35 Replies Latest reply on Sep 27, 2013 1:54 PM by Jared Conway

# Big difference between 2D and 3D Computational Domain

hello thank you in advance anyone who can help me in this problem that is going bugging me.

I am studying a particular type of valve (Tesla's Valve) has conducted a very simple I want to know the pressure difference between inlet and outlet, before sliding the fluid in one direction and then in the other (I use 2 configurations), the fluid is water and there are no particular boundary conditions.

The problem is that when I study in 2D Computational Domain, the values are similar to those of a scientific comparison but when I use 3D computational domain (that of my greatest interest), the values of pressure decrease dramatically.

I don't understand why this difference of values should be similar, I do not know what to do to find the cause.

If someone would be so kind as to help me I would be very grateful, I could provide more specific details.

• ###### Re: Big difference between 2D and 3D Computational Domain

hi, the file you uploaded doesn't look like it has your flow simulation studies in it. that would be the most helpful in understanding what might be causing any troubles.

my first thought is, comp domain size, doesn't matter here because you have an internal analysis. it is what it is. however have you checked what the mesh looks like in both studies?

the next thing, what kind of assumptions have you decided upon for this analysis? with a 2d analysis it assumes this part is infinitely wide. with 3d you have the wall effects. is that ok? in that case, your 3d could be right and your 2d wrong. you could try ideal walls to eliminate this and add an assumption to try bringing the 2 results together.

the last is, have you looked at percentage differences between the 2d and 3d, in both directions? the one that you marked as "wrong" has the least percent difference between the 2 solutions. it might be helpful here to include your reference results and the experiment that they did.

• ###### Re: Big difference between 2D and 3D Computational Domain

I uploaded the file containing the simulation, I hope you can open it, I tried to put all the walls as ideals (except the caps). Unfortunately, the situation is not improved, apparently worst, the values are still very different. probably omit some of the settings.

This should be developed for my thesis, I am very worried because the results are not what you expected ...

thanks again for your kind cooperation.

• ###### Re: Big difference between 2D and 3D Computational Domain

also, before we go much deeper, lets make sure we check the other things i mentioned:

1. what are the assumptions in both your model and the equations

3. check the mesh between the studies, also, have you improved the mesh, are the results converged?

4. something i didn't think about before, you look like you're looking at one specific parameter. have you looked at other parameters and the flow in general. if you're doing the thesis on this, you're obviously the expert in this, is everything else working? if you make a change, do the trends in the results look like they make sense? unless you spend a LOT of time making sure the physical and simulated tests match, there may always be a delta but between studies in the simulation, you should be able to find some trends.

• ###### Re: Big difference between 2D and 3D Computational Domain

hi

between the various tests have also tried to densify the mesh, or to vary the values of mash in correspondence of the narrow channels, the result is always the same.

L 'equation that characterizes the main conduit is the diocidity ie the ratio between the pressure difference between reverse flow and direct one, as an attachment file in excel.

To clarify the operation of the calcolatroe I also tried to calculate the pressure drop in a simple pipe, and also in this case the values of DeltaP in 2D are significantly higher than those of 3D.

• ###### Re: Big difference between 2D and 3D Computational Domain

hi, can you post your reference for the pipe example? if you were doing 2d, remember it assumes an infinitely wide pipe with no walls and also square. if you compare to a 3d pipe, they aren't equal comparisons.

this is why i recommended doing this test. until you can get that to match, you're probably going to have trouble matching your other geometry. but once you figure out the missing peice of the puzzle, you can apply the model that you're actually testing to get more exact results. without knowing the history/background and seeing the reference for your model (equations, testing data, experiment explanation) it is very difficult to know where you may have missed the mark in the setup such that you're getting results that don't check out.

if you're concerned with accuracy, you could double check the sample problems in the documentation. if they run correctly, then it is likely not a software issue but rather a setup issue which brings us back to where we are before. we need more information about the physical case to confirm your setup for the simulated case.

• ###### Re: Big difference between 2D and 3D Computational Domain

hi.

