neXt generation numerical Techniques for deterministic REActor Modelling (X-TREAM)

By essence, nuclear reactors represent multi-physic systems, in which different fields of physics are inter-related. In best-estimate approaches, such interdependent fields are usually restricted to the physics of neutron transport, the physics of fluid dynamics and heat transfer. The simultaneous modelling of these three fields is important both in steady-state conditions as well as in transient conditions. The addition of other fields (fuel behaviour, chemistry, material physics, etc.) might also be necessary in some specific situations.

Typically, the modelling of each physics is done in a segregated manner. One code is used per field at a time, assuming that the other fields are “frozen”, and iterations are performed to resolve the interdependences until convergence on all physical fields is achieved. Such a solution strategy is the result of the use of legacy codes developed for modelling each physical field separately, codes that were à posteriori coupled to model multi-physics problems. The main advantage of such an à posteriori coupling lies with the absence of internal code modification when the code is externally coupled to another software. This results in the preservation of the Verification and Validation (V&V) work for each code. Nevertheless, achieving tight convergence on all physical fields might be difficult and numerical oscillations are often encountered. In addition, when considering transient calculations, the non-linearities are often never fully resolved, resulting in inconsistent coupling algorithms.

In this project, a monolithic coupling approach relying of the so-called Jacobian-Free Newton Krylov (JFNK) method is studied. This technique considers solving the entire multi-physics problem as one single problem, in which all the non-linearities are inherently taken into account without any need to iterate between each physics.

Illustration of a workbench developed for testing the Jacobian-Free Newton Krylov (JFNK) method in the case of a coupled one-dimensional BWR model – the animation shows how the coupled solution evolves during the application of the JNFK algorithm

Sebastian Gonzalez-Pintor, Post-Doc, Subatomic and Plasma Physics, Department of Physics
Manuel Calleja, Post-Doc, Subatomic and Plasma Physics, Department of Physics
Anders Ålund, Fraunhofer-Chalmers Centre

​The Nordic Thermal-Hydraulic Network (NORTHNET)
This page is only available in english

Published: Thu 09 Jan 2020.