Pump Recirculation and Suction Pressure

[Armen75]

Theoretical question:
On a centrifugal compressor.
assume normally operating at a suction of 10 bar (constant, very large volume tank) and a discharge of 100 bar. Flowrate of 100m3/h.

If I have a blocked outlet and a recirculation line which is grossly oversized (worst case scenario), what will be the pressure at the suction of the pump?

With no Delta Pressure in the system, the pump's suction and discharge pressures will equalise, but at what:
-suction pressure (ie 10 bar)
-discharge pressure (ie 100 bar)
-somewhere in the middle?? (how to determine ?)

I also have figured in that practically the pump will go off its curve (on the right) and hopefully shut itself off prior to shaking itself to peices....

I am trying to determine if a heat exchanger in the suction side needs overpressure protection, and if so, how much.....


many people are throwing in their 2-cents, but all I have is a bunch of pennies!

 

[BigInch]

If someone closes the outlet valve off the heat exchanger, or it becomes plugged, you've probably got close to full pressure on it. If someone makes a mistake and closes the discharge to the mainline quickly, then the recirculation valve opens, full pressure could reach the exchanger. At least until the pump gets out on its curve again. Deciding to use a relief valve or not requires that you consider all unsteady state cases that can happen. Use a relief valve!

If you have high pressure sources connected to low pressure piping, that's a potentially dangerous item that is not protected and a safety issue that should come up on any HAZOP review. If flow is blocked in the low pressure piping, pressure will go high. Its only a matter of time as to when that will happen. Use a relief valve.

In a "large" recirculation line, large enough to prohibit reaching sound velocity, you will have pump discharge pressure on the inlet and pump suction pressure at its outlet. The flowrate around and through the loop formed by the pump and recirculation line will be whatever it has to be to have a pressure loss equal to the current pump's pressure increase. As you have surmized, the pump will go to that flowrate wherever it is on its curve and its discharge pressure will adjust as it does so, in either direction. Its new pressure differential will decide the flowrate in the pipe. It keeps iterating until it arrives at a equilibrium. Before it reaches the end of its curve, you could have any pressure all the way up to full shutoff pressure in that loop somewhere as flow in the loop is setting up, depending on the rate of fill and velocity changes occruing at points inside it.

**********************
"The problem isn't finding the solution, its trying to get to the real question." BigInch

http://virtualpipeline.spaces.live.com/

 

[Armen75]

Yes, HAZOP done and flag raised however the feasability of overpressure came up when trying to calculate a relief rate.

Considering it is a liquid, therefore incompressible, if there were to be an overpressure, the relief valve will open and burp some liquid... since the source of the fluid is below relief pressure, the lack of fluid will drop the pressure and close the PSV... a closed PSV, fresh liquid introduced into the system and we start all over.....
do you keep doing this over and over?

how big does the PSV need to be?

 

[BigInch]

First, refer to a liquid as "slightly compressible", never incompressible. If steel is compressible, you can be sure that liquid is. See the liquid's bulk modulus property.

You need to protect low pressure piping, therefore you must set the relief valve to the allowable pressure + transient allowance for that low pressure piping.

In the case you are trying to set the relief valve pressure for the pump shutoff pressure + 10 %, or something similar that is higher than the heat exchanger or the low pressure piping allowable, you can't. You would then have to have a downstream pressure control valve between high pressure source and low pressure piping and equipment. It should be set to control to a pressure <= low pressure allowable. The relief valve would then go between the control valve outlet and the low pressure piping inlet, while still being located on high pressure piping; before you make the spec change transition to low pressure piping and equipment.

Size the relief valve for,

The maximum flow of the pump to that relief valve. If the recirculation line is large, so pressure drop inside is insignificant, the pressure drop would be 0 from the pump to the relief valve, so then you should take the flowrate below the head equivalent of that relief valve pressure, as shown on the pump curve, for your relief valve sizing flow.

