# Internal Airflow through a building

Submitted by Anonymous on August 15, 2010 - 11:12

When using CAEDIUM Builder to simulate air flow through a building, how do I create the geometry to represent individual rooms and ensure that there is a single face representing a wall as opposed to the two super-imposed faces I get when I either subtract or intersect?

Thanks

[Edit: Based on an email exchange]

## No Zero-Thickness Walls

[Edit November, 7, 2011: Caedium RANS Flow can now deal with zero thickness (or double sided) walls]

We don't currently support zero-thickness double-sided walls in Caedium,

instead you'll need to use a box with a nominal thickness to represent the

wall and then subtract it from your air flow solid model.

## Solver singularity

I've managed some sort of workaround for now, which is so far so good.

Unfortunately, I am unable to run the solver successfully. A brief synopsis of the problem setup - an IT room with computers to study different ventilation strategies. I have 3 inlets and 2 outlets and want to just run an airflow simulation first before introducing heat transfer.

At present, I keep getting the following error messages:

-----

Warning:

RANS Flow failed

Solver singularity: DILUPBiCG: Solving for epsilon: solution singularity

Warning:

RANS Flow failed

Solver singularity: DILUPBiCG: Solving for Ux: solution singularity

Warning:

RANS Flow failed

Solver singularity: DILUPBiCG: Solving for Ux: solution singularity

Warning:

RANS Flow failed

Solver singularity: DILUPBiCG: Solving for epsilon: solution singularity

-----

I have played around with my initial conditions to no avail (using both the fixed value, Turbulent Intensity & Length Scale for U, k & e in the flow Physics as well as reference values).

What am I missing?

## Possible solver singularity causes

Glad to hear you've got your geometry configured.

Possible causes of solver singularities:

Substance->Properties->Reference.Substance->Solver->Relaxation Factors. For more on relaxation factors try "Relaxation Factors in RANS Flow Solvers"