Odd pump behavior - off the curve

[Rputvin]

I've got an interesting set of circumstances that is confounding all of the engineers at my company and I'm looking for someone that may have some idea of how to troubleshoot this.

Less than a month old, we ran off at our facility, but don't record flow values, only pressure and current on the motors. Closed loop, running a stamped stainless close-coupled centrifugal pump at ~73 GPM and ~25psi differential pressure. The pump curve for this pressure should be giving us ~120gpm of flow (selected slightly oversized), and we're sitting short of the customer's required flow rate of 80 GPM. The system pressure is almost dead-on the calculated drop for the system.

We're moving city water between 75 and 100°F. Using standard pressure gauges at multiple points in the system, before and after the pump. Using IFM Efector flow meters with proper installation runs, one after the pump discharge before branching, and one on each branch before collecting to the return header, all flow meters downstream equal the discharge meter before the header.

We're on the curve through about 40 GPM, then we start getting deviations and we're sitting 40% or so off the curve. The pump manufacturer built us another pump and ran it off through the entire curve to make sure their stamping and impellers hadn't drifted since the curve was established - their pump matched the curve values, albeit with much more ideal and controlled conditions than what I'm dealing with.

Our first step was to change the impeller to the largest available trim, which improved the numbers slightly to where I'm at now, but still significantly lower than expected flow.

The system is pressurized with a diaphragm expansion tank, we started at 12psi and have increased it to 20psi. There's an air separator installed, which is operating as best as we can tell, it does not have a strainer in it. The make-up line is city water connection regulated to 17 psi. The rest of the system is the customer's piping that I'm short on details on, but should follow a standard flow/pressure drop relationship and not figure into the conundrum.

The pump will make deadhead (shutoff/no flow) pressure. As we open the throttle valve (butterfly valve used for isolation mainly) we see a 1-2psi drop in suction pressure. The motor is drawing near nameplate amperage, suggesting we're to the far right on the curve. Required NPSH is all of 2.5psi


Any ideas as to what might cause a pump to... not pump? We're thinking suction restriction is a possible culprit, but there's nothing there to impede flow. Would a small amount of air in the system wreak this kind of havoc?

Any help is appreciated!

 

[LittleInch]

The pump curve would be nice to see but have you checked that the rotation is correct. I mean really checked yourself.

A strewn sketch or diagram would be nice to see but a very good description of your issue

Remember - More details = better answers
Also: If you get a response it's polite to respond to it.

 

[Rputvin]

Here's the pump curve. The triangle is the design point, 80GPM at 50ft TDH, which intercepted the smaller impeller curve. We're seeing right about 48-53ft of TDH, give or take, depending on conditions at time of reading. The original thought was a little system variance put us a few GPM shy and it was cutting it too close, stick in the bigger impeller and away we go. Well, flow increased by about 4 GPM moving to this impeller.

I haven't physically been on site, but I believe the rotation was checked when our service team was on site to do the commissioning and performed the impeller swap.

We've swapped gauges and verified against a calibrated master to ensure we're not off by 3psi or some amount on a gauge.

 

[Rputvin]

And here's the system diagram excerpt. I removed most of the brand/model info.

We're running a simple loop, removing heat from a reactor (non-nuclear). The customer's somewhat complex piping is more or less the big circle. As I said, I'm short on details.

We're moving forward with sticking a bigger pump in place, trying to get them running with enough flow to meet their needs. I'd still really like to know if we can figure out what's wrong either before we start cutting into the piping, or purely academically in case we run into this scenario again.

 

[LittleInch]

Ok. Bleeding obvious questions.

Is the standby pump isolated or sitting in the non return valve?

What is the 3m filter?

What is the chemical shot thing?

Where is your flow meter on that diagram?



Remember - More details = better answers
Also: If you get a response it's polite to respond to it.

 

[Rputvin]

Solid lines = things my company built. Dashed lines = everything else, field piping, customer devices, etc.

The standby pump is checked to prevent backflow. The higher pressure on the discharge side would force the check valve shut. We've run both pumps simultaneously with similarly underwhelming results.

3M filter is a fairly simple bag filter with a differential pressure switch to signal a dirty filter. The chemical feeder is a 2 gallon drum that allows easy checks of the water health and dosing of chemicals as needed to combat untreated city water coming into the system - mostly biocides and rust inhibitor. Both were isolated during measurements, they consume a total of ~3 GPM combined when open, flowing from high to low pressure. So our target is technically 83-84 GPM on the pump discharge, but as I said we're oversized and should be pushing >100 GPM, even if the piping conditions give us added inefficiency and loss to move 10-20% off the curve.

The flow meter is downstream of the heat exchanger in the customer piping. It exits our pump skid and the first transition is to a straight pipe run with 10 diameters upstream and 5 downstream of the flow meter. From there it flows into a header to feed the individual sections of the reactor. Post-reactor flow meters are on every branch, then the branches collect into a return header and loop back to the pump skid.

