Tools / Fire Pump Curve & Annual Test / Field guide
The Annual Fire Pump Flow Test, Explained Step by Step — and Why Each Step Is Done
A fire pump sits idle for months and is asked to perform perfectly the one night it matters. The annual flow test is the only time anyone checks that it still can. This guide assumes no background: it explains what the three test points prove, shows how to run the test step by step with the reason behind each step, and teaches you to judge whether a result can be trusted. Every section can be listened to.
Start here: the whole test in one minute
Think of a fire pump as a booster for the building’s water supply. When the sprinklers open, the street’s pressure is not enough, so the pump adds pressure. Most of the year it just sits there. The weekly or monthly run proves it starts. The annual flow test proves it still performs.
The test is simple in idea:
- Run the pump at no flow (called churn), then at its rated flow, then at 150 % of rated flow.
- At each point, record the pressure the pump adds and the flow, along with speed and electrical readings.
- Compare each point with the pump’s own curve. Since the 2017 edition NFPA 25 expects at least 95 % of the reference at every point.
Why anyone cares
If the pump has quietly lost output, the sprinkler system it feeds will not get the water it was designed for in a fire. The annual comparison turns slow wear into a number you can see years in advance.Why the test exists
A pump idles without being challenged. A weekly or monthly run starts it with no water going through, and that catches starting and engine problems. But a pump can start perfectly and still have worn parts inside, a restricted suction, or a driver that cannot pull its full load. Only moving real water at real flow shows that.
That is why NFPA 25 requires the annual flow test and compares the result with the original acceptance curve (or the nameplate). A trend matters more than a single year: a pump that loses 2–3 % a year is telling you when it will need rebuilding.
Who uses the result
- Building owners and facility managers, who are responsible for keeping the pump ready.
- Inspectors and fire marshals, who review the report.
- Insurers, who care whether the system can deliver what it was designed for.
The curve and the three points that matter
Every listed fire pump comes with a performance curve: pressure on one axis, flow on the other. With no water going through, the pump makes its highest pressure. As more water flows, the pressure it adds falls. The annual test checks three places on that curve:
- Churn (no flow): the highest pressure the pump makes. It should not exceed 140 % of rated pressure (NFPA 20).
- Rated (100 %): at rated flow the pump should deliver its rated pressure.
- Overload (150 %): at 1.5 × rated flow the pump must still deliver at least 65 % of rated pressure. Most pumps sit well above this; 65 % is a floor, not a target.
NFPA 25 §8.3.3 has the owner test the pump at churn, 100 % and 150 % every year (variable-speed pumps are tested at more points). Since the 2017 edition, each point is judged at no less than 95 % of the flow and pressure on the original unadjusted field test curve or the nameplate.
Plan the test: how the water gets out, and what to bring
To measure 150 % flow, a lot of water has to go somewhere. NFPA 25 describes three ways to measure it:
- Hose streams: hoses from the test header to nozzles, with a pitot gauge in each stream. Each nozzle’s pitot reading becomes a flow, and the flows are added.
- Flow meter to a drain or tank: water routed through a calibrated meter.
- Closed loop: water recirculated from the pump discharge back to its suction through a flow meter. NFPA 25 limits how many years in a row this may be used, so an open flow test is still needed periodically.
Gauges decide everything
NFPA 25 expects the pressure gauges and transducers to be calibrated annually to ±1 % and flow meters to ±3 %, with a current calibration label. A 3 psi gauge error is invisible on the dial and obvious on the curve.What to bring
- The acceptance test curve, the nameplate data and last year’s report — the baseline.
- Calibrated suction and discharge gauges, a pitot gauge or flow meter, a tachometer, and a meter for voltage and current.
- Hoses and nozzles rated for the flow, a safe discharge plan, and a camera.
Steps 1–4: prepare, check, and run churn — what to do, why, and what goes wrong
Plan the test, tell the right people, and bring the baseline
Do this: Schedule the test with the building owner and tell the monitoring company, the fire department if required, and the insurer or AHJ if they want to witness. Bring the pump's nameplate data, the original acceptance test curve, and last year's report.
