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.

moving source in heat transfer problem

Please login with a confirmed email address before reporting spam

Hi:

The attached images in the pdf file are from a model I'm working on. The axiymmetric model file is also attached.

Basically, it's a model of a small disc of metal that is moving in a hole in a material. This small disc is meant to represent the cutting tip of a drill, and due to cutting action, it is releasing heat. This disc is therefore made to be a kind of heat source -- by assigning a "pair boundary heat source" to the identity pair at the boundary between the disc and the surrounding cylinder of material. I have a moving mesh set up to allow the disc (drill bit) to move into the hole in the material at a chosen velocity.

The model appears to work fine, however I am confused by 2 things, as explained in the pdf file.

First, as the drill tip moves into the material via the moving mesh, it generates heat at each time point and I see the results (which to some extent make sense), but at each position of the drill tip, is the Comsol formulation solving for the temperature distribution using the original (time = 0) conditions of temperature in the surrounding cylinder, or does each solution use the temperature distribution created in the previous time step(s)?

The second question concerns the default "insulation" condition that is within the "pair boundary heat source" definition. See the pdf file for the detailed question.

Thanks very much, John


5 Replies Last Post 2016年6月8日 GMT-4 02:40
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 8 years ago 2016年6月6日 GMT-4 15:20
Hi John

You have a nice model there, and COMSOL is playing you a little trick with the mesh (automatic Assembly boundary meshing new in v5.2 I believe). Furthermore you have not checked your heat diffusivity and the mesh size + time stepping so you get a funny impression of local cooling and little remnant heat.

I would suggest to start as following:

1) set the mesh to "Finer" or even "Extra fine"
2) set the time depending solver node set the time stepping to "0.1 sec" per steps range(0,0.1,5) or even finer (you have previously 1 sec, or 1 mm/time_steps at your speed, far too coarse steps for a heat diffusivity of 0.2[mm^2/s].
3) in the Time dependent solver node under "Results while solving" select the first "3D Plot Group" and select "Time steps taken by solver"
4) under the Time dependent Solver sub-node (far down), in the Time Stepping tab, set the time steps of your BDF solver to "Intermediate" and not the default "Free"
5) now run your model

There are two-three issues here:
1) the time steps: you must allow the heat to flow across the boundary, else you move it too fast and you do not get correct heat transfer.
2) your mesh must be fine enough to resolve the heat flow gradient, your heat diffusivity alpha = k/rho/Cp of the hollow cylinder material is only 0.2[mm^2/s] for time stepping of Dt=0.1 sec you should have a mesh size h << sqrt(0.2[mm^2/s]*Dt[s]).
3) the boundary layer mesh has a new "automatic physics controlled mesh feature I discover now in v5.2, you see it meshes very fine the destination, a typical case for "solid-Contact" physics, if you inverse the Identity Pair boundaries and re-mesh you will get a very coarse mesh on the hollow surface. But you are here in thermal physics, and because your heat diffusivity on the outer hollow cylinder is so low you need a finer mesh on the outside.
You should set a manual mesh and mesh the inner cylinder boundary very fine and leave the mesh coarser towards the external radius.

By the way it's worth to mention this to "support" as they have obviously not though of your case and the mesher should behave differently in this HT case, the improvement brought here in V5.2 for solid contact physics is worsening your case, as you r alpha ratio between the two materials are in the opposite ratio to this meshing logic.

--
Good luck
Ivar
Hi John You have a nice model there, and COMSOL is playing you a little trick with the mesh (automatic Assembly boundary meshing new in v5.2 I believe). Furthermore you have not checked your heat diffusivity and the mesh size + time stepping so you get a funny impression of local cooling and little remnant heat. I would suggest to start as following: 1) set the mesh to "Finer" or even "Extra fine" 2) set the time depending solver node set the time stepping to "0.1 sec" per steps range(0,0.1,5) or even finer (you have previously 1 sec, or 1 mm/time_steps at your speed, far too coarse steps for a heat diffusivity of 0.2[mm^2/s]. 3) in the Time dependent solver node under "Results while solving" select the first "3D Plot Group" and select "Time steps taken by solver" 4) under the Time dependent Solver sub-node (far down), in the Time Stepping tab, set the time steps of your BDF solver to "Intermediate" and not the default "Free" 5) now run your model There are two-three issues here: 1) the time steps: you must allow the heat to flow across the boundary, else you move it too fast and you do not get correct heat transfer. 2) your mesh must be fine enough to resolve the heat flow gradient, your heat diffusivity alpha = k/rho/Cp of the hollow cylinder material is only 0.2[mm^2/s] for time stepping of Dt=0.1 sec you should have a mesh size h

Please login with a confirmed email address before reporting spam

Posted: 8 years ago 2016年6月6日 GMT-4 21:31
Ivar:

Thanks very much for your very complete and thoughtful response. Very helpful! Hopefully this will let me sort out the problem. Heat transfer is not my favorite topic (I'm more comfortable with solid mechanics problems), but sometimes one needs to solve what one needs to solve! I'll keep you posted.

