Preparing System Curves 5

[PBW2]

We are having a discussion in our office on the development of pump system curves; in particular for lift stations with triplex pumps. In developing the system curve for a triplex setup, with two pumps running there seems to be a couple of schools of thought:
1. Ignore the head through the individual discharge pipe and determine head for the common header and forcemain based on total (2 pump) flow.
2. Determine head for discharge for (1) pump based on pumps flow and add head for additional flow (2nd pump) through the common header and forcemain. The head for the individual pump discharge remains constant above the pumps capacity.

Also the question on how to handle additional flows from separate pump stations on a common forcemain has come up. The consensus is to add the flow as additional head to the system curve. However, what if there are (5) pump stations? Is it realistic to determine the system curve based on all (5) running at capacity? Anyone have any code references.

Thanks in advance for comments/suggestions/discussion.

 

[BigInch]

I don't understand question 1 or 2.

What you're asking about, "head through the individual discharge pipe". What does that mean?

And the part about "add head for additional flow (2nd pump) through the common header and forcemain". No idea of what you're trying to suggest.

If all pump stations can operate together, it is realistic to design the system for that condition. If all pump stations won't operate all together now, in five years, THEY WILL.

http://virtualpipeline.spaces.msn.com

 

[RWF7437]

You develop the system head curve by ASSUMING a range if flows , from the smallest expected flow to the largest. Including all of the pipes and fittings in the system is the safest and most conservative way to do this. Unless the "headers" are very small or very long their effect on the total system head is usually neglible. You should end up with only ONE system head curve. Plot this as Head ( feet of fluid)as the ordinate and flow ( gpm ?) as the abcissa (x axis).

You then plot the pump curves on the system head curve for each pump and find the point(s) of intersection. These are the operating points for each of the pumps running individually.

When more than one pump is running ( in parallel) you must add the pump curves together for those pumps which are on. Two identical pumps running in parallel will roughly double the flow, though not quite as you'll see when you do this.

good luck

 

[rcooper]

A non-conventional approach I use with multiple pumps in parallel is to calculate the system curve as described by RWF7437 and plot this on the pump curve. I then divide the flow rate by 2 on the system curve and plot this on the pump curve. For three pumps I would divide the system curve by three and plot it on the pump curve.

The intersection heads are identical to RWF7437, but you are looking at the influence of the system on an individual pump. The advantage of this method is you can see how the efficiency, NPSHr, flow rate compared to BEP flow rate and power all compare for all pumping conditions. To identify the total flow rate with three pumps running you would find the intersect of the system curve/3 and the pump curve, and multiply this by 3.

Another advantage of this method is you can compare the performance of various different pumps easily on a single graph.

 

[BigInch]

Think you need to calcuate the system flowrate at 1 pump flow, two pump flow, three pump flow, etc. Calculating a system curve and dividing by some integer doesn't produce a correct solution, although it might be somewhat close from time to time, if static head is more important than friction head loss. If friction head loss is more severe, errors will increase.

http://virtualpipeline.spaces.msn.com

 

[RWF7437]

RCooper's solution is interesting and often useful but only works, I believe , if all the pumps are identical.

 

[cvg]

doesn't even work then, because as you fire up additional pumps, velocity increases and head loss varies with square of the velocity. So unless you have a very large diameter, smooth and short header and transmission pipeline, two pumps probably does not give double the flow of one pump.

 

[RWF7437]

Yes cvg,

That was said earlier:

"Two identical pumps running in parallel will roughly double the flow, though not quite as you'll see when you do this."

 

[BigInch]

Agree it was implied, but not exactly said.

http://virtualpipeline.spaces.msn.com

 

[quark]

Don't mix up system and pump curves, initially. Construct the system curve based on maximum possible flow and corresponding pressure drop, then check various pump combinations.

Calculating individual flowrates and pressure drops is a bit difficult and you can construct a system curve without doing this by using Parabola Method. However, you should be cautious that parabola method assumes a constant friction factor for all flowrates (or Reynolds numbers). I gave a detailed procedure in one of the past threads in piping forum. This procedure deals with what cwg mentioned in his post.

