Flow Measurement and Hydraulics

[EmmanuelTop]

We are running a Gas/LNG plant here in Bioko island, and there is one interesting issue related to parallel pump operation and flow measurement/control from the day No.1 of plant operation. I have attached all the information I have in the Excel workbook uploaded below. It also includes a simple process model prepared in HYSYS, where all piping isometrics, pump curves (Q-H), and control valve data are incorporated. Model outputs match closely with actual performance data.

I would appreciate if you can take a look at these materials and help us figuring out the following issues:

1. Why there is such fluctuation of flow across the flow control valve in common discharge line, as compared to individual flows delivered by each pump? All flows are measured by Pitot tubes, both at individual pump discharge and in the common discharge line (upstream of FCV12002). You can see the range of fluctuations of the sum of flows from two pumps operating is much smaller than the fluctuations observed at the common FCV. The same thing happens when we close the FCV and control the flow through FCV bypass line - flow fluctuations remain at 150 m3/h! This indicates the control valve is definitely not the culprit of our problem, although it has been designed for much smaller dP than it is in practice. In addition, there is an offset of Pump (1) + Pump (2) flow versus total flow measured upstream of FCV.

2. The Instrument guys and us did some testing, and these are the results:

- Instrument maintenance checked out the flow transmitter FT12002 including blowing down the impulse lines.
- FC12002 in manual output steady, valve feedback fluctuating 0.5% (witnessed movement at valve), unsteady flow around 150 m3/hr.
- Flow controlled via bypass valve and FV12002 closed. No improvement.
- Flow controlled via bypass valve, FV12002 closed and upstream block valve of FV12002 closed. No improvement.

 

[CJKruger]

EmmanuelTop,

Very good information to review. I don't see anything obvious, but here are my comments anyway:

- Pump curves seems reasonable for parallel operation.

- Common control valve dP seems reasonable.

I would look at the following:

- Minimium flow valves FC150-152. Are you sure one of them are not leaking or something ?

- Is the standby pumped blocked in ? Probably not, maybe do it for a while and see if things improve.

- On our lean amine pumps we had some problems with cavitation. Are your pumps running noisy ? Does the oscillation reduce if the tank is fuller ?

- Oscillating frequency seems roughly 10 cycle per minute. Feels to fast for a control loop and too slow for cavitation.

- Seems one pump is not on its curve. Maybe have a look at the test results (if available). Else it may be spilling back to the tank somehow.

- I don't know pitot tubes. Do you think flow turbulence in the pipe can do it ? What kind of straight runs do you need for a pitot tube ?

- If all else fails, can you just dampen the signal and forget about it ?

 

[BigInch]

Does the pump piping and header configuration actually look something similar to the model diagram? (I don't use Hysys so those numbers on the pipe don't mean anything to me) I notice that your pumps, even though they are in parallel operation, do not operate at the same point, I would imagine due to the nonsymetrical arrangement of the suction and discharge lines. Is the flow graph the result of dynamic Hysys simulation (I don't know if Hysys does dynamics), or is it steady state only?

My initial suspicions would involve some oscillation going on between the two pumps, which are not actually remaining on the same operating points shown, but rather more drifting to towards each other's operating points, then drifting back as some unequal and intermitant load distribution is going on. Of course I may be all wet.

Have you tried a dynamic sim, with all the control actuators accurately tuned and duplicating actual field measured values, entered those into the model and seen no indication of oscillations, even when making small changes to one pump's flowrate to try to set up some oscillations to see if they deteriorate, remain stable or increase? These may be some slight semi-stable oscillations between pumps, resulting in those longer term time periods that you are seeing.

**********************
"Pumping accounts for 20% of the world’s energy used by electric motors and 25-50% of the total electrical energy usage in certain industrial facilities."-DOE statistic (Note: Make that 99% for pipeline companies)

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

 

[pmover]

i cannot explain the differences in flow (total vs individual pump); however, focusing on the flow element, instrumentation (check calibration), & pump vent/drain piping/valves to ensure integrity of the entire system is apparent.

