenges with respect to computational time and accuracy. In this paper, scaling laws for different error contributions are assessed and practical procedures for error estimation are proposed. First results on a 1D and a 2D test case are discussed.
Iteratively solved Monte Carlo (MC) codes are frequently used for plasma edge simulations. However, their accuracy and convergence assessment are still unresolved issues. In analogy with the error classification recently developed for coupled finite‐volume/Monte Carlo (FV‐MC) codes, we define different error contributions and analyse them separately in a simplified non‐linear MC code. Three iterative procedures are examined: Random Noise (RN), where different seeds are used in each iteration; Correlated Sampling, where particle trajectories remain correlated between iterations; and Robbins Monro, where averaging is used during the simulation. We show that, as in FV‐MC codes, RN is the most efficient iterative procedure provided averaging is used to decrease the statistical error. In addition, we conclude that the accuracy can be assessed using the same techniques as in FV‐MC codes.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.