parallel pump problems 1

[hitchhiker]

Dear fellow members,

I'd like to ask your expert opinion on my current situation. I'm currently handling left-over projects that have 5 booster pumps in paralell delivering liquids from storage tank (produced water) to a pig launcher. 3 of the booster pumps are old existing pumps and another 2 are fairly recently bought with matching H/Q characteristic with the existing (not really identical). These pumps fail frequently and they took turns (imagine that!). so operation guys came to my department (project) to look for solutions and hopefully create something to solve their predicament.

the first thing when I look up when I step in is the calculation by FEED contractor. The old pump itself basically reused from another project, rated at 600 gpm @ 735 ft, water temp 180 F. the newer 2 pumps also have the same rating, 600 gpm @ 735 ft to suit the old one. So before they actually bought the 2 new pumps, FEED contractor have studied the system and made recommendation for 5 pumps in parallel. So the combined pump curve only intersect the system curve at 4 & 5 pumps operating. At single, double and triple pump operation the system is well below the pump curve (way to the right). So I wonder does this possible?

by right, operating 1,2 or 3 pumps will have cavitation problem, since it will push operation way to the right curve (it doesnt even intersect the system curve!)and while the pumps do have problems and regulary in-out service for maintenance, there were no report of cavitation damage on impellers. there are erosion damage on the casing, however (it turns out there were sand sediment in the fluid that was not considered). there are even report on casing leak and damage. I still wonder if this is because of the sand or other things.

So back to the my question, can parallel pumps designed this way? that the system curve only intersect the combined curve of 4 & 5 pumps?

 

[hitchhiker]

clay87,

there is only one common recirculation line from the discharge header as you can see in my previous post. each individual pump doesn't have dedicated miniflow line. let say the old pump (stronger) dominates the new (weaker), and with recirculation line the flow from the new pump goes back to the suction tank. How does this line ensure that it will make the new (weak pump) will pump above its minimum flow? since the suction tank itself have continuous incoming stream as well?

ccfowler,

unfortunately i don't have temperature reading data yet. I will have to push the o/m guys to provide it. I agree with you, that looking at this condition it'd be beneficial to have temperature reading. I only have pressure data and I will post it on monday.

 

[LittleInch]

So long as the reciculation line goes back into the same tank that the inlet line comes from, then there is no NET effect, i.e. flow out of the tank into the pump equals the flow through the recycle line. It's only when the two are from different locations that you run into problems.

If you can, measure temp, pressure and running amps from each pump and motor when it's working which will tell you a lot about how equal the pumps really are. Whilst it is tempting to use the isolation valves as control valves to equal up the flow, this is a short term measure only and cannot be recommended for any long term operation. When you've worked out which ones are stronger than the other, fitting RO's may be the thing to do.

As ccfowler points out, if the actual piping layout follows that of the diagram then you have inbuilt problems as overall resistance for each pump should be the same for inlet and outlet piping length, no of elbows etc from the common inlet point to the common outlet point. Sometime didn't think about this when they designed it, but unfortunately this isn't uncommon.

You have provided much useful info, but I await what the set pressure is of your outlet back pressure control valve before being able to understand how the system works.

Keep it up.

My motto: Learn something new every day

Also: There's usually a good reason why everyone does it that way

 

[BigInch]

Recirc off the header will allow control of group flow, but not allow control of individual units.

Independent events are seldomly independent.

 

[Artisi]

An option might be to fit an orifice plate to the discharge of each unit sized to force the pump to run at a point on its curve close to the required flow rate (seems this might be 600gpm) which hopefully is somewhere near to BEP.

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.)

 

[ccfowler]

Artisi, that is an excellent recommendation. At least it would tend to force some semblance of balance in the mix of problems present without introducing any great costs. Brilliantly elegant and inexpensive!

Valuable advice from a professor many years ago: First, design for graceful failure. Everything we build will eventually fail, so we must strive to avoid injuries or secondary damage when that failure occurs. Only then can practicality and economics be properly considered.

 

[hitchhiker]

Guys, thanks for all the very insightful replies. I'm currently working with operation to obtain more data. Will let you know the reading once I get it.

 

[hitchhiker]

Guys, while waiting for the data, I got another question. Is there any guide or recommended practice on number of pumps in parallel? Like for instance in my current case I have 5 pumps. Personally to me, 5 is too much because the flow addition normally quite marginal, but yet I've seen 8-10 pumps or more designed in parallel in my current company. any comment? have you ever design such system as well?

 

[Artisi]

The number of pumps in parallel is unlimited if the system is correctly designed.

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.)

 

[LittleInch]

Multiple parallel pumps work best when the system curve changes a lot depending on how many units are flowing, e.g. a/c units or multiple outlets. For a static system curve they don't work very well as you note as additional pumps make very little additional flow, just a reduced flow through each pump.