Why: Without a baseline the numbers have nothing to be compared with. NFPA 25 judges each point against the original unadjusted field test curve or the nameplate. A running pump and water discharging at the test header can also trigger supervisory or flow signals, so the people who monitor the building should not be surprised.
If it goes wrong: No acceptance curve on file? The nameplate can be used, but a "good year" without a baseline hides slow wear. Start building one now.
Set up the flow path and check the instruments
Do this: Choose how flow will be measured: hose streams through nozzles with pitot gauges, a flow meter discharging to a drain or tank, or a closed loop back to the pump suction. Secure hoses and nozzles, point the discharge somewhere safe, and attach a tachometer and an electrical meter. Confirm every gauge and the flow meter carries a current calibration label.
Why: NFPA 25 expects test pressure gauges and transducers calibrated annually to ±1 % and flow meters to ±3 %. The test can only detect a decline bigger than the instruments' own error, and flow measurement is the number most often wrong. Hoses at 150 % flow carry enormous force.
If it goes wrong: An uncalibrated gauge can make a healthy pump look worn or a worn pump look healthy. Closed-loop testing is limited by NFPA 25 in how many years it can be used in a row.
Pre-start checks
Do this: Confirm the water supply is what the pump expects: tank level, suction valves open, public main pressure. Check the pump room for leaks and, for a diesel, ventilation and fuel. Confirm the controller is in automatic. Record the suction gauge reading and the nameplate.
Why: The pump can only deliver what its supply can give it. A low tank or a partly closed suction valve looks exactly like a worn pump on the curve. Drawing from a public main, the pressure at the pump suction flange must not drop below zero while the pump runs.
If it goes wrong: If a suction valve is not fully open, fix that first. Do not interpret a test taken on a compromised supply.
No-flow (churn): start it automatically and record
Do this: With the discharge closed, start the pump the way an emergency would — let the controller start it by lowering the sensed pressure (not by hand). Once stable, record suction pressure, discharge pressure, speed, and (electric) voltage and amperage on every line, all at the same moment.
Why: Churn is the highest pressure on the curve and the first test point. Starting automatically proves the controller will start the pump on the night it matters. Voltage, amps and rpm tell you what the driver is doing. NFPA 25 summaries give a minimum total run of 10 minutes for an electric pump and 30 minutes for a diesel during the annual test.
If it goes wrong: Do not hold churn longer than needed: water in the casing heats up with no flow through it.
Steps 5–8: the flow points
Rated flow (100 %): open valves, wait, record
Do this: With the pump still running, open test header valves one at a time until the pitot readings add up to the rated flow. Wait for everything to settle, then record the same set of readings: suction, discharge, speed, volts, amps, and the flow.
Why: This is the point the pump is rated at: rated flow at rated pressure. Opening the valves gradually and waiting avoids reading a transient. Everything must be recorded together because the pump's output changes with every adjustment.
If it goes wrong: A flow reading that creeps or jumps points to air in the pitot or meter, or an unstable supply. Wait, bleed, and re-read.
Overload (150 %): the point that finds weak pumps
Do this: Open more valves to reach 1.5 × rated flow, stabilise, and record the same readings. If your supply or header will not let you reach 150 %, test to the maximum allowable discharge and record how far you got and why. Never raise the engine speed above rated to make the point.
Why: A healthy pump still produces at least 65 % of rated pressure at 150 % flow. This is where wear, suction restrictions and driver limits show up, because friction losses and driver loads are highest. NFPA 25 specifically does not allow raising engine speed beyond the pump's rated speed to achieve rated performance.
If it goes wrong: Low pressure here can be real wear — or a restricted suction, a plugged gauge line, or trapped air in the flow meter. Check the measurement before blaming the pump.
Extra points, repeats and variable-speed pumps
Do this: If a point looks wrong, stop and find out why before moving on, or repeat it. For a variable-speed pump, NFPA 25 also tests at 25, 50, 75 and 125 % of rated flow in addition to churn, 100 % and 150 %.
Why: A test with a bad point is not a test; it is a question. Repeating a reading is cheap compared to writing up the wrong problem. Variable-speed pumps behave differently across the range, so more points are needed.
If it goes wrong: If you detect an electrical or mechanical problem while testing, stop until a qualified person has identified and corrected it.
