Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.

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.

Inlet / Outlet Mass Flow in HMNF

Please login with a confirmed email address before reporting spam

In an attempt to understand inlet and outlet conditions I have constructed a simple model (attached below). It is an axisymmetric tube, L/D=100, Pin/Pout=10, Nitrogen gas, inlet of Pin & 273.15, Ma=0, outlet of Hybrid. I then calculate the flow using Normal Mesh and the default solver. To check conservation, I took a line integral on the inlet and a line integral on the outlet, both of the z component of velocity times density.

w2*hmnf.rho (kg/(m*s))
0.05070548906795929 0.05475974936528191

The difference is almost 10%. I would like to get some feedback as to how that discrepancy can be reduced.

T.C. Lilly


4 Replies Last Post 17 avr. 2014, 10:16 UTC−4

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 14 avr. 2014, 14:12 UTC−4
Received from my Technical Sales Manager:

I noticed that you had a question on our Discussion Forum, and I thought I should get back to you with some help.

As a first possible approach to get the overall mass balance to work out, I would consider setting up a denser mesh, to see if this will improve your results. You could even set up a mesh sensitivity analysis, where you let your mesh size vary as a parameter, and then define an integration coupling variable, to see how the inlet-outlet vary as a function of mesh size.

If you still can't get a satisfactory solution, I would encourage you to contact our support team, and they can have a closer look at your model.
Received from my Technical Sales Manager: I noticed that you had a question on our Discussion Forum, and I thought I should get back to you with some help. As a first possible approach to get the overall mass balance to work out, I would consider setting up a denser mesh, to see if this will improve your results. You could even set up a mesh sensitivity analysis, where you let your mesh size vary as a parameter, and then define an integration coupling variable, to see how the inlet-outlet vary as a function of mesh size. If you still can't get a satisfactory solution, I would encourage you to contact our support team, and they can have a closer look at your model.

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 14 avr. 2014, 14:12 UTC−4
My reply:

Thank you very much for the suggestion. I am not familiar with the sensitivity analysis physics package, but I can run the denser meshes manually and see what the answer comes back as. I will report back when I have done that.

-TCL
My reply: Thank you very much for the suggestion. I am not familiar with the sensitivity analysis physics package, but I can run the denser meshes manually and see what the answer comes back as. I will report back when I have done that. -TCL

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 17 avr. 2014, 10:08 UTC−4
Thank you very much for your reply. I have taken your suggestion and run with it. Here are the values (inlet and outlet mass flow) for various mesh settings using the Normal solution as an initial condition (to save time):

Normal:
Stationary Solver 1 in Solver 4: Solution time: 141 s (2 minutes, 21 seconds)
Physical memory: 1.96 GB
Virtual memory: 2.25 GB
w2*hmnf.rho (kg/(m*s)) w2*hmnf.rho (kg/(m*s))
0.05071 0.05476

Fine:
Stationary Solver 1 in Solver 2: Solution time: 77 s (1 minute, 17 seconds)
Physical memory: 2.41 GB
Virtual memory: 2.69 GB
w2*hmnf.rho (kg/(m*s)) w2*hmnf.rho (kg/(m*s))
0.0506 0.05473

Finer:
150 0.8000000 0.012 150 150 150 3.2e-012 9.7e-015
Stationary Solver 1 in Solver 2: Solution time: 3637 s (1 hour, 0 minutes, 37 seconds)
Physical memory: 4.3 GB
Virtual memory: 4.98 GB
Updated maximum iterations to 1E6, as maximum was reached with 150
2237 0.8000000 0.013 2237 2237 2237 4.8e-012 1e-014
Canceled
Stationary Solver 1 in Solver 2: Solution time: 62187 s (17 hours, 16 minutes, 27 seconds)
Physical memory: 11.91 GB
Virtual memory: 12.55 GB
Failed to converge after running overnight… Cancelled simulation, but the program then froze and would not give back GUI control. After waiting for 30 minutes, I had to kill the process via the program manager.
Next, I down-checked the Newtonian damping factor to 0.1, instead of 0.8 (default) and turned on pseudo-timestepping.
128 0.1000000 0.001 128 128 128 2.8e-011 7.9e-014
Stationary Solver 1 in Solver 2: Solution time: 2846 s (47 minutes, 26 seconds)
Physical memory: 4.22 GB
Virtual memory: 4.91 GB
w2*hmnf.rho (kg/(m*s)) w2*hmnf.rho (kg/(m*s))
0.05043 0.05511

