Hi there,

I've just been made aware of this forum by my administrator so hoping someone out there can help. I'm using EFD Pro to model a rack containing many enclosures with IT equipment in them.

An air conditioner has been attached to the rack to aid in the cooling of these enclousures. I was wondering if there is a preferred way of modelling the air con as there doesn't seem to be any guidelines on such an application. Also what appropriate settings I should apply to the inlets / outlets.

Hello

If I may broaden the topic slightly to more than just the A/C then my first stab at the problem would be to:

(1) define vol flow rate and temperature on inlet (2) Static pressure (altitude?) on outlet. (3) adiabatic rack walls (4) define gravity

My sub-racks would be simple solids (rectangular boxes) and I would apply individual heat loads to each of them.

My second stab would have more complex sub-racks and I would think about whether there was swirl introduced by the A/C. I would also add some heat transfer through the walls and turn on the Radiation model.

...

I would instrument up the current rack for temperature (I like iButtons) and try to match the model to measurements before I make any serious recommendations.

Cheers Nick

Chris,

Usually such air conditioning units aim to control the temperature of the air going into the rack against a varying ambient, so one simple approach is to just assume it does its job and model from the exhaust of the air conditioner, treating this as a fixed flow and (sub-ambient) temperature.

The downside of using an air conditioning unit on the entrance to the rack is that although the temperature going into the rack is reduced, the total total power consumption and heat going into the room are increased. The unit can help if the problem is just one rack, but you would not want to fill a data center with these.

Hope that helps

John Parry, Research Manager, Flomerics Ltd.

Hi,

I wonder, if EFD is the right tool for complex internal flows with heat transfer and high streamline curvature. Using a standard k-eps model will give you just some results ...!?

Fab

Hi Fab,

EFD uses a low-Re form of the k-e model. Probably this is the best that's achievable for electronics type geometries where the turbulence is strongly controlled by wall effects.

For geometries that are relatively flat (uniform height flat pack components independent research has shown the Spalart-Allmaras model to give good results, but this is an increasingly less common situation due to the increasing use of heat sinks.

John Parry, Research Manager, Flomerics Ltd.

While you didn't really provide tons of info here another approach is to use a disabled component and apply a negative power to it and put an internal fan (curve) in the appropriate spot to drive the flow through the disabled componet (A/C radiator). You can also attach a presure drop vs. flowrate characteristic to the diabled component with directional properties using the porous media option. I have done coolers with great success this way. You can always apply the heat rejection from the device to some other component in the area if in fact the heat is rejected in the room. You can have quite complex models that will solve in what I consider reasonable time.

[DineshramBalaji]

Hi ,

I am very new to OpenFoam. I need to cool several servers, around 200. Considering that the servers are switched off, How can I model the forced cooling of the server?

[Boris_M]

Hi DineshramBalaji,

If you are using OpenFoam then you are in the wrong forum.

Regards,
Boris