Close down gently and put everything back
Do this: Close the test valves gradually, return to churn, then shut the pump down following the manufacturer's procedure. Return the controller to automatic, reopen any valve you closed, and check the pump room for leaks, heat and noise.
Why: Closing gradually avoids pressure surges in the system. Leaving the controller or a valve in the wrong position is exactly how a pump fails on the night it matters.
If it goes wrong: Confirm the controller shows normal and no alarms before you leave.
Safety
Test water is high-pressure and high-volume. Keep people clear of the discharge, check hoses and nozzles are rated and secured, and protect the ground from erosion. Follow your site lock-out and notification procedure, and stop the test if you detect an electrical or mechanical problem.Steps 9–11: transfer switch, compare, and report
Transfer switch test (if there is one)
Do this: With the pump running at peak load, simulate a power failure so the automatic transfer switch moves to the alternate source. Record voltage, amperage, speed, pressures and flow for at least two minutes on alternate power, then confirm it transfers back to normal after its time delay.
Why: A fire pump is only reliable if its power is. The transfer switch is part of the annual test because it is the part nobody sees move until it must.
If it goes wrong: Coordinate with the building before simulating a power failure and follow the manufacturer's and controller instructions.
Calculate and compare
Do this: For every point: subtract suction from discharge to get net pressure; correct flow and pressure for speed if the driver was off rated rpm; then compare with the reference at that point. The result should be at least 95 % of the reference flow and pressure.
Why: This is the whole reason for the test. Net pressure is what the pump itself added. The affinity laws put readings on a rated-speed basis. NFPA 25 (2017 and later) expects at least 95 % of the original unadjusted field test curve or nameplate at each point. Below that, the owner is notified and the cause investigated and corrected.
If it goes wrong: Compare against the correct reference. Do not move the baseline to make this year pass.
Record, photograph and file
Do this: Write the report: date, pump and driver details, every raw reading and corrected value, instrument serial numbers and calibration dates, who witnessed, and what you found. Photograph the gauges at each point and the test header. File it with the acceptance curve and last year's report.
Why: Next year's test needs this year's curve. A report without calibration details can be challenged. Photographs show which gauge produced which number.
If it goes wrong: Notify the owner immediately if any point is below the 95 % expectation or a limit was broken.
The maths, explained in plain language
Net pressure: what the pump actually added
The discharge gauge reads the pressure leaving the pump. But some of that pressure was already there when the water arrived. Subtract the suction gauge to get the pump’s own contribution. Both gauges are read at the same moment.
Worked example. A pump rated 750 gpm at 100 psi, with churn 118 psi and 78 psi at 150 %. At rated flow the discharge reads 102 and the suction 3, so the net is 102 − 3 = 99 psi. At 150 % (1,125 gpm) discharge 70 and suction 2 give a net of 68 psi.
Speed correction: putting readings on a rated-speed basis
Pumps follow the affinity laws: flow changes in proportion to speed, and pressure changes with the square of speed. A diesel running at 1,700 rpm when its rated speed is 1,760 produces less flow and pressure than at rated speed. To compare fairly, scale the readings up: flow × (1,760 ÷ 1,700) and pressure × (1,760 ÷ 1,700)². For example 700 gpm at 90 psi becomes 725 gpm at 96.5 psi. NFPA 25 does not permit raising engine speed above rated to make a point.
The 95 % check
Compare each corrected point with the reference at the same flow, and divide. In the example: churn 117 ÷ 118 = 99 %; rated 99 ÷ 100 = 99 %; overload 68 ÷ 78 = 87 %. The first two pass; the overload point is below the 95 % expected, so the test says: notify, investigate, and correct.
Why measurement quality matters so much
The same healthy pump can look perfect or look failed, depending only on how well the test was done. This table starts with a pump exactly on its curve at the 150 % point and applies common measurement problems one at a time.
Some mistakes hide a decline (forgetting suction pressure flatters a weak pump) and others invent one (an air bubble in the flow meter, an uncorrected diesel speed). Neither tells you anything about the pump. That is why calibration labels, simultaneous readings and speed correction are part of the procedure, not paperwork.
Reading the result: what to look for
Check the test before you blame the pump. These patterns and first suspects are common — not diagnoses.