We tend not to trust flow meters, but all of them are setup perfectly according to the install guides, and have a published error of less than 2% (+/- 0.8% + 0.5% of flow value) and are from a reputable company.

 

[LittleInch]

OK, You seem to have covered all the bases.

Only thing left to check as far as I can see is to check the actual voltage on the motor connectors when running to see if you're getting 460V and 60 htz when operating at the max flow.

Looking at the information supplied though it seems to me that in reality you've just got a pump that has the wrong pump curve shape ( far too flat at duty point) and also far too susceptible to small changes in pressure.

You say ~25 psi differential. I can see on the diagram a pressure gauge? rather than transmitter on the pump discharge, but not sure where you're taking inlet pressure from?

If we take this 25 as correct, 25 psi is 58 ft (1psi = 2.31ft of water) head. Even with your max impellor size that's only 90 pgm, not the 120 you state.

If you allow say 3psi for extra losses between your inlet pressure gauge, wherever it is, and the pump inlet, that gives you 65 ft differential and a flow of about 60 GPM.

To have such a small differential pressure (+10%) to make such a difference in flow (-35%) will result in the position you find yourself in, IMHO.

So what to do?
1) buy a larger pump with a steeper pump curve at your duty point so you're not so badly affected by small changes in differential pressure. At the duty point I would try and make sure that +/- 5 psi )~10 ft) only makes +/- 10 GPM difference. your current pump 10ft makes 40 GPM difference.
2) buy a VFD or invertor and run the motor at 70 htz or 80 htz. However if the motor is rated at only 2HP (you don't say what the nameplate of the motor says?) then you might as well just get the next size up pump.
3) buy a PD pump of some sort where you don't have to worry about the pressure, if its the flow rate you really want to achieve. At that size of pump there are many pulseless PD options ( Screw, Gear, sliding vane...)

Remember - More details = better answers
Also: If you get a response it's polite to respond to it.

 

[quark]

How did you check the flow?

 

[LittleInch]

To be fair to the OP, he mentions a number of times where the flow figure comes from and the accuracy he is obtaining.

Remember - More details = better answers
Also: If you get a response it's polite to respond to it.

 

[MJCronin]

Are you sure that the impeller IN THE PUMP ... is the same as that shown on your pump curves ?

Could someone somehow have changed out the expected impeller for a smaller one ???

MJCronin
Sr. Process Engineer

 

[quark]

Yes, my mistake.

 

[Rputvin]

The 25 psi I originally mentioned was a round number. We're seeing between 20.8 and 22.9psi - 48 to 53 ft. As I said, it was cutting the ragged edge for the original configuration, but moving to the bigger impeller, which has since been verified by the manufacturer, should have put us at the very least above 80gpm.

A gauge was added to the pump suction during install to troubleshoot the flow numbers.

You may be correct, and error, drift, tolerance, etc is just all stacking the wrong way on me, but it doesn't explain the lack of performance after making the impeller swap that should have added nearly 40gpm at the same pressure and we saw around a 5gpm increase.

Positive displacement is a bad plan in this situation. Mostly due to a few circumstances with the reactor that I can't really get into.

One of the planned options was to over-spin a 3hp with a VFD, but there's a large concern that we're going to do more work and get nowhere given the changes we've already made and the results we've gotten.

We're moving forward with a new pump. It requires a lot of piping changes, electrical changes, etc. I posted here in hopes there was something we were overlooking to avoid the expense of replacing everything.


MJCronin - the original impeller was removed by my lead service tech and swapped for the new one, measured and verified at the 200mm max size for the case per the manufacturer. The test pump built for verification was made with the same parts lot. The original setup spec'd a 1.5hp motor, but our set of pumps was built with 2hp due to availability. The larger trim requires 2hp, so there was a bit of convenience.

 

[LittleInch]

Ok, looks like I'm running out of ideas. The only thing I can suggest further which you've probably done is to check every last connection and valve to make sure there isn't some flow going somewhere it shouldn't. if the pump should be putting out more flow, but you're not spotting it on the flow meter D/s the heat exchanger where can it possibly go?

Through the NRV - but you've tried isolating the second valve
Through the other filters etc
Through some other small line not shown on the diagram

Otherwise pump and motor
check rotation
check impellor clearances / wear rings / sealing
check power supply and frequency / voltages
Check seals / stuffing boxes for excessive frictional losses from the motor ( heat gun would be good when its running to spot heat where it shouldn't be) 2HP isn't much and it wouldn't take much to reduce shaft power.

Let us know if you find anything....



Remember - More details = better answers
Also: If you get a response it's polite to respond to it.

 

[Artisi]

Can you run the existing pump/s in the system on an isolated section of the pipework to verify if the pumps are performing to curve, if they are then the problem is the system and not the usual call "the pumps aren't performing", alternately have an off-site test undertaken.

It is a capital mistake to theorise before one has data. Insensibly one begins to twist facts to suit theories, instead of theories to suit facts. (Sherlock Holmes - A Scandal in Bohemia.)

