[MrFuglsang]

Hi

I'm thying to solve the turbulent flow for a open channel using the LESModel oneEqEddy. I created a rectangular box with wallFunctions at the three fixed walls, rigid lid at the open surface and cyclic conditions at the ends, to simulate a infinitely long channel.
The flow is driven by a pressure gradient anded in pimpleFoam UEqn.C

tmp<fvVectorMatrix> UEqn
(
fvm::ddt(U)
+ fvm::div(phi, U)
+ GradPz
+ turbulence->divDevReff(U)
);

dimensionedVector GradPz
(
"GradPz", p.dimensions()/dimLength,vector(0.0,0.0,0.0)
);

scalar SlopeGradPz
(
readScalar(transportProperties.lookup("SlopeGradPz "))
);

The strange thing is that the result changes when I change the the number of processors in the decomposeDict, and that the result of a 100 long channel calculated by 4 processors, gives the same average flow as a 25 long of one processor. The strange thing is that a 100 long channel on 5 processors is the same as a 20 long, but not the same 100 long of 4 processors. The SlopeGradPz is the same for all four cases.

My ddtSheme type is steadyState and the end time in both cases are 350 with deltaT of 0.1.

How can I make sure that the result is not influenced by the decomposition of the problem?

[MrFuglsang]

The endTime was not enough... the stability was fin each processor and the average the result just created a picture of 4 x 25 meter or 5 x 20 meter, but extending the time to 1000, the two 100 meter channels turned out the same result.