Thanks again,

John


Hi John

You have a nice model there, and COMSOL is playing you a little trick with the mesh (automatic Assembly boundary meshing new in v5.2 I believe). Furthermore you have not checked your heat diffusivity and the mesh size + time stepping so you get a funny impression of local cooling and little remnant heat.

I would suggest to start as following:

1) set the mesh to "Finer" or even "Extra fine"
2) set the time depending solver node set the time stepping to "0.1 sec" per steps range(0,0.1,5) or even finer (you have previously 1 sec, or 1 mm/time_steps at your speed, far too coarse steps for a heat diffusivity of 0.2[mm^2/s].
3) in the Time dependent solver node under "Results while solving" select the first "3D Plot Group" and select "Time steps taken by solver"
4) under the Time dependent Solver sub-node (far down), in the Time Stepping tab, set the time steps of your BDF solver to "Intermediate" and not the default "Free"
5) now run your model

There are two-three issues here:
1) the time steps: you must allow the heat to flow across the boundary, else you move it too fast and you do not get correct heat transfer.
2) your mesh must be fine enough to resolve the heat flow gradient, your heat diffusivity alpha = k/rho/Cp of the hollow cylinder material is only 0.2[mm^2/s] for time stepping of Dt=0.1 sec you should have a mesh size h << sqrt(0.2[mm^2/s]*Dt[s]).
3) the boundary layer mesh has a new "automatic physics controlled mesh feature I discover now in v5.2, you see it meshes very fine the destination, a typical case for "solid-Contact" physics, if you inverse the Identity Pair boundaries and re-mesh you will get a very coarse mesh on the hollow surface. But you are here in thermal physics, and because your heat diffusivity on the outer hollow cylinder is so low you need a finer mesh on the outside.
You should set a manual mesh and mesh the inner cylinder boundary very fine and leave the mesh coarser towards the external radius.

By the way it's worth to mention this to "support" as they have obviously not though of your case and the mesher should behave differently in this HT case, the improvement brought here in V5.2 for solid contact physics is worsening your case, as you r alpha ratio between the two materials are in the opposite ratio to this meshing logic.

--
Good luck
Ivar


Ivar: Thanks very much for your very complete and thoughtful response. Very helpful! Hopefully this will let me sort out the problem. Heat transfer is not my favorite topic (I'm more comfortable with solid mechanics problems), but sometimes one needs to solve what one needs to solve! I'll keep you posted. Thanks again, John [QUOTE] Hi John You have a nice model there, and COMSOL is playing you a little trick with the mesh (automatic Assembly boundary meshing new in v5.2 I believe). Furthermore you have not checked your heat diffusivity and the mesh size + time stepping so you get a funny impression of local cooling and little remnant heat. I would suggest to start as following: 1) set the mesh to "Finer" or even "Extra fine" 2) set the time depending solver node set the time stepping to "0.1 sec" per steps range(0,0.1,5) or even finer (you have previously 1 sec, or 1 mm/time_steps at your speed, far too coarse steps for a heat diffusivity of 0.2[mm^2/s]. 3) in the Time dependent solver node under "Results while solving" select the first "3D Plot Group" and select "Time steps taken by solver" 4) under the Time dependent Solver sub-node (far down), in the Time Stepping tab, set the time steps of your BDF solver to "Intermediate" and not the default "Free" 5) now run your model There are two-three issues here: 1) the time steps: you must allow the heat to flow across the boundary, else you move it too fast and you do not get correct heat transfer. 2) your mesh must be fine enough to resolve the heat flow gradient, your heat diffusivity alpha = k/rho/Cp of the hollow cylinder material is only 0.2[mm^2/s] for time stepping of Dt=0.1 sec you should have a mesh size h

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 8 years ago 2016年6月7日 GMT-4 04:29
Hi John

Well diffusion problems such as chemistry concentration or HT are somewhat different than "solid" particularly as you do not have any 2nd time derivatives, so no standing waves, only critically damped steps propagations. But once you master HT you mater also chemistry diffusion problems too ... ;)

There are finally not that many different PDE behaviour to master, so Multi-physics is not that difficult, if you see and use all the commonalities.

Have you checked the Application Library "Friction Stir Welding" model ?

--
Good luck
Ivar
Hi John Well diffusion problems such as chemistry concentration or HT are somewhat different than "solid" particularly as you do not have any 2nd time derivatives, so no standing waves, only critically damped steps propagations. But once you master HT you mater also chemistry diffusion problems too ... ;) There are finally not that many different PDE behaviour to master, so Multi-physics is not that difficult, if you see and use all the commonalities. Have you checked the Application Library "Friction Stir Welding" model ? -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 8 years ago 2016年6月7日 GMT-4 20:15
Ivar:

I was pondering a little more on what you wrote, and I think one tricky thing that may arise at some stage relates to the fact that one of the issues I'd like to investigate using this model (or one like it) is the advance speed of the drill (heat source in this problem). In this problem, this is the speed of the moving mesh. So although I get what you mean about adjusting the time step and mesh size etc. to be appropriate for the physics at hand (heat diffusion), I guess this issue of the mesh size etc. could become trickier as the advance speeds of the drill tip increases.

By the way, what was your opinion on the insulation condition seemingly existing all along the pair boundary while at the same time heat transfer also occurs at the pair boundary.

I'll take a look at that model in the application library.

Thanks again, John


Hi John

Well diffusion problems such as chemistry concentration or HT are somewhat different than "solid" particularly as you do not have any 2nd time derivatives, so no standing waves, only critically damped steps propagations. But once you master HT you mater also chemistry diffusion problems too ... ;)

There are finally not that many different PDE behaviour to master, so Multi-physics is not that difficult, if you see and use all the commonalities.

Have you checked the Application Library "Friction Stir Welding" model ?

--
Good luck
Ivar


Ivar: I was pondering a little more on what you wrote, and I think one tricky thing that may arise at some stage relates to the fact that one of the issues I'd like to investigate using this model (or one like it) is the advance speed of the drill (heat source in this problem). In this problem, this is the speed of the moving mesh. So although I get what you mean about adjusting the time step and mesh size etc. to be appropriate for the physics at hand (heat diffusion), I guess this issue of the mesh size etc. could become trickier as the advance speeds of the drill tip increases. By the way, what was your opinion on the insulation condition seemingly existing all along the pair boundary while at the same time heat transfer also occurs at the pair boundary. I'll take a look at that model in the application library. Thanks again, John [QUOTE] Hi John Well diffusion problems such as chemistry concentration or HT are somewhat different than "solid" particularly as you do not have any 2nd time derivatives, so no standing waves, only critically damped steps propagations. But once you master HT you mater also chemistry diffusion problems too ... ;) There are finally not that many different PDE behaviour to master, so Multi-physics is not that difficult, if you see and use all the commonalities. Have you checked the Application Library "Friction Stir Welding" model ? -- Good luck Ivar [/QUOTE]

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 8 years ago 2016年6月8日 GMT-4 02:40
Hi John

Indeed the issue of time steps, speed and mesh size is crucial for a correct flux transfer (for any diffusion equation of your type), you will notice that COMSOL has mange to improve the overall flux transfers and global energy conservation for their FEM models, so the total amount of heat is rapidly correct once you start to control the mesh size, but the amount of heat in your drill and the part going to the bulk will continue to vary until you get small enough mesh and small time steps.

It's only that the automatic "physics induced" mesh on identity/contact boundaries is tailored for "solid contact" physics and the default mesh is not ideal for your purely HT diffusivity case, and for the particular heat diffusivities of your two materials.

Now COMSOL time dependent solver is aware of this and will use small steps even if you ask for larger ones for storage (except if you take full control). Basically for such a model the mesh optimization is crucial.

Finally, the "fallback" Insulation condition is "normal", look at your two identity edges, one is long the other short, the flux identity is only where both edges are "in common", the long edge where the drill is not should be isolated, or have some other boundary condition you may add (radiation exchange, convection ...)

Still your "drill" is very short (along the axis) I would have expected a "long cylinder", heating up and extracting some heat, even if the contact region might be only at the tip.

--
Good luck
Ivar
Hi John Indeed the issue of time steps, speed and mesh size is crucial for a correct flux transfer (for any diffusion equation of your type), you will notice that COMSOL has mange to improve the overall flux transfers and global energy conservation for their FEM models, so the total amount of heat is rapidly correct once you start to control the mesh size, but the amount of heat in your drill and the part going to the bulk will continue to vary until you get small enough mesh and small time steps. It's only that the automatic "physics induced" mesh on identity/contact boundaries is tailored for "solid contact" physics and the default mesh is not ideal for your purely HT diffusivity case, and for the particular heat diffusivities of your two materials. Now COMSOL time dependent solver is aware of this and will use small steps even if you ask for larger ones for storage (except if you take full control). Basically for such a model the mesh optimization is crucial. Finally, the "fallback" Insulation condition is "normal", look at your two identity edges, one is long the other short, the flux identity is only where both edges are "in common", the long edge where the drill is not should be isolated, or have some other boundary condition you may add (radiation exchange, convection ...) Still your "drill" is very short (along the axis) I would have expected a "long cylinder", heating up and extracting some heat, even if the contact region might be only at the tip. -- Good luck Ivar

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.