**********************
"The problem isn't finding the solution, its trying to get to the real question." BigInch

http://virtualpipeline.spaces.live.com/

 

[Armen75]

I see your point, and agree the safest oute is to have a large PSV for pump flow but is it really necessary:

the acknowledging the possibility of overpressure brings up other issues.

ISSUE 1: If the relief valve is set to open on low pressure piping design pressure of 15 bar, once it opens and relieves a few liters, will it starve of liquid and stop pumping? (source tank does not have the pressure to fill into the system at relief condition)

ISSUE 2 : the pump is a providing head, it the inlet can see outlet pressure (ie 100) instead of its normal 10, wouldn,t that make the "new" outlet an unrealistic 190.... and then round and round??

 

[BigInch]

The pump is running, the pipe is large diameter, how's it going to starve? It may modulate, but in that case, you should get a non-modulating type relief valve, where lift pressure is slightly lower than return seating pressure.

2) No, when the suction pressure tries to go higher, the differential between discharge and suction drops, so flow in the recirc drops (that's feeding suction pressure) and then suction goes lower. When suction goes lower, the differential of the pump increases and any elevated discharge pressure would reduce. I wouldn't discount the possibility of some fast transient situation that might tend to do what you say, but it could only be sustained for a few seconds, and as soon as the pump differential rose again, its flow would drop too; also the typical pump motor would run out of power shortly thereafter anyway.

But if you can do it, I want the design. I think it would be a good start on a perpetual motion machine.

**********************
"The problem isn't finding the solution, its trying to get to the real question." BigInch

http://virtualpipeline.spaces.live.com/

 

[JJPellin]

I think I disagree. I can't say for sure because you are trying to make a decision about a real system by describing a ridiculous theoretical system. You would probably get a better answer if you described the actual system you are looking at. But, based on what you describe, I would say, no, you cannot overpressure the suction exchanger by opening an excessively large spill-back line. I could offer another theoretical example. Imagine a vertical turbine pump hanging in an open sump. If you open a full size spill-back to the sump, the pump will run off the end of the curve. It cannot build head since the discharge is wide open to atmosphere. The pump will stonewall off the end of the curve and the flow will stabilize at some level beyond the end of the published curve. The suction pressure to the pump cannot increase since it is an open sump. The only way to increase the suction pressure would be to raise the level in the sump. Since the fluid coming back to the sump cannot be any greater than the fluid removed from the sump by the pump, the level cannot go up. Matter cannot be created or destroyed. The pressure in the piping will equalize to atmospheric pressure which is basically suction pressure (allowing for some submergence).

If you take the same system and enclose it, you get the same result. The pressure in the system will equalize to suction pressure and not much more. The only way that it could reach pressures higher than that would be if the piping involved were extremely long relative to the speed of sound in the fluid being pumped. In this case, you could get a pressure pulse, traveling at the speed of sound and producing local pressure much greater than suction pressure. But, a PSV wouldn’t be good protection against a sonic pressure pulse. BigInch is extremely knowledgeable in the operation and behavior of pipeline systems. But pipeline systems tend to have very long pipes where the transients are more important. And he is correctly applying his real world experience to your poorly described theoretical system.

But, once again, this is all terribly theoretical. You didn’t say where the spill-back line ties into the suction relative to the very large tank or the exchanger. You didn’t say what the piping system was like between the tank, the spill-back tie-in point, the exchanger and the pump. In the end, I don’t think it really changes the answer. I don’t think you need a theoretical PSV in your theoretical system. And if you did decide to place one there for some circumstance that I have not considered, I think it would only need to be very small. As you say, in a closed loop system, you don’t have to burp out much fluid to drop the pressure a lot. And in the total spill-back case you describe, your system almost becomes a closed loop. Except if it is still open on the supply side to a suction source at lower pressure, and if there is not check valve in that supply line, then it can’t pressure up . . . I think . . . theoretically.


