Page 403 - Thermal Hydraulics Aspects of Liquid Metal Cooled Nuclear Reactors

366 Thermal Hydraulics Aspects of Liquid Metal Cooled Nuclear Reactors

can be successfully used to evaluate the influence of 3-D phenomena that can be
modeled accurately with coarse meshes (such as bulk thermal stratification). Thus,
STH codes with 3-D modules can play a key role as an intermediate step between
0-D/1-D STH codes and coupled or full CFD approaches.

7.2 Multi-scale coupling algorithms

This section will cover the principal points of interest for the development of a mul-
tiscale coupling scheme using existing codes, with the aim of describing commonly
used approaches while providing an overview of their advantages and disadvantages.

7.2.1 Domain decomposition versus domain overlapping
The first step in the development of a multiscale code coupling consists in the search
of an appropriate modeling scale for each part of the domain of interest; in most cases,
a PIRT-like process leads one to select the coarsest scale capable of representing all
the local phenomena that may affect the global behavior of the system. This process
will in turn lead to the identification of
l one or more “fine” domains, covering areas where local effects may occur, for which the
natural choice is subchannel or CFD codes;
l a “coarse” domain, covering the rest of the reactor or loop (and possibly additional circuits,
such as secondary loops), which should be modeled by a system code.
Once this choice has been made, the actual computational domains of each code
should then be chosen. Again, two choices are available:
l One may choose to “fit” the computational domain of each code to the fine and coarse
domains identified above, so that each of them is assigned to a single code. In this domain
decomposition approach (Fig. 7.2, left), the interactions between codes take place exclu-
sively at the boundaries between coarse and fine domains, which usually lead to a simpler
design for the coupling algorithm. However, this choice can also lead to a tighter coupling
between the codes; in particular, the overall pressure field, which is strongly coupled in
incompressible systems, must be shared between several codes in this approach. This can
in turn make convergence more difficult in the final coupling algorithm.
l Instead, one may choose to leave the complete domain of interest in the coarse computational
domain. In this domain overlapping approach (Fig. 7.2, right), the coarse results obtained by
the system code in the “fine” computational domain must be overlaid by the results obtained
at the CFD/subchannel scale to obtain an overall coupled calculation. Compared with the
decomposition approach, this method has both advantages and disadvantages:
Because the overlapped domain is still computed by the system code, exchanging cou-
pling data at the boundaries of the fine domain may not be sufficient. One needs to guar-
antee that system-scale calculations occurring inside the overlapped domain will not
affect the parts of the system model outside the overlapped domain; instead, the coupling
should ensure that the outside part of the system-scale calculation is consistent with the
results obtained with the CFD/subchannel code in the overlapped domain(s). For this
property to hold, the coupling algorithm may need to affect the system code inside