Page 412 - Thermal Hydraulics Aspects of Liquid Metal Cooled Nuclear Reactors

Multi-scale simulations of liquid metal systems 375

momentum and heat source terms in the system model to “drive” these quantities to their
desired values.
Starting from these interfaces, a first implementation of the desired multiscale cou-
pling can be developed. In most cases, this implementation will be specific to a given
application (a specific reactor or experiment or even a single transient). In the long
term, the development of a more generic algorithm applicable to a wider variety of
situations can bring considerable added value:
l In a specialized coupling code, the coupling algorithm itself is part of the model; thus, it is
hard to argue that the validation obtained by comparing a coupled model with a given exper-
iment can be “extrapolated” to a new application, as this new application will contain, by
necessity, a different version of the coupling algorithm. By comparison, a generic coupling
algorithm can be applied as is to both validation experiments and reactor applications and
can therefore be considered as validated when performing the latter.
l A specialized coupling algorithm will often contain “hard-wired” constants (such as mesh
numbers) adapted to a given input deck. By contrast, a generic coupling algorithm needs to
perform error checking on its input data, for instance by checking that the coupling bound-
aries have compatible elevations in the STH and CFD codes. In the case of liquid-metal flow,
this might be very important. Very small inconsistencies in the elevations might introduce
remarkable pressure inconsistency due to the hydrostatic pressure (high density) and poor
coupled code convergence. This kind of input checking can be crucial in avoiding a range
of elementary but sometimes hard-to-catch errors.
l Finally, improvements made to a generic coupling algorithm immediately benefit all the
models to which this algorithm can be applied. By comparison, specialized coupling algo-
rithms must be upgraded individually with the new features.
At CEA, these considerations led to the integration of several specialized couplings
(developed for the PHENIX reactor, the TALL-3D lead-bismuth experiment, and
the ASTRID reactor case) into a unified “coupling code” called Multiscale ASTRID
Thermal-HYdraulics Simulation (MATHYS) (Gerschenfeld et al., 2017). The switch
to MATHYS made the development of new coupled models much more straightfor-
ward and allowed for the detection of several errors in the coupled models; this devel-
opment also made it possible to argue that the validation obtained by confronting
MATHYS to experimental data could be extrapolated to reactor cases.

7.3 Development and validation of multi-scale
approaches

This section aims to provide a nonexhaustive overview of the application of multiscale
coupled models to liquid-metal experiments. The eventual aim of these model-
to-experiment confrontations is to lead to a better understanding of liquid-metal reac-
tor transients, which should lead to a reduction of the conservative margins used in the
safety assessment of these reactors. To the knowledge of the author, multiscale
calculations have been applied to reactor safety analysis in the following cases: