Why Minimum Flow Exists and What Happens Below It

Why this matters

A centrifugal pump does not idle. Shaft power goes in whether or not liquid goes out, and when the flow falls the power does not fall with it in anything like proportion - it stays in the casing as heat and as violent internal flow. That is why every pump has a minimum flow, and why the number is set by the pump rather than by the process. It is also why the two most common ways of running a pump badly, throttling it down to hold a level and leaving a discharge valve nearly shut, are the two that feel most harmless. There are exactly two limits that create a minimum flow, they come from different places, they are computed differently, and the one that binds is not the same pump to pump. This article runs the same comparison against two pumps that answer it oppositely.

Do not create the condition in order to see it

The obvious field test, throttling a running pump toward shutoff to hear what happens, is the one instruction this article will not give. On a hot service the liquid in a dead-headed casing reaches its boiling point in minutes, and a casing that has flashed holds vapor at pressure behind a seal and a joint that were sized for liquid. So: do not throttle a pump below its published minimum flow to observe the symptom, and never close a discharge valve fully on a running pump. If the pump has already been run there and has to be opened, isolate it, lock and tag the valves and the driver disconnect (29 CFR 1910.147), prove the driver dead against a known live source (29 CFR 1910.333(b)(2); NFPA 70E-2021, 120.5), let the casing cool below 140 F checked with a non-contact infrared thermometer, and vent to a routed drain with nobody standing at the outlet before a joint is broken. A hot casing vented into an open room flashes to steam at the fitting.

Where the power goes when the flow does not

Shaft power splits two ways: the part that leaves with the liquid as pressure and velocity, and the part that does not. The second part is heat, and its size is set by efficiency.

At best efficiency the losses are a modest share of a large number. At a quarter of best efficiency flow, efficiency itself has collapsed, so the losses are a large share of a somewhat smaller number, and all of it goes into a shrinking stream of liquid. At true shutoff, efficiency is zero by definition - no flow leaves, so no useful work is done - and every bit of shaft power becomes heat in the liquid sitting in the casing.

Two independent consequences follow, and they are the two limits.

Limit one: the thermal limit

The temperature rise across a pump, per pass, is head times the loss fraction divided by efficiency, converted into heat units. For water, with head in feet:

Rise in F = head x (1 - efficiency) / (778 x efficiency)

The 778 is the mechanical equivalent of heat in foot-pounds per Btu, and the formula as written assumes a specific heat of 1.0 Btu per pound per degree F, which is water. A fluid with a lower specific heat heats faster for the same power, so oil or glycol changes this number and the fluid's own specific heat replaces the 1.0.

The efficiency in that expression is the efficiency at the flow you are actually at, not at best efficiency point. That distinction is worth more than the formula.

The rise matters in two different ways depending on whether the liquid leaves. In a system with fresh liquid arriving every pass, one pass of rise is usually trivial. In a dead-headed or heavily recirculating pump, the same liquid takes the rise again and again, and the question becomes how long until it reaches saturation at the pressure it is sitting at.

Limit two: the mechanical limit

Below a certain flow the liquid stops filling the impeller passages cleanly. Flow separates and reverses at the impeller eye and again at the vane tips, and those reversing vortices carry local pressures low enough to form and collapse vapor. It produces damage that looks like cavitation and is not: recirculation damage lands on the pressure side of the vane inlet, while classic suction cavitation damages the low-pressure side just behind the leading edge. That is the teardown discriminator.

The field discriminator is better still. Both sound the same, and only one responds to suction pressure. If a pump crackles like gravel and raising suction pressure or lowering suction lift does nothing at all, you are hearing recirculation from running too far left, not a suction problem, and the fix is flow, not suction head.

There is a second mechanical cost at low flow that has nothing to do with vapor. In a single volute casing, the pressure around the impeller is uniform only near best efficiency; away from it the pressure distribution goes lopsided and imposes a steady radial load on the shaft, smallest at best efficiency and largest at shutoff. That load bends the shaft once per revolution, loads the bearings, and works the seal faces. A double volute or a diffuser casing largely cancels it, which is why the low-flow radial load argument is much weaker on those and why you cannot carry a single-volute rule of thumb onto a diffuser pump.

There is no field formula for the mechanical limit. It comes from the pump manufacturer, and they usually publish two numbers with different meanings: a continuous minimum flow you may run at indefinitely, and a lower intermittent minimum for short excursions. Do not use the second where the first applies.

The gate

Compute the thermal limit, read the mechanical limit off the manufacturer's data, and the pump's minimum flow is the higher of the two. They are different kinds of number from different sources, and they bind in different applications.

Outcome one: the thermal limit binds

A hot water transfer pump. Liquid at 200 F, gauge suction 10 psig, motor nameplate 15 hp, and the curve gives head near shutoff of 250 ft, shutoff shaft power of 9 hp, and efficiency at 25 percent of best efficiency flow of 22 percent. The casing holds about 3 gallons, which is about 25 lb of water.

Rise per pass at the reduced flow. 250 x (1 - 0.22) / (778 x 0.22) = 195 / 171.2 = 1.14 F per pass.