The starting point of your system curve depends upon the static component of the pressure drop and this is your pressure drop corresponding to zero flow (in other terms, this head is necessary minimum value for all flowrates)

Once you draw the system curve with the maximum possible flowrate, draw the compound pump curves on it (keeping head constant and adding up flowrates, for similar pumps) and the intersection gives you the operating point for any setup. The operating flowrate, i.e the intersection of pump curve and system curve, divided by the number of working pumps gives you flowrate through each pump.

 

[quark]

If you can figure out from my explanation, the procedure is at thread391-82973

 

[BigInch]

quark,

Not to detract from your parabolic curve approximations, which I have used from time to time, but you yourself do recoginze some limitations in its application.

I suggest that exact system curves are relatively easily put together (with a spreadsheet), especially if using the very accurate and computationally efficient and easily implemented Churchill pipe friction equation. Using Churchill, it is quite easy to construct system curves, even for a pipeline system flowing multiple products.

http://virtualpipeline.spaces.msn.com

 

[quark]

BigInch,

No doubt that you are correct. However, I am more vulnerable to pure mathematical figures than to scattered graphs:-(. Your advice is much valuable in terms of accuracy.

 

[rcooper]

A bit of care with the parabola method is needed if you are pumping through a system comprising some lamina flow components (such as a sand filter). You will require three points to determine the formula for head where;

h=Aq^2 + Bq + C

where A is the turbulent headloss component, B is the lamina component and C is the static head

 

[BigInch]

quark,

Its not too confusing and, since you're a nice guy (in fact we're both nice guys), I'll share my system curve spreadsheet with you all. This one is using an old method I had implemented of a 1-cell contained 3x iteration of Colebrook-White friction factor equation. Use the rapidshare link below to download it if you like.

http://rapidshare.com/files/40950094/Pipeline_System_Curve_Colebrook.XLS.html

http://virtualpipeline.spaces.msn.com

 

[quark]

BigInch,

That is very good. I never tried that out of sheer laziness though I had my first piping spreadsheet 8 years back. Thanks for your continuous efforts towards accuracy. If the email id on virtualpipeline is yours, I would like to send you a spreadsheet I prepared that has a comparison of friction factors (equations) and pipe sizing by 2K and 3K methods. You will get katmar's version of Churchill as a bonus and that equation seems to be a perfect fit even in transition region. Let me know.

Thanks again.

rcooper,

I would like to know more about Bq and its simultaneous occurance along with Aq[sup]2[/sup]. Can you please explain?

 

[BigInch]

quark,

Yes Churchill works through all flow regimes. And yes, my e-mail is on my web page in the contacts section. I would be pleased to hear from you.

If you would like to share with everyone,

http://www.rapidshare.com

makes it quite easy to do so. You don't even have to make an account if you don't want to, or even log in for that matter.

http://virtualpipeline.spaces.msn.com

 

[rcooper]

quark,

Take a system where you pump from a tank, through some pipes, down through a pressure sand filter and through some pipes to a reservoir.

Your tank has level So
Your reservoir has level S1

Therefore you static head is S1-So=C

Your pipe work friction and fittings loss is given by (K+fL/d)v^2/(2g) - assuming one pipe diameter and the friction factor does not vary with velocity. i.e. for turbulent flow the headloss is proportional to the squart of the flow rate.

The flow through the sand filter is lamina, so the head loss is proportional to the flow rate, so the headloss for this section of the works is given by Bq.

So if you have a system which involves both turbulent flow and lamina flow the head loss is proportional to both q and q^2.

As I said above, this happens if you are flowing down through a bed of sand. If you are flowing up through a bed of sand, the sand could fluidise, at which point the headloss through the sand is independant of the flow rate.

 

[quark]

Good point! Similar is the case with all packed beds. The Ergun equation for turbulent flows is of the form Av+Bv[sup]3[/sup], where v is superficial velocity. I enjoyed this discussion.

BTW, how do you guys manage these wide posts without having to change the resolution? I have IE6 and there is no horizontal scroll bar appearing.

 

[BigInch]

Sorry, I caused the wide screen when I posted the view of the speadsheet.

IE has no hscroll. You have to use full screen view, or grab the right edge and stretch IE wider.

http://virtualpipeline.spaces.msn.com