What a Flexible Connector Does and Does Not Break

Why this matters

Two flexible connectors go on a pump, the hum upstairs does not move, and the shop is now holding an invoice it has to defend. The connectors were not defective and the install was not sloppy. A connector interrupts one path, the one running along the metal of the pipe wall, and it is blind to the other one, the pressure pulsation travelling in the fluid inside. On a pump complaint the second path is at least as often the culprit, and it walks straight through the connector and out through every rigid hanger on the run. This card follows one basement pump set where that is exactly what happened, and the number that separated the two paths was a single third-octave band, not the overall level.

Before the first flange bolt moves

Breaking into a pumped, pressurized, possibly hot line is the hazard here and it comes before any diagnosis. Shut the pump down and lock and tag it under 29 CFR 1910.147 for the mechanical and stored-energy hazards, close and lock both isolation valves, relieve the trapped section through a drain and confirm zero at a gauge on that section rather than at the header, and let the fluid cool below scald temperature before loosening a bolt, because hot system water escapes on the first turn of a flange fastener and reaches skin before anyone can step back. Where the pump is on a variable frequency drive, de-energize, lock, tag and prove dead under 29 CFR 1910.333(b)(2), wait the discharge time the drive's manual states, and verify the bus is discharged with an instrument before opening the enclosure.

Pressure thrust deserves its own sentence. A pressurized elastomeric connector generates an end thrust that tries to extend it, and its control rods are what restrain that thrust. Never slacken or remove control rods on a connector that has not been isolated and depressurized, and never let the connector itself carry thrust the piping anchors were supposed to take.

The complaint and the first set of readings

A basement pump set under an occupied suite. A steady hum, present whenever either pump ran, unchanged by weather and unchanged by time of day. Two visits had already replaced a coupling and rebalanced a motor.

The readings that started the third visit, all at one marked position in the suite, microphone at 1.2 m above the floor, 5-minute equivalent-continuous, referenced to 20 micropascals:

The margin gate, applied per figure rather than once. The overall A-weighted margin is 2 dB, which is under 3, so there is no valid source level on the overall figure at all - only an upper bound of 35 dB A-weighted, and that is what went on the record. The 160 Hz band margin is 15 dB, which is at or above 10, so the combined band reading is the source's own band level within a few tenths and the background correction was declined and said to be declined. One measurement set, two different verdicts, because the gate is applied to the figure you intend to report and not to the visit.

That split is the reason the diagnosis was possible. Anybody working from the overall A-weighted number had a 2 dB margin and no case.

The band that named the path

The pump nameplate reads 1,760 rpm at rated load, so shaft rotation is 29.3 Hz, which lands in the 31.5 Hz third-octave band. The impeller carries six vanes, so blade pass is six times 29.3, or 176 Hz, which lands in the 160 Hz third-octave band whose upper edge is just above 178 Hz.

The 31.5 Hz band showed a 2 dB margin over background. The 160 Hz band showed 15 dB. The energy was at blade pass, not at shaft rate.

That single fact decides the job, because the two frequencies come from different mechanisms and leave the pump by different routes. Shaft-rate energy is mechanical, it comes off the rotating assembly as a force at the mounts and the nozzles, and it is what isolators and connectors exist to interrupt. Blade-pass energy is hydraulic, generated as each vane passes the casing cutwater, and it exists as a pressure pulsation in the water. Reading the spectrum back to the mechanism is its own subject and a sibling card owns the method; the point here is only that the answer routed the work away from the connectors.

The two paths, and which one a connector sees

   pump casing
       |
   [ nozzle ]
       |                     path A: pipe wall metal
   [ flexible connector ]  <-- interrupted here
       |
   ---- pipe run ------- rigid hanger ---- slab
       |
   ( water inside the pipe )
       |                     path B: pressure pulsation
       |                     passes straight through
   ---- continues along the whole run --------

Path A, structure-borne, along the wall of the pipe. Force from the rotating assembly and from casing motion enters the nozzle and travels as vibration in the metal. A flexible connector puts a soft element in that metal path and, if nothing bridges it, reduces what crosses.

Path B, fluid-borne, inside the pipe. The pressure pulsation travels in the water at the speed of sound in water, past the connector, along the entire run, and it excites the pipe wall everywhere it goes. Every rigid hanger, every sleeve packed solid, every clamp to a stud is a fresh injection point into the structure. A connector is a mechanical decoupler in path A. It is a piece of pipe in path B.

