How to Find a Restriction From Upstream and Downstream Readings

Why this matters

The usual way to hunt a restriction is to start at the pump and open ports going downstream until something reads wrong. On a six-component loop that is up to five port openings, five chances to introduce air, and an afternoon. There is a better ordering, and it comes from one property of series flow: the same flow passes through every segment, so every clean segment must report the same flow ratio. The segment that reports an impossible one is the restriction. That test lets you open ports in the order that halves the remaining suspects rather than in the order the pipe happens to run, and it turns five openings into two.

Before the first port on this particular walk

The general rules for opening a live port apply here and are covered by the pressure-drop reading card. Two hazards belong specifically to a restriction hunt, because you are opening ports on both sides of something that may be nearly plugged:

• A plugged component holds the whole system differential across itself. The downstream port can read near zero while the upstream side is at full pressure and full temperature. Confirming "the loop is relieved" at one gauge tells you nothing about the other side. Gauge each side independently before you break either fitting.
• Opening a port on the downstream side of a plug and then clearing the plug releases the upstream pressure through the opening you just made. Close and cap every port you have opened before you touch the restriction itself.
• You will introduce air at every port on a closed loop. Have the high-point vents and the make-up arrangement ready before you start, because a loop you air-bind while diagnosing it looks exactly like a worse restriction and you will chase your own work.
• Shut the heat source and let the loop fall below 120 F before opening anything, or pipe every discharge into a bucket or floor drain, stand to the side of the opening, and wear a face shield and heat-rated gloves.

What you need on paper before you open anything

The test needs an expected drop per segment at a known flow. Best to worst source:

1. The design submittal or commissioning report. Component drops at design flow, which is exactly what you want.
2. A baseline set of readings from a visit when the system worked. Just as good for locating, because the test only needs the segments to be internally consistent, not correct in absolute terms.
3. Component curves plus a pipe-friction estimate. Slower, workable.
4. Nothing. Then the test degrades to judgement: which segment's drop is grossly out of proportion to its physical length, fitting count and known component type. A short strainer producing more drop than 60 ft of pipe is still an obvious answer. Say on the report that the finding is qualitative.

Write the segments and their expected shares down before the first reading. Working example, a heating branch designed for 45 GPM against 36 ft of total loop head:

Step 1: the whole-machine reading, which locates nothing

Take the pump differential first, corrected for gauge elevation, with one instrument. It reads 41 ft against a design of 36 ft.

This reading cannot tell you where anything is, and that is the point of taking it first: it is the cheapest reading available and it decides whether the rest of the walk is worth doing.

• Differential up, flow down means the loop got more resistive. Something is restricting. Continue.
• Differential down, flow down means the pump lost capability, not the loop. Stop the walk and go look at the pump: worn wear rings, a trimmed or damaged impeller, reversed rotation, low speed. The operating-point movement card covers that read.
• Differential near design means the deficit is not in this loop at all, and you are on the wrong side of a heat exchanger.

Here it is up, so a restriction exists somewhere in six segments.

Step 2: split the loop by design share, not by geography

Do not go to the next component. Go to the tap nearest the point that splits the expected drop in half, because that is what balances the two halves of the test.

• Group A, S1 to S3: 2 + 6 + 8 = 16 ft expected at design flow
• Group B, S4 to S6: 11 + 6 + 3 = 20 ft expected at design flow

One reading, at the balancing valve outlet, gives Group A directly. Group B comes by subtraction from the 41 ft total, so you get both halves for the price of one port.

Measured: Group A = 28.0 ft. Group B = 41.0 - 28.0 = 13.0 ft.

Now convert each to the flow ratio it implies, using the square-law relationship (drop rises with the square of flow, so flow ratio is the square root of the drop ratio):

• Group A: square root of (28.0 / 16) = square root of 1.75 = 1.32
• Group B: square root of (13.0 / 20) = square root of 0.65 = 0.81

Group B says flow is at 81 percent of design. Group A says flow is at 132 percent of design. Both cannot be true, because the same water goes through both. The half that reports the impossible ratio is the half carrying resistance it was not designed to have. Six suspects became three, on one port opening.

