Results for forces and pressure in symmetry case

Hi all,

This is my very first post here in SimScale and I hope anyone of you can help me with my issues. :smiley:

Well, first I will introduce you to my case. I am simulating the flow around a loudspeaker which is symmetric, actually it is a revolution. In my first simulations the air flow comes from the front (positive Y direction). See the figure.

image

First I started simulating the half of it due to its symmetry, then I realised that it would also make sense to simulate only a quarter of it in order to do some fast tests while I was trying to find the proper mesh and simulation parameters. Besides that, since it is a very simple case, I decided to try to mesh the whole loudspeaker in order to compare the results obtanied before in the symmetrical approximations. And here comes the problems… :persevere:

As I am interested in the forces generated due to te wind I used the Result Control Item “Forces and moments” and I figured out that the software is not accounting for the symmetry boundary condition when it comes to compute the forces and moments. As you can see in the following extracts from the log files.

full case

forces forces10 output:
    sum of forces:
        pressure : (0.00825357707279 87.5149908496 -0.0124717628806)
        viscous  : (-0.000623983437007 -0.380113391217 8.91044105778e-05)
        porous   : (0 0 0)
    sum of moments:
        pressure : (0.00366623415021 9.70446114237e-08 0.00358251252754)
        viscous  : (-4.85506514759e-06 -1.54834589139e-06 2.53780385519e-05)
        porous   : (0 0 0)

----------------------------------------------------------------------------

half case

forces forces10 output:
    sum of forces:
        pressure : (0.063647549237 84.2962878434 -107.410562012)
        viscous  : (5.27459856009e-05 -0.310844783919 -0.058027945383)
        porous   : (0 0 0)
    sum of moments:
        pressure : (11.5147196512 0.00198231280144 0.011050221433)
        viscous  : (-0.010047362284 3.77202129623e-06 2.28325651371e-06)
        porous   : (0 0 0)

----------------------------------------------------------------------------

quarter case

    forces forces10 output:
        sum of forces:
            pressure : (54.0351529408 43.4167688264 -54.0068510305)
            viscous  : (0.0347102611718 -0.15713904296 -0.0346544658095)
            porous   : (0 0 0)

I can understand that the forces in X and Z directions vary between cases since symmetry is not handled as such, but I don’t understand the noticeable differences in the force in the Y direction (wind direction) in the three cases. Can anyone explain me where they come from? I also attach the pictures of the pressure distribution in all the cases, which is related to the resulting force.

Cananyone explain me why the max pressure in the full case is less than half the max pressure in the symmetry cases? Did I make any mistake defining the symmetry BC’s?

Any hint will be very much appreciated! :grinning:

Best regards,

Alex

Hi Alex!

Firstly welcome to SimScale! Regarding your simulation and in particular where you are obtaining your data from, when mean wind forces I assume that would be after the speaker is being “used” and the wind velocity on the outlet of the megaphone? Where exactly? At the front outlet? A meter in front of the outlet? I’m not very clear on this.

Your assumption about symmetry are likely correct as in this particular case the values can’t simply be obtained by running a quarter or half case unless its for some rough calculations that you just multiply as accordingly to how you’ve “cut” the geometry but I wouldn’t think that would be a good assumption :stuck_out_tongue:. So doing the full simulation would be best since it is a simple geometry anyway like you’ve mentioned.

So in general the first paragraph basically brings me to your possible first area where the data might be wrong, collection and obtaining of the data. The second thing which is more obvious is the fact that your general cell size is very very coarse. If you refer to all of your pressure pictures you can see “blocks” at the outlet of your megaphone and without even checking the criteria, it is very likely that it is much too coarse.

Fixing this has two approaches. The first is to decrease your general cell size to say 0.1m per cell. That means you take the bounding box encompassing the entire megaphone and dividing it by 0.1 to obtain the number of cells that can then be input under “Bounding Box geometry primitive” in the x, y and z direction. The second approach would be to simply assign a region refinement to the background bounding box, set it to refine internally, and increase the refinement level to say 2 or 3. Both of which will give you the accuracy you should need. Of course assuming that the way you’ve collected your data is correct in the first place.

So this solves the immediate problems I see, now lets move on to your simulation.

Your simulation itself is generally alright except that the sides should be a “slip” wall instead of “no-slip” as you don’t want the flow to interact with your results. With that being said, do increase your background bounding box size back in your mesher so we can negate these effects as much as possible.

