The modelling of neutron transport typically relies on two rather opposite approaches: the probabilistic approach, and the deterministic approach. The probabilistic approach or Monte Carlo approach relies on tracking the individual lives of neutrons, and requires a large computing power for nuclear reactors. The deterministic approach, on the other hand, is based upon fast running algorithms, that solve the problem at hand in only an approximate manner. The purpose of the present project is to combine both approaches in order to obtain fast running methods (thanks to the deterministic route) and accurate results (thanks to the probabilistic route). One promising hybrid method is the so-called response matrix method. This method was originally derived in the early seventies in a pure deterministic sense. In the proposed project, the computation of the collision probabilities required for applying the method will be carried out using a probabilistic solver. Due to the level of details of the simulations to be performed, the development of enhanced visualization tools will be necessary as an aid to development. This will require the construction of an adequate data management system and results visualization capabilities.
In the projects initial phase in 2016, the project will aim at investigating the feasibility of the proposed hybrid method, and at demonstrating the usefulness of the fine-scale results obtained, compared to traditional coarse-mesh approaches. The project will also result in the specification of a system architecture description for visualizing the results of the hybrid calculations. Examples of implementations will also be given.
Professor vid Chalmers, Physics, Subatomic and Plasma Physics
Funding Chalmers participation during 2016–2017