Why Soft Foot Defeats an Alignment That Measured Good
Why this matters
An alignment report is a true statement about two shaft ends and nothing else. Four numbers, describing where one centerline sits relative to another. A machine can be twisted like a wrung towel and still put those four numbers exactly where you want them, because the coupling sits between the two machines and the twist is distributed across four feet several inches away from it. So the failure this article is about is not an alignment that went out of tolerance. It is the machine that passed, was signed off, and kept killing a bearing anyway, while everyone involved kept re-reading the one document that could not contain the answer. The useful content here is the list of things that document does not contain.
What the coupling reading measures, and its reach
Offset and angle in two planes. That is the whole instrument. It is measured at the coupling, which is the point on the machine furthest from the feet and closest to the middle of the set.
That geometry matters. Pulling one corner of a frame down does not move the shaft end by the same amount; it twists the whole structure, and the shaft end takes a small share of that twist, often a share so small it hides inside the acceptance band. The coupling is a poor detector of frame distortion by design, and that is not a defect in the method, it is the method measuring what it was built to measure.
The negative space: what an alignment report does not contain
This is the section worth keeping. A signed alignment report, as normally written, is silent on all eight of these, and each one of them can be the reason a machine that passed still fails:
- Whether each foot was flat on the base before any bolt was torqued. The report starts after the machine is bolted down.
- Which foot carried the correction, and therefore which foot is now doing structural work it was never meant to do.
- The shim stack: how many pieces, what condition, whether they are burred, corroded or bridging a hollow.
- Whether the base itself is flat, and where it is not.
- The pipe strain state, meaning whether the nozzles are pulling the machine and in which direction.
- Hold-down bolt torque, and whether it was even across the four feet.
- Whether the frame is in tension between feet, which is the actual definition of soft foot and the thing the whole exercise is about.
- The condition the readings were taken at: hot or cold, pipe bolted or free, how long since shutdown.
A report carrying those eight alongside the four numbers is a diagnostic record. A report carrying only the four numbers is a receipt, and a receipt cannot be argued with or learned from.
What the distortion does where the coupling cannot see it
Bearing housing bores go out of round. A bearing sits in a bore machined round in a frame that is now not the shape it was machined in. The clearance around the bearing stops being uniform, load concentrates, and the bearing runs hot at one position on the housing rather than evenly.
A motor's air gap goes eccentric. The gap between rotor and stator is small and it is the most distortion-sensitive dimension on the machine. Pull the frame out of shape and the rotor sits closer to the stator on one side than the other, which produces a steady sideways magnetic pull toward the narrow side. That pull is a real bearing load, it exists whenever the motor is energised, and it does not appear in any mechanical measurement of the stopped machine.
Internal running clearances inside the driven machine move too. The soft foot article that owns this library's mechanism explanation covers that consequence in detail, and it is not re-derived here.
The machine goes out of alignment when it heats, even where the thermal allowance was calculated correctly. A frame carrying built-in stress does not grow the way an unstressed one does, so a good cold alignment with a good thermal allowance can still land wrong hot.
The one signal the alignment job itself gives you
Soft foot has a fingerprint on the alignment work: the job will not converge. You correct, re-read, and the numbers have moved somewhere you did not send them. Three passes like that is a diagnosis, not bad luck. Stop correcting the coupling and go check the feet.
The converse is the trap this article exists for. A converging, first-pass, comfortably-in-tolerance alignment does not rule soft foot out. It only rules out the case where the distortion happened to move the shaft ends a lot.
Worked example: one soft foot, a passing report, and a motor bearing on a five-month cycle
The record. Alignment signed off at 1.2 mils offset against an acceptance figure of 3.0 mils from the coupling manufacturer's table for this machine at its speed. Converged in two passes. No complaint about the alignment work at all.
The complaint. The motor's drive-end bearing had been replaced three times in about sixteen months, roughly every five months, always the same end.
The soft foot check, with its measurement correction printed. With the machine isolated, locked and the rotor blocked, an indicator was set on each foot and each hold-down bolt loosened one at a time. Three feet moved under 1 mil. The fourth lifted 5.2 mils on the first pass.
That first figure was taken with the indicator's magnetic base sitting on the same baseplate as the machine, which flexes as the bolt releases, so the reading was the difference between two movements rather than the movement of the foot. Repeated with the stand on an independent structure clear of the machine and the baseplate, the same foot lifted 4.5 mils. The 0.7 mil difference was the baseplate moving under the stand. The corrected 4.5 is the number, and using the 5.2 would have over-shimmed the foot by 0.7 mils in the direction that creates a new soft foot on the diagonal.