Johnny Pellin

 

[BigInch]

To me, it sounds exactly like a pump test loop. You just have to look at the pump curve to see what will happen. At startup/shutdown and no or lower flows, there will be a steep pressure gradient in the recirculation loop, if the exchanger is located near the pump discharge, it will see the high end of the gradient. If located near the suction, it will see the lower end of the pressure gradient. Once flow is established, the pressure gradient will reduce to that shown on the pump curve at the stabilized flowrate, which could be zero (or lower) if past the pump curve's intersection with the X axis. At low pump flows and, if someone closes the HX outlet valve, or it becomes plugged, the HX will be exposed to shut-off pressure. A PSV between hi/lo pressure ratings is the only safe way to go.

**********************
"The problem isn't finding the solution, its trying to get to the real question." BigInch

http://virtualpipeline.spaces.live.com/

 

[JJPellin]

BigInch,

But, he describes a very large supply tank. He does not say if the minimum flow line goes back to the tank (which it should) or the suction line. He does not say if it comes in before of after the heat exchanger. He does not say how the valve line up is arranged. In any real system that was not built as a test loop, it would make much more sense to design it so that it was impossible to overpressure the suction exchanger (which is quite easy) rather than add a PSV to protect it. Without further description (or a diagram) we are both guessing. Perhaps the original poster can enlighten us with more information.

You describe a situation where the flow could stabilize with a pressure gradient below zero. I don’t understand how this could be possible. With a negative gradient, the flow would reverse (or attempt to reverse, given a check valve). I can’t picture a way to describe this as stabile. The true closed loop would be the only situation where the suction could pressurize above the pressure in the large tank. And if the large tank was not connected to the loop, why was it mentioned. If it was a closed loop, as the OP and I suggested, a very small PSV could relieve the pressure since there would be no additional sources of fluid to replenish the volume removed through the PSV. This concept is used for water cooled exchangers all over our plant. For a blocked in case, with expansion of the fluid, either heated up by the process on the other side of the tubes, or a blocked-in fire case, a thermal PSV is added to protect the exchanger. But no matter now large the exchanger, the PSV’s are always tiny (usually 1”).


Johnny Pellin

 

[BigInch]

sorry, 0 @ 14.5 psia

**********************
"The problem isn't finding the solution, its trying to get to the real question." BigInch

http://virtualpipeline.spaces.live.com/

 

[Armen75]

Thank you all,
I am glad to see that this mirrors the debate we are having here.

I was trying to "simplify" and generalise to what I thought would be the worst case scenario, best described as a "pump test loop" (thank you BigInch).

The REAL system consists of the following:
It is an lean amine sytem pumping from the stripper to the absorber. the equipment sequence is
- A small pump (with it's own recirc) normally at 5 bar, shutoff at 10 bar.
- the main pump recycle return
- the cooler
- the main pump (from 5 to 90 bar)
- the min flow recirculation take-off with control valve
- the flow control valve
- the On-Off emergency valve.
- injection into the absorber.

It would be nice to return the fluid back to the source (Stripper) but we can't do that due to the risk of starving a large pump if the small pump is turned off.

The cooler is designed at the lower pressure due to cost, and changing is not a prefered option.

although there are manual valves downstream of the cooler, there is no risk of causing shut-off pressures since closing these valves would also starve the pump and cause no flow.

Therefore, the "real situation" would be the closing of the On-Off valve. That would cause a no-flow situation. The first small pump would go into recycle and there are no issues with shut-off for that one. The Main pump will also go to recycle.
With a properly sized recirc valve (minimal excess margin above manufacturers recommended minimum flow)there should be enough pressure drop to prevent an overpressure.

Our concern is what happens when that recirculation valve opens: is there a wave of high pressure that goes through the line until the new (higher) flow rate is established.
Or, what happens if the valve is too big. what pressures can the system settle to.
Or, if a pump has run-off the pump curve on the right, and head is close to zero, where inlet outlet pressures are close to the same, are they the same at 10, or 55?

