Selecting centrifugal pumps for occaisional psarallel operation? 2

[JoeySoap]

Hello all, recently, when I was asked a question by a young process engineer, I struggled to answer it concisely. The question is- When selecting 2 x 100% pumps for a given duty, if they may sometimes be run in parallel, we select a pump curve where the single pump case is close to BEP but that the flow/head curve rises to the left hand side of the curve to ensure that at the increased head with both pumps running, both pumps should be running stabily. To what degree should the curve rise? At what slope do we consider the pumps could be possibly unstable when run in parallel operation? I couldn't answer this question properly. Thank you.

 

[BigInch]

I would think that the question is not very relavent when discussed in the context of 2 x 100% pumps, as it would be extremely rare that you would ever want to run that configuration. Its doubtful the system would resemble the normal operation curve at 2 x flowrate, so there could possibly be many more instability questions arising from the increased steepness of the system curve. In fact, the system curve could increase so much, that running 2 x 100 % pumps might require more head than the pumps could deliver, thereby slowing the 2 x target flow to much lower levels. To make a short story out of this, you can't tell, unless you examine the system curve at a 2 x flowrate and see what the intersection with the 2 x pump curve looks like. The more perpendicular the two curves intersect, the better the control response would be, but that's not to say that they must be anywhere near perpendicular to provide a good enough control response. Frankly, I don't know what the minimum intersection angle should be, they certainly don't need to be anywhere near perpendicular. I can imagine that the wholistic answer would also require some input about how closely you need to control the flow point too.

The question is however entirely relavent when discussing 3 x 50% configurations, as well as other configs, 1 x 100% + 2 x 50%, where you obviously could indeed have many desirable occasions to actually run such configurations, so it does warrent some thought, I'm just not sure how much. , but I will say, it depends just as much on your system response as it will your pump curves, and on your acceptable process flowrate, pressure tolerances too.

"We have a leadership style that is too directive and doesn't listen sufficiently well. The top of the organisation doesn't listen sufficiently to what the bottom is saying." Tony Hayward CEO BP
"Being GREEN isn't easy." Kermit

http://www.youtube.com/watch?v=hpiIWMWWVco

http://virtualpipeline.spaces.liv

 

[JoeySoap]

Thanks that's very helpful.

 

[CH5OH]

important for stable operation is increasing head with decreasing flow.consider two pumps running parallel with such a curve(total flow is constant):
if one pump its flow is increased, its pressure drops, so other pumps will increase its flow and flow will settle between the pumps.
consider two pumps with curve with falling pressure and falling flow:
if one pump its flow is increased, its pressure increases,so other pump will decrease flow and decrease pressure.when pump 1 is somewhere at end of curve and its pressure dropped to shut off pressure of 2nd pump,then it all reverses(oscillating).I think API has a standard on prefered curve rise towards shut off pressure,important is to select a pump with rising curve (as opposed to flat or falling curve)towards shut off, when parallel operation is required.

 

[BigInch]

What about something like this,

The algebraic difference of slopes x shutoff head x density of product (units of pressure) should be more than the accuracy of your PT.

That defines it from a controls sensitivity perspective, but there remains some system hysteresis that still isn't resolved. How fast the system curve changes is the result of how fast the head or pressure changes to a unit change of flowrate. That is dependent on where you are on the system curve, the pipe material, length, pipe diameter and product bulk modulus, so that could effectively be anything. There may be no other perspective to examaine this problem from other than from the controls perspective above. Let's see if anyone else has some ideas.

"We have a leadership style that is too directive and doesn't listen sufficiently well. The top of the organisation doesn't listen sufficiently to what the bottom is saying." Tony Hayward CEO BP
"Being GREEN isn't easy." Kermit

http://www.youtube.com/watch?v=hpiIWMWWVco

http://virtualpipeline.spaces.liv

 

[CH5OH]

The algebraic difference of slopes x shutoff head x density of product (units of pressure) should be more than the accuracy of your PT.

given that answer to the young operator should make him somewhat reluctant to ask any more difficult questions (lol)
what is PT?

 

[BigInch]

Good one. But rising head with falling flow is only half the "X".

PT = Pressure Transmitter.

"We have a leadership style that is too directive and doesn't listen sufficiently well. The top of the organisation doesn't listen sufficiently to what the bottom is saying." Tony Hayward CEO BP
"Being GREEN isn't easy." Kermit

http://www.youtube.com/watch?v=hpiIWMWWVco

http://virtualpipeline.spaces.liv

 

[Artisi]

First up, I would send the "young process engineer" away to think about the problem and do a bit of research as this is a VERY basic pump application problem.

A simple solution to a simple problem is;

On a graph sheet draw on the pipe system curve to beyond the anticipated flow for the 2 units running in parallel, now draw on the single pump curve and then the 2 pumps in parallel which is the flows added together horizontally at the same head.

Where the single pump and the pumps in parallel cross the system head curve is the operating points for the pumps in single and parallel operation.

 

[rmw]

Not only basic but very fundamental if you are going to mess with pumps at all.

rmw

 

[JoeySoap]

Thanks all, that's interesting. I think my answer to the young guy wasn't far off. Let me expand slightly, the duty in question is a storm water sump that will normally be emptied by one pump on an on/off level control. There is a single 100% standby pump. We all know that operators may soemtimes run both pumps if they think it will empty the sump a bit quicker in the run up to a shift change/coffee break etc... Hence the question about selecting the pump for an occaisional parallel operation. The pump selection we made is for single pump operation at around BEP flow matched to the system resistance/static discharge head and in the datasheet we asked the vendor for a 10% continuous head rise to shut off. I asked the guy to look at the system resistance for flows between 100% and 200% and to compare that against the pump curves to see where (if anywhere) the pump curve matched the revised system curve. I said that it was possible that we'd not have anything like 200% flow, we'd be to the left of the curve where one pump could partially close the discharge non return valve on the other pump since the curves on the two pumps will not be identical but this was why it was important to have a continuous head rise to shutoff, otherwise if the curve was very flat the pumps would be unstable. I hope I didn't mislead the young guy too much, he hasn't come back to me yet!

 

[BigInch]

That's good enough. Let him run with it.

"We have a leadership style that is too directive and doesn't listen sufficiently well. The top of the organisation doesn't listen sufficiently to what the bottom is saying." Tony Hayward CEO BP
"Being GREEN isn't easy." Kermit

http://www.youtube.com/watch?v=hpiIWMWWVco

http://virtualpipeline.spaces.liv

 

[JoeySoap]

Thanks BigInch, good enough for me too then!

 

[stanier]

Suggest the young engineer visits

http://www.mcnallyinstitute.com

and has a read.

Also he/she could download this free ware and get to see what happens under different scenarios/pump curves

http://www.pumpsystemsmatter.org/content_detail.aspx?id=110

 

[JoeySoap]

Thanks, that's helpful, I'll have a look at those myself too. I'd forgotten about McNally, his papers are very good as I recall.