CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   Complex problem: Open Channel Flow (http://www.cfd-online.com/Forums/fluent/63233-complex-problem-open-channel-flow.html)

 lett April 1, 2009 12:21

Complex problem: Open Channel Flow

Hi guys,

I am modelling (Fluent 6.3) the open channel flow (river flow to the dam). I am interested in steady solution, in the velocity profile near the outputs BC and in the volume flow rates at inflows of turbines (= outputs BC).

The domain is 192 x 131 x 9.5 m (x, y, z):
http://img5.imageshack.us/img5/1381/geometry.png

Well, I cant get a steady solution! And I am trying really hard.. :confused:

During running I get TVL warning (turbulent viscosity limited to viscosity ratio of 10e5 in ... most of my cells). Solution is a little converged after +/- 50 iterations (reziduals 1e-1) and then turbulence goes up and down and immediately the continuity goes too much up... and crash..

Crashed solution looks like:
water level is strongly unbalanced (I can see water in sky BC, i.e. 2 meters above water level) and I can see islands of high velocity in domain routing to sky, in the rest (in horozintal cuts) is zero velocity (even at input BC! (vertical))

I have already tried:
a) improving, improving and improving my mesh (I started at 1.6M and ending at 6.2M = top of my computer)
b) increase the viscosity of air, 10x - 1000x
c) setting up my model differently (more below)
d) increase TLV ratio from 10e5 to 10e8

BC:

INPUT:
a) pressure-inlet
direction vector 1,0,0
intensity 5%, diameter 2 m
open channel - free surface 199.5 m, bottom 192 m (bottom of the domain), velocity magnitude 0,0818 m/s

I tried too:
b) mass flow inlet
direction vector 1,0,0
intensity 5%, diameter 2 m
open channel - free surface 199.5 m, bottom 192 m
phase1 (air) mass flow-rate 26 kg/s
phase2 (water) mass flow-rate 80,000. kg/s

SKY:
pressure inlet
gauge pressure 0, normal to boundary
intensity 5%, diameter 2 m
volume fraction of phase2 = 0

OUTPUTS:
2x separated pressure outlets:
gauge pressure 0, backflow normal to boundary
intensity 5%, diameter 2 m
backflow volume fraction of phase2 = 1 (outputs are below water level)

WALL:
no slip

TURBULENCE MODELS:
a) realizable k-eps, standard wall function
I tried, too:
b) RSM

OPERATING CONDITIONS:
oper. pressure 101325 Pa
gravity: 0, 0, -9.81 m2/s
specific density 1.225 kg/m3

SOLUTION CONTROLS:
I tried: PISO, SIMPLE, Coupled (unsuccesfully - I think because not enough computer power (for 6.2M mesh))

Pressure: PRESTO!, tried too: body force weighted
Everything else: 1st order

I tried to play with URFs.

INITIALIZE:
Init: volume fraction of phase2 = 1
k = 2.5e-5
eps = 1.47e-7
x-velocity = 0.0818 m/s

then patch: upper region (air region)
volume fraction of phase2 = 0

I always check the init before solving

VOF set up:
using Implicit Body force and Implicit VOF scheme

Any ideas? Did anyone have that same problem? Any experiences with VOF and Open Channel Flow?
(If you need more info, let me know!)

I would guess that the problem is in too poor mesh, but I cant make it smoother.. If you think that it is that problem, I will have to make smaller domain..

Thank you very much. (Sorry for my English ;))
Lett.

 panda April 2, 2009 04:51

The convergence problem may be related with your solver selection. I recommend you use SIMPLE with relative low uder-relaxation factors (cannot be too low to achieve a premature convergence).

By the way, have you used the Open Channel boundary condition for the pressure outlet?

 lett April 2, 2009 07:51

I tried SIMPLE and I tried low URFs.

