How to Use a Performance Curve

Why this matters

A performance curve is the only piece of manufacturer documentation that tells you what a machine can do across its whole range instead of at one rated point. Most techs use it once, at selection, and never again. That is backwards. The curve earns its keep in the field, because it is the one tool that turns two pressure readings into an answer to the question everybody argues about: is the machine bad, or is the system making it look bad? The shop that cannot answer that replaces good pumps and good blowers, twice, and still has the complaint.

Before you open a port or pull a guard

Installing a gauge tap means breaking into a pressurized circuit. Isolate the section with valves you have verified hold, relieve pressure through a drain or vent, and confirm zero at an installed gauge before you turn a fitting. On any hot-water or steam circuit, let the isolated section cool first: liquid above its atmospheric boiling point flashes to vapor the moment the fitting cracks, and it does not wait for you to get your hand out of the way.

Pulling a coupling guard or a belt guard is a separate job with a separate rule. Open the disconnect, apply your own lock and tag, and confirm the shaft is stationary before the guard comes off. That is 29 CFR 1910.147, and the reason it is written for exactly this moment is that a motor a helper energizes while your hand is on a coupling does not need long to take the hand.

The complaint, and the number that started it

A building's top floor lost pressure after a booster pump was replaced. The old pump had died, the replacement was ordered against the original schedule, and the installer confirmed it matched: 120 gallons per minute at 60 feet of head, same speed, same nominal size. Two weeks later the top-floor tenant was still complaining, and the shop was one phone call away from pulling the new pump out as defective.

The number the installer had, and the only number anyone had taken, was that the pump ran and the motor did not trip. That is not a measurement.

What a performance curve actually puts on one sheet

A curve sheet for a centrifugal machine carries four things stacked on the same horizontal axis, which is always flow.

• Head or pressure, falling from left to right. Shutoff, at the far left, is what the machine develops against a closed valve with no flow. Runout, at the far right, is the most flow it can pass at almost no resistance.
• A family of curves, not one curve. Each line in the family is one impeller diameter, or one wheel size, or one speed. The sheet is for the casing; the line is for what is inside it.
• Efficiency contours, drawn across the family. The peak is the best efficiency point, and the machine is happiest within a band around it.
• Power, usually a separate set of lines at the bottom, and on a suction-limited machine a required inlet pressure line as well.

The single most important idea on the sheet is that the machine does not choose its operating point. The curve tells you every point it is capable of. The piping, ductwork, filters, valves and fittings draw a second curve, rising from left to right, that says how much resistance the system imposes at each flow. Where the two cross is where the machine actually runs. Change the system and the machine walks along its own curve to a new point without anything being wrong with it.

Turning two gauge readings into a point on the curve

Field readings come in pressure units. The curve is usually drawn in feet of head. For cold water, 1 psi is about 2.31 feet of head, and that conversion is the whole bridge.

We put one gauge on the suction tap and the same gauge, moved, on the discharge tap, so any gauge error cancels. Discharge read 78 psi. Suction read 32 psi. The differential is 78 minus 32, or 46 psi. Converted, 46 times 2.31 is about 106 feet of head.

That is the reading that ended the argument. The pump was supposed to be doing 60 feet. It was doing 106. A pump developing nearly twice its design head is not weak.

Now read the curve backwards. Find 106 feet on the vertical axis, run across to the ordered impeller line, drop down to the flow axis: about 55 gallons per minute. Against a design of 120, the pump was moving about 46 percent of the flow it was bought for.

The three explanations the point killed

Wrong rotation. A centrifugal pump running backwards still moves water, just badly, and develops well under its rated head. Ours was over its rated head. Eliminated by the same number that started the investigation, without pulling a lead.

Wrong impeller shipped. A smaller trim than ordered produces less head at every flow, so its whole curve sits below the ordered one. Our measured point, 55 gallons per minute at 106 feet, landed on the ordered trim's line, not below it. A pump sitting on its own published curve is a pump doing exactly what it was built to do.

Gauge error. Reading both ports with one gauge already removed relative error between the two taps. We also checked the reading against a second gauge and it agreed within a couple of psi, which is well inside anything that would move the flow read-off meaningfully.

