[andreaRocca]

Hello Foamers,



I'am working with foam-extend 4.1 on Francis turbine flow.

The problem it is well know and already successful resolved using commercial codes as CFX, Numeca.

Foam-extend was attracting me already during my PhD time, but I was ESI addicted

Now, I have tried to simulate my turbine using the segregated solver fe4.1. The post-processing is still on going but results seem promising and the simulation converged very very good. So I'am quite happy with that.



Then I moved to the coupled solver, the one I'am really interested in. Unfortunately, I can not even get a piece of result.


At this point some details:


Configuration: 360° GV + 360° RU + DT.

Mesh: fully hexa multiblock (AG5+Omnis) and check mesh ok.

GV/RU : GGI
RU/DT GGI with bridgeOverlap set to true; (apparently there is some imperfection at the interface, which at CAD level it even not noticeble, btw before doing again the geometry and the mesh I want to be sure about numerics.)


I'm using MRF and coupledKEpsilon turbulence model + wall modeling.



BCs where U has inletOutlet p has outletInlet.

GV_INLET I used the rotatingWallTangentialVelocity


Running the case in parallel over 72 cores and 2 nodes and using scotch as decomposition method (actually I was not able to use metis or patch constrained). The patches at the interfaces are grouped into globalFaceZones.



Moving to fvSchemes, I started with slightly dissipative numerics moving to "second order upwind". At the beginning omega it is 1/10 of its real value.



fvSolution ... here it comes. I've tried AMG + VCycle +SAMG+ ILUC0 (+ many internal settings according to the information I've found in the papers suggested in foam-extend-4.1 release), BiCGStab, GMRES.



No way.



So, I'am not sure I'am doing all correctly in the setup. In particular I feel not comfortable with the fvSolution and decomposition method. Anyone could give me some tips about that in similar applications or even some references would be very appreciated.



Thank you!