Stirred reactor simulation - unsatisfying results


#1

Hi all,
I am struggelling with one simulation I want to do. In this four rod geometry, only the upper layer seems to move at high velocity, while you would think that also in the center of the volume there is more movement.


I think the MRIZone should also be placed higher (touching the fluid surface) but this gave bad meshing results…
I simulated another geometry in a similar way and this gives nice results.

I am quite new to SimScale and simlations in general, so I appreciate every help!
Thanks in advance!
You can find my project here: https://www.simscale.com/projects/achristiaens/stirred_bioreactor_-_marine_impeller/


#2

Hi @achristiaens and thanks for reaching out to us!

You have plenty of models inside your projects. Would you tell us the names you are referring to - makes it a lot easier! Once we know that the @PowerUsers_CFD and I will have a look at that and tell you what might be wrong.

Edit: What we need to fix first is your mesh because you have way to many illegal cells and the mesh does not look very good at the moment.

All the best!

Jousef


#3

Hi Jousef,
thanks in advance to look at my model!
The first printscreen I added was from simulation “Turbulent (WITH CellZone) fourrod agarose” based on “Hex-dominant parametric fourrof + agarose (WITH cellzone)”.
The second prinscreen I added was from simulation “Turbulent (WITH CellZone) agarose” based on “Hex-dominant parametric marine_3D-printed + agarose (WITH cellzone) - y”.
Apologies for the long names :slight_smile:

Is the problem in my mesh or in my CAD-design? Because for the “good” result I worked with a subtracted volume, while for the bad result I started from the reactor itself.

Best regards,
An-Sofie


#4

Hi An-Sofie (@achristiaens)!

The second one you posted looks good but we could further improve the mesh and see if this will have an impact on the results as it is too coarse for now I assume. For the other case I would like to know from you which part of the geometry you want to be the MRF zone? The whole container where the rods are in or the top part or the geometry? Once I know this we may proceed with the meshing operation to get rid of the illegal cells and create a proper MRF zone.

Best,

Jousef


#5

Hello @jousefm!

Indeed, I picked a coarser mesh for now just to see if it works and I will refine the mesh soon. For the first picture, the fluid that is mixed is in the bottom part so this is where the MRF zone should be.
afbeelding


#6

Hi @achristiaens!

Neat project you have here. I agree with what Jousef has spotted and namely your ,mesh does have a lot of illegal cells that would effectively not be usable in the following geometries:

  • Hex-dominant parametric 2
  • Hex-dominant parametric marine_3D-printed + agarose (WITH cellzone)
  • Hex-dominant parametric four rod (WITH cellzone!)
  • Hex-dominant parametric fourrof + agarose (WITH cellzone)

The other meshes not listed here all are alright with no illegal cells. You might probably want to identify the source of such of errors first before moving on the simulation.

Do let us know how it goes with coarse mesh and some context on what you’re exactly trying to find out/optimize/design along with how you are going to validate the results (if needed) would allow us to better help you!

Cheers.

Regards,
Barry


#7

Hi @jousefm and @Get_Barried,

I could fix the illegal cells in “Hex-dominant parametric marine_3D-printed + agarose (WITH cellzone)” following this Illegal cells inspection and Treatment.
However, this does not work for the two four rod meshes. I think the problem there is in my CAD-design:
afbeelding
solid_0 is the fluid volume that I want to simulate being stirred.
solid_1 is the impeller that should mix
solid_2 is the cellZone.

Problem here is that I don’t know where to put the cellZone. I thought it needed to be like in the above geometry (touching the surface of the fluid). But if I let create a mesh for this, it gives this result:
afbeelding

If I put the cellZone lower (see “emptycylinder_fourrod_celzone_low”) the mesh looks fine but again has illegal cells.
afbeelding

I hope this explanation is clear. So my major problem is: where to put the cellzone?

Thank you guys!
An-Sofie


#8

Hi @achristiaens,

looks like there is some double skin geometry going on that is causing the illegal cells:


here I have hidden the outer walls to show the second outer wall.

this causes this odd behaviour in the mesh:

Hope this helps,
Darren


#9

Hi Darren,
I have solved the problem with the mesh now. It was a mistake in my CAD-design.
afbeelding
Now my result looks like this:
afbeelding
However I would expect some higher speeds in the middle. Also the convergence plot does not look too good:


So I think there is a mistake again that I am overlooking.

The mesh I am talking about is called Hex-dominant parametric fourrod (WITH cellzone) and the simulation Turbulent (WITH CellZone) fourrod.

