Centrifugal and Reciprocating Pumps

[Thealanator]

I have been trying to understand the flow vs. head business in the case of a centrifugal pump feeding a reciprocal pump.

Since a reciprocating pump is a constant-flow device, the flow rate in the centrifugal pump is fixed. Therefore, the centrifugal discharge head is the sum of the suction head and the differential head at the particular flow rate. If NPSP (A) is greater than NPSP (R), the reciprocating pump is happy. If the flow rate is at or near BEP flow, the centrifugal pump is happy.

Is my logic correct?

 

[jistre]

The centrifugal pump will not see constant flow here. It's a constant head device. If the two pumps are in series, it's impossible for the centrifugal pump to flow any fluid if the intake valves of the PD pump are closed. There's nowhere for it to go. The centrifugal pump should (in theory) happily keep developing a steady head, and the flowrate will change based on the backpressure at the outlet developed by the opening and closing of the PD pump valves. The centrifugal pump's instantaneous efficiency will be all over the curve, as it flowrate should be pretty much constantly changing.

 

[jistre]

Ignore that last part about efficiency. Brainfart.

You should be right on the BEP. The centrifugal pump will not have constant flow, but it also shouldn't care what part of its TDH is split between static head and velocity head.

 

[dcasto]

Thealanator, you are correct. Between the PD pumps suction bottles, lines, valves and such, the centrifugal will be smooth, UNLESS you have a 30" diameter, 15 foot stroke PD pump, runing 1 rpm, then it will act per jistre states.

 

[phidias]

i am a little unsure in what you are asking but if you look at a pump curves of a particular pump and you put a cent pump pump in series wih an identical pump it will do twice the differential press across the both pumps as each pump can raise the pressure by x amount (same value press not flow) depending on the flow/pressure required. (eg no flow max pressure max flow no pressure follow pump curves). if you put a cent pump in parallel they will do twice the flow and the same head depending on pipework. in series the first pump flow is the max flow at the discharge of the receving pump but the presure will be twice the first pump size and application is a large factor - nsph. a cr pump is a cent pump that adds to its press as the fluid travels through its stages it is a in series pump.

 

[phidias]

so if you are feeding a recproicating pump the head flow is constantly changing. and proformance of second pump will always change

 

[phidias]

yes so the npsh is imposible to calculate well at least i cant but a may be posible? the only way i can think of is to take the lowest or worst case point of the second pump and spec the firsst pump to not fall bellow this set point press and flow

 

[JJPellin]

All you have to do is install an appropriately sized accumulator near the suction of the PD pump. The accumulator will take the variations in flow and the centrifugal pump can delivery a smooth and continuous flow at the rate that the PD pump will pump it. This also helps limit the affect of acceleration head on the NPSH available to the PD pump, so it is less likely to cavitate. You affectively decouple the performance of the two pumps and each can be evaluated using standard methods for a pump of that type.

Johnny Pellin

 

[BigInch]

I think its all easier to see, if you go back to basics and make a combined system diagram showing TOTAL head vs flow. Its not a waste of time, since you get not only a clear view, but you can include the diagram in the system operating manual later on.

[/url]

Draw the load on the Combined Diagram (CD).
Draw the NPSHA curve.
Add each centrifugal differential head to the NPSHA as you transfer its curve to the CD to show the total discharge head (TDH).
Draw the family of Positive Displacement Pump's "Curves" from each Centrifugal's Total Discharge point.
Now from any flowrate, go up to the Centrifugal's TDH, then follow the Recip Family Curve at that point to its intersection with the Load Curve to see what the Recip Pump will have to do to get you to the combined system operating point.

Take flowrate 1, go up to Load Curve. Recip Pump is not rquired to run and could be bypassed. Actually a CV would need to be installed to run at flowrate 1 or less and knock the centrifugal's TDH down, or the operating point for the centrif alone would move to the intersection of load and Centrif curve (near flowrate 2).

Take flowrate 2, go up to Load curve. Recip must provide a little bit of head to run at flowrate 2.

Take flowrate 3, go up to Load Curve. Recip must provide significant head to run at flowrate 3.

To see when the system could be run with the centrifugal bypassed, draw a second diagram placing first the recip head, then add the centrifugal head on top of that.

http://virtualpipeline.spaces.msn.com

 

[dcasto]

This getting way to complicated. The reason for doing this is that you must need lots of head, if not, then why a PD pump? The PD pump is going to do the work, the centrifugal is just there to make sure you have NPSH for the PD, don't over complicate the centrifugal pump.

Cetrifugal pumps feeding PD's is not that complicated. The centrifugal pump is selected based on NPSH available and its differential is the NPSH required by the PD at the PD's rate (heck add 100% more, it won't hurt anything. Pipe the discharge of the Cent to the suction of the PD, put a recycle valve from the the discharge of the PD to the suction of the centrifugal keep the flow constant if the feed source fluctuates.

 

[BigInch]

I'd put the recycle from the discharge of the Centrifugal to the suction of the centrifugal. Pressures not so high that way, it does the same thing and probably easier on the valve taking that pressure cut.

http://virtualpipeline.spaces.msn.com

 

[Thealanator]

Interesting points, one and all.

I am a bit confused by the diagram, but maybe I can add some assumptions to my "case":

RPM's, bore and stroke on the reciprocating pump are fixed, so flow rate cannot change (hence my confusion over a family of curves).

The reciprocating pump discharges to vaporizers and into receiver tanks whose pressure runs between 2500 and 5000 psig.

 

[BigInch]

That's what you needed to specify, if you only want to look at one recip line. Since the class of recip pumps are not fixed flowrate devices as you said, they have a family of curves for all possible flowrates. Until you clarified, the recip pump could have had a variable speed (diesel) driver or a VFD on an electric motor, or as decasto suggests... a recycle line from the centrifugal's discharge to the centrifugal's suction. All of those would give the recip a family of curves it could follow.

http://virtualpipeline.spaces.msn.com

 

[dcasto]

To clarify your controls, what would you do if the reciever pressure started to exceed 5000 psig, shutdown the whole process? Would it not be better to take un vaporized material back to the liquid source and recycle a little until the downstream system catches up?

 

[BigInch]

Recycling from either the recip or the centrifugal to the CENTF suction would have the same effect, nothing as far as the pump discharge is concerned. Flow through the pumps stays the same, so discharge head stays the same. The system curve would tend to lower though, but not too much, if the discharge header was a big one and operating at 5000 psig.

If you recycled from recip discharge to recip suction, you would be able to hold centrifugal discharge pressure, but the high pressure liquid arriving from the recip's discharge would tend to increase interstage pressures anyway, so possible of little benefit. Hence I would look for the driving cause of the process going over 5000 psig and deal with that. If that didn't work, then yes, shutdown the whole system.

http://virtualpipeline.spaces.msn.com

 

[dcasto]

The OP's appears easy enough, ther may be other things, like fluid properties, that allow upsets, so BigInch's proposal my be just as valid as mine or others. Point is, there is no magic, not patent medicine, no snake oil, its easy. A set of curves might be better in explaining to some people, but I just know how simple it is in real life. The OP appears to have it right for his situation.

 

[BigInch]

True.

I just always go back to basics and line up whatever type of flow element curves I have with differential head references to the system curves referenced to absolute head. Works EVERY time.

http://virtualpipeline.spaces.msn.com

 

[Thealanator]

To answer the control question - if the receiver pressure reaches the upper limit, shut the pumps down and when pressure reaches the lower limit, start the pumps up.

 

[dcasto]

Oh, snap acting process controls. Ever looked at VFD's and recycle loops.

Then again, if the pump is 20 GPM, not much to save in energy.