AnsweredAssumed Answered

Is my flow simulation setup correctly?

Question asked by Scott Rosoff on Aug 3, 2015
Latest reply on Aug 10, 2015 by Amit Katz

Like • Show 0 Likes0
Comment • 7

Hello,

I'm trying to find the head loss/pressure drop across an open channel for a specified flow rate. The picture below shows how the system looks.

Basically water flows through the left face, and has to go around a bunch of these glass rods, and exits out the other face.

Setup

Left face is an inlet with specified volumetric flow rate
Right face is specified to be environmental pressure
Surface goals are set up to be the average static pressure at inlet, and average static pressure at outlet
The rods are glass, so I set up a wall condition on them to have a roughness of 5 microinches (found this number online)
The base and "roof" of the system are steel, so they have a roughness of 200 microinches

Problem

I then run an internal analysis, and find that the pressure difference/head loss between the inlet and outlet is about .82 psi, which is way too high (it should be around .20 psi). This number doesn't change much if I change the roughness of the glass rods. However, if I specify 0 roughness for the glass rods, the head loss is around .20 psi, which is strange.

So, my questions are: Am I doing something wrong in my setup here? How can I diagnose this issue of having too high of a number? Is it possible that the glass rods aren't even recognized, and water's flowing straight through them?

Thanks,

Scott

No one else has this question

Outcomes

Amit Katz Aug 3, 2015 11:19 AM

How does your mesh look around the tubes? It seems to me like you're starting out with too much detail. I would try to simulate just one of these bundles, maybe even just a 2D analysis and increase in complexity from there.

Actions

Chris Michalski Aug 3, 2015 11:32 AM

I would tend to agree with Amit - at least use symmetry along the central plane if not a 2D simulation to start. Then use the fact that you just cut your number of cells in half to make sure your cylinders are refined enough (using a local mesh).

With curved surfaces being approximated as rectangular cells your mesh resolution is going to make a lot more difference in drag than your surface roughness on those cylinders.

Actions

Scott Rosoff @ Chris Michalski on Aug 6, 2015 7:38 PM

Hi Chris,

I created a local mesh to refine it around the cylinders a lot more, as well as testing using half-symmetry, and then quarter-symmetry (and reduced the inlet flow accordingly). I used a 3d comp. domain because the 2d was giving me some computing/memory issues for whatever reason. From these two tests, I get numbers that are way off from 6" of head loss again. However, there's an issue when I display the trajectories that could be related to this problem: the trajectories don't go all the way through the system, only up to a certain point. See below pics for the setup:

The local mesh around cylinders:

Half symmetry:

Trajectories that stop short:

Any idea of what's going on, or what my options are from here?

Thanks,

Scott

Actions

Chris Michalski @ Scott Rosoff on Aug 7, 2015 7:58 AM

If you're looking to compare effects of roughness you need a LOT finer mesh. If you look at the mesh without the parts I couldn't even tell that those are round pins and not square/rectangular. Until your model looks just like reality the roughness is irrelevant. On flat surfaces it would be useful, but when your circle looks like a square the rectangular approximation makes your roughness equal to your mesh roughness.

In the Flow Trajectory tab what is the length limit of the paths shown? How big is this model actually? (.001" rods, 1" rods, ?)

When you setup symmetry did you use fully 1/2 or 1/4 of the model? The highlighted yellow region does not appear to be 1/2 of the thickness.

Actions

Scott Rosoff @ Chris Michalski on Aug 7, 2015 2:15 PM

I'm not trying to compare the results of the different roughnesses, just a constant roughness of 5 microinches on the cylinders. How can I make the mesh recognize the curvature even more, because the mesh levels are already pretty high, and Solidworks can barely calculate the mesh/run the simulation as it is. The local mesh settings around the cylinders are currently:

Solid/Fluid Interface

Small Solids Features Refinement Level: 5
Curvature Refinement Level: 5
Curvature refinement criteriod: 0.389 rad
Tolerance Refinement Level: 5
Tolerance refinement criterion: 5.85in

Refining Cells

Just "Refine fluid cells" and "refine partial cells" are checked and are Level 5

Actions

Chris Michalski @ Scott Rosoff on Aug 7, 2015 2:26 PM

How many cells are you generating currently?

I would likely alter your basic mesh settings so that the mesh around the rods has the same size in both directions (look like squares instead of long thin rectangles). You are creating a lot of cells in the bulk flow region that you may not need if the flow around the rods is your primary interest.

Unfortunately, it might also benefit to put reference planes in the center of your rods so that you know they are all resolved the same. It would be a royal pain to add them, but the more effort you put into setting up the initial mesh the easier it is for the software to get the answer you are looking for. Many of my models can take several full days to create all the BCs, ICs and local meshes but when it takes a day for each configuration and you're comparing a dozen changes it gets easier to justify the time up front.