Also do note about your simulation control you want to save the last timestep which is the most important so lets say you have an end time of 1500, you want to save say every 750 timesteps in order to obtain a middle data point and a final data point. Convergence wise it seems quite sufficient for the number of the timesteps so do maintain that end time.

There are a couple more things we can do to increase accuracy but lets work on these larger issues first.

Do let me know if you need further clarification on anything else.

Cheers.

Regards,
Barry

Hi Barry!

Thanks for your quick response! I’m surprised! :smiley:

Well, first of all, I appreciate yout hints regarding the improvement of the accuracy of my mesh, I will take it into account. However, it’s something I’m not concerned about at this moment since my concerns go into another direction. As per your words, I understand that symmetry BC’s just don’t work

Your assumption about symmetry are likely correct as in this particular case the values can’t simply be obtained by running a quarter or half case unless its for some rough calculations that you just multiply as accordingly to how you’ve “cut” the geometry but I wouldn’t think that would be a good assumption :stuck_out_tongue:. So doing the full simulation would be best since it is a simple geometry anyway like you’ve mentioned.

I don’t understand why I can’t use the benefits of symmetry in this case, I have a (very) symmetrical geometry undergoing symmetrical loads, why shouldn’t I use symmetric BC’s?

As I was just starting to simulate this case I wanted to get quick results, that’s why I didn’t want to make use of a heavy mesh and I preferred to use a split mesh and leverage symmetry BC’s. Besides that, the fact of using a smaller symmetrical mesh gives me the opportunity to make use of a finer mesh, as you suggested, but only in a part of the overall system, reducing the computation time to get the results. However, it seems that the results I get this way are somewhat “weird”.

Regarding my data, I don’t know what you mean when you say that my data may be wrong. Maybe my description of the case was a little vague, I will give more details about it. In this case, the wind comes from the front of the speaker (positive Y direction) with a velocity of 40m/s (see the picture below). This simulation is to verify that the product will pass a wind tunnel test with this wind velocity as an objective. I have to say that this speaker will not be tested alone but it will be tested a 8-speaker stack mounted on a steel frame (the geometry can be already found in my project folder). However, as I already said, this first simulation is to find out the proper set up for the final case. Besides that, relevant information can be retrieved by means of this simple case without the need of running the global case with the 8-speaker system.

For the next simulations I will give a “slip” condition to the boundary walls, as you suggested, and maybe I could also increase the X and Z sizes of the bounding box, as you also suggested.

But first of all, I would like to really understand why I can’t use symmetry BC’s in this case to get the correct pressure field. In the end, my interest, beyond what I already said, is to simulate just the half of the 8-speaker system in order to minimize the cell number and computing time whereas obtaining accurate results.

I hope this helps you understand my simulation and why I am so interested in getting good results by means of the use of symmetry BC’s.

Thanks again for your help.

1 Like

Hi Alex,

This is true and normally I wouldn’t have issues with it, however from my limited understanding of the case I assumed that you wanted to somehow measure airflow outside and away from the speaker and having just a symmetry might negate how the flow progresses that you would normally see with a full geometry. Now that I think about it however, your simulation is a steady-state so you don’t expect much changes for the resultant airflow as it exits the nozzle of geometry and with your clarification of what the intended purpose of the simulation is, symmetric boundary conditions will be no problem in this case.

As mentioned, I wasn’t clear on the specific data point you wanted to capture so I was suggesting double checking what you want to deduce and find out so as to know where we can tackle.

A wind tunnel test? As in making sure flow exits the front of the speaker completely and doesn’t ingest back into the inlet at that speed?

Very interesting, I look forward to that!

Cheers.

Regards,
Barry

Hi Barry!

I really appreciate your hints and your interest in my case.

I think about it however, your simulation is a steady-state so you don’t expect much changes for the resultant airflow as it exits the nozzle of geometry and with your clarification of what the intended purpose of the simulation is, symmetric boundary conditions will be no problem in this case.

I agree with you, it should be no problem and I think I have found something that might be causing the difference in the pressure, I hope that’s the main cause althoug the pressure difference between the full case and the half case is very big. See the pictures below, now I reoriented the full case by clipping by another plane and I found an important difference.

full case (clipped by the Z plane)

half case

quarter case

As you can see, the position within the bounding box is a little different, I can’t remember why I located the speaker nearer the inlet in the symmetrical cases, I have done a lot of mesh tests :sweat_smile:

How important is to compare images with the same orientation :flushed:

Now I’m gonna simulate all options making sure that the speaker is situated in the same point and I will get back to you!

The main goal of the wind tunnel test is to verify that the whole stack of speakers, actually the frame of it, can hold the wind loads.