Three ways a connector is bridged into uselessness

Even in path A, the connector was doing less than the schedule implied, and all three reasons were visible once anyone looked.

Control rods bolted metal to metal. The rods were fitted correctly for thrust and installed with steel washers hard against the flange plates. That is a rigid bar across the soft element. Control rods need the resilient bushings and washers the connector manufacturer's instruction sheet specifies, set to the gap that sheet gives, or the connector is a decorative sleeve.

Installed in tension to close a gap. The discharge connector had been stretched to make up a short spool. An elastomer preloaded in tension is stiffer dynamically than the same elastomer at its neutral length, so its published performance no longer describes it. Set the piping to the connector's face-to-face dimension; do not use the connector as a make-up piece.

The first hanger too close and rigid. The pipe was rigidly clevis-hung to the slab under the tenant, roughly a metre past the connector. Whatever crossed the connector, plus everything path B delivered, went into that slab through a solid rod. Piping downstream of a connector needs support of its own, and on the first several hangers that support needs deflection.

What the fix was, and the numbers after it

Resilient hangers with the specified deflection on the first several hangers off both pumps, resilient sleeving where the pipe penetrated the slab, and correctly bushed control rods on both connectors. Nothing was done to the pumps.

Same position, same microphone height, same 5-minute equivalent-continuous basis, referenced to 20 micropascals:

The margin is now 1 dB, which is under 3, so there is no source level to report in that band. What went in the report is an upper bound: the pump set's contribution at that position in the 160 Hz band is no greater than 31 dB unweighted, written with one inequality and no interval. That is a weaker statement than a measurement and a stronger outcome, because the source has fallen into the residual and there is nothing left to argue about. The complaint stopped.

Sibling-rule check. Every level above carries its quantity, weighting, bandwidth, time basis, reference and position. The background margin gate is applied per reported figure, and the three zones are used as the sibling procedure states them, including refusing a source level where the margin was under 3 dB in both the before and after sets. No corrected figure is compared against an uncorrected one; the before and after band figures were both taken against the same background under the same conditions. The blade-pass frequency is derived from vane count times measured shaft speed rather than assumed, and nameplate speed is treated as the speed at rated load.

When the connector really is the whole answer

Where the complaint's energy sits at shaft rate or at a mechanical order rather than at blade pass, path A is the path and a correctly installed connector plus a resiliently supported first span is usually the fix. Where the equipment is a reciprocating machine, path B carries far more energy and a connector is nowhere near sufficient on its own. Where the line is gas rather than liquid, path B is weaker relative to path A because the fluid is a poorer transmitter, and connector work goes further. Where the connector is on a duct rather than a pipe, the same two-path split holds but the fluid-borne half has its own name and its own card, and duct breakout is a third path neither one covers.

One condition inverts the recommendation entirely: a connector installed on a line whose thermal movement or anchor forces the piping design gave it is not an acoustic component at all, and removing or softening it to chase a hum introduces a mechanical failure. If the connector is doing expansion duty, changing it is a piping design question for the engineer of record.

How to verify a connector installation

Check the face-to-face dimension against the manufacturer's neutral length with the system at operating temperature and pressure, since a connector correct when cold can be extended when hot. Check that each control rod has its resilient bushings and that the specified gap exists at the plates, which you can see by eye. Then run one hand along the pipe on the far side of the connector while the machine runs and compare with the same hand on the pump casing: a working path A interruption gives an obvious drop across the connector, and no drop means it is bridged by rods, by tension, or by the first hanger. Finally, walk the run and count rigid contacts with structure, because that count is what path B is worth, and it is the number most connector installations never establish.

References

• Connector manufacturer's installation instructions, which own the neutral face-to-face length, the control-rod bushing arrangement and the pressure-thrust rating
• 29 CFR 1910.147, control of hazardous energy, for isolating and locking the pump before the line is opened
• 29 CFR 1910.333(b)(2), electrical safe work practices, where a drive enclosure is opened
• ASHRAE Handbook chapters on sound and vibration control, for resilient pipe support practice
• See related: Why Duct-Borne Noise and Duct Breakout Are Different Faults; What a Vibration Isolating Connection Can and Cannot Do; How to Add and Subtract Noise Levels Correctly