 

[BigInch]

Yes, you get a high pressure shot. The question then becomes, does the pump sustain that pressure wave by continuing enough of a flowrate to keep that pressure up and force a pressure up of the cooler, or will the pump's flowrate fall off enough that the pressure wave is unsustained and essentially dissipated by the flow of the liquid behind it failing to keep up with the pressure front, ie. is it a dissipating pressure wave feed by only the expansion of the volume behind it, or is it a continuous constantly feed pressure wave. You must look at what the pump will do after the recirc opens up. If it continues to run at normal speed, it could be sustained, and probably will. You might see a pressure wave, followed by repressure as the pump continues. If it trips, maybe nothing to worry about, if it does trip.

Closing valve downstream of the cooler, or plugging of the cooler may eventually starve the pump, but what pressure will the cooler reach before the pump starves out and ... shuts down, burns up, etc. and discharge pressure is dissipated?

**********************
"The problem isn't finding the solution, its trying to get to the real question." BigInch

http://virtualpipeline.spaces.live.com/

 

[JJPellin]

This is similar to many amine systems we have in our refinery. They will generally never let ups recirculate all the way back to the supply vessel. In fact, in lean amine, it is much more typical for the minimum flow line to be a forward loop that dumps into the rich amine header to avoid velocity, turbulence and performance problems with the amine stripper.

In the system you describe, it is impossible to overpressure the suction exchanger unless the valves between the exchanger and the pump are closed. And, even then, the pressure spike will be very brief and probably not high enough to exceed the design rating of the exchanger. In all other cases, the flow in the loop will stabilize to suction pressure from the booster pump since there is still a wide open line from the stripper by way of the booster pump. The most pressure the exchanger can see is the dead-head pressure from the booster pump.

All of your comments are predicated on the idea of an oversized minimum flow line. If you eliminate this possibility, then the question is moot. The minimum flow spill-back should be restricted by the use of reduced trim in the spill-back valve or installation of an appropriate orifice plate such that it cannot pass enough flow to allow the pump to drop off the end of the curve. With an imposed pressure drop in the recirculation line, the pump flow is limited to some value just above the required minimum flow for the large pump. This serves at least three purposes. It prevents the possibility of overloading the suction side of the pump by high flow. It prevents damage to the large pump from operation at excessively high flows. And, it protects the piping system from velocity above design limits for amine service. As you are well aware, high velocity in amine piping greatly accelerates corrosion and reduces line life.

I believe we have at least 10 amine loops of the type that you describe. None of them have PSV protection on the suction side for the purpose of protecting against pressure resulting from a pump running in full spill-back. Your piping design group should be able to model the system and tell you very accurately what the pressure would be in full spill-back with your actual system configuration. Let them model it and you will have your proof.


Johnny Pellin

 

[Armen75]

I can not thank you enough for your answers;

Our conclusion is roughly what you both described; By limiting the capacity of the recirc valve (specifying a max Cv) and imposing a minimum pressure drop into the system, we are staying away from any continuous high presure scenario.
We are going to keep the PSV on that suction line but more as a small thermal relief line mainly due to the heat input due to the booster pump. That might also help dissipate a pressure wave from a recirc valve failure.

Practically, this makes a lot of sense and I feel comfortable with that.

 

[BigInch]

A limiting value for Cv is a commom misunderstanding, or not full understanding of what happens when there is little or no flow. Cv has nothing to do with limiting flow. Cv specifies a pressure drop across the valve when it has a certain flowrate. At NO FLOW, there is NO pressure drop and full pressure is on both sides of the valve. Thus, if the cooler were plugged, or shut off by its downstream valve, full pressure is at the cooler during a no flow situation.

**********************
"The problem isn't finding the solution, its trying to get to the real question." BigInch

http://virtualpipeline.spaces.live.com/

 

[Armen75]

Yes, I agree that specifying a max Cv implies the sytem is operating at minumum flow, and am comfortable doing so as a no-flow scenario is not really credible.
The only valves between the cooler and the pump suction are the pump maintenance valve which are 18 inches and will not be inadvertantly closed.
As for the cooler plugging, although possible, I think it would be initially identified by it's effect on the smaller booster pump upstream during normal operation.