Reassuring
Every point at 95 % or more
The corrected readings overlay the reference. File the report so next year has a baseline, and watch for a slow downward trend.
Look closer
All points low by about the same percentage
Often points at a measurement or speed error, or general wear (wear rings, impeller). Check gauge calibration and the speed correction first.
Look closer
Fine at churn, falls away toward 150 %
Suspect something limiting the supply: a strainer, a partly closed suction valve, air entering, or a low water level. Also re-check the flow measurement at the high point — a trapped air bubble in a flow meter has produced a false failure.
Red flag
Churn above 140 % of rated
Check the gauge and speed first, then the driver. An over-speed engine or a mismatched impeller can both raise churn pressure.
Red flag
Below 65 % of rated at 150 % flow
A hard NFPA 20 limit. Verify the readings, then bring in the manufacturer or a fire pump service company.
Look closer
Diesel speed sagging as flow rises
Correct the data to rated speed, then look at the engine side: fuel, cooling, governor setting, battery and charging.
Look closer
Electric amps high or voltage low
Pressures may look fine while the motor is working too hard. Compare with the nameplate and with last year.
Look closer
Pressure fine but a gauge line is suspect
A plugged gauge line or pet cock, a clogged relief-valve pilot screen, or a relief valve set too low all show up as "low pump" readings. Check them before pulling the pump.
Reassuring
Slow, steady decline year over year
A pump falling 2–3 % a year is telling you when it will need rebuilding. That is exactly what the annual comparison is for.
Mistakes that quietly ruin a test — and why they matter
- Ignoring suction pressure and plotting discharge as if it were net pressure — it flatters a weak pump.
- Comparing with the wrong baseline. NFPA 25 uses the original unadjusted field test curve or the nameplate; moving the baseline to make this year pass hides the decline.
- Skipping speed correction on a diesel or variable-speed pump, which invents a failure.
- Raising engine speed above rated to land on a point — not permitted.
- Reading before things stabilise, or reading suction and discharge at different moments.
- Uncalibrated gauges or flow meter, or a meter with trapped air.
- Starting the pump by hand, so the automatic start is never proved.
- Silently skipping the 150 % point when the header cannot reach it, instead of testing to the maximum allowable discharge and recording why.
- Not keeping the report. Next year’s test needs this year’s curve.
What it looks like in the field
Watch it done
Watch · 4 min
Annual Flow Test Requirements for Electric Fire Pumps
NFPA's own channel walks through the annual flow test requirements for an electric-driven pump. Courtesy of NFPA LiNK · open on YouTube
Watch · 10 min
Understanding Fire Pump Performance Curves: A Simple Guide - Part 1
How to read a fire pump performance curve per NFPA 20 — the same curve this tool checks your readings against. Pause it next to the diagram above. Courtesy of Irfan CFPS · open on YouTube
Watch · 3 min
How to conduct a fire pump churn test
A Certified Fire Protection Specialist explains the churn test, how often it is done and best practices. The churn run is the weekly or monthly check, and the first point of the annual test. Courtesy of Flameguard Fire Protection Philippines · open on YouTube
Check yourself
Check yourself
10 questions
Glossary
- Churn (shutoff)
- The pump running with the discharge closed and no flow. The highest pressure on the curve, and the weekly or monthly no-flow run.
- Rated capacity
- The flow, in gpm, the pump is listed for, at its rated pressure. 100 % on the curve.
- 150 % point
- The overload test: 1.5 × rated flow, where the pump must still deliver at least 65 % of rated pressure.
- Net pressure
- Discharge pressure minus suction pressure. The pump's own contribution, and what is plotted.
- Acceptance test curve
- The performance recorded when the pump was first accepted (the original unadjusted field test curve). Annual results are compared with it or with the nameplate.
- Nameplate
- The pump's plate showing rated flow, pressure and speed — the other acceptable reference.
- Test header
- Hose valves or a manifold that discharges test water outside the building, used with pitot readings or a flow meter to measure flow.
- Pitot / flow meter
- The two ways to measure test flow: a pitot gauge on each discharge nozzle, or a calibrated meter in the test loop.
- Affinity laws
- Flow varies with speed; pressure varies with speed squared. Used to correct readings taken off rated speed.
