Why a Bigger Pump Does Not Fix a Restriction

Why this matters

"It is not moving enough water, put a bigger pump on it" is the most expensive sentence in hydronics, and it is usually said by someone who is right about the symptom. The reason it fails is not that a bigger pump does nothing. It is that resistance shares are set by the geometry of the loop, not by the size of the pump, so a fouled component takes the same fraction of the pump's work after the upsize as it took before. You do not shrink the restriction by pushing harder. You enlarge it, and you pay for the enlargement forever.

Before you go near a pump that is starved

• Do not close a discharge valve on a running pump to "see what happens." A rotodynamic pump run at or near zero flow has nothing carrying its heat away and the casing temperature climbs fast; on a wet-rotor circulator the pumped fluid is also the bearing coolant, so dead-heading destroys it. On a positive-displacement pump, closing the discharge raises pressure until the relief valve or the weakest joint gives, and that failure sprays hot fluid.
• A fouled strainer housing holds full system pressure and full system temperature. Isolate both sides, open the drain into a bucket with the hose already seated, confirm zero on a gauge, and only then break the cover. Shut the heat source and let the loop drop below 120 F first, or wear a face shield and heat-rated gloves and stand clear of the opening.
• Before pulling a pump or a coupling guard, lock the disconnect open and verify the motor is dead with a meter proved on a known live source before and after (NFPA 70E-2021, 120.5). The mechanical isolation and the stored pressure fall under 29 CFR 1910.147; if the work takes you inside an energized panel, the electrical isolation duty is 29 CFR 1910.333(b)(2).

One gate, run against two systems

Everything turns on a single question, and it has a measurable answer:

Does one component's measured pressure drop account for a share of the total that is far out of line with its design share, or do all the segments report the same flow ratio?

The method for answering it is the segment walk, which has its own card in this library. Here is what the two answers do to your recommendation.

Case A: the deficit is concentrated

A closed heating loop. Design is 45 GPM at 34 ft of pump head. The resistance coefficient of the loop as designed is:

k = head / flow squared = 34 / (45 x 45) = 0.0168

Today it moves 36 GPM at 41 ft of pump differential. That loop's current coefficient is:

k = 41 / (36 x 36) = 0.0316

The loop is now about 1.9 times as resistive as designed, and a segment walk puts nearly all of the excess across one plugged strainer.

Ask the upsize question properly: what head would it take to push the design 45 GPM through the loop in its current condition?

head = 0.0316 x (45 x 45) = 64 ft

So to buy back 25 percent more flow (36 to 45 GPM) you need 56 percent more head (41 to 64 ft). Hydraulic work goes as flow times head, so compare the fouled loop against itself: 45 x 64 = 2880 units against today's 36 x 41 = 1476 units. You are paying about 1.95 times the hydraulic work for 25 percent more flow, and you are paying it every hour the pump runs.

What the extra head actually does at the restriction

Here is the part that decides the argument. Split the 64 ft into the loop the designer built and the fouling laid on top of it:

• Loop as designed, at 45 GPM: 0.0168 x 2025 = 34 ft
• Excess from the fouling, at 45 GPM: (0.0316 - 0.0168) x 2025 = 30 ft

Thirty of the sixty-four feet, 47 percent of the pump's work, is spent inside one strainer. Now do the same split at today's 36 GPM: the fouling accounts for (0.0316 - 0.0168) x 1296 = 19 ft out of 41 ft, the same 47 percent share to within the rounding on these figures.

The share does not move. Both terms scale with the square of flow, so the fouling's cut of the pump's work is a property of the geometry and stays at 47 percent whatever pump you bolt on. The upsize does not reduce what the restriction takes. It nearly doubles it in absolute terms, from 19 ft worth to 30 ft worth, and every foot of that lands as heat and velocity inside the component that was already failing.