Extra Fine:
163 0.1000000 0.00098 163 163 163 3e-011 1.3e-013
Stationary Solver 1 in Solver 2: Solution time: 10288 s (2 hours, 51 minutes, 28 seconds)
Physical memory: 8.09 GB
Virtual memory: 9.57 GB
w2*hmnf.rho (kg/(m*s)) w2*hmnf.rho (kg/(m*s))
0.05249 0.05742

Extremely Fine:
240 0.1000000 0.00099 240 240 240 2.7e-011 9.1e-014
Stationary Solver 1 in Solver 2: Solution time: 43360 s (12 hours, 2 minutes, 40 seconds)
Physical memory: 14.34 GB
Virtual memory: 18.97 GB
w2*hmnf.rho (kg/(m*s)) w2*hmnf.rho (kg/(m*s))
0.05286 0.05851

Since this was not as productive as I had hoped, are there any suggestions on how to get mass balance? I tried to make this setup as absolutely simple as possible. It is an axisymmetric tube, L/D=100, Pin/Pout=10, Nitrogen gas, inlet of Pin & 273.15, Ma=0, outlet of Hybrid. Pictures of the conditions are at the bottom. I then calculate the flow using Normal Mesh and the default solver. To check conservation, I took a line integral on the inlet and a line integral on the outlet, both of the z component of velocity times density.

Is there another example I may look to for a high pressure ratio pipe flow, a capillary flow maybe, which conserves mass?

-TCL
Thank you very much for your reply. I have taken your suggestion and run with it. Here are the values (inlet and outlet mass flow) for various mesh settings using the Normal solution as an initial condition (to save time): Normal: Stationary Solver 1 in Solver 4: Solution time: 141 s (2 minutes, 21 seconds) Physical memory: 1.96 GB Virtual memory: 2.25 GB w2*hmnf.rho (kg/(m*s)) w2*hmnf.rho (kg/(m*s)) 0.05071 0.05476 Fine: Stationary Solver 1 in Solver 2: Solution time: 77 s (1 minute, 17 seconds) Physical memory: 2.41 GB Virtual memory: 2.69 GB w2*hmnf.rho (kg/(m*s)) w2*hmnf.rho (kg/(m*s)) 0.0506 0.05473 Finer: 150 0.8000000 0.012 150 150 150 3.2e-012 9.7e-015 Stationary Solver 1 in Solver 2: Solution time: 3637 s (1 hour, 0 minutes, 37 seconds) Physical memory: 4.3 GB Virtual memory: 4.98 GB Updated maximum iterations to 1E6, as maximum was reached with 150 2237 0.8000000 0.013 2237 2237 2237 4.8e-012 1e-014 Canceled Stationary Solver 1 in Solver 2: Solution time: 62187 s (17 hours, 16 minutes, 27 seconds) Physical memory: 11.91 GB Virtual memory: 12.55 GB Failed to converge after running overnight… Cancelled simulation, but the program then froze and would not give back GUI control. After waiting for 30 minutes, I had to kill the process via the program manager. Next, I down-checked the Newtonian damping factor to 0.1, instead of 0.8 (default) and turned on pseudo-timestepping. 128 0.1000000 0.001 128 128 128 2.8e-011 7.9e-014 Stationary Solver 1 in Solver 2: Solution time: 2846 s (47 minutes, 26 seconds) Physical memory: 4.22 GB Virtual memory: 4.91 GB w2*hmnf.rho (kg/(m*s)) w2*hmnf.rho (kg/(m*s)) 0.05043 0.05511 Extra Fine: 163 0.1000000 0.00098 163 163 163 3e-011 1.3e-013 Stationary Solver 1 in Solver 2: Solution time: 10288 s (2 hours, 51 minutes, 28 seconds) Physical memory: 8.09 GB Virtual memory: 9.57 GB w2*hmnf.rho (kg/(m*s)) w2*hmnf.rho (kg/(m*s)) 0.05249 0.05742 Extremely Fine: 240 0.1000000 0.00099 240 240 240 2.7e-011 9.1e-014 Stationary Solver 1 in Solver 2: Solution time: 43360 s (12 hours, 2 minutes, 40 seconds) Physical memory: 14.34 GB Virtual memory: 18.97 GB w2*hmnf.rho (kg/(m*s)) w2*hmnf.rho (kg/(m*s)) 0.05286 0.05851 Since this was not as productive as I had hoped, are there any suggestions on how to get mass balance? I tried to make this setup as absolutely simple as possible. It is an axisymmetric tube, L/D=100, Pin/Pout=10, Nitrogen gas, inlet of Pin & 273.15, Ma=0, outlet of Hybrid. Pictures of the conditions are at the bottom. I then calculate the flow using Normal Mesh and the default solver. To check conservation, I took a line integral on the inlet and a line integral on the outlet, both of the z component of velocity times density. Is there another example I may look to for a high pressure ratio pipe flow, a capillary flow maybe, which conserves mass? -TCL