I can’t wait to simulate again the simple cases to be able to start the simulation of the whole system! :grin:

Thanks again!

Regards,
Alex

Finally, everything makes sense! :grinning::grinning::grinning:

Here you are the results I got when the position in thebounding box is the same in all cases (full, half and quarter geometries).

FULL CASE

HALF CASE

QUARTER CASE

As you can see, pressure fields are the same in all cases!

Besides, here you are the forces generated due to the wind in all 3 cases:

full case

forces forces10 output:
    sum of forces:
        pressure : (0.00825357707279 87.5149908496 -0.0124717628806)
        viscous  : (-0.000623983437007 -0.380113391217 8.91044105778e-05)
        porous   : (0 0 0)
    sum of moments:
        pressure : (0.00366623415021 9.70446114237e-08 0.00358251252754)
        viscous  : (-4.85506514759e-06 -1.54834589139e-06 2.53780385519e-05)
        porous   : (0 0 0)

----------------------------------------------------------------------------

half case (new)

forces forces10 output:
    sum of forces:
        pressure : (60.4848872382 43.7769499342 -0.00414707762603)
        viscous  : (0.0397436511769 -0.189406795305 -0.000107988218761)
        porous   : (0 0 0)
    sum of moments:
        pressure : (0.00105349763548 4.11542417013e-07 6.04559003522)
        viscous  : (-1.74950006388e-06 2.46336318257e-05 -0.00683926760319)
        porous   : (0 0 0)

----------------------------------------------------------------------------

quarter case (new)

orces forces10 output:
    sum of forces:
        pressure : (30.230585227 21.8761399642 30.2326819049)
        viscous  : (0.0199233908974 -0.0950959477436 0.0201493297246)
        porous   : (0 0 0)
    sum of moments:
        pressure : (-3.02551435717 0.000818213967908 3.02393830547)
        viscous  : (0.00340518591207 -1.46244726998e-05 -0.00343273773321)
        porous   : (0 0 0)

Here you can find that forces in the Y direction are exactly divided by 2 from one case to the following one, which makes totally sense to me.

So problem solved so far! :smiley:

Let’s see if more problems show up in this project…

Cheers

2 Likes

Hi Alex,

Things are looking good! Remember do ensure your mesh is more fine at the empty domains within and around your speaker. It should help with accuracy and acceptance of your results as valid.

Do keep me updated!

Cheers.

Regards,
Barry

Hi Barry!

Here I go with more results after following tour suggestions regarding the refinement.

Original situation

1st refinement

2nd refinement

Result force comparison

centered quarter (new)

forces forces10 output:
    sum of forces:
        pressure : (30.230585227 21.8761399642 30.2326819049)
        viscous  : (0.0199233908974 -0.0950959477436 0.0201493297246)
        porous   : (0 0 0)
    sum of moments:
        pressure : (-3.02551435717 0.000818213967908 3.02393830547)
        viscous  : (0.00340518591207 -1.46244726998e-05 -0.00343273773321)
        porous   : (0 0 0)

----------------------------------------------------------------------------

centered quarter (new_refined)

forces forces10 output:
    sum of forces:
        pressure : (30.5763305387 22.3407652994 30.5597004226)
        viscous  : (0.0174370075665 -0.0771660539596 0.0181454247264)
        porous   : (0 0 0)
    sum of moments:
        pressure : (-3.02828207157 0.000828241651288 3.03043746709)
        viscous  : (0.00285316063203 -4.4053224206e-05 -0.0029176657417)
        porous   : (0 0 0)

----------------------------------------------------------------------------

centered quarter (new_refined_2)

    sum of forces:
        pressure : (31.1750362891 22.6925740435 31.1580805624)
        viscous  : (0.0172395971639 -0.0724233235703 0.017752807426)
        porous   : (0 0 0)
    sum of moments:
        pressure : (-3.04948988534 0.000841997544967 3.05169520933)
        viscous  : (0.00272665221567 -3.00254234555e-05 -0.00276790026688)
        porous   : (0 0 0)

Besides, the most refined case had a better convergence than the original one, so the case was running for less time despite the greater number of cells (around 200.000 cells more, which is not much but it’s around 25% more cells, more or less).

Now I’m trying to improve my results by making the bounding box a little bigger, as you also suggested, keeping the cell sizes as in the last refinment I did. Also, I will test the same case with “slip” boundary condition on the walls to see the difference.

Any other hint or suggestion? :yum:

Thanks for everything, Barry!