Thanks in advance for your help,
An-Sofie


#10

Hi @achristiaens, I copied the project and mad a few alterations to numerics, convergence seems to be much tighter now. One simulation is running longer to try get force and area average convergence (see plots in post processor).

https://www.simscale.com/projects/dlynch/stirred_bioreactor_-_marine_impeller/

I looked at a slice of velocity, firstly I noticed it was unnaturally ‘square like’ for a rotating zone, I then overplayed the mesh edges with the results and noticed it coincided with large cells, therefore I think you should refine the inner mesh.

Not so sure I would expect higher speeds in the outside, lower speeds in the centre. Although I would expect to see a more gradual degradation in velocity towards the centre.

Hope this helps,
Darren


#11

Wow, great! I will take a look at what you exactly did tonight. Thank you for the tips, I will keep you updated on the progress!


#12

Hello, I (we) am (are) getting there. Now I have some more questions about the postprocessing.

First of all, I want to calculate the volume average speed U. I know this must be possible (based on this explanation: [Result Control] Calculation methods on specific entities and their differences) but I cannot find how. Is it something that I have to specify before simulating?

Secondly, I want to calculate the shear stress from the speed. I think this has something to do with the calculator in the post-processing environment, but I cannot figure out how to do it exactly. This is where I got stuck:

I hope I didn’t look over similar questions in this forum but I think not…
Thanks in advance again guys,
An-Sofie


#13

Hi An-Sofie,

Unfortunately, yes this is something you have to define before simulating under the “Results Control”. However, don’t fret! The results are still there though extracting them will require some research.

Yes the calculator is the right function you want to use. The post-processor online is exactly the same as offline but I would firstly recommend you to post-process offline using Paraview where it will not only run better but you will have more options available to you in order to extract data. Like I mentioned, extracting your desired data (shear stress from speed) will require some research. Are you referring to shear stress on the blade or the fluid? I would assume the blade. The shear stress formula dictates the need to obtain force and area.

Area is simple enough, you can extract the blade surface area via your CAD tool or directly through ParaView itself. Let me know if you want to extract it through ParaView, though I would just typically get that value from CAD as it would be more “reliable”.

Force on the other hand is where your speed should come in and unfortunately this is a little more complicated. If we just look at velocity, you will realize that you cannot translate velocity into force due to F=MA. However, your blade is rotating so I believe you can use Centripetal Acceleration to deduce your force.

@jousefm @1318980 Any ideas on how to extract the needed data from the surface in ParaView? I was thinking surface normals.

Regards,
Barry


#14

I forgot to mention: I want to calculate the shear stress in the fluid, not on the blade.


#15

This explains shear stress from the velocity gradient tensor I think:

and the velocity gradient tensor can be calculated using same techniques here, in paraview:
http://www.iesensor.com/blog/2016/01/24/using-paraview-to-visualise-vorticity-lambda2-vorticity-q-criterion/

I have not tried this, so any experience you gain would be really appreciated if you could share it here. Weather this technique works, and produces expected results?

Cheers,
Darren

Edit: Also, I am interested in finding out why the shear stresses are interesting to you, and what it means in your design. Thank you.


#16

Hi @achristiaens, did you give this a go? do it help at all?

Looking forward to hearing from you,
Darren


#17

Hi :slight_smile:
Unfortunately, I did not get any further with this. I have no problems running the simulation again, so could you please give me more info:
under “Results control”, should I add scalar control? If yes, I tried, but still do not understand how I can calculate volume average from this…

I wish to calculate these velocities and shear stresses because I culture human bone cells in this setup and want to understand the forces that they endure, as part of my thesis project!

Can you help me with this velocity average?

PS: I use ParaView now :slight_smile:


#18

I think for fluid shear stresses follow the links I posted above, using paraview offline. As for average velocity, I imagine you could do something with integrate in respect to volume then multiply by the total volume, you might need to balance the units to get a m/s.m3 then multiply by total volume m3.

Hope this gets you somewhere,
Darren


#19

Hi Darren,

I was able to calculate the volume average velocity by first calculating the volume of each cell (see https://www.paraview.org/Wiki/Python_calculator_and_programmable_filter), then extracting an Excel-file with all data, and then I just did sum_i(volume_i*magnitudeU_i)/sum_i(volume_i) for all cells of the mesh. Might not be the ideal way, but it works.

All of the above is done in ParaView and Excel!

For the shear stress I will start trying now :slight_smile:


Mixing of powder and liquid in a tank
#20

A post was split to a new topic: Post-Process Results