Dear all,
I am currently using Elmer to do some heat conduction simulations with a bi-phasic medium, but can't manage to have it working as I would like. For my particular analysis, I need to determine the normal fluxes crossing the internal boundary from one phase to the other (and signed so I know if the surface is locally receiving or lossing heat). Ideally, I would need this normal flux to be expressed as a distributed field, so I can treat it like any other variable in the Elmer simulation.
For this my current strategy is to:
- Compute the temperature field on the entire domain (phase 1 + 2) using the HeatSolver. This involves a mix of Dirichlet and Neumann boundary conditions.
- Re-compute the temperature in phase 2 only, this time activating the Calculate Loads and Calculate Weights options. For this, I enforce the temperature computed in step 1 as a Dirichlet boundary condition.
- Get the distributed boundary flux by computing Nodal Loads/Nodal Weights in Paraview (Following Section 12.2 of the Solver manual).
I have attached a short working example, with the mesh of a simple geometry and everything expressed in SI units. In this particular example, the FluxSolver tells me that the boundary flux should be about +150 W/m2 in the lower boundary and -150 W/m2 in the upper one, which is also what I physically expect. However, the computation Nodal Loads/Nodal Weights in Paraview gives me a numerical value of 15 000, two orders of magnitude above.
I'm not sure what I'm doing wrong here, and how to fix my situation. Maybe the values given by the nodal loads are not expressed in W/m2, as I expect. Does anyone have an idea?
Thanks a lot!
Kévin
EDIT - SOLUTION OF THE PROBLEM
The area associated with the boundary nodes is computed using Calculate Boundary Weights, and not Calculate Weights. In post-processing Nodal Loads / Nodal Boundary Weights give me the proper result.
Compute internal boundary heat flux with Calculate Loads
Compute internal boundary heat flux with Calculate Loads
- Attachments
-
- Elmer_Forum.zip
- (7.04 KiB) Downloaded 164 times
Last edited by kfourteau on 02 Jun 2021, 10:58, edited 1 time in total.
-
- Site Admin
- Posts: 4832
- Joined: 22 Aug 2009, 11:57
- Antispam: Yes
- Location: Espoo, Finland
- Contact:
Re: Compute internal boundary heat flux with Calculate Loads
Hi
Nodal loads are in Watts. And for internal boundary it should be zero.
You could define FluxSolver and have that be active only on one side.
-Peter
Nodal loads are in Watts. And for internal boundary it should be zero.
You could define FluxSolver and have that be active only on one side.
-Peter
Re: Compute internal boundary heat flux with Calculate Loads
Hi Peter,
Thanks for the quick response.
Nodal loads are in Watts, but I compute the Nodal Loads / Nodal Weights variable (as in the Section 12.3 of the ElmerSolver Manual). I would expect it to be in W/m2.
The Calculate Loads option is only activated in the second step, when I re-compute the temperature field in Body 2 only. This way the "internal boundary" between Body 1 and 2 becomes the external boundary of Body 2, and Calculate Loads should apply.
I am not sure FluxSolver is the way to solve my problem. For that I would need to project the computed flux onto the normal vector of the interface, and I am not sure of to do that in a complex bi-continuous structure where the normal vector is no longer simply in the z direction.
Kévin
Thanks for the quick response.
Nodal loads are in Watts, but I compute the Nodal Loads / Nodal Weights variable (as in the Section 12.3 of the ElmerSolver Manual). I would expect it to be in W/m2.
The Calculate Loads option is only activated in the second step, when I re-compute the temperature field in Body 2 only. This way the "internal boundary" between Body 1 and 2 becomes the external boundary of Body 2, and Calculate Loads should apply.
I am not sure FluxSolver is the way to solve my problem. For that I would need to project the computed flux onto the normal vector of the interface, and I am not sure of to do that in a complex bi-continuous structure where the normal vector is no longer simply in the z direction.
Kévin