Refined turbulence modelling for power generation industry

Direct numerical and large eddy simulations in nuclear applications

“…Thus, one finds a wide spectrum of channel dimensions, of characteristic non-dimensional numbers, and flow types in the same application. Examples are analyses of passive decay heat removal within a complete model of the European Fast Reactor by means of the FLUTAN code by Weinberg et al (124), investigations of thermal mixing in the upper plenum of the PHENIX reactor with the TRIO-VF code by Roubin & Astegiano (103), or investigations of local problems, like the thermal barrier in a pump housing with the N3S code by Archambeau et al (2). Most investigations are combined turbulent momentum and heat transfer tasks in three-dimensional geometry.…”

“…Therefore, in trying to achieve more universal models with better prediction capabilities one also considers the application of full second order closure models basing on RANS. Current experience with these models is that second order models require very fine grids for practical applications (2); as a consequence wider applications can only be expected on future parallel high performance computer systems.…”

“…For a more quantitative analysis of thermal striping it is necessary to treat the turbulent convective heat transfer coupled with the thermal conduction in the solid structures. This is achieved in the applications by Ushijima & Tanaka (121) to the upper plenum flow including the instrumentation plug, in the applications of the finite element code N3S using a dynamic Smagorinsky model to the flow in the thermal barrier in a pump shaft (2), and in the applications of the STAR-CD code using unstructured grids and no SGS model to the flow through a T-junction, Simoneau et al (114). Of course, with this task LES is superior to any RANS based calculation because the dynamics of interest are a result of the large scale structures and are without additional modelling a direct outcome of a simulation.…”

“…On the other hand, the large flexibility and accuracy to correctly reproduce complex geometries is an advantage of the new code generation with unstructured grids. Convincing examples are those by Simoneau et al (114) who use STAR-CD for a Tjunction, those first demonstrations of the capabilities of TRIO_U, but obviously in the structured mode, by Barsamian et al (9), or those with the finite element code N3S (2). Proceeding to unstructured grids needs special care with the widely used dynamic model because extended test filters have to be applied (61).…”