Why the coupling never saw it. With the correction shimmed in and the machine re-bolted, the coupling reading moved from 1.2 to 0.6 mils of offset. So a 4.5 mil frame distortion showed up at the coupling as a 0.6 mil change: 0.6 / 4.5 = 13 percent. Eighty-seven percent of the distortion went into twisting the frame, and the report only ever had access to the 13.
The evidence that mattered, taken at the motor. Air gap was measured with feeler gauges through the motor's inspection ports at four positions around the stator. The first set of readings disagreed with itself until they were retaken at a single marked axial depth, because a gauge pushed in further reads a different part of a slightly tapered gap and that reads exactly like eccentricity. At the marked depth: 30, 30, 34 and 26 thousandths of an inch.
Mean gap = (30 + 30 + 34 + 26) / 4 = 30 thousandths. Largest deviation from the mean = 4 thousandths. That is 4 / 30 = 13 percent eccentricity. The commonly used field acceptance is that gap variation should stay within about 10 percent of the mean, and the motor manufacturer's own figure governs. This motor was outside it, on the axis you would predict from which foot was short, with the narrow side toward the low corner.
What that means for the bearing. An eccentric gap pulls the rotor steadily toward the narrow side whenever the motor is energised. That is a constant, unidirectional, non-rotating load added to whatever the bearing already carries, and a bearing under a steady preload runs hot and fails at a fraction of its rated life without ever producing the fatigue signature people look for.
The repair and the confirmation. The base under the short foot was corrected and the machine re-bolted. Re-measured air gap came to 30, 30, 31 and 29 thousandths: mean 30, largest deviation 1, which is 1 / 30 = 3.3 percent, comfortably inside the 10 percent figure. Alignment was re-read and re-signed, and this time the report carried the eight items in the list above alongside the four numbers. The bearing has run since without replacement.
What getting this wrong looks like. The version that actually happens is three bearing replacements, each one closed out with "bearing failed, replaced, alignment checked and within tolerance." Every word of that is true and it never once approaches the cause, because the check performed was the one measurement in the building that could not detect the fault. The tell that should have redirected it after the first replacement is that the same end failed every time. A cause that picks the same bearing three times is a steady directional load, and steady directional loads come from geometry, not from bad parts.
What would change the reading
A small motor with no inspection ports cannot have its air gap measured, so the surrogate is a vibration reading dominated by twice line frequency that disappears the instant power is removed while the machine is still coasting at almost the same speed. Vibration that vanishes with the power is electromagnetic in origin; vibration that coasts down with the speed is mechanical.
A heavy, stiff fabricated base distributes a short foot differently from a light frame, so the same foot error moves the coupling by a different fraction. The 13 percent above is this machine's number, not a constant, and the honest use of it is as an argument that the fraction is small rather than as a figure to apply elsewhere.
A machine on a slide base or rails, common on belt drives, has a fifth surface in the load path that can be the thing that is not flat, and it wears.
A vertical machine puts its own weight into the mounting flange rather than across four feet, so the distortion mode is different and the flange flatness specification from the manufacturer is what applies.
How to verify you got this right
Check for soft foot before you take a single coupling reading, with the indicator stand on an independent structure, one bolt at a time. Then, after the alignment is signed, take one piece of evidence from somewhere other than the coupling: air gap on a motor that permits it, a bearing housing temperature at each end at the same load and ambient, or a vibration reading with the power-off test above. A single measurement taken at the coupling can only ever tell you about the coupling.
And write the eight items into the record. The reason the machine above took sixteen months is not that anybody lacked skill; it is that each visit inherited a document that had already answered the only question it was capable of answering.
References
- Motor manufacturer's air gap eccentricity limit, which owns the acceptance figure for the specific machine
- Coupling or machine manufacturer's alignment tolerance table for the acceptance figure at your machine's speed
- 29 CFR 1910.147 for mechanical isolation and blocking the rotor before feet or coupling are touched, with 29 CFR 1910.333(b)(2) and NFPA 70E-2021, 120.5 for proving the motor dead
- See related: Why Soft Foot Defeats an Alignment; How to Check Shaft Alignment and Work Out Why It Drifted; What Thermal Growth Does to an Aligned Machine