Why Soft Foot Defeats an Alignment

Why this matters

The machine that "will not hold alignment" almost never has an alignment problem. It has a frame that changes shape when you torque the bolts, so the readings you took describe a machine that stops existing the moment it is bolted down. That is soft foot, and calling it "a gap under a foot" is what makes shops treat it as a shimming nuisance instead of the structural problem it is. The tell is a job that will not converge: you correct, re-read, and the numbers have moved somewhere you did not send them, three times in a row, until somebody decides the machine is just like that.

Isolate first, and treat the fixes as their own hazards

Open and lock the motor disconnect and prove the terminals dead with a live-dead-live check on a known live source (29 CFR 1910.333(b)(2); NFPA 70E-2021, 120.5). Then work under 29 CFR 1910.147: confirm zero rotation, block the rotor, and isolate, lock, relieve and drain any hot or pressurized line into the machine before a foot bolt is loosened, because a strained line unloads into the machine when a restraint releases.

The three common fixes each carry a hazard the lockout does not address. Loosening one hold-down bolt on a machine leaves it on three, and a frame under torsion can spring several mils when the fourth releases, so keep hands and fingers out from between the foot and