I concur with BigInch regarding the piping arrangement. If the inlet/outlet piping is not symmetrical (i.e. equal resistance across each pump), then the pump(s) with greater resistance will likely have less flow.

I note the pipe size for the common recycle flow back to the tank is the same pipe size exiting each pump flow control valve. You might want to investigate if there is any impact on pump/flow control operation resulting from any back-pressure. something to ponder/investigate . . .

any or a combination of factors may result in the oscillations.

good luck!
-pmover

 

[BigInch]

I think the pump with greater resistance has less flow for a certain time, since in reality it is a dynamic system, a slight perturbation will cause the operating point of one pump to move a little on that pump's curve. If the pump curve and system curve shapes are relatively flat, the flow will change significantly, with little change in pressure. The change in flow of that pump is added or deducted from the other pump, which causes the 2nd pump's operating point to change. If the 2nd pump's flow changes too much, the change must be added/deducted from the 1st pump's flow, hence the oscillations passing back and forth between both of them occur.

This can also happen if a control valve's parameters are not tuned properly to the system and the resulting time constant of the control valve causes system instabilities. The control valve actuator's time constants can either cause the system to be under-responsive and too much drift from the set point, over-responsive moving too fast to the set point and overshooting, or somewhere inbetween in a weak control band. If the parameters are set too far away from the right values for a given system, set points are too slow to return, or overshoots fly out of the controller's range and the system will probably shut down on too low or too high flow.

Whether you have a control valve that causes these effects, or they are caused by the interactions of the pump's flow change accelerations interacting with and the pipe fluid accelerations are immaterial. One or both could be the cause.

If you are interested in the theoretical aspects of how such effects can be shown to occur, I have attached a spreadsheet of a simplified dynamic analysis of a system with two pumps and its control system. In the original analysis I have set up there, you can see that an oscillation between flowrates of 40 and 110 is obtained at pump 1 and an oscillation between 95 and 150 at pump 2.

**********************
"Pumping accounts for 20% of the world’s energy used by electric motors and 25-50% of the total electrical energy usage in certain industrial facilities."-DOE statistic (Note: Make that 99% for pipeline companies)

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

 

[pmover]

BigInch,

yes, i concur with thoughts regarding the control system. interesting analysis as well, but how can a system resistance have 300' head with 0 flow? i am puzzled.

btw, the pump curves are not flat (furnished by OP) with a good steady increase in head with decreasing flow.

i'm not familiar with using pitot tubes for flow measurement in liquid systems. i wonder how repeatable and strong the flow signals are and whether or not the flow signal is oscillating and its impact on the control system.

obviously, when two separate flow measurement devices are not equal to the total flow measurement device, investigating the flow measurement devices and the entire piping system for integrity purposes is first and foremost.

pmover

 

[BigInch]

Pumping against a static pressure of 130 psig. For example, if you have to pump to a tank who's maximum fluid elevation is 300 feet above your pump centerline, when the pump starts at 0 flow, there's 300 feet of head it must start against. Or another ex., if you have to pump water into a steam generator that operates at 800 psig at all flowrates, the minimum head needed to do so is 800*144/62.4 = 1846 ft. As flow increases from 0, head increases due to frictional resistance. And you should check to see if the driver (not so much of a problem with electric drives, but engine start torques can be low) has sufficient starting torque to start against large static heads, or if that backpressure will overcome the pump starting torque and try to reverse flow through the pump. I had a system where a diesel engine had to start and build up pressure to some 1400 psig before the discharge check would open. That 1400 psi was at zero system flow, due to a 4000 ft rise (of diesel SG=0.825) across the next mountain chain summit. It needed a recirculation system to prevent overheating during warm-up and the ramp to 1400 psi, since there was no cooling system flow at all during that time. We could not even count on any tiny amount of line pack flow to help with cooling.

The pump curves if flat, don't help, but actually it is the angle of intersection between the pump curve and the system curve that is of interest. The closer that is to 90º, the stronger the feedback signal you can get, ie. in terms of one varible's value affecting the other; a unit change of head producing a unit change in flow & v/v. So, the shape of both curves are important. Also important not to have regions, for example, where the pump curve has rising head and rising flow, because when plotted against the system curve there may be no intersection, thus no one point solution. Some centrifugals have that characteristic at lower flowrates. Reciprocal pumps can discharge with rising head and flow, but their "curves" are pretty much straight lines with positive pronounced slope and they are less likely to closely approximate the system curve, so you can still find intersections.

Thinking of pump systems in curve intersection terms, graphically shows why there can be lots of wide flow and pressure variations during startup against an automatic control valve that is trying to balance system flow with pump flow when both are small, as both pump and system curves are likely to be flat(ter) during low flows and intersections, the operating point, at those low flows are harder for the equipment to pinpoint.

**********************
"Pumping accounts for 20% of the world’s energy used by electric motors and 25-50% of the total electrical energy usage in certain industrial facilities."-DOE statistic (Note: Make that 99% for pipeline companies)

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

 

[BigInch]

And that spreadsheet represents a very simplified solution, since I determined the flowrate of the 2nd pump simply by algebraicly subtracting any flow change at the first pump to the second pump. That might not be the case, as that flow change might take time to set up at the second pump. I grossly ignored a potential highly complex hydraulic feedback occuring between the two pumps by assuming that it is a simplistic and instantaneous algebraic function. A very good transient simulation can demonstrate if any of these problems exist. Fortunately, most can be at least toned down, if not eliminated, by proper choice of the controller parameters, but that's often accomplished by an increase in the time needed to arrive at stability. Ask an instrumentation engineer for details.

**********************
"Pumping accounts for 20% of the world’s energy used by electric motors and 25-50% of the total electrical energy usage in certain industrial facilities."-DOE statistic (Note: Make that 99% for pipeline companies)

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

 

[RAHSEED]

1.YOUR VALVE HAS SOME HYSTERISIS ISSUES,GET YOUR TECHS TO CALIBRATE AND CHECK REPEATABILTY.
2.CHECK YOUR PUMPS ONLINE SEPERATELY (CURRENT DRAW)
3.MAKE SURE YOUR DP TX IS CONFIGURED FOR SQUARE ROOT EXTRACTION.
4.MOUNT YOUR DP BEFORE THE FCV(AT LEAST 5 PIPE ID)
5.DAMPEN YOUR TX 3-4 SECS.

IVAN

 

[786392]

Dear RASHEED, Kindly avoid using 'BLOCK CAPITALS' in forums.

Best Regards
Qalander(Chem)

 

[EmmanuelTop]

OK

To bring some refreshment to this topic, here are the field measurement results:

- Both pumps have rock-steady discharge pressure
- Flow variations at FT12002: as high as 175 m3/h
- Amine velocity inside the piping: 1.5 m/sec

I would say this is purely a measurement problem because I cannot think of anything else that can cause such huge fluctuations of flow.

Any further hints on this?
Thanks,

 

[786392]

Dear
If there is not any variation(s) on changing of pump's set combination(s;then check

1)Tapping's for no blockage(s)at root take-off or up to the sensing element.

2)sensing element's electrical connections or

3)electronic cum pneumatic reading/converting systems

Best Regards
Qalander(Chem)

 

[EmmanuelTop]

CJ Kruger & Others,

The transmitter: quite odd, but the signal cannot be dampened. The only transmitter in the plant where signal damping is not an option.

Can this be the sole cause of flowrate measurement fluctuations?

 

[JLSeagull]

If you need dampening and it is not available in that transmitter, then swap transmitters. Perhaps the HART communicator or AMS software can access dampening.