Motor current supported all three eliminations at once. On a centrifugal pump, power falls as flow falls, so a machine pushed to the left on its curve draws less than nameplate, not more. Ours was drawing under nameplate full-load amps. A tech who read that as "the motor is weak" would have had it exactly upside down.

What was going on

Nothing was wrong with the pump. The system curve had steepened, so the crossing point had walked up and left. A loaded suction strainer and a balancing valve that had been throttled during commissioning and never reopened were adding resistance the original schedule never accounted for. The pump answered that resistance the only way a centrifugal machine can: by producing more head and less flow.

The failure mode of getting this wrong is expensive and self-perpetuating. Replace the pump with a bigger one and the new machine also lands on the steep system curve, delivers less than schedule, and burns more energy doing it. The shop then has two pumps, one complaint, and a customer who has paid for a second install that fixed nothing.

Confirming it

We cleaned the strainer and opened the balancing valve to its commissioned position, then re-read the same two taps with the same gauge. Discharge 62 psi, suction 36 psi, differential 26 psi. Converted, 26 times 2.31 is about 60 feet of head, which is the design point, and the curve read at 60 feet gives about 120 gallons per minute.

Motor current rose toward nameplate as flow came up, which is the direction a centrifugal machine's power moves and the second confirmation that we were reading the right physics. Top-floor pressure came back the same afternoon.

The same method on a fan, where density is the trap

A fan curve is read identically, with static pressure in place of head and cubic feet per minute in place of gallons per minute, and one extra step that catches people out.

Published fan curves are drawn for standard air: roughly 0.075 pounds per cubic foot, which is dry air near sea level at ordinary room temperature. A fan is a constant-volume machine. At lower density it still moves the same cubic feet per minute at a given point on its curve, but the static pressure it develops and the power it draws both scale directly with density. At 0.060 pounds per cubic foot, which is 80 percent of standard, the same fan at the same speed delivers its cataloged airflow at 80 percent of the cataloged static pressure and 80 percent of the cataloged power.

Skip the correction and a fan running perfectly at elevation, or on hot air, reads as underperforming against a sea-level curve, and someone increases the sheave to fix a fault that does not exist.

One fork that matters before you generalize this. All of the above is centrifugal behavior. A positive-displacement pump or blower does the opposite: restrict its discharge and it holds flow while pressure and power climb until something gives, which is why those machines carry a relief device as a design requirement rather than an accessory. Never diagnose a positive-displacement machine by throttling it and watching what happens.

Where the curve you are holding is the wrong curve

• It is drawn for one speed. On a variable-speed drive the curve you have describes only full speed. The affinity relationships carry it: flow moves in proportion to speed, head with the square of speed, power with the cube. Drop to 80 percent speed and you have 80 percent of the flow, 64 percent of the head, and about 51 percent of the power. That last figure is why variable speed pays, and it is also why comparing a field reading taken at part speed against a full-speed curve makes any machine look broken.
• It is a shop-test curve, not a field promise. Published curves come from a test stand under standardized conditions, and acceptance standards allow a tolerance band around the published line. A field point sitting a little off the line is normal; a field point sitting on a different line in the family is a different machine.
• The fluid is not what the curve assumed. A head curve for water is very close to correct for any thin fluid of similar density, but power scales with specific gravity, and a viscous fluid degrades head, flow and efficiency together. A glycol mix is heavier and thicker than water, and both effects run the same direction on the motor.
• The trim is not what the tag says. Impellers get trimmed in the field and rarely get documented. If the measured point will not land anywhere on the family, measure the impeller before you argue with the sheet.

References

• 29 CFR 1910.147, control of hazardous energy, for locking and tagging a drive before removing a coupling or belt guard
• Hydraulic Institute standards for pump performance testing and published-curve tolerance
• AMCA standards for fan performance rating and the standard-air density basis of published fan curves
• Manufacturer documentation for the specific curve family, impeller trim, and required inlet pressure
• See related: Reading Pump Curves Reference; How to Read a Spec Sheet Without Drowning