Wet filtration test

Asked by Ali Abdallah

Hello,

I want to run some wet filtration tests, so I created a sample of spheres with a specific density, fixed the sample and then I need to deposit some fine particles under water flow.

I was trying to see what ar the tools that I can use to do this coupling between solid and liquid and I find:
-CFD-DEM coupled simulations with Yade and OpenFOAM
-YADE 1D vertical VANS fluid resolution
-Lattice–Boltzmann method (LBM) but only available for 2D
-Pore-scale finite volumes method (PFV)

I am wondering which one of these tools can be the most efficient way to do my simulation, and is it possible to use one of these tools with non-spherical particles? Knowing that I just only care about the number of fine particles that can pass the sample and maybe measuring the permeability.

Thank you for your help.
Best regards,
Ali

Question information

Language:
English Edit question
Status:
Solved
For:
Yade Edit question
Assignee:
No assignee Edit question
Solved by:
Robert Caulk
Solved:
Last query:
Last reply:
Revision history for this message
Robert Caulk (rcaulk) said :
#1

Hello,

It is not clear what you are trying to simulate, or what you need from your model.

How well do you need to resolve capillary action and forces on these arbitrary shapes? Is it quasistatic? What is the porosity of the simulation?

-rc

Revision history for this message
Ali Abdallah (ali96) said :
#2

Hello Robert,

I am trying to study the filters used for the protection of the dams and dikes.
So basically I am creating samples having the same PSD as the one found on the site and I am trying with different porosity ( for now I am working with dense one). I am also investigating the effect of the shape of the particles on the filtration process, I am working for now with spheres and later I will try to create samples from polyhedra...
I was reading the two papers published for PFM and I understand that they are using Delaney tessellation to obtain finite element volume of the pore space which only work with spheres.
Honestly, I am not so familiar with fluid computations, so I have no idea how it is possible to resolve capillary action and forces on arbitrary shapes.

So to sum up, I will create a sample that emulates the filters found on the dikes, then I will create a thin layer of fine particles with a specific size. With the flow introduced I want to see if these fine particles can pass the filter.

Thank you Robert for helping me.

Revision history for this message
Robert Caulk (rcaulk) said :
#3

It sounds like quite an expensive simulation, considering the size distribution of the particles.

PFV was created for stoke's flow in dense granular packings. It seems that since you have extremely small particles moving freely through the system, then it is not a dense packing. If the small particle migrations can be modeled more analytically, then you could try to model their movements as a sort-of "concentration" in a species transport sense which depends on the macro fluid fluxes of the triangulated space between your larger particles.

I'm not expert on LBM, but it seems you are not trying to simulate fluid surfaces so I don't think it is applicable.

CFD-DEM would certainly resolve the problem but it still seems like it is probably going to be extremely expensive.

Cheers,

Robert

Revision history for this message
Ali Abdallah (ali96) said :
#4

Thank you very much for your clarification and sorry for my late reply.

In the end, I decided to work on the Pore-scale finite volumes method (PFV), as it seems very simple to model.
I did a small model and it worked fine. However, I have some small questions about the parameters if you can help I really appreciate that:

for the two parameters: flow.defTolerance=0.3 and flow.meshUpdateInterval=200, I took their values from the oedometer example. I understand that they are responsible for the threshold and the maximum number of timesteps for retriangulation of the mesh.
My question is: Can I directly use these two values or I should do a calibration to choose the optimum values knowing that my filter is fixed and only the fine particles are moving.

For the flow.useSolver I saw that the value should be 3 for Cholesky factorization (via Eigen3 interface) and 4 for multicore CPU. In my case, I usually use parallel computation ( -j X ( x can be between 1 and 16 depending on the number of particles).
My question is: Is multicore CPU has the same meaning as parallel computation? if yes this means I should put the value equal to 4 right?

Finally is there any criteria to define the timestep? I used to increase the density of the fine particles as a tool to increase the time step when I used to do dry filtration tests under gravity. Do you think it is okay to do the same thing here?

Thank you very much Robert for helping me.

Best regards,
Ali

Revision history for this message
Best Robert Caulk (rcaulk) said :
#5

Hello,

>>Can I directly use these two values or I should do a calibration to choose the optimum values knowing that my filter is fixed and only the fine particles are moving.

It is just a trade off of accuraacy, and in some cases stability. If you have alot of deformation, you should be remeshing more frequently. If the problem is quasistatic, then you dont need to do it very often. Just remember that force vectors will not update until the mesh is updated. So you if you have particles moving alot based on certain pressure/viscous forces, then the orientation of those forces will be outdated quickly if you arent remeshing enough. defTolerance is trying to detect when the deformation warrants a new remesh, but it is a relatively expensive operation too.

>>My question is: Is multicore CPU has the same meaning as parallel computation? if yes this means I should put the value equal to 4 right?

Yes.

>>Finally is there any criteria to define the timestep?

See our paper [1] where we discuss this. It is really dependent on your regime: stiffness, viscous, or some combination of the two. It is a difficult problem to solve - there will always be some trial and error involved unless you solve for the eigenvalues of the conductivity matrix. Which is quite expensive on its own.

>> I used to increase the density of the fine particles as a tool to increase the time step when I used to do dry filtration tests under gravity. Do you think it is okay to do the same thing here?

Yes you can do that assuming inertial effects are negligible in the system.

Cheers,

Robert

[1]https://linkinghub.elsevier.com/retrieve/pii/S0010465519303340

Revision history for this message
Ali Abdallah (ali96) said :
#6

Thanks Robert Caulk, that solved my question.