The efficiency correction, printed. Using best efficiency point efficiency of 78 percent instead would give 250 x 0.22 / (778 x 0.78) = 55 / 607 = 0.09 F per pass. That is smaller by more than a factor of 12, and it is the number a tech gets by grabbing the efficiency off the top of the curve. Use the efficiency at the flow you are at.

The saturation temperature, at the pump's own suction pressure. 10 psig is 24.7 psia, and water boils at about 240 F there. Margin from 200 F is 40 F, not the 12 F you would get by using atmospheric pressure and 212 F. Draw the suction tank down and that margin shrinks with the suction pressure, which is the same correction the seal flush article applies to a seal chamber.

Time to saturation if the discharge is closed. Heat needed = 25 lb x 1.0 Btu per lb per F x 40 F = 1,000 Btu. Shutoff power 9 hp x 2,545 Btu per hour per hp = 22,905 Btu per hour. Time = 1,000 / 22,905 = 0.0437 hours = 2.6 minutes.

The shutoff power correction, printed. Using the 15 hp nameplate instead of the curve's 9 hp shutoff figure gives 38,175 Btu per hour and 1.6 minutes. That correction runs conservative rather than flattering here, and it still matters: the nameplate is the motor's rating, not the pump's absorbed power at that point, and the two are not interchangeable in either direction.

The bound, written as a bound. This balance puts every Btu into the liquid and none into the casing, the base or the air, so the real time is longer than the computed one. The result is > 2.6 minutes, an inequality with one sign, not 2.6 plus or minus anything.

What that means operationally. A closed discharge valve on this pump is not a condition anybody walks over and notices. Its minimum flow is set by the thermal limit, and it needs a recirculation path sized so the rise stays well inside the 40 F margin under all suction conditions, not just today's.

Outcome two: the mechanical limit binds

A cold water booster. Liquid at 60 F in an open system with fresh water every pass, gauge suction low enough that saturation is near 212 F, head at the reduced flow 180 ft, efficiency at 25 percent of best efficiency flow 25 percent.

Rise per pass. 180 x (1 - 0.25) / (778 x 0.25) = 135 / 194.5 = 0.69 F per pass, against a margin to saturation of about 150 F, in a system where the liquid leaves and does not come back. The thermal limit is nowhere near binding and no thermal recirculation line is needed.

And yet. The manufacturer's continuous minimum flow for this pump is 40 percent of best efficiency flow. The pump is being run at 25 percent to hold a downstream level. It is below the mechanical limit by a wide margin and it is being damaged continuously by internal recirculation, in a way no temperature reading will ever reveal.

The two outcomes side by side. Same arithmetic, same two limits, same 25 percent of best efficiency flow. On the hot pump the thermal limit binds and the mechanical one is comfortably clear. On the cold one the thermal figure is negligible and the mechanical limit is being violated by 15 points of best efficiency flow. Checking only the limit you know how to calculate gets one of these two right, and there is no way to tell in advance which one you are standing in front of.

What getting it wrong looks like on the cold pump. Nothing, for a long time. Then a crackling noise that the shop diagnoses as cavitation, followed by work on the suction side that changes nothing, followed eventually by an impeller with eroded vane inlets on the pressure side and a shaft that has been taking a radial bending load for years.

How minimum flow is actually provided

• A fixed orifice recirculation line back to the source. Simple, nothing to fail, and it runs all the time, so you pay for that flow at every operating point including full load.
• An automatic recirculation valve that opens the bypass only as forward flow falls. No standing loss, one more device in the pressure boundary to maintain.
• A control scheme that simply will not let the pump run below the flow, by staging pumps or by limiting the speed floor on a drive. Nothing added to the pipe, but it depends entirely on a setting somebody can change.

Whichever is fitted, the bypass has to return somewhere with enough volume and enough heat sink to absorb it. A bypass returning into a small suction line just recirculates hot liquid past the pump inlet and defeats itself.

How to verify you got this right

Measure the flow. Not the valve position, not the amps, not the discharge pressure - the flow, with a meter, at the worst condition the pump sees, and compare it against both limits with the numbers written down. Then check that the recirculation path is actually flowing: a bypass line at room temperature on a hot pump is a bypass that is not passing anything, and a bypass valve that somebody closed during a repair looks exactly like one that is working.

If the pump is on a drive, check the speed floor as well as the flow, because a drive with a low minimum speed setting will happily walk the pump below its minimum flow while every other reading in the room looks calm.

References

• Pump manufacturer's published continuous and intermittent minimum flow figures for the specific pump, which own the mechanical limit
• ANSI/HI 9.6.3, the Hydraulic Institute guideline on rotodynamic pump operating regions, in the edition your engineering specification adopts
• 29 CFR 1910.147 for isolating the pump and its stored pressure and heat, with 29 CFR 1910.333(b)(2) and NFPA 70E-2021, 120.5 for proving the driver dead
• See related: What Cavitation Is and How It Announces Itself; What a Seal Flush Plan Is Actually For; What the Affinity Laws Let You Predict and Where They Stop