Chiller system where supply is its own complete loop and return is its own complete loop

[BronYrAur]

I am looking at some drawings of a chilled water system serving a bunch of data center CRAC units. There are a lot more CRAC units than I have shown on the attached sketch, but the gist is there.

I am confused because the supply piping is a complete loop all the way around and so is the return. If you follow the flow in my sketch. Water leaves the pumps and can go either right or left to get to the chillers. Water leaves the chillers and can go either right or left to get to the CRAC units. Finally, water leaves the CRAC units and can either right or left to get back to the pumps.

To borrow a line from This is Spinal Tap, "there's a fine line between stupid and cleaver." I am trying to figure out which this is. I regularly see complete loops like this on compressible systems, such as compressed air, to help equalize the pressure. But I have only seen it once before on a hydronic system, and it was a data center as well (coincidence???)

I'm trying to wrap my mind around the pros/cons of such a system. Is it something that you would recommend on a new system design? Seems like a lot of additional piping cost. What is the benefit?

 

[Compositepro]

You do not have loops you have manifolds feeding parallel flow paths. That simply reduces pressure drop to make flow through all of the parallel paths more uniform. There is no way for flow to return to the pump without first going through a chiller and then a CRAC unit (whatever that is)

 

[Frank1344]

Hello,
I would recommend to post your thread in the HVAC/R engineering group if haven't done yet.
Good luck.

 

[BronYrAur]

I specifically chose the fluid mechanics forum as opposed to HVAC because this seems like more of a flow issue.

 

[LittleInch]

Well it seems pretty logical to me. What's the alternative? one pump to one chiller to one CRAC unit? Any one of those goes down and you've lost cooling.

This way any pump and any chiller can feed any CRAC unit. Large redundancy and turndown.

Petty efficient piping system as it's really a header system fed from both ends to even out flow. You probably need a smaller feed line than a single large line

The trick is having flow control on each of these so you don't starve one pump/chiller/ CRAC unit and overload another.

But I would say the redundancy trumps it all.

A fire water loop works the same way - nearly always two different routes to get to a monitor incse one section springs aleak and is shut off between isolating valves. Same principle.

See also ring mains for electricity supply. Chosen so that it saved on copper after the war ( a little off the wall I know but again the same principle - more than one flow path available)

Remember - More details = better answers
Also: If you get a response it's polite to respond to it.

 

[BronYrAur]

The arrangement just strikes me as very odd. Typically, I see systems like the one I have attached here. And actually, most that I see are direct-return and not reverse-return. It is extremely rare that I see the arrangement in my original post.

 

[Compositepro]

There are two problems with your second drawing.

The first is probably a minor oversight you made when you drew it. There is a lack of flow symetry with some pumps having a longer flow path than the others. The return flow to your pumps should come from the right side of the header.

The second issue is that the peak flow though some sections of pipe will be the total flow of all the pumps. In the arrangement that you do not like, the peak flow in any section of pipe will be half the flow of all of the pumps.

 

[LittleInch]

Did your first drawing start as three pumps and chillers and then they added three more? If so makes more sense.

A lot depends on the actual distance the main pipes run for.

But still not that odd.

Remember - More details = better answers
Also: If you get a response it's polite to respond to it.

 

[BronYrAur]

It's not that I don't like it. In fact it seems pretty slick. I just have almost never seen it that way, which leads me to question it.

I know I only showed 5 CRAC units on the original post. I threw in a 6th one on this post. There are in fact a lot more. But in terms of pipe sizing, the main headers would only need to be sized for half the total flow, correct?

In other words, if I fictitiously split the system in half (as shown here), I would have 2 separate systems, each sized for 3 chillers/pumps/CRACs. But by leaving them connected and feeding both directions, I still just size each half for 3 units, correct?

So the advantage of having all of those pipes physically connected is mostly redundancy and allowing all pumps to feed all units and all chillers to do the same, correct? Is there any other HYDRONIC advantage to having the pipes looped as opposed to having them severed and capped the way I do here?

 

[LittleInch]

No I think you've got it sussed

Remember - More details = better answers
Also: If you get a response it's polite to respond to it.

 

[georgeverghese]

The setup on the first post is best. If you have a 5+1 sparing arrangement, the each half (of 3 pumps) would be sized for (6/5)/2 = 6/10 of total flow in terms of flow and pressure drop.