I do not know if you can open the last attachment that contains the calculation for the conduit, then you may need a rebuild. but if it's better I can also attach the hose to the simple test.

unfortunately there is not much literature on this subject and what there is data calculated with fluent and include a study taking into account the symmetry of the duct, ie a study 2D

but there is really cheap

• ###### Re: Big difference between 2D and 3D Computational Domain

i opened your model, i can see the simulation setup now but only the 2d case so it is hard to see if you've made a setup error between 2d and 3d.

like i said, i'd go with getting yourself comfortablewith the results of a square pipe in 2d and 3d. once you have that then you can move to your model.

one other comment, comparing simulation with simulation is kind of moot. how do you know the previous simulation was correct? do you know all your boundary conditions are exactly the same? do you know that if they had run 3d that it wouldn't come up with different results? right now your 2 2d simulations match so simulation vs simulation sounds good. why use flow if you have fluent?

• ###### Re: Big difference between 2D and 3D Computational Domain

I do not know much about CFD therefore mistakenly thought that the 2D results could be a way to lighten the simulation calculations but in the end 2D and 3D must somehow be similar.

The article with which I compared my results is the one attached [2.2 numerical models], they use fluent not me

I enclose also the simple tube which I used to do the test, creating two configurations where I simply changed the computational domain, but I think not much use...

Now it gets complicated.

I thank again you for your kind support

• ###### Re: Big difference between 2D and 3D Computational Domain

2d is infinite

3d has boundaries

but you're correct on its application

thanks for hte article as well. if you look at section 2.3 they show that they tested against physical and it was within 10% (figure 3). so by association, their model is within 10% yours may be within 10% of physical.

now you've moved to 3d and you're having trouble. are you sure your 3d model matches the one tested exactly?

regarding your pipe test. what are you comparing it against?

• ###### Re: Big difference between 2D and 3D Computational Domain

good morning

I think that this last is the right way, in fact my 3D values don't differ that much from the experimental results of the article, I think that I shall in this way.

I would ask you the last courtesy, can  I see the mesh in 3D or is only possible with the cut plot on a plane?

thank you very much for the valuable assistance

• ###### Re: Big difference between 2D and 3D Computational Domain

use the "mesh" item in the results tree

• ###### Re: Big difference between 2D and 3D Computational Domain

hi

ok thanks, I saw that I had to select which cells choose to display .... Selecting fluid cells show me the mash of the fluid within the volume, while solid cells is it?

In case you want to make a kind consideration I am attaching the mash to view, I would be grateful.

One more thing to summarize ...

2d is infinite, no roughness, no wall effect

3d has boundaries, eventually roughness, wall effect, Reynolds number effect

... anything else?

thanks again

as you have already understood'm new to the forum, how can I do to increase your score? Is there a way

• ###### Re: Big difference between 2D and 3D Computational Domain

Hi Filippo, there are some good definitions of what the different cells are in the help file. But really solid cells are anything that aren't fluid.

the only comment on your summary is that 2d can have roughness on the walls that are involved. just not on the end walls.

also, not sure what you mean by reynolds number effect for 3d.

on the points, just make sure to choose the most "correct" answer and thats good enough.

• ###### Re: Big difference between 2D and 3D Computational Domain

from a quick glance of the jpg. What do you think of these mash?

I know it is not easy to be a picture, but it seems there are no obvious errors ...

• ###### Re: Big difference between 2D and 3D Computational Domain

looks uniform which is good

looks like you have more than 2-3 cells across a channel which is good

defintely a good starting point

• ###### Re: Big difference between 2D and 3D Computational Domain

ok well. What would you advise me to do to improve the study of the flow?

Perhaps, for example, I would think to try to densify the mash only near the walls. There is a choice to do it? May be right?

• ###### Re: Big difference between 2D and 3D Computational Domain

don't worry about improving the mesh at the wall. the software is built to handle the wall in an efficient way. improve the mesh overall and use the adaptive meshing to do a bit of a mesh convergence check. if you improve the mesh and the results don't change you have "the answer". if it still doesn't fit the physical, you will have to confirm that what you have is an exact replica of the physical test. i might also suggest varying some of the parameters slightly to see how the system behaves from a trend perspective. if the trends are right, it might just be that your model doesn't match physical.

• ###### Re: Big difference between 2D and 3D Computational Domain

excuse the question, what do you mean by "mesh convergence check", you mean act on command "auto" in the window where you can manage the size of the cells?