Which brings the second-order problems, all of them real:

• Velocity through the restricted passage rises with flow, and erosion rate rises faster than velocity does. Common design practice caps copper at roughly 8 ft per second in cold water lines and lower, around 4 to 5 ft per second, in hot recirculating lines, because erosion-corrosion accelerates with both velocity and temperature. Confirm the limit against the tubing manufacturer's data for the alloy and service you have.
• Net positive suction head margin closes from both sides. Suction-side friction rises with the square of flow, so the head available at the impeller eye falls, while the head the pump requires rises with flow. Run far enough right on the curve and you cavitate a pump that was fine yesterday.
• The motor can overload. On most radial-flow pumps power rises continuously with flow, so a selection that moves the operating point right also moves it toward the motor's limit. Some pumps are built with non-overloading curves where power peaks and falls; do not assume yours is one without reading the curve sheet.

Case B: the deficit is distributed

Same complaint, different building. A wing was added last year, the loop was extended, and the segment walk comes back clean: every segment's measured drop is consistent with the same flow ratio, no component out of line with its design share. Nothing is fouled. The loop is simply longer than the pump was chosen for.

Now more head is the correct answer, and the sizing method is the same arithmetic run forwards. Measured today: 38 GPM at 39 ft.

k = 39 / (38 x 38) = 0.0270

The added wing needs the loop to carry 46 GPM. Head required:

head = 0.0270 x (46 x 46) = 57 ft

Select a pump whose curve crosses that point, and check where 46 GPM sits relative to its best efficiency point rather than only checking that the curve reaches 57 ft.

Two checks before you order, because "distributed" does not automatically mean "buy a pump":

1. Check the new pipe size against velocity, not just against the drop. A run that is one size small is a restriction that happens to be long, and it will not show up as one component out of line because it is distributed by construction. Compute velocity in the new segments and compare against the erosion limit above.
2. Check whether the added load actually needs that flow. Flow requirement follows from the heat you must move and the temperature difference you can accept. Widening the design temperature difference across the loop reduces required flow proportionally, and required head falls with the square of that reduction. On a loop where the terminals can tolerate it, this is the cheapest fix on the list because it costs a control setting rather than a machine.

The machine families this reasoning covers

Everything above is about rotodynamic machines, which is centrifugal and axial pumps and fans: head falls as flow rises, and the operating point is where the falling machine curve meets the rising system curve.

Positive-displacement machines do not work this way. A gear, vane, lobe, piston or progressive-cavity pump moves a nearly fixed volume per revolution regardless of pressure, so a restriction downstream does not reduce its flow much. It raises pressure until the relief valve opens or a component fails. Fitting a larger positive-displacement pump to a restricted line does not buy flow, it raises the pressure the line sees, and the relief valve setting is the only thing bounding it. Diagnose those by pressure and by relief-valve behaviour, not by the curve reasoning in this article.

Confirming which case you are in before you order anything

• Run the segment walk and require agreement, not just a suspect. Compute the implied flow ratio from each clean segment as the square root of measured drop over design drop. If four of them land within a few percent of each other and one is wildly outside, that one is your restriction and you are in Case A. If all of them agree including the suspect, you are in Case B.
• Re-measure after clearing the concentrated resistance, before quoting a pump. In Case A, pull and clean the element, then take the pump differential again. If it drops back toward the design value and flow recovers, the case is closed and the pump was never the problem.
• In Case B, verify the new operating point on paper before the pump ships. Plot the required point against the candidate curve and against its best efficiency point. A pump that reaches the head but sits far off its best efficiency point runs loud, wears its seals and bearings early, and gives up the savings you bought it for.
• Log the pump differential and the strainer drop as a standing pair on every maintenance visit. Case A always starts as a slow trend, and a shop with two numbers per visit catches it at 10 percent before it becomes a no-heat call at 47 percent.

References

• Hydraulic Institute standards for centrifugal pump application, including net positive suction head margin and operation away from best efficiency point
• Tubing manufacturer data for maximum design velocity by alloy, temperature and service
• OSHA 29 CFR 1910.147, control of hazardous energy for mechanical isolation and stored pressure
• NFPA 70E-2021, 120.5, verification of an electrically safe work condition before working on the motor circuit
• See related: How to Find a Restriction From Upstream and Downstream Readings; Reading Pump Curves Reference; Cavitation and What It Tells You