Page 402 - Thermal Hydraulics Aspects of Liquid Metal Cooled Nuclear Reactors

Multi-scale simulations of liquid metal systems 365

coupling interface with functionalities capable of piloting each code and exchanging
data with an outside “supervisor” code.
– A new numerical scheme must be developed for the overall coupled calculation. This
scheme must describe how interfaces and data (1-D and 3-D) transfer between different
scales should be treated (for instance, at the boundary between a system and a CFD
domain); it should also converge to a consistent “multiscale” solution, with no residual
inconsistencies between the solutions obtained by each code in their respective domains.
Energy- and mass-conservation properties are also highly desirable. Once developed, this
coupling scheme should be verified and validated to the same level as the initial codes.

Several examples of implementation of these strategies have been deployed to this
day. From the most to the least common, these include

l multiscale models of a complete loop/reactor coupling a CFD code (used to model a specific
part of the circuit, such as the hot/cold pool) with a system code (used to model the rest of the
circuit) (Bandini et al., 2015);
l multiscale models of a complete SFR/LFR core coupling a subchannel model of the inside of
the subassemblies to a CFD model of the interwrapper region (Conti et al., 2015). In some
cases, this type of model has been combined with a system/CFD coupling in order to produce
a three-scale model of a complete reactor (Gerschenfeld et al., 2017);
l single-scale models of a complete reactor primary circuit in a CFD code using integrated
“coarse” (porous or 1-D) models for components with complex geometries such as the core,
heat exchangers, and pumps.
In general, the ability to reuse the capabilities of existing codes has proved to be a
strong attractant for the multiscale approach; small supervisor programs imple-
menting a code-to-code coupling can allow one to obtain results in a relatively short
amount of time. However, few points should be noted:
The development and verification of a coupled numerical scheme is a potentially difficult
l
task and can require important or unexpected modifications in the underlying codes.
l Data averaging and reconstruction can be a challenging task. While the 2-D velocity and
temperature profiles computed at the CFD scale can be averaged and passed to the system
code, the reverse process includes reconstructing 2-D profiles from 0-D STH-scale values.
This point is treated in more detail in Section 7.2.2.
l Although the development of a coupled model of a given experiment or reactor case can be
undertaken “ad hoc,” the implementation of a V&V strategy will usually require the devel-
opment of a generic coupling, which can be used without modifications to model both reac-
tor cases and the experiments used to validate them. This capability is considered essential to
ensuring that the validation studies can be extrapolated to the reactor applications.
l Finally, the new effects predicted by the single-scale or the multiscale approach will need to
be validated against a suitable validation matrix. The realization of the experiments compris-
ing this matrix and their exploitation by the code developers often turn out to be more time-
consuming as the development of the coupling itself.
Finally, it should be noted that most STH codes now contain 3-D modules (such as
CATHARE, ATHLET, or RELAP). Compared with CFD codes, these modules often
suffer from limitations (such as restriction to structured meshes and the absence of
large-scale parallelism). This makes it difficult to successfully reproduce some of
the most complex 3-D effects (such as jet behavior in a reactor pool); however, they