Read the direction carefully: the restriction is in the group whose implied ratio is higher, because extra resistance inflates that group's drop.

Step 3: split the guilty half again

Group A is S1 (2 ft), S2 (6 ft) and S3 (8 ft). The nearest tap to a half-split of 16 ft is after the strainer, which isolates S1 at 2 ft against S2 plus S3 at 14 ft. That is a lopsided split, but the tap exists and the alternative is drilling.

Measured across S1: 19.0 ft. By subtraction, S2 plus S3 = 28.0 - 19.0 = 9.0 ft.

• S1: square root of (19.0 / 2) = 3.08
• S2 plus S3: square root of (9.0 / 14) = 0.80

S2 and S3 agree with Group B's 0.81. S1 claims flow is at three times design. The strainer is the restriction, found on the second port opened rather than the fifth.

Step 4: turn the located segment into a number you can quote

Three things come out of the numbers you already have.

The actual flow. The clean segments agree at a ratio of about 0.80, so flow is 0.80 x 45 = 36 GPM, a 20 percent shortfall against design. Note the base: 0.80 is a fraction of the 45 GPM design flow, not of anything else.

How far gone the component is. Its expected drop at the actual flow, if it were clean, is its design drop scaled by the flow ratio squared: 2 x (0.80 x 0.80) = 1.28 ft. It is producing 19.0 ft. Its resistance coefficient is therefore 19.0 / 1.28 = about 15 times its design value. That is a plugged element, not a dirty one, and it says so on the report without any adjectives.

A cross-check that the whole picture closes. If the strainer were clean, the entire loop at 36 GPM would drop 36 x (0.80 x 0.80) = 23.0 ft. Measured total is 41.0 ft. The excess is 18.0 ft, and the strainer's own excess is 19.0 - 1.28 = 17.7 ft. Those agree within about 0.3 ft, which is inside the resolution of a gauge reading feet of water on a psi dial. If they had disagreed by several feet, a second segment is also fouled and you have more work to do.

When the walk says there is no restriction

Sometimes every segment agrees. All six report a ratio near 0.86, nothing is out of line, and flow is still short. That is a real and useful result, and it means the loop is behaving exactly as its geometry says it should. The shortfall is then either on the pump side or built into the design, and the answer is a selection question rather than a cleaning question. The card on why a bigger pump does not fix a restriction covers that fork, including the check that stops you buying a pump for a pipe-sizing problem.

Reading tolerances, because they decide whether the split is real

The test compares two numbers derived through a square root, and the square root is forgiving in one direction and not the other.

• A 10 percent error in a measured drop becomes about a 5 percent error in the implied flow ratio, because the square root halves proportional error. That is why ratios of 0.80 and 0.81 count as agreement.
• The forgiveness disappears on small drops. A segment with a 2 ft design drop measured on a gauge whose smallest useful division is about 1 ft of water is being read at roughly half-scale precision. Group small segments together rather than trying to read them alone, unless, as here, the answer is so far out that precision does not matter.
• Every reading in the walk must be at the same operating state. If a two-way control valve modulates between your Group A reading and your total, the subtraction is between two different systems and the ratios will disagree for a reason that has nothing to do with a restriction. Hold the system in a fixed state, or use a differential gauge so each pair is simultaneous.
• Elevation correction applies to every pair, not just the pump. A segment that runs up and over carries a static column in its raw reading, and 3 ft of rise is 1.3 psi that has nothing to do with friction.

References

• Manufacturer submittal and commissioning documentation for component pressure drops at design flow
• Hydraulic Institute guidance on system resistance and series-circuit behaviour
• OSHA 29 CFR 1910.147, control of hazardous energy including stored pressure, applied to each side of a restricted component independently
• See related: How to Read a Pressure Drop Across a Component; Why a Bigger Pump Does Not Fix a Restriction; How an Operating Point Moves