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 17 avr. 2014, 10:16 UTC−4
OK… I may have found my own answer. It, as it usually is, may be PICNIC (problem in chair, not in computer). I found the “surface integration” check box within line integration (I assume that means completing the revolve). Now the answer seems to work out. Below are the new answers using extremely fine as the initial solution. I feel a bit silly, but I had assumed that the Axisymmetric model did the revolve when you took the line integral across the boundary.

Extremely Fine:
240 0.1000000 0.00099 240 240 240 2.7e-011 9.1e-014
Stationary Solver 1 in Solver 2: Solution time: 43360 s (12 hours, 2 minutes, 40 seconds)
Physical memory: 14.34 GB
Virtual memory: 18.97 GB
w2*hmnf.rho (kg/s) w2*hmnf.rho (kg/s)
8.26522e-5 8.27247e-5

Normal:
51 0.8000000 0.00099 51 51 51 81 0.00051 9.9e-007
Stationary Solver 1 in Solver 3: Solution time: 566 s (9 minutes, 26 seconds)
Physical memory: 2.23 GB
Virtual memory: 3.01 GB
w2*hmnf.rho (kg/s) w2*hmnf.rho (kg/s)
7.97197e-5 8.00929e-5

My next question is: how do I evaluate multiple cut lines so that I can see the mass flow as function of position along the pipe? I can create a data set of cut lines along the pipe, but I don’t know how to evaluate the integrals at each of them. When I try, it wants to evaluate as if the data set is one long line, not several individual lines.
OK… I may have found my own answer. It, as it usually is, may be PICNIC (problem in chair, not in computer). I found the “surface integration” check box within line integration (I assume that means completing the revolve). Now the answer seems to work out. Below are the new answers using extremely fine as the initial solution. I feel a bit silly, but I had assumed that the Axisymmetric model did the revolve when you took the line integral across the boundary. Extremely Fine: 240 0.1000000 0.00099 240 240 240 2.7e-011 9.1e-014 Stationary Solver 1 in Solver 2: Solution time: 43360 s (12 hours, 2 minutes, 40 seconds) Physical memory: 14.34 GB Virtual memory: 18.97 GB w2*hmnf.rho (kg/s) w2*hmnf.rho (kg/s) 8.26522e-5 8.27247e-5 Normal: 51 0.8000000 0.00099 51 51 51 81 0.00051 9.9e-007 Stationary Solver 1 in Solver 3: Solution time: 566 s (9 minutes, 26 seconds) Physical memory: 2.23 GB Virtual memory: 3.01 GB w2*hmnf.rho (kg/s) w2*hmnf.rho (kg/s) 7.97197e-5 8.00929e-5 My next question is: how do I evaluate multiple cut lines so that I can see the mass flow as function of position along the pipe? I can create a data set of cut lines along the pipe, but I don’t know how to evaluate the integrals at each of them. When I try, it wants to evaluate as if the data set is one long line, not several individual lines.

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.