Actions

Amit Katz @ Scott Rosoff on Aug 10, 2015 8:32 AM

Scott, your mesh looks very rough. Can you post a screenshot of a mesh only cut plane, and please be sure to toggle off "use cad geometry" in the cut plane definition. I think you may be surprised at what your solids actually look like.

Actions

7 Replies

Amit Katz Aug 3, 2015 11:19 AM
Mark Correct
Correct Answer
How does your mesh look around the tubes? It seems to me like you're starting out with too much detail. I would try to simulate just one of these bundles, maybe even just a 2D analysis and increase in complexity from there.
Like • Show 1 Like1
Actions
Chris Michalski Aug 3, 2015 11:32 AM
Mark Correct
Correct Answer
I would tend to agree with Amit - at least use symmetry along the central plane if not a 2D simulation to start. Then use the fact that you just cut your number of cells in half to make sure your cylinders are refined enough (using a local mesh).
With curved surfaces being approximated as rectangular cells your mesh resolution is going to make a lot more difference in drag than your surface roughness on those cylinders.
Like • Show 1 Like1
Actions
- Scott Rosoff @ Chris Michalski on Aug 6, 2015 7:38 PM
  Mark Correct
  Correct Answer
  Hi Chris,
  
  I created a local mesh to refine it around the cylinders a lot more, as well as testing using half-symmetry, and then quarter-symmetry (and reduced the inlet flow accordingly). I used a 3d comp. domain because the 2d was giving me some computing/memory issues for whatever reason. From these two tests, I get numbers that are way off from 6" of head loss again. However, there's an issue when I display the trajectories that could be related to this problem: the trajectories don't go all the way through the system, only up to a certain point. See below pics for the setup:
  
  The local mesh around cylinders:
  Half symmetry:
  Trajectories that stop short:
  
  Any idea of what's going on, or what my options are from here?
  
  Thanks,
  Scott
  Like • Show 0 Likes0
  Actions
  - Chris Michalski @ Scott Rosoff on Aug 7, 2015 7:58 AM
    Mark Correct
    Correct Answer
    If you're looking to compare effects of roughness you need a LOT finer mesh. If you look at the mesh without the parts I couldn't even tell that those are round pins and not square/rectangular. Until your model looks just like reality the roughness is irrelevant. On flat surfaces it would be useful, but when your circle looks like a square the rectangular approximation makes your roughness equal to your mesh roughness.
    
    In the Flow Trajectory tab what is the length limit of the paths shown? How big is this model actually? (.001" rods, 1" rods, ?)
    
    When you setup symmetry did you use fully 1/2 or 1/4 of the model? The highlighted yellow region does not appear to be 1/2 of the thickness.
    Like • Show 1 Like1
    Actions
    - Scott Rosoff @ Chris Michalski on Aug 7, 2015 2:15 PM
      Mark Correct
      Correct Answer
      I'm not trying to compare the results of the different roughnesses, just a constant roughness of 5 microinches on the cylinders. How can I make the mesh recognize the curvature even more, because the mesh levels are already pretty high, and Solidworks can barely calculate the mesh/run the simulation as it is. The local mesh settings around the cylinders are currently:
      
           Solid/Fluid Interface
      Small Solids Features Refinement Level: 5
      Curvature Refinement Level: 5
      Curvature refinement criteriod: 0.389 rad
      Tolerance Refinement Level: 5
      Tolerance refinement criterion: 5.85in
      
           Refining Cells
           Just "Refine fluid cells" and "refine partial cells" are checked and are Level 5
      Like • Show 0 Likes0
      Actions
      - Chris Michalski @ Scott Rosoff on Aug 7, 2015 2:26 PM
        Mark Correct
        Correct Answer
        How many cells are you generating currently?
        
        I would likely alter your basic mesh settings so that the mesh around the rods has the same size in both directions (look like squares instead of long thin rectangles). You are creating a lot of cells in the bulk flow region that you may not need if the flow around the rods is your primary interest.
        
        Unfortunately, it might also benefit to put reference planes in the center of your rods so that you know they are all resolved the same. It would be a royal pain to add them, but the more effort you put into setting up the initial mesh the easier it is for the software to get the answer you are looking for. Many of my models can take several full days to create all the BCs, ICs and local meshes but when it takes a day for each configuration and you're comparing a dozen changes it gets easier to justify the time up front.
        Like • Show 0 Likes0
        Actions
  - Amit Katz @ Scott Rosoff on Aug 10, 2015 8:32 AM
    Mark Correct
    Correct Answer
    Scott, your mesh looks very rough. Can you post a screenshot of a mesh only cut plane, and please be sure to toggle off "use cad geometry" in the cut plane definition. I think you may be surprised at what your solids actually look like.
    Like • Show 0 Likes0
    Actions

Is my flow simulation setup correctly?

Outcomes

Related Content

Recommended Content