- Closed-loop test
- Water is recirculated from the pump discharge back to its suction through a flow meter instead of being discharged outside.
- Transfer switch (ATS)
- Switches the pump between normal and alternate power; tested annually by simulating a power failure at peak load.
- Controller
- The panel that starts and stops the pump, senses system pressure, and records events.
Frequently Asked Questions
What does the annual fire pump flow test actually prove?
It proves the pump can still do what it was bought to do: hold a sensible pressure with no flow (churn), deliver its rated flow at its rated pressure, and still deliver 150 % of rated flow at no less than 65 % of rated pressure. Since the 2017 edition NFPA 25 judges each point at no less than 95 % of the flow and pressure on the original unadjusted field test curve or the nameplate, so slow wear shows up as a trend instead of a surprise.
What is net pressure and why do I subtract suction?
The pump adds pressure to the water that arrives at its inlet. The gauge on the discharge reads suction plus what the pump added, so the pump's own contribution — the number that goes on its curve — is discharge minus suction. Both gauges must be read at the same moment.
When do I need a speed correction?
Whenever the driver was not running at its rated speed during the test, which is common with diesel engines and variable-speed drives. Flow scales with speed and pressure with its square (the affinity laws), so readings are adjusted to rated speed before they are compared with the curve. NFPA 25 does not allow raising engine speed above rated to make a point.
What is the 95 % rule?
Since the 2017 edition, NFPA 25 evaluates the churn, rated and overload points at no less than 95 % of the flow rates and pressures on either the original unadjusted field test curve or the pump nameplate. If a point comes in under that, the owner is notified and the cause is investigated and corrected. It is a requirement on the results, so measurement quality matters.
The test header cannot flow 150 % of rated capacity. What now?
NFPA 25 says that if the available supply does not allow 150 % of rated capacity to be flowed, the pump is tested to the maximum allowable discharge under the available conditions. Do not skip the point silently: record how far you got and why, and discuss it with the authority having jurisdiction before test day.
How is the flow measured?
NFPA 25 describes three arrangements: hose streams through nozzles with pitot gauges, a flow meter discharging to a drain or tank, and a closed loop that recirculates water to the pump suction through a flow meter. Closed-loop testing is limited in how many years it may be used in a row. Flow meters are calibrated annually to within 3 %, and test pressure gauges to within 1 %.
How long does the pump run during the annual test?
NFPA 25 summaries give a minimum run of 10 minutes for an electric-motor pump and 30 minutes for a diesel-driven pump during the test. The separate weekly (diesel) and monthly (electric) no-flow runs in section 8.3.1 have their own minimum run times.
Who can run the test?
Someone qualified — typically a fire protection contractor or pump service company working with the building owner, often with the AHJ or insurer present. The report belongs to the qualified person who ran the test and should carry the instrument calibration details.
Sources and further reading
- NFPA 20, Standard for the Installation of Stationary Pumps for Fire Protection — performance limits (NFPA)
- NFPA 25, Inspection, Testing, and Maintenance of Water-Based Fire Protection Systems — Chapter 8, fire pumps (NFPA)
- QRFS — The Fire Pump Flow Test: NFPA 25 requirements (test points, §8.3.3.7 readings, §8.3.3.6 flow methods, §8.3.7.2 acceptance, §8.3.3.9 transfer switch, §8.3.3.2.2.2 calibration)
- Rotaflow — How to interpret a fire water pump annual flow test result (the 95 % criterion, speed correction)
- Fire Engineering — Testing Building Fire Pumps, Part 2 (procedure, readings at each point, causes of low performance)
- MeyerFire — Why low pressure at the 150 % test for a fire pump? (trapped air in a flow meter, suction friction)
- LifeSafetyWiki — NFPA 25 Decoded, section 8.3.3 annual flow test
Accuracy and limits
This guide explains the test as described in NFPA 20, NFPA 25 and the practitioner sources above. Section numbers and some thresholds differ between editions (for example the 95 % evaluation was introduced in the 2017 edition of NFPA 25), so check the edition adopted where you work. The authority having jurisdiction, the pump manufacturer and the qualified person who runs the test decide the requirements for your project, and the test report remains that person’s responsibility.Check your own pump in the tool → · Also: reading a hydrant flow test →