This is normal when you run the pumps parallel, the capacity will not increased in proportion to the number of pump. Depending on the pump and system characteristic curve, if the pump curve is relatively steep compare to your system then the third and forth pump will add capacity to the system. But if pump curve is flat compare to system curve then your third and forth additional pump will add very small capacity to the system.

   Multiple centrifugal pumps running in parallel will produce a total flow that is additive at any particular head regardless of the shape of the individual pump's head-flow curve. If each of the pump branches are unthrottled (isolation valves wide open) and all the pumps have the same head-flow characteristic and are running at the same speed, the n  pumps will have n different possible system resistance curves, so it depends on how many of the pumps are running at a particular time where the running pumps will be situated on their common head-flow curve. If the pumps are designed for operation in a 4-pump system that provides 100% of required sytem flow, then 3, 2 and 1 pump operation will deliver 75%, 50% and 25% of maximum required system flowrate. In such a case, the design flowrate of all pumps should be the flowrate that will deliver total required sytem flowrate when 4 pumps are operating. System resistance curves will then be lower when 3, 2 and 1 pumps are operating depending on branching flowrates through the idle pump branches (presumably controlled by non-return valves like swing check valves to prevent idled pumps from "breaking-away" and running in reverse rotation). For a 4-pump system, fractional pump flowrates for 4, 3, 2 and 1 pump operation might be 100%, 115%, 125% and 140% of rated (best efficiency) flow for the 4 identical pumps. Since axial hydraulic thrust loading, vibration levels, etc are sensitive to flow fraction, then 3, 2 and 1-pump operations are all off-design conditions and subject to degraded performance conditions that need to be accounted for in the detailed pump design.

PumpDude

You will find that vanstoja is right with a couple possible exceptions.

1)    If the intake of the four pumps is one common line or source when the third and fourth pumps start you may be starving them.  You need to review the intake configuration.

2)    The four pumps are additive.  The system curve (friction loss etc.) will change with Q.  If one pump will produce 10 GPM into a common system two pumps may only produce 19 GPM due to the system limits or changes.

I have made a spreadsheet in the past that calculates four pumps connected in parallel.  One pump by itself would produce a little over 900 GPM through a 11” pipe, 11000 feet long.  With all four pumps operating total Q was about 3000 GPM due to the increased system losses.  This sheet was a prediction before the pumps were installed.  The actual operation was very close to my prediction.

Good Luck
D23

You can draw out the combined pump curves for 2,3 and 4 pumps just by adding the flows at each given head point.
Then if you draw on your system curve it will be clear what the problem is - too much system friction. You are trying to push too much liquid down the pipe because the pumps develop insufficient head to do it. You will see that the curves show each pump adds progressively less additional head - the combined curve gets flatter. If the system had negligible friction (e.g. pumping into an elevated reservoir through a really big pipe), the system curve would be flat and the flow would be proportional to the number of pumps running. However in your case, increasing flow also requires increasingly more head to deliver it.
 Either reduce system resistance, look at speeding up the pumps or fitting bigger impellers (and probably motors). Or look at running the pumps in two parallel trains of two pumps each.

Cheers

Steve McKenzie

Steve's reply to you say's it all, there is 1 point that you need to understand so that your understanding of the problem is clear.
The head loss component of the total head on the pumps increases as you bring more pumps online, due to the fact that the pipeline friction loss increases as the square of the flow rate change, ie., double the flow rate increases the friction loss by a factor of 4.    
So lets assume that 1 pump will deliver 100 gpm and the friction head in the pipeline is 2 feet loss, the second pump increases the flow to 200 gpm the friction head is now 8 feet loss, the 3rd pump increases flow to 300 gpm the head loss is now 18 ft the 4th pump increases the flow to 400 gpm and the friction loss becomes 32 ft loss.

As Steve has pointed out, because of the increased friction loss plus your static head(read total head)it is forcing the pumps left along the performance curve reducing your flow rate.
I suggest you draw the combined pump and system curve so you see the real situation, it's always an interesting exercise and tells all.
    

International College
Naresuan University
Phitsanulok
Thailand

I think the above have adequately answered your question. If dynamic losses in the system are high (proportional to Q squared) the additional flow from 3 or 4 pumps is  relatively small.

But that does not mean that running 3 or 4 pumps in parallel --- is pretty much useless…….

I have worked on a number of stations with 6 pumps in parallel. - The advantages of using larger number of pumps are:

1) Allows the use of fixed speed rather than variable flow pumps to match variable demands.
2) Lower starting currents.
3) Lower voltage.
4) reduced stand by capacity (eg if you have only one duty and a stand-by the satnd-by capacity is 100%. If you have 3 duty and one stand-by the stand-by capacity is 33%).

Disadvatages include increased capital costs.


It is  a question of selecting the right number of pumps for the system and operating conditions taking account of available voltage, voltage drop, capital and running costs etc.

Running a multi-pump system for a minimal increase of flow when compared to the capital cost and the power input costs does not make good ecomonic sense. As pointed out by Bris, a careful economic study should be undertaken to evaluate "total" costs over time.

International College
Naresuan University
Phitsanulok
Thailand