Boundary Conditions

[samwicks]

Is it possible to specify non-uniform boundary conditions? Currently the only method I'm aware of for establishing the correct flow entering a domain is to model the region before the domain. This is computationally prohibitive in many circumstances. It would be ideal to be able to 'link' two different models such that the outflow of one is the inflow of the other, whether intelligently linked or simply copy'n'pasted, but a simple 'velocity magnitude (x,y,z) vs point on plane (x,y,z)' across a face would be suitable.

[derrek.cooper]

I agree.. Something we have discussed internally for awhile. What sort of application are you running? We have looked at a few areas - 1 is to provide fully developed flow BCs, the other is a bit more generic and allows mapping from one model to the next. Ideally, we will have both as they are two similar, yet different applications.

If you could provide some context, it would help in the decision making process.

[samwicks]

Within the context of Data Centre modeling, where the entire room is a closed CV. Mapping one model to another would be useful: currently, node counts are high for a level of detail that can resolve each individual rack within a server cabinet, prohibiting such analysis for large data centers. It would be better if each cabinet was a separate analysis, specifying BCs for a larger analysis which resolves the flow at the scale of the room.This is probably the application requiring the most detail from such a feature.

The feature would also be useful for flow through perforated tiles. Currently the free area ratio function is good, as is the pressure head function, but the velocity distributions coming through the tile aren't realistic for floor tiles with large free area ratios that have non-uniform geometries, or have adjustable dampers. I can accurately model flow rates across the entire tile, but jet velocities are unrealistic. Modeling the tile in detail, then incorporating that into a general model would greatly improve accuracy.

At the most basic level, simply specifying a non-uniform velocity distribution on a plane would be sufficient to enable everything I'd need it for (although running different scenarios of the same entire-configuration would be laborious, and accurate results would be difficult for an entirely internal flow driven by fans with no external BCs) but real linkage between models would be ideal, especially if you could post-process the whole thing.