I've done many tests significantly changed the thickening cells, but the result is not aprezzabile, when I check the "auto" the results differ very little

• ###### Re: Big difference between 2D and 3D Computational Domain

mesh convergence is the act of improving the mesh to the point where the solution doesn't change. ie, you have not only a converged solution (Which will happen 99% of the time in flow) but also one that is converged with respect to the mesh (the results aren't dependent on the mesh).

auto is the automatic meshing mode, it is good for general meshing, at minimum you should have a min wall or min gap

manual is nice because you can make improvements manually based on what you know about the geometry.

adaptive is in the calc control options, this adapts the mesh to the results.

• ###### Re: Big difference between 2D and 3D Computational Domain

good day

I'm still here ....

about last post, I'm trying to improve the results of the calculation, but in a model, by varying the values of the mesh I get results that vary both up and down and do not converge to any value.

Reading the last things that I have written seems to understand that it would be wrong to use automatic meshing mode in the menu "Initial Mesh", but to improve the mesh convergence must act on adaptive meshing calc menu control options, but unfortunately I can not find it ...

thanks

• ###### Re: Big difference between 2D and 3D Computational Domain

I'm sorry but I can not find it here

• ###### Re: Big difference between 2D and 3D Computational Domain

refinement tab

• ###### Re: Big difference between 2D and 3D Computational Domain

I've only this...

• ###### Re: Big difference between 2D and 3D Computational Domain

hi filippo, i'd recommend checking out the help on this section. here's a trick that you can use while you learn the interface. choose the reset option. choose the option to reset everything and set it to level 6 or higher. that will automatically setup solution adaptive.

• ###### Re: Big difference between 2D and 3D Computational Domain

ok I'll do it!

I therefore begin by setting the initial mesh on automatic setting, reset refinement and see what changes. might be right?

• ###### Re: Big difference between 2D and 3D Computational Domain

solution adaptive refinement starts with your intiial mesh and then during the solution process will refine it.

• ###### Re: Big difference between 2D and 3D Computational Domain

hi

I have tested changing refiniment, but the result does not change.

Instead, I've noticed that changing end conditions, from "if one is satisfiad" to " if all are satisfiad" I notice a little change.

putting in final conditions if all are satisfated you make mistakes?

• ###### Re: Big difference between 2D and 3D Computational Domain

you're going to have to elaborate on what "result does not change" means.

the result may not change, that would be good!

but you should check that refinement process is happening by checking the mesh. also, you should hceck why the solution completed. if it doesn't complete because the refinements are completed, then you're stopping the solution early.

• ###### Re: Big difference between 2D and 3D Computational Domain

hi
i'm here again with another question always regarding my valve....

I have the inlet mass flow (5 l/min), the outside pressure condition ( 1 atm ambient), and I would like to know the inlet pressure and the outlet pressure of the flow of my conduct.

In the boundary condition I put 5 l/min on the inlet face, 1 atm on the outlet face,  beliving that 1 atm in the ambient pressure. Dosen't it ?

But when I set goal 1 static pressure on the inlet surface and static pressure on the outlet pressure, i think that the program return me inlet and outlet pressure, but the result is not what I expected.
is different from what they read the gauges on my material model .

I do not mean that the pressure on the face of output is 1 atm because that is just the pressure I want to know, 1 atm is the ambient
how can I do ?

thanks a lot

• ###### Re: Big difference between 2D and 3D Computational Domain

Hi Filippo, you might want to make a new thread for this new issue. I'm having a tough time invisioning what you're trying to do, in that thread, maybe you could post a picture of your model. For example i can't tell if you're doing internal or external analysis. You also mention that the results aren't what you expect. Are you saying they don't match what you're seeing on the screen like a cut plot, or the values you've calculated in advance. If the latter, how did you come up with those values? Remember that in flow simulation everything uses absolute pressure, not gauge.

The only thing I can comment on is if you have a pressure opening, you are setting the pressure. If you have a mass flow or volume flow rate or velocity on the other end, your output will be the pressure at that side. So you can calculate the pressure drop. If you don't know whta the pressure is at that outlet, you need to add the next component in your assembly.

• ###### Re: Big difference between 2D and 3D Computational Domain

Ok I'll try to open a new thread

yes I don't know what is the pressure at that outlet of valve, i would like to know. now i think to put o long pipe, with the out side ambient pressure.

• ###### Re: Big difference between 2D and 3D Computational Domain

if you don't know what it is, you can't set it, you have to add the next copmonent or make an internal/external analysis where the pressure "spills out" into an environment