[kar]

Hello,

I was trying to solve a parabolic flow in 2D square-like domain.

* In blockMeshDict and boundary file I specified just "patch" for inlet and outlet,
* In /0/U I set zeroGradient on in/out, (0 0 0) on walls + empty
* In /0/p I set fixedValue on in/out, zeroGradient on walls + empty.

But.... solution obviously diverge. Aren't those BCs sufficent to describe a problem?
I think, I don't understand how to set BCs in OpenFOAM, for example - why should I use physicalType? Maybe you can advice something useful at least for that poiseuille flow?

[kar]

Oh, I see a parabolic velocity distribution now and it feels so good! The problem was - I forgot to take into account a fact, that for icoFoam pressure must be divided by density. So I tried to solve for 1000 times greater pressure!

Maybe I'll see somewhere else why physicalType should be specified...

[kar]

I checked solution more carefully and .. trouble doesn't stop - I have almost exactly 2x greater U values comparing with analytical!

So in SI:
nu = 1e-4,
geometry: dx=0.01, dy=0.1, dz=0.001,
flow is in y direction,
delta p = 0.016 (divided by ro)

According to wikipedia, Umax = 0.01. But solution gives 0.02!! How come?

[ngj]

Hi Karlis

The Poiseuille-flow I could find on wiki (please add link in the future), was for a circular pipe flow. Your setup is for a 2D case, i.e. Poiseuille flow between two flat plates. The ratio between u_max in these two cases is exactly a factor of 2.

/ Niels

[kar]

Thank you, Niels!
I was thinking in a quick way - take a plane of cylinder and calculate for it.. Actually different problem is solved in such way.