Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
Turbulent air flow through a continuously variable transmission enclosure
Posted 14 févr. 2014, 14:01 UTC−5 Fluid & Heat, Computational Fluid Dynamics (CFD), Geometry, Mesh 3 Replies
Please login with a confirmed email address before reporting spam
Hi there,
I am a student working on my senior ME project. I am hoping to model turbulent air flow through the casing of a CVT enclosure. The pulleys of the CVT will be rotating at the optimal engine RPM of 3000, and we have inlet and outlet ducts located at the top and bottom respectively. Eventually I hope to perform a chemical species transport to measure how much mixing is occurring but for now I am just trying to get the turbulent air flow to show.
I have attached some photos of the geometry of the casing with the pulleys inside of it. As you can see, I made a solid geometry and performed a boolean difference to remove the area where the pulleys are. I am attempting to use the Rotating Machinery, Turbulent k-e module to simulate this.
This is quite a big computation to perform so I have made some simpler geometries to make sure that I am applying the physics and geometric conditions correctly.
My problem is meshing.
So far I have just been trying the standard physics induced mesh and I have only been changing the overall mesh quality. When using the finest mesh settings, I get an error stating that a few cavities have been made inside of the geometry. If i use a normal mesh setting COMSOL has problems creating the mesh for the boundary layers.
I am aware this could be a number of problems such as import settings, the geometry I chose, or just lack of meshing knowledge.
I was hoping someone could take a look at the mph file and maybe give some suggestions as to settings to use when creating a custom mesh.
I have attached the photos for clarification and the mph file. Also, note that this file was created with an academic license.
Thanks for your time!
Zachary
Edit: Note that I am using a step function with a time dependent solver to ramp up the RPM of the pulleys from 0 to 10 in 10 seconds. My solver is from 0 to 10 in time steps of 0.1. Earlier I stated that the RPM of the pulleys is 3000. In the simulation I changed it to 10 just so I could see if it worked.
I am a student working on my senior ME project. I am hoping to model turbulent air flow through the casing of a CVT enclosure. The pulleys of the CVT will be rotating at the optimal engine RPM of 3000, and we have inlet and outlet ducts located at the top and bottom respectively. Eventually I hope to perform a chemical species transport to measure how much mixing is occurring but for now I am just trying to get the turbulent air flow to show.
I have attached some photos of the geometry of the casing with the pulleys inside of it. As you can see, I made a solid geometry and performed a boolean difference to remove the area where the pulleys are. I am attempting to use the Rotating Machinery, Turbulent k-e module to simulate this.
This is quite a big computation to perform so I have made some simpler geometries to make sure that I am applying the physics and geometric conditions correctly.
My problem is meshing.
So far I have just been trying the standard physics induced mesh and I have only been changing the overall mesh quality. When using the finest mesh settings, I get an error stating that a few cavities have been made inside of the geometry. If i use a normal mesh setting COMSOL has problems creating the mesh for the boundary layers.
I am aware this could be a number of problems such as import settings, the geometry I chose, or just lack of meshing knowledge.
I was hoping someone could take a look at the mph file and maybe give some suggestions as to settings to use when creating a custom mesh.
I have attached the photos for clarification and the mph file. Also, note that this file was created with an academic license.
Thanks for your time!
Zachary
Edit: Note that I am using a step function with a time dependent solver to ramp up the RPM of the pulleys from 0 to 10 in 10 seconds. My solver is from 0 to 10 in time steps of 0.1. Earlier I stated that the RPM of the pulleys is 3000. In the simulation I changed it to 10 just so I could see if it worked.
3 Replies Last Post 28 févr. 2014, 20:45 UTC−5
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Dear Zachary,
The reason for the meshing problem seems to be that the imported geometry contains a sliver face. You can remove this face by adding a "Delete Sliver Faces" node after each of the two "Import" nodes. Type 0.003 in the field "Maximum face width" to find the sliver face.
Sincerely,
Bjorn Bretz
COMSOL
The reason for the meshing problem seems to be that the imported geometry contains a sliver face. You can remove this face by adding a "Delete Sliver Faces" node after each of the two "Import" nodes. Type 0.003 in the field "Maximum face width" to find the sliver face.
Sincerely,
Bjorn Bretz
COMSOL
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi Bjorn,
Thank you for your response. I am running the simulation right now and have not run into any problems yet. It looks like I need to spend some more time reading about importing and geometry in general.
Zachary
Thank you for your response. I am running the simulation right now and have not run into any problems yet. It looks like I need to spend some more time reading about importing and geometry in general.
