What a Wear Ring Clearance Costs You as It Opens

Why this matters

A centrifugal pump has a deliberate leak inside it. The impeller eye runs inside a close-fitting ring, discharge pressure sits on one side of that ring and suction pressure on the other, and some liquid goes round and round instead of out the nozzle. The ring is sacrificial and it is supposed to wear. What catches shops out is how the wear presents: the pump gets quietly worse at delivering water while the discharge gauge barely moves and the amp clamp barely moves, so the two readings a tech takes first are the two least sensitive to the thing that is actually happening. The complaint arrives as "it is not moving enough water" from a customer who has been living with it getting slowly worse for two years, and by then the casing bore has usually gone too, which changes the repair from a set of rings to a decision about the pump.

Isolate before any of this becomes hands-on

Nothing in the field half of this needs the pump opened. The teardown half does, so: isolate suction and discharge, lock and tag the valves and the driver disconnect under 29 CFR 1910.147, prove the electrical side dead with a live-dead-live check against a known live source (29 CFR 1910.333(b)(2); NFPA 70E-2021, 120.5), vent and drain the casing to a routed drain with nobody standing at the outlet, and confirm zero on a gauge rather than on the absence of noise. Read casing temperature with a non-contact infrared thermometer and let anything above 140 F cool before bare hands touch it. Lifting a casing half or a rotating element is rigging: use rated slings inspected before the lift under 29 CFR 1910.184, keep hands off the load and out of pinch points, and never stand under it. If you are installing a gauge to get a reading, install it on a relieved, isolated line - a gauge fitted into a live pressurized port is how people get sprayed.

The ring is a resistance, and the leak is what the resistance allows

Two things sit in series inside the pump: the impeller, which raises pressure, and the ring gap, which lets some of that pressure push liquid backwards to the eye. Total flow through the impeller is the flow you deliver plus the flow that leaks. Open the gap and the leak grows, so at the same impeller flow you deliver less.

Three consequences follow and they do not arrive together:

• Delivered flow falls first and falls most. The impeller is unchanged, so it still makes head, but you get less of the flow at the nozzle.
• Power falls a little. The impeller is still pumping roughly the same total volume, just recirculating a bigger share of it. Motor current drops, but nothing like proportionally. That is why an amp clamp is a poor instrument for this fault: it moves in the right direction and it moves far too little to be a signal.
• The energy you spend per unit of water delivered rises, which is the honest way to state the loss, because it is the only one of the three that tracks the wear closely.

There is a fourth cost that has nothing to do with efficiency. A running ring with liquid in the gap behaves like a hydrostatic bearing and adds real stiffness to the rotor. That stiffening exists only with the pump at speed and the gap flooded, and it falls off as the gap opens. A rotor that was well supported at its rings when new can, at heavily worn clearance, run closer to a resonance it used to sit well clear of. The field symptom is vibration that has crept up over years with no change in alignment, balance or bearings.

The leak also does something to the suction. It re-enters at the eye as a jet running the wrong way against incoming liquid, and that disturbance degrades the pump's suction performance. The size of the degradation is a pump-by-pump number the manufacturer owns; the direction is not in doubt.

How fast the leak grows

Leak through a plain annular ring grows faster than the clearance does. A common working figure is that leak rises with roughly the 1.5 power of the diametral clearance, and that exponent is derived for a plain cylindrical ring whose loss is set by the entrance and the ring length. A serrated ring, a labyrinth ring or a stepped ring behaves differently, so treat 1.5 as an order-of-magnitude check and let the pump manufacturer's clearance table govern the accept-or-reject call.

The commonly used renewal trigger is around twice the as-new diametral clearance, and that is a trade convention, not a rule with a legal home. Your pump's manual owns the real number, and pumps in hot or abrasive service are often specified with a larger as-new clearance to start with, so copying a clearance from a different pump is a way to build a fault in.

The measurement that actually detects it

Not the discharge gauge. Not the amp clamp. The detection is a flow reading compared against the pump's own published curve at the head the pump is actually making, and both halves of that comparison have to be right.

• Differential head, taken from suction and discharge gauges on the same pump at the same moment, converted to feet of the liquid being pumped.
• Delivered flow, measured. A clamp-on transit-time meter is the usual field option because it needs no line break.
• The published curve for this impeller diameter at this speed, which for a drive-fed pump means reading the speed off the drive rather than assuming nameplate.

Worked case: a transfer pump that was down thirty points and only eight of them were rings

The signal. A cooling water transfer pump the plant described as "gutless for the last couple of years." No noise complaint, no leak, no vibration alarm.

Readings taken on the running pump. Suction and discharge gauges gave a differential of 46 psi. On water that is 46 x 2.31 = 106 ft of head. The drive display read full nameplate speed, so the full-speed curve applies. A clamp-on meter on a straight run of the discharge line read 78 percent of the pump's design flow.