My motto: Learn something new every day

Also: There's usually a good reason why everyone does it that way

 

[hitchhiker]

@artisi,
true, but is there any "rule of thumb"? we can have many, but too many?

@littleinch,
even at dynamic dominated system would make the addition more insignificant, with steeper system curve, am i right?

Those I see here with more than 5 pumps in parallel usually on water injection system (produced water) to wells in multiple location. It is largely determined by static since each well would have certain minimum pressure. the pipeline to the wells would give friction head but considerably small compared to the min well pressure. The feed to the pumps depends on how much water produced (normally by surge/water tank level, as you've seen in above system). Given the incoming feed fluctuate througout the day, it is logical to use parallel pumps. But as we've discussed,in static dominated system,up to certain number of pumps it doesn't give benefit because of marginal addition.

I'm fairly new in this company, looking at this practice triggers my curiosity.

 

[Compositepro]

Little Inch, I don't understand your assertion. Take, for example, flood water pumps around New Orleans. Every pump is in parallel and the system curve for each pump is identical and static. There is no limit to the number of pumps in parallel.

 

[LittleInch]

Parallel pumping means that pumps have a common discharge header and one pipe - See the diagram earlier on in this post. What I believe you refer to is multiple pumps with their own discharge pipe. That is not parallel pumping, but pumps mounted together. Not the same thing.



My motto: Learn something new every day

Also: There's usually a good reason why everyone does it that way

 

[Compositepro]

In my example the pumps are pumping from the same place and to the same place but with effectively VERY large inlet and outlet pipes. The point is that one can design a system to work or not to work. I still do not understand what you were trying to say about system curves. Perhaps you could elaborate.

 

[Artisi]

Compositepro; fully agree with your comment/s, exactly what I was hoping to infer with my earlier post re "correctly designed".

hitchhiker: No rule of thumb just common sense, if increasing flow only increases total head by a small amount then you can keep adding pumps until the total of the system exceeds the useful operating head / conditions of the individual pumps - however, if increasing flow results in a large increase of total head then it stands to reason that the number of pumps could very quickly become limited by the total head being imposed on each pump unit.
Again, the system, the pumps,and the control system must be engineered for each application.

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.)

 

[LittleInch]

Composite pro,

A system curve exists for a set run of pipe plus any fixed items which create head loss such as filters or heat exchangers. However the shape of the system curve can change if you add or subtract those fixed units also in parallel. You see this in ac systems where additional pumps in parallel work well when combined with additional chillers.

For systems like you describe, the system curve is fairly flat, but has a high static element compared to the pump delivery head. For instance if the pump duty head is say 30m, with a static lift of say 20m, but with a very large pipe so the system curve takes a long time to get to 30m, equivalent to several pumps, then adding extra pumps increases flow more or less in proportion to the number of pumps. On the other hand if you look at the curves the OP posted in his second post, then you can see the additional flow for the fourth pump is very small compared to it's nominal capacity and a fifth would add virtually no extra flow because the system head at that flow rate has equaled the pumps delivery head.

I think we all agree that if you design it right it will work, but once you reach an optimum number of centrifugal pumps, adding more has little effect unless you change the system curve at the same time by e.g. opening another valve or flow path creating a lower head loss for the same flow rate.

My motto: Learn something new every day

Also: There's usually a good reason why everyone does it that way

 

[clay87]

As mentioned, there is no rule of thumb for 'how many' parallel pumps that I know of. Parallel pumps (with a common discharge header) are usually specified to minimize pump size and/or provide some sort of redundancy. For reasons the OP has observed, using additional pumps to handle changing 'demand' has its limitations.

 

[BigInch]

3 or 4 pumps will pretty much make any vfd system redundant. All reasonable flowrates can be reached with pumps running at very good overall efficiencies, using only control valves, and with nearly full discharge head available if necessary.

Independent events are seldomly independent.

 

[hitchhiker]

another thing, would you recommend individual control valve at each pump discharge or 1 common control valve at the discharge header like the design above? Which one will work better?

 

[BigInch]

Eitehr CV configuration produces pretty much the same effect when looking at what the system will output overall, however 1 CV requires all pumps to be outputting the same discharge pressure at all times, whereas multiple CVs would allow for individual control of recirculation (internal or external) for each pump and would be more effective at paralleling pumps with different characteristic curves into a common header. If you have the same curves for each pump, there would be no real difference if you wanted to put 1 CV for all pumps, or 1 CV for each pump. One possible exception would be that a problem with the 1 common CV for all pumps would shut the whole station down. In that case, I'd probably consider using at least 2 CVs between header and downstream pipeline, in parallel, 1 operating and the other blocked off when on standby.

Independent events are seldomly independent.

 

[hitchhiker]

Thanks. What about the recirculation line? Ideally it's better to control from individual discharge goes to common recirculation header (separate recirculation line from each pump) or take common recirculation line from the discharge header as shown on P&ID?