Open Channel Boundary for the pressure outlet? No I havent. Should I? I wrote that outputs are below water level... So I cant see a reason why to use Open Channel BC instead of ordinary pressure outlet.

 zhaopeng April 2, 2009 12:09

I see your geometry,are you sure that it is a continuous zone,you'd better check this by a single phase flow.

Although VOF model can be used to perform a steady-state calculation,it is usually difficult to get a convergence solution.then you can try to switch to unsteady solver.

 lett April 2, 2009 13:24

Yes, I am sure that my domain is one continues zone.

If I switch to unsteady solver, it takes too much time.. I tried... (It was running more than 3 weeks and still too far from steady solution)

Do you have next idea?

 zhaopeng April 2, 2009 14:21

I don't think it is proper to set gauge pressure to 0 at outlet.There are also two questions.Is the Z-coordinate of bottom 192m in Fluent,because you set the bottom level to 192m.And did you examine the phase distribution after patch ?

 panda April 2, 2009 20:28

I think you should check your outlet boundary conditions again and it seems that the open channel boundary conditions best suit this problem, you may find more details in the Fluent manual about the settings.

 lett April 3, 2009 04:41

2zhaopeng:

1) yes, the Z-coordinate of bottom is 192 m
2) yes, I did examine the phase distribution after patch

What value do you suggest for gauge pressure at outlet? I still think that my problem is not in that.

2panda:

Why should I use the Open Channel Boundary (OCB) for pressure outlet at the outputs? When I choose OCB I have to set free surface and bottom. And in my case the lowest Z-coordinate of outlet is 192.5 m and the highest is 198.6 m (water is from 192.0 - 199.5). So all of the surface of outlets is covered by water.
Second option is to set bottom level and gauge pressure at pressure outlet BC.

Thank you both for your time and opinion.

 zhaopeng April 4, 2009 09:13

I think you should define a hydrostatic pressure on outlet,maybe this advice is wrong,you'd better have a try.I still advise you to use unsteady solver,you can reduce the time step size when the interation is stable.

 panda April 5, 2009 20:51

your outlet bc still seems strange to me. is it a pipe-like outlet such that it is alway covered with water? At least, you may have to consider the hydrostatic pressure.

 lett April 6, 2009 04:36

Hi, thank you for not giving up :-)

I cant use the unsteady solver, believe me it takes too much time (even with a large time step).

2panda: yes, we can say, it is a pipe-like outlet always covered with water.

Right now, I am trying a hydrostatic pressure on outlets. Any suggestions? What value should I use? Should I use UDF for that?
I am starting with a constant value of 10.000 Pa.

 zhaopeng April 6, 2009 07:28

1 Attachment(s)
Hope this UDF can help you,it may need to be modified.

 lett April 6, 2009 12:11

zhaopeng, I used your UDF (changed for my conditions), still cant get a converged solution :-(

http://img222.imageshack.us/img222/2338/crash1.png
http://img222.imageshack.us/img222/7365/crash2.png

Any suggestion? :-(

 zhaopeng April 6, 2009 23:51

I review the geometry you uploaded,it seems that the inlet was not seperated to air-inlet and water-inlet,how did you define the phase mass flow-rate in the mass flow inlet boundary?

 lett April 7, 2009 05:22

well, for input I have used:

A) pressure inlet

and

B) mass flow inlet - here:
direction vector 1,0,0
intensity 5%, diameter 2 m
open channel - free surface 199.5 m, bottom 192 m
phase1 (air) mass flow-rate 26 kg/s
phase2 (water) mass flow-rate 80,000. kg/s

 zhaopeng April 7, 2009 05:40

Did you divide the inlet to two faces and define the boundary separately in Gambit?

 lett April 7, 2009 09:24

No, I didnt.
Should I? I dont think its necessary.

 lett April 9, 2009 11:56

Please, is there yet anyone, who have experiences in Open Channel Flow and want to help me?

 All times are GMT -4. The time now is 04:14.