Zachary
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi there,
I've got the simulation running on a computing cluster using a 256GB node. I didn't modify the tolerance for the time stepping COMSOL uses to compute the solution, might this have been a bad idea? The simulation has been running for about 48 hours and as you can see the time steps are extremely small.
Here's a portion of the batch.log file:
Time-Dependent Solver 1 in Solver 1 started at 26-Feb-2014 15:23:28.
Time-dependent solver (BDF)
Current Progress: 0 %
Memory: 6899/7093 16006/16270
Number of degrees of freedom solved for: 8330414.
Symmetric matrices found.
Scales for dependent variables:
Angular displacement (comp1.rmspfomegard1): 1
Angular displacement (comp1.rmspfomegard2): 1
Nonsymmetric matrix found.
Scales for dependent variables:
Pressure (comp1.p): 0.53
Velocity field (Spatial) (comp1.u): 1
Nonsymmetric matrix found.
Scales for dependent variables:
Turbulent dissipation rate (comp1.ep): 5.6e-07
Turbulent kinetic energy (comp1.k): 5.3e-07
Nonsymmetric matrix found.
Step Time Stepsize Res Jac Sol Order Tfail NLfail LinIt LinErr LinRes
Inconsistent pointwise unidirectional constraints found.
7 constraints are merged/removed.
Inconsistent pointwise unidirectional constraints found.
7 constraints are merged/removed.
More inconsistent pointwise unidirectional constraints found...
0 0 out 98 80 98 0
Group #1: 25 26 25 0 0
Group #2: 25 5 25 46 6.7e-05 1.2e-06
Group #3: 48 49 48 132 4.3e-06 4.1e-09
1 6.7113e-08 6.7113e-08 102 84 102 1 0 1
Group #1: 26 27 26 0 0
Group #2: 26 6 26 48 0.00022 4.7e-06
Group #3: 50 51 50 138 5.7e-06 6.7e-09
2 1.3423e-07 6.7113e-08 106 88 106 1 0 1
Group #1: 27 28 27 0 0
Group #2: 27 7 27 50 0.00019 3.5e-06
Group #3: 52 53 52 144 5.8e-06 5.7e-09
The solution is going to be used qualitatively to figure out if 'sufficient' mixing is happening inside of the case. What would be a good tolerance to use so that the computation time is reduced but the solution is not completely wrong?
I've attached the mph file if this helps.
Thanks,
Zach
I've got the simulation running on a computing cluster using a 256GB node. I didn't modify the tolerance for the time stepping COMSOL uses to compute the solution, might this have been a bad idea? The simulation has been running for about 48 hours and as you can see the time steps are extremely small.
Here's a portion of the batch.log file:
Time-Dependent Solver 1 in Solver 1 started at 26-Feb-2014 15:23:28.
Time-dependent solver (BDF)
Current Progress: 0 %
Memory: 6899/7093 16006/16270
Number of degrees of freedom solved for: 8330414.
Symmetric matrices found.
Scales for dependent variables:
Angular displacement (comp1.rmspfomegard1): 1
Angular displacement (comp1.rmspfomegard2): 1
Nonsymmetric matrix found.
Scales for dependent variables:
Pressure (comp1.p): 0.53
Velocity field (Spatial) (comp1.u): 1
Nonsymmetric matrix found.
Scales for dependent variables:
Turbulent dissipation rate (comp1.ep): 5.6e-07
Turbulent kinetic energy (comp1.k): 5.3e-07
Nonsymmetric matrix found.
Step Time Stepsize Res Jac Sol Order Tfail NLfail LinIt LinErr LinRes
Inconsistent pointwise unidirectional constraints found.
7 constraints are merged/removed.
Inconsistent pointwise unidirectional constraints found.
7 constraints are merged/removed.
More inconsistent pointwise unidirectional constraints found...
0 0 out 98 80 98 0
Group #1: 25 26 25 0 0
Group #2: 25 5 25 46 6.7e-05 1.2e-06
Group #3: 48 49 48 132 4.3e-06 4.1e-09
1 6.7113e-08 6.7113e-08 102 84 102 1 0 1
Group #1: 26 27 26 0 0
Group #2: 26 6 26 48 0.00022 4.7e-06
Group #3: 50 51 50 138 5.7e-06 6.7e-09
2 1.3423e-07 6.7113e-08 106 88 106 1 0 1
Group #1: 27 28 27 0 0
Group #2: 27 7 27 50 0.00019 3.5e-06
Group #3: 52 53 52 144 5.8e-06 5.7e-09
The solution is going to be used qualitatively to figure out if 'sufficient' mixing is happening inside of the case. What would be a good tolerance to use so that the computation time is reduced but the solution is not completely wrong?
I've attached the mph file if this helps.
Thanks,
Zach
Attachments: