Hi,
I'm doing a coupled heat fluid flow simulation with a fluid channel flowing through a solid with a heat load on it's surface (liquid cooled heatsink), the only path for the heat is into the fluid.
My solution conserves fluid flow, the heat flux into the surface agrees with my Heat Flux boundary condition but the heat energy isn't conserved, i.e. the difference between the integrals of V.rho.c.T over the inlet and outlet is significantly higher (up to 20%) than the heat load?
Does this suggest that my fluid mesh at the noslip boundary is not good enough? The error does decrease if I make it finer but I only have the option of tetrahedral or prism elements (I only have GMSH for meshing). With a "fine" mesh using 8 cores my solution time with linear elements is >8hrs and increasing the order to quadratic would be prohibitive and would exceed my available memory ;(
My tolerance settings in both the NavierStokes and the Heat equation are already very tight.
The Reynolds number for the flow is ~1000
Are there any settings that would reduce the energy imbalance error?
Any suggestions?
Thanks in advance, Regards
Mark
Conjugate heat transfer problem

 Posts: 208
 Joined: 26 Aug 2009, 18:20
 Location: Peterborough, England

 Posts: 208
 Joined: 26 Aug 2009, 18:20
 Location: Peterborough, England
Re: Conjugate heat transfer problem
Hi,
Just a quick update.
I have run an all linear tet mesh problem, GMSH mesh file size 170Mb (650000 fluid tets 2330000 solid tets), this runs to completion on 8 processors in 20 mins.
The mesh is converted to 2nd order 10 node tets in GMSH (incomplete elements) producing a GMSH file of size 565Mb, this runs (again 8 processors) but refuses to converge, I eventually aborted run after 34 hours ;(
Comments/observations encouraged.
Regards
Mark
Just a quick update.
I have run an all linear tet mesh problem, GMSH mesh file size 170Mb (650000 fluid tets 2330000 solid tets), this runs to completion on 8 processors in 20 mins.
The mesh is converted to 2nd order 10 node tets in GMSH (incomplete elements) producing a GMSH file of size 565Mb, this runs (again 8 processors) but refuses to converge, I eventually aborted run after 34 hours ;(
Comments/observations encouraged.
Regards
Mark

 Site Admin
 Posts: 3453
 Joined: 22 Aug 2009, 11:57
 Antispam: Yes
 Location: Espoo, Finland
 Contact:
Re: Conjugate heat transfer problem
Hi Mark
Are you sure that your heat is not convected away at the outlet, or is not eaten up to warm the system in a transient case?
Peter
Are you sure that your heat is not convected away at the outlet, or is not eaten up to warm the system in a transient case?
Peter

 Posts: 208
 Joined: 26 Aug 2009, 18:20
 Location: Peterborough, England
Re: Conjugate heat transfer problem
Hi Peter, you asked
Q: "Are you sure that your heat is not convected away at the outlet"?
Indeed it is, in the model the heat cannot leave by any other means. The problem is that the warm fluid leaving the system has more energy than that supplied via the heat flux BC, by 20% more in my linear tet model.
In my 2nd post with the second order tets that wouldn't converge, it was the the fluid flow that would not converge,the heat equation was never started.
Regards
Mark
Q: "Are you sure that your heat is not convected away at the outlet"?
Indeed it is, in the model the heat cannot leave by any other means. The problem is that the warm fluid leaving the system has more energy than that supplied via the heat flux BC, by 20% more in my linear tet model.
In my 2nd post with the second order tets that wouldn't converge, it was the the fluid flow that would not converge,the heat equation was never started.
Regards
Mark
Re: Conjugate heat transfer problem
Hi All,
I have similar problems. The case is the air flow around heater (500degC). The walls of channel are modelled by solid bodies.
The convective flux is calculated by 'convective flux' operator, the fluxes on the solid faces by 'boundary sum' on 'temperature load'.
The flux on internal faces of the channel walls is 4974 W, on external faces 4769 W, so difference is 205 W.
Net convective flow is 1438 W. The outflow from heater 3821 W, perfect radiation outflow 26W, so inside the channel are an extra 2617 W.
Best regards, ark.
I have similar problems. The case is the air flow around heater (500degC). The walls of channel are modelled by solid bodies.
The convective flux is calculated by 'convective flux' operator, the fluxes on the solid faces by 'boundary sum' on 'temperature load'.
The flux on internal faces of the channel walls is 4974 W, on external faces 4769 W, so difference is 205 W.
Net convective flow is 1438 W. The outflow from heater 3821 W, perfect radiation outflow 26W, so inside the channel are an extra 2617 W.
Best regards, ark.
 Attachments

 results.dat
 (6.11 KiB) Downloaded 27 times

 case.sif
 (5.24 KiB) Downloaded 33 times
Re: Conjugate heat transfer problem
Hi,
I have used heat source (body force) instead of the fixed temperature on heater and the difference decried to about 1100 W.
What is the formula implemented in Elmer for calculating the heat exchange between a fluid and solid?
Here is my mesh (not very 'fluidic'):
I have used heat source (body force) instead of the fixed temperature on heater and the difference decried to about 1100 W.
What is the formula implemented in Elmer for calculating the heat exchange between a fluid and solid?
Here is my mesh (not very 'fluidic'):
Re: Conjugate heat transfer problem
Hi,
by switching off the radiation I have found that radiation is not included in boundary sum on heater in my original case.
by switching off the radiation I have found that radiation is not included in boundary sum on heater in my original case.