Hi CFDEM members,
I am investigating fluid flow through a porous material with specific geometric structure. I build a number of particles on LIGGGHTS and use coupling to simulate fluid flow through such media. This is quite similar to Ergun test case in tutorial. I try to see how the size of simulation domain effects on result. It is interesting that for small number of particles, (720 particles), pressure drop is stable (see graph), however it becomes highly fluctuating at the beginning and stabilised gradually for the larger scale of simulation (2880 particles). Note that the porous structure is ensured to be uniform and the same in both scales. Input, boundary conditions (U,p), CFD meshing (cell) and particles sizes are the same. The averaged pressure drops is quite similar in two cases (as shown in graph attached).
Could anyone explain why increase geometric scale of simulation results in fluctuation of pressure drop? How the fluctuation of pressure drop is generated? I think when CFD (finite volume method) solves momentum equation in each cell and interpolate for all in domain, it should not be dependent much on the scale of the whole domain but characteristics of cell (cell size, pressure gradient, void fraction …)
Looking forward to your ideas,
Thanks and regards,
| Attachment | Size |
|---|---|
 pressure_drop.jpg | 19.42 KB |
jtvanlew | Tue, 02/24/2015 - 18:26
i work with packed beds too
i work with packed beds too so this is an interesting investigation for me. I'm 100% less of a numeric expert than others on this forum, so if the problem is coming from some computational source I'll not likely help. But if it's physical maybe I'll help. I've got some questions for your simulation...
What's the Reynolds number for the two cases (presumably the same)?
You keep the cell size the same so in both cases you have the same number of particles/cell?
Which drag correlation are you implementing?
At what locations are you measuring pressure? Do you have a fluid inlet and exit region before/after the packed bed?
NTT1508 | Wed, 02/25/2015 - 00:18
thank for interest
Hi Jon,
Thank for your interest. happy to share with you.
In this investigation, both cases use laminar flow with turbulent off. I hope input velocity is small enough U=0.001m/s. Wall boundary is slip. Pressure is set as zero at outlet.
What I am doing with geometry is just double the each size of domain in 1st case to make 2nd case. At the 1st case, mesh is 5(W)x5(H)x20(L) for the whole domain 3.750x3.750x20mm. Total number of particles at 1st case is 720. And in the 2nd case, the domain is double but cell size the same, so mesh is 10(W)x10(H)x20(L) for the size of domain: 7.5x7.5x20mm. As you see the flow length is constant (L=20mm), only the cross section is 4 time bigger. The number of particles in 2nd case is 4 time bigger too (2880). So the number of particle/cell is constant. Diameter of particle is 0.5mm (5E-4m).
Flow from bottom to the top. In 20mm length of flow, there is only 10mm length in the middle contain particles. 5 mm length for the input and output.
My force models including DiFeliceDrag; gradPForce and viscForce. I measure pressure right at the points before and after particles zone.
As you see on the graph, in bigger domain, pressure drop is a kind of sine wave gradually stabilised to the value of smaller size. Only problem is that in bigger domain, pressure drop is varying much before stabilised. I have checked motion of particle. It is interesting to see the particles moving upward and downward like a sine wave. That possibly results in sine wave type of pressure drop in the bigger domain. My confusions are below.
1) Do you think the length of flow is not long enough? I heard some suggestions about the length of flow to let it fully developed.
2) Why change of domain size makes particles position unstable while we keep void fraction is constant?
3) When particles motion, it makes void fraction and Ksl changed. And therefore, pressure drop changes with respect to solving Navies-Stock equation. But I cannot understand the sine wave motion of particles.
To be honest, I am a material engineer, not good at fluid mechanics. Your idea is really appreciated.
Thanks,
Regards,
jtvanlew | Sat, 02/28/2015 - 21:15
I'm guessing your kinematic
I'm guessing your kinematic viscosity is something on the order of 1E-4? In which case, your particle Reynolds number is probably something like Re = 0.005 which is really small. What timestep are you using? I think for such a small Reynolds # flow, satisfying just the Courant # requirement isn't enough. You often need really small timsteps for 'creeping' flow to prevent the diffusive terms from messing things up. When there isn't a good deal of flow, errors at the outlet can propagate upstream. If you drop the timestep in your CFD by like, 2 orders of magnitude, what happens?
In my opinion, with how small your velocity is, the length of inlet/outlet region is fine. But what if you try a velocity that's maybe U = 0.1 m/s? If my hunch is true, then you'd see more stable results with a higher Reynolds number. But then at teh same time, with higher velocity you _might_ need to have a longer outlet region.
what is confining the particles in the DEM domain? why are they allowed to oscillate along the axis of fluid? do you have primitive walls?
One last thing to mention is that pressure drop is linear with length of the packed bed. so if your packed bed is twice as long, say, the pressure drop should be twice as high. You should find different values for the two different size beds you're analyzing (unless you're looking at Delta-P/L).
elham.nasiri | Mon, 03/16/2015 - 08:21
High velocity and scattering of particles
Hi all
I'm simulating the sediment transport by cfdemSolverInterFoam that I wrote. But I did not succeed yet!
I run a simple channel with a particle box with using this solver. The velocity increased in beginning and end of particle box and the direction of vectors was abnormal(please see the figures in Problem with new Solver post)!
I don't know its reason!
Looking forward to your ideas,
Thanks
Elham
NTT1508 | Thu, 03/19/2015 - 01:38
Distribution of particles
Hi Elham, Can you make a new post for your problem? We and others can discuss there.
Hi Jon, Many thanks for your ideas. I have tried with different U, but the same problem. I think it due to movement of particles that results in change of mentumexchange. I used primitive walls. Yes theoretically the pressure drop should be linear to length. I have doubled length and found some difference. that might be due to numerical process.
Anyway Jon, I have a question for you that, if we keep the porosity constant but change the position or particles in the system (different distribution), do you think the pressure drop varying? I think the pressure drop should change with arrangement of particles because the effective flow paths changes.
thanks,
Nathan,