 

[Rputvin]

There's nowhere for the flow to go. Even if we had a leak or a stuck valve 40gpm of flow is hard to miss, unless we have several smaller leaks. I find that hard to believe given the scrutiny this has been given.

Rotation has been checked. The pump case is stamped stainless and doesn't use wear rings. One pump impeller is slightly tight and rubbing, but the other isn't and both perform the same. When it was opened no blockages, debris, etc was noted that could be impeding flow through the vanes. The motors and pumps were gone through when they were pulled to service the impellers.

The one thing that I don't believe was checked was actual motor speed and incoming frequency. But if we were under-spinning we wouldn't be on the curve at lower flows and wouldn't make the rated shutoff pressure.

The other trouble with this is the customer's facility is a few hundred miles away and our next visit is to perform the changeover to the new pumps. I doubt they'll want to spend another half day trying to track things down, they're not running production and it's costing them money to be down.

Artisi - due to the nature of the system we can't do a bypass or a bucket test or anything. The branching circuits split from a 2.5" pipe to over a dozen lines of varying size, none of which can individually handle the flow. Doing a physical check of flow would have been best, but we can't pump to an open bucket since the circuit is pressurized and we'd quickly be running dry.

A third pump was made at the manufacturer's facility and was verified that the performance curve is accurate. No, it wasn't done with the actual pumps, but if this company can't make 3 "identical" pumps from the same parts lot we've got different issues. We could have swapped in that pump to the circuit, but our customer is getting increasingly impatient and doesn't want to spend days on what-if investigation and wants it fixed so they can move on.

 

[LittleInch]

The very last thing I can see is the integrity of the Plate Heat Exchanger.
And all those city water inlets and outlets need to be isolated during your testing I think or disconnected to make sure they're not passing.

What actual pressures are you getting into the pump, not the differential?

The thing I was taking about voltage was that at higher current levels you'll get higher voltage drops but agree the frequency one is a bit left field.

If there was some sort of cross over going on or leak into the void and then a drain out? difficult to see if happening with 20-40 GPM I'll grant you and would need the city water supply to be constantly feeding water in. Stranger things have happened....

BTW what does the temperature Sensor control?

But remember your current 80 odd GPM is still only 26 psi differential compared to your 21 to 23. If your pressure gauges on the pumps were inboard of the valves then there's very little chance of them being affected by other things.

Nope - time for a bigger / steeper pump curve I think.

Please keep us informed as to progress - these sorts of posts are the most interesting and we can all learn from anything you eventually find out or don't find out....

Remember - More details = better answers
Also: If you get a response it's polite to respond to it.

 

[quark]

Isolate the bladder tank, stop circulation and check for air at the vent. There can be flow discrepancy when there is air in the system.

 

[Rputvin]

The plate heat exchanger is a standard setup. It's pressure tested as part of our factory runoff.

As far as I know the city water is not influencing the system. I was told they were isolated before recording the values.

Suction pressure is 19 psi. I don't have the numbers in front of me, but we're charged to 20.1 psi I believe and when a pump starts running we drop to 19.3psi if memory serves.

All of the branches are independent and all have flow meters installed. Shouldn't be any cross flow to anything else. And if there was we'd see a discrepancy between upstream and collective downstream flow meters.

The temp sensor is operating controls on the opposite side of the heat exchanger to modulate cold water flow. Over-cooling is a concern for the reactor, so it needs to be monitored and modulated to ensure we're not causing condensation or allowing thermal gradient to cause undue stress in a few key areas.


quark - this was done when our tech was originally on site. The customer claimed it was fully purged of air, but their venting was inadequate. The system was re-vented and is at a "satisfactory" level. We'll never get it 100% bled, but we're not getting surging, no variation to the numbers when everything is steady, no noise from the pipes that can indicate air, no movement in the flexible hoses on start/stop, etc. The increase in the bladder pressure was an attempt to see if there was still any large pockets trapped, but once the pressure was increased the vent didn't let any additional air out of the system.

Is it possible that air in the system can cause behavior like this? We know there's some air in there, just form the way it's designed and every branch doesn't have venting, with multiple elevation changes as different systems route along the reactor. But the balancing is steady enough that any air in the system isn't moving. I'd understand pressure increases due to the restrictions air causes, but would air in some other part of the system cause the pump to drop flow and no longer match the curve as you move along to the right on it?

 

[lilliput1]

Part of the pump flow is recirculated through the side stream filter (and shot feeder if valves are open). If pump puts out 80 gpm and 10 gpm goes throught the side stream filter then only 70 gpm will go to the heat exchanger. The 10 gpm goes from pump discharge to side stream filter then back to pump suction.

 

[Rputvin]

lilliput - the side stream items were closed off. They take about 3gpm combined when opened. I'm missing 40 gpm from the rated performance of the pump, 50% of my current performance. That's not making it through a 3/4" line when the system is setup in 2.5"