Automatic generation of a mesh for the building model

[AleksanderKamienski]

Hi,

I would like to write a program that automatically converts a CFAST fire simulation input file to an FDS fire simulation input file. I need to automate the conversion process. I do a lot of Monte Carlo simulations. In the CFAST model, the room is represented as a cuboid. For simplicity, the walls in the FDS model could be represented as a 4 cm thick obstacle, I would define the thermal properties later.

The key problem is the automatic adjustment of the computational mesh to the geometry, which can be different because the user creates the building geometry in a CAD-like program. He can add obstacles in the rooms, doors and windows in different places on the walls, with different heights and widths. The heights of the rooms can be different depending on the building. I wonder if a possible modification of the FDS program to add the possibility of mesh stretching in three coordinate directions would solve this problem because it would allow for the appropriate adjustment of the mesh so that the mesh does not deform the building model. Is such a feature even possible to make in the FDS program and how time-consuming is it?

I thought about representing each room as a separate mesh and using mesh alignment to connect individual rooms in the FDS model.

I want the computational mesh to be relatively large, in some places even with dimensions of 50cmx50cmx50 cm or 1mx1mx1m so that FDS simulations do not take very long. Key places from the simulation point of view - fire areas, openings, walls on the border of overlapping meshes would have lower resolution - for this purpose I wanted to use mesh stretching.

This would ensure that all building elements are covered with a mesh - so that a large mesh does not deform the building model while limiting the number of cells in the mesh and limiting the simulation time - I would like FDS simulations to be calculated relatively quickly even for large buildings.

[drjfloyd]

Converting this to a discussion

[drjfloyd]

Stretching in only two-directions is a limitation of the FFT solver. It is not something that can be changed. Use of multiple meshes within each compartment could be another route to achieving the effect.

Overall what you are looking to do is an interesting idea. I'd caution about going to coarse with meshes. At some point if you make the meshes coarse enough you would be better just using a zone model in terms of the quality of results you would obtain.

[AleksanderKamienski]

Thanks for your reply. I guess we'll continue using the zone model then.