Problem with constraintIB validation

[Sagar G Nayak]

Hello,

I am trying to validate constraintIB code by simulating flow through a packed bed of spheres. I am comparing the pressure difference across the bed predicted by IBAMR with that predicted by Ergun equation. IBAMR is giving a much higher pressure difference than expected (~7 times) and I am unable to figure out why. Can someone help me with this? I have attached the figure. Its a 3d channel with only one layer of spheres in z-direction.

Problem2: If I simulate a flow though constricted channel, the velocity rise at the throat is much higher than expected (figure attached). A_in = 1.2, V_in = 1, A_out = 0.3. V_out should be ~4 units but it is almost twice. I think both problems are related.

I would appreciate any help. Thanks

[Boyce Griffith]

For IB methods that use regularized delta functions, the numerical size of the immersed structures is slightly larger than their physical size. I would suggest running at a higher resolution to see how the results depend on grid resolution, and also using a narrower delta function (perhaps BSPLINE_3 or the poorly named DISCONTINUOUS_LINEAR) at the current resolution.

--
You received this message because you are subscribed to the Google Groups "IBAMR Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ibamr-users...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/ibamr-users/5fdb7e58-8cac-4232-85b4-2da9de4983cen%40googlegroups.com.
<Screenshot from 2024-12-30 16-58-09.png><Screenshot from 2024-12-28 19-23-53.png>

[Sagar G Nayak]

Thank you very much Dr. Griffith. That actually helped.
But when I increased the grid resolution, a patch of high pressure appeared at a corner near the inlet (figure attached). Can you please comment on that? Also, when I switch-on AMR (max_levels = 2), it refines the mesh near the left top and left bottom corners even though there are no structures at those places. I suppose it is because of velocity gradients. How can I overcome that?

Thanks again,

[Boyce Griffith]

Sagar —

Did you ever sort out this pressure issue?

(It is probably obvious, but I am working my way back through my inbox. 😬)

— Boyce

To view this discussion visit https://groups.google.com/d/msgid/ibamr-users/97f00fc5-bc73-4cd6-b1c1-338a7eafbc41n%40googlegroups.com.
<Screenshot from 2025-01-09 12-17-22.png>

[Sagar G Nayak]

Thank you for the reply, Dr. Griffith.
Yes, I have figured out the problem. Apparently, I was not running it for enough time.

But, still the pressure drop predicted by IBAMR is much higher than that predicted by Ergun equation. For the packing I am using, Ergun equation predicts pressure drop to be 2800Pa but IBAMR is giving 9000Pa. Do you have any suggestions on this?

(In the figure, pressure is in dyne/cm^2)

Thanks

[Boyce Griffith]

Can you give me a little more detail on the problem setup — both physical problem and numerical setup? From your original email, it sounds like this is a 3D computation. What is the boundary condition in the z direction? How well resolved are the structures? Is there a 2D version of the problem?

— Boyce

To view this discussion visit https://groups.google.com/d/msgid/ibamr-users/04a161ab-d659-4cc6-8f3d-5dd2b600879en%40googlegroups.com.
<Screenshot from 2025-09-13 17-57-45.png>

[Sagar G Nayak]

Thanks for the reply Prof. Griffith

I am trying to study something similar to the problem in the section 4.2 of this paper (10.1016/j.ijmultiphaseflow.2016.05.001). Ergun equation computes the pressure drop during the fluid flow through packed beds. It computes pressure drop per unit length of the packed bed as a function of void fraction, diameter of the particles constituting the bed, fluid velocity, density and viscosity.

The paper talks about pressure rise in the domain due to clogging of particles. However to keep things simple, I constructed a packed bed of stationary spheres as shown in my last email. There is only one sphere in the z direction.

From your original email, it sounds like this is a 3D computation. What is the boundary condition in the z direction?

Yes, it is a 3d domain with one particle width (0.2cm) in z direction. Except left and right boundaries all others are no-slip walls i.e there are walls in y and z direction. In the above result, I divided 3.6x1.2x0.2 domain into 144x48x8 grids, with number of levels =1. However, I also tried using AMR (levels=2), but did not get much improvement.

Is there a 2D version of the problem?

 2d version of this problem is not possible since cylinders would entirely block the flow. 

Thanks

[Boyce Griffith]

On Sep 13, 2025, at 1:41 PM, Sagar G Nayak <nayaks...@gmail.com> wrote:

Thanks for the reply Prof. Griffith

I am trying to study something similar to the problem in the section 4.2 of this paper (10.1016/j.ijmultiphaseflow.2016.05.001). Ergun equation computes the pressure drop during the fluid flow through packed beds. It computes pressure drop per unit length of the packed bed as a function of void fraction, diameter of the particles constituting the bed, fluid velocity, density and viscosity.

The paper talks about pressure rise in the domain due to clogging of particles. However to keep things simple, I constructed a packed bed of stationary spheres as shown in my last email. There is only one sphere in the z direction.

From your original email, it sounds like this is a 3D computation. What is the boundary condition in the z direction?

Yes, it is a 3d domain with one particle width (0.2cm) in z direction. Except left and right boundaries all others are no-slip walls i.e there are walls in y and z direction. In the above result, I divided 3.6x1.2x0.2 domain into 144x48x8 grids, with number of levels =1. However, I also tried using AMR (levels=2), but did not get much improvement.

I wonder if these are the same boundary conditions as the paper and/or the derivation of the Ergun equation. It sounds like the channel height is small, and if that is different from the original setup that you are comparing against, that could explain the fairly large difference in pressure drop — particularly if it seems like you are getting a reasonably well converged pressure drop.

— Boyce

To view this discussion visit https://groups.google.com/d/msgid/ibamr-users/aad144c8-f307-4df4-b9ef-0e9dfa1c4706n%40googlegroups.com.