Concrete block under fire - possible problems with mesh/timestep and convergence

gemada

I'm trying to model a concrete block in a fire situation (thermal transient analysis), to get the temperatures on the face not exposed to the fire, but I'm having problems with the convergence or the results.
- MODELING: 3 blocks with very similar geometry, total thickness of 14cm and thickness of block walls variyng between 30mm (nominal value), approximately 25mm (to take in account block imperfections) and 20mm (to compare with 30mm and see the influence of wall block thickness in temperature increase);
- MESH AND PROPERTIES: mesh of the 25mm block is less refined; thermal properties of materials are constant (concrete and air inside the block);
- RESULTS: 30mm block (with refined mesh) gives much lower temperatures than experimental results (measured in the not exposed face); 25mm block gives results very close to the experimental ones; 20mm block doesn't exist but was modelled to compare thermal isolation with 30mm block and tresults between 20mm and 30mm block were relatively close;
- CONVERGENCE PROBLEMS: if I try to reduce time step to 1 minute for the refined blocks (blocks 30mm and 20mm), Mecway gives convergence error; if I try to use thermal properties that vary with temperature, the same thing happens.
I would expect that the 30mm block, with a more refined mesh, would be more close to experimental results...
I've attached the files, in case someone can give a clue of whats wrong in my models.
Thanks

Victor

I'm having a look now but a couple of concerns:
- Density of air is set to 1000 kg/m^3 instead of 1.
- Are you sure convection in the cavity isn't significant?
- The CCX solver is much faster.

Victor

It looks to be correct given the model definition because there's good agreement with these settings:

bloco-concreto-30mm.liml
final temperature on the opposite face to the radiation:

hex8 internal solver 2 min time step (original):       80-137 C
hex8 CCX 2 min time step:                              81-137 C 
hex8 CCX 1 min time step:                              80-136 C
hex20 internal solver 2 min time step:                 slow
hex20 CCX 2 min time step:                             82-138 C

gemada

Thanks a lot for the reply Victor. As always support is great here!
- Density of air is set wrong (it should be 1); Thanks
- I´ve seen two ways of modelling: witthout and with convection in cavity...I´ll try both ways and I´ll put them here in the forum;
- CCX solver is much faster!! Thanks a lot for the tip!!

Some updates to my model:
- 30mm block gives more accurate temperatures if I reduce timestep (from 2 to 1min) and generate solution at every 1 time step...I'll try to experiment further with timestep parameters and meshes;
- Using CCX, the initial temperatures are a bit weird, specially on the hot face ...for example: in 30mm block model they begin with more than 120C (node 6084...) but they should begin at 20C (see the attached file)...could it be something related to time step settings and or mesh?

Victor

CCX doesn't generate output for t=0 so the first time step is already 1 or 2 minutes in where it's supposed to be hotter.

disla

If you are planning to reduce the time step, you could also reduce the number of elements. The brick is symmetric and BC also makes this a 2D problem.

gemada

Thanks a lot Victor and Disla!
I'll try with symmetry.