The comparator correction, made visible. The tempting subtraction is 100 - 78 = 22 points down against design. That is the wrong comparator, because the pump is not being asked to run at its design point. At 106 ft of head, the published curve for this impeller passes 108 percent of design flow. So the deficit against what the pump should be doing where it is actually operating is 108 - 78 = 30 points of design flow, not 22.

The error terms, with their basis and character. The clamp-on meter's spec is percent of reading, an independent random spread, so at a 78 reading it is worth 2 percent of 78 = 1.6 points. The published curve carries its own tolerance, also treated as an independent spread, worth about 3 percent of the 108 figure = 3.2 points. These are independent, so they combine in quadrature and not by addition: the square root of (1.6 squared plus 3.2 squared) is 3.6 points, against the 4.8 you would get by adding them. The deficit is 30 plus or minus about 3.6 points. It survives its own uncertainty comfortably, which is the only reason it is worth acting on.

The current reading, and why it was not used as evidence. Motor current read 91 percent of the value logged at commissioning. Current is not power, because power factor moves with load, so a 9 point current drop is not a 9 point power drop and cannot be converted into one with a clamp meter alone. It was recorded as a supporting observation and explicitly not entered into any arithmetic. What it did establish is direction: down a little, while flow was down a lot, which is the signature of an internal leak rather than of a pump that has stopped making head.

The teardown, and the part the field readings could not tell anyone. Impeller ring diametral clearance measured 2.6 times the as-new figure in the manual. At the 1.5 power that is on the order of 4 times the as-new leak. The published curve already contains the as-new leak, so only the GROWTH counts against it. If the as-new leak was around 2 points of design flow, four times it is about 8 points of leak, of which the extra 6 points is what the curve was not expecting. So the rings account for something like 6 points of the 30. The casing bore, which on this pump is the casing itself rather than a renewable ring, was worn oval, and the impeller vane inlets were eroded thin at the leading edges.

So the split was: rings roughly 6 points, everything else the balance. That is the finding worth carrying away. A field reading tells you honestly and quantitatively that the pump is 30 points down against its own curve. It tells you nothing about how that 30 splits between ring leak, casing bore, and an impeller that no longer matches the curve it is being judged against. Anyone who reads a low flow and orders rings is guessing at the split.

The repair decision that followed. Rings alone would have recovered about a fifth of the loss and the customer would have paid to reopen a pump that was still down twenty-odd points. The casing bore needed machining and a fitted casing ring, and the impeller needed replacement. That combination is where the arithmetic usually turns: once you are into the casing and the impeller together, the comparison worth putting in front of an owner is total hours for the repair against total hours for a replacement rotating element, plus the years each buys.

What getting it wrong looks like. The common version is a shop that pulls the pump, finds worn rings, fits new rings to a worn casing bore, and hands back a pump whose clearance is barely better than what came out because half the gap lives in the bore. It runs, it is slightly better, and it is back inside a year with the customer's confidence spent.

What would change the call

An abrasive service. Rings in silt, sand or scale wear on a schedule rather than a lifetime, and the right answer is often a harder ring pair and a planned renewal interval, not a search for a cause.

A pump that has been run far from its best efficiency point for years. Then internal recirculation, not ring wear, may be doing the damage, and putting rings in without moving the duty point re-runs the same wear. That is a different mechanism and it has its own article.

No published curve, or no confidence in it. If the impeller has been trimmed at some point in its history, or the drive is running below nameplate speed, the curve you are comparing against is not this machine's curve and the 30 point deficit is not real. Establish speed and impeller diameter before you trust the comparison.

How to verify you got this right

Take the same three readings, on the same ports, at the same duty point, and keep them together as a set: differential head, delivered flow, and speed. A single set is a snapshot. Two sets a year apart is a wear rate, and a wear rate is what lets you schedule the repair rather than react to it. Log the ratio of motor power to delivered flow if you have a power reading available, because that ratio rises steadily with internal leak and is the one number that moves early enough to plan around. If your only comparison is against a design flow figure on a nameplate rather than against the curve at your measured head, write in your notes that the comparator is uncorrected, because it will overstate or understate the deficit depending on which side of design the pump is sitting.

References

• Pump manufacturer's clearance table and cross-sectional drawing, which own the as-new and maximum clearances for the specific pump
• ANSI/HI 9.6.1, Hydraulic Institute guidance on NPSH margin, in the edition your engineering specification adopts, for the suction consequences of ring leakage
• 29 CFR 1910.147 for isolation of the pump and its stored pressure, 29 CFR 1910.333(b)(2) with NFPA 70E-2021, 120.5 for proving the driver dead, and 29 CFR 1910.184 for sling inspection before lifting a casing or rotating element
• See related: What a Pump Curve and a System Curve Do Together; How an Operating Point Moves; Why a Bigger Pump Does Not Fix a Restriction