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.

Basic pipeflow

Please login with a confirmed email address before reporting spam

Hello
I started using Comsol to solve fluid mechanic cases recently, using the incompressible laminar flow module. However, when I try to compare the velocity profile with analytical solution, I have approximately 3-4% difference in the maximum velocity. The case is water laminar pipe flow(3D) and I'm looking into the developed velocity at center line;its analytical solution can be found in all fluidmechanics textbooks.
I think for such a simple case the error shouldn't be as large as that(3-4%), and wonder if I should do any particular implement for this type of problem?
Thank you.

6 Replies Last Post 17.11.2010, 11:12 GMT-5

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 08.11.2010, 16:36 GMT-5
Possible reasons: #
- insufficient mesh, i.e. use boundary mesh and finer meshes
- low tolerance
- artificial diffusion used

if you upload your example file we can help maybe better.

--
Comsol 4.0a
Ubuntu 10.04.1
Possible reasons: # - insufficient mesh, i.e. use boundary mesh and finer meshes - low tolerance - artificial diffusion used if you upload your example file we can help maybe better. -- Comsol 4.0a Ubuntu 10.04.1

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 09.11.2010, 15:14 GMT-5
Thank you for the advice. I've already tried these but cannot eliminate the error. Besides, the velocity result looks fine in 2D example (flow between two infinite large parallel plates).
The example file is attached. I treat -pz*(r^2)/4/viscosity as analytical velocity at center line of the pipe.
Many thanks.
Thank you for the advice. I've already tried these but cannot eliminate the error. Besides, the velocity result looks fine in 2D example (flow between two infinite large parallel plates). The example file is attached. I treat -pz*(r^2)/4/viscosity as analytical velocity at center line of the pipe. Many thanks.


Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 14.11.2010, 16:41 GMT-5
Hi, I haven't used the 3D CFD modules, but I've run the 2D models a fair bit. Looking at your pipe mesh, it is too coarse near the tube walls. Also, I would be tempted to start off with a radial mesh and then sweep that around azimuthally, and then sweep along pipe length.

Regards, John
Hi, I haven't used the 3D CFD modules, but I've run the 2D models a fair bit. Looking at your pipe mesh, it is too coarse near the tube walls. Also, I would be tempted to start off with a radial mesh and then sweep that around azimuthally, and then sweep along pipe length. Regards, John

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 15.11.2010, 10:30 GMT-5
Thank you John. The mesh was truely too coarse near the wall. I tried to adopt boundary layer mesh for the 2D cross section, and extrude it to a 3D pipe, and that works. The problem is that, to achieve a accurate result for such a simple structure, I've adopted 800,000 DOFs and must use a server machine with 96GB memory. That's absolutely not a wise way to build complicated models. I'm considering use linear elements to reduce the # of DOF, but that might result in larger error.

About the radial mesh and sweep azimuthally, I couldn't find it in the mesh lists. Are you also using Comsol 3.5a?

Best regards

Cheng
Thank you John. The mesh was truely too coarse near the wall. I tried to adopt boundary layer mesh for the 2D cross section, and extrude it to a 3D pipe, and that works. The problem is that, to achieve a accurate result for such a simple structure, I've adopted 800,000 DOFs and must use a server machine with 96GB memory. That's absolutely not a wise way to build complicated models. I'm considering use linear elements to reduce the # of DOF, but that might result in larger error. About the radial mesh and sweep azimuthally, I couldn't find it in the mesh lists. Are you also using Comsol 3.5a? Best regards Cheng

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 16.11.2010, 10:40 GMT-5

Thank you John. The mesh was truely too coarse near the wall. I tried to adopt boundary layer mesh for the 2D cross section, and extrude it to a 3D pipe, and that works. The problem is that, to achieve a accurate result for such a simple structure, I've adopted 800,000 DOFs and must use a server machine with 96GB memory. That's absolutely not a wise way to build complicated models. I'm considering use linear elements to reduce the # of DOF, but that might result in larger error.

About the radial mesh and sweep azimuthally, I couldn't find it in the mesh lists. Are you also using Comsol 3.5a?

Best regards

Cheng


That was my first impression with the model too. Mesh was too coarse. Nevertheless, if you use streamline diffusion you can use first order elements theoretically. Also, use symmetry in your model, this way you will have 1/2 of your elements.


--
Comsol 4.0a
Ubuntu 10.04.1
[QUOTE] Thank you John. The mesh was truely too coarse near the wall. I tried to adopt boundary layer mesh for the 2D cross section, and extrude it to a 3D pipe, and that works. The problem is that, to achieve a accurate result for such a simple structure, I've adopted 800,000 DOFs and must use a server machine with 96GB memory. That's absolutely not a wise way to build complicated models. I'm considering use linear elements to reduce the # of DOF, but that might result in larger error. About the radial mesh and sweep azimuthally, I couldn't find it in the mesh lists. Are you also using Comsol 3.5a? Best regards Cheng [/QUOTE] That was my first impression with the model too. Mesh was too coarse. Nevertheless, if you use streamline diffusion you can use first order elements theoretically. Also, use symmetry in your model, this way you will have 1/2 of your elements. -- Comsol 4.0a Ubuntu 10.04.1

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 17.11.2010, 11:12 GMT-5
Since streamline diffusion is defaultly applied, do you mean I can acquire accurate result with linear elements?
Anyway, I think it's rare to solve more than 500,000 DOFs on most PCs and error exists for many cases. I want to know how is the estimated error for common fluid problems that you have dealt with?

Many thanks.
Since streamline diffusion is defaultly applied, do you mean I can acquire accurate result with linear elements? Anyway, I think it's rare to solve more than 500,000 DOFs on most PCs and error exists for many cases. I want to know how is the estimated error for common fluid problems that you have dealt with? Many thanks.

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.