Best regards,

Alex

1 Like

Hi Alex!

The last and most refined run looks good. As you can see, the “blocks” of data are now gone and you have a very smooth transition of pressure gradients. Although your results do not deviate that much but maintaining accuracy while keeping computational costs low is always an important consideration.

All goo so far, do let me know of the results when you’re done running the refined and larger bounding box!

Cheers.

Regards,
Barry

Hi again!

Here you go the results of a bigger bounding box (it’s not much bigger, but it makes a difference).

Original bounding box ( P )

Bigger bounding box ( P )

Bigger bounding box ( U ) (“no-slip” wall condition)

Bigger bounding box ( U ) (“slip” wall condition)

As you can see, in this case using a “slip” wall condition just makes a very little difference since it only affects the near wall area but not the loud speaker itself.

EDIT: Btw, why am I not able to see the velocity profile by cells, but I can only see it by points? If I change the visualization to cells I can only see the components but not the magnitude, that’s why I attached the points visualization, not the cells.

Hi Alex,

Looking at this, then I would recommend sticking with the parameters of your last figure, the “Bigger bounding box ( U ) (“slip” wall condition)”.

Velocity is a vector quantity and as such requires a direction and magnitude hence I suspect that the solver assigns the data to points which is feasible in the case of dealing with velocity. Pressure on the other hand is a scalar quantity meaning it only has a magnitude and no direction, hence the solver can compute the average pressure for all points within the cell to generate the average pressure of the cell, something which is not really applicable for magnitude as the solver dosen’t know which is your wanted direction of the velocity magnitude.

Cheers.

Regards,
Barry

Hi Barry,

Looking at this, then I would recommend sticking with the parameters of your last figure, the “Bigger bounding box ( U ) (“slip” wall condition)”.

That’s exactly what I was going to do! :upside_down_face:

Velocity is a vector quantity and as such requires a direction and magnitude hence I suspect that the solver assigns the data to points which is feasible in the case of dealing with velocity. Pressure on the other hand is a scalar quantity meaning it only has a magnitude and no direction, hence the solver can compute the average pressure for all points within the cell to generate the average pressure of the cell, something which is not really applicable for magnitude as the solver dosen’t know which is your wanted direction of the velocity magnitude.

I don’t know wether that’s true or not… Now I’m talking from memory and I might be wrong, it’s been a while since my last simulation in OpenFOAM, but I think I was able to watch cell velocity results in magnitude in Paraview. Now I am using the post processor provided by SimScale, wich I guess is a lightweight version of it, and it gives me troubles when trying to do that. Could you (or anyone else) confirm that in Paraview it’s possible to display correctly the velocity field? Now I’m unable to do that because I just started working on a new job and I don’t have a proper laptop to do that at the moment, it is actually a 32 bits computer…:persevere:

Best regards,

Alex

Hi Alex,

I wouldn’t mind post-processing for you in paraview in this case. Gotta put my shiny new 32GBs of RAM i just bought a month ago to good use :stuck_out_tongue:

Can you refer to me which simulation and run is the one you want to check?

Cheers.

Regards,
Barry

Hi Barry,

Sorry for the late replay, I just saw you wrote me.

If you want to try your flaming new machine you can try the last of my simulations, named “Copy of Incomp_speaker_centered_man_mesh_quart_ref_bbox_in0.56” (I know, maybe I “overnamed” it, my fault :blush:).

Btw, a while ago I managed -unintendedly and with some luck, I suppose- to display the magnitude velocity field by cells, not points. However, when I tried to display the color bar it disapeared and turn to blue!

That’s a mistery… I think I discovered this malfunctioning in my previous job place some months ago, but I had no time to report or comment…

Enjoy your post processing! :yum:

Hi Alex,

ParaView does have cell velocity results and you see in 1. I’ve also posted for point for both pressure and velocity as well as in 5 where you can see the mesh.

1

2

3

4

5

Cheers.

Regards,
Barry

3 Likes

Hi Barry!

What a nice postpro! I love the color scheme for pressure! One question about it, is it possible to change the color scheme in the SimScale post-processor? I have been trying to but with no succes so far… :persevere:

Btw, the first simulation of the whole speaker system just finished and I was playing a little with it and now I can visualize the velocity field with no problem even with the color bar…

1 Like

Hi Alex,

I don’t believe so. I always do my post-processing offline on ParaView. Much much faster for me.

Cheers.

Regards,
Barry

I also prefer Paraview, it’s more complete, but now I’m unable to use it until I get a decent computer, so I can only post process in SimScale

1 Like