Discussion Forum

June 4, 2010 7:19am UTC

Navier Stokes equation with periodic boundary?

Hello,

I want to model flow in an open stream over repeating streambed geometry (sinusoidal geometry of ripples). My background is to use the calculated pressures on the streambed for further modelling in porous media.

I use Navier-Stokes equation with slip-bc on top (open water) and streambed, inflow velocity as left bc, and constant pressure value as outflow boundary on the right side.
The model runs quite good, but I’m searching for a solution where the pressure and velocity field over one ripple is equal to any other ripple in the system. That means a periodic pressure distribution for an “infinite” long streambed is necessary.

I tried it with periodic bc, with right bc as source and left as destination, but there was no change in the pressure distribution. But I’m not sure if a periodic bc is the right application for this.
Are there any other possibilities?

Find my model attached.

Thanks for your help,

Nico

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse

June 4, 2010 12:40pm UTC in response to Nico T

Re: Navier Stokes equation with periodic boundary?

I have to correct myself. I have done a similar model before. You are right at setting periodic boundary condition but you have outlet as source and inlet as destination. For p you set p as a source and p+dp as destination but for other parameters source and destination are the same. dp then decide the velocity.

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse

June 5, 2010 2:02pm UTC in response to Nico T

Re: Navier Stokes equation with periodic boundary?

Hi

for me I would try "noslip" on the bottom "riple" surface, as I beleive this is more realistic, and open (no visquaos stress or f=p0=1[atm] for the top surface, for me it converges better.

And define p0=1[atm] in your constant and use that in the BC. Jut as well plot p-p0 (or perhaps use a gauge pressure of 0[atm]) to get the pressure changes

you have a Reynolds number up to 38000 too

Then your inlet velocity is interacting with the slip/noslip ripple surface boundary condition. I would rather use an average velocity or a volumic/surface density velocity, by having COMSOL defining the velocity distribution itself, only restraining the integral (or average) of the velocity over the inlet edge/surface

Finally, why not have periodic boundary conditions, I believe it should work too, but I'm always suspicious about those, so I like to simulate two periods (as you have) and I split the volume vertically in the middle to compare the natural continuity boundary in the middle of my simulation with the expected identical boundary variable at the periodic boundary, have a try.

Have fun Comsoling
Ivar

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse

June 8, 2010 2:34pm UTC in response to Nico T

Re: Navier Stokes equation with periodic boundary?

Hi

I have tried out different sequences for "fun" in V3.5a. I come to that if you enter (as input and output) a velocity profile as a Vx_ave*sqrt(2.24*(y-Y0)/length), where y is the vertical coordinate of the inflow edge (I was in 2D) Y0 the lower vertex where I expect Vx_ave=0 and length is the total inlet edge length, and the 2.24 is just to make the integrated value equal to the average, and then I pass over 5 identical ripples, I notice that my velocity profile is rather nice and periodic (if you exclude the inlet and outlet boundaries.

I noted that with "s" I got some funny velocity profil shapes, it does not seem to run from 0 to 1 ?

For this I also set the inlet pressure top vertex to p=p0=1[atm] and an initial value of the pressure to p0, and velocity initial condition Vx=Vx_ave. My Reynolds cell number wvere also more compliant with what I expected

Howevever, I failed to set up a "periodoc boundary condition" So obviosly , its easier to run over a few periodes of "ripples", and let the system stabilise. You can always then extract the velocity profile via a extrusion coupling variable into another geoemtry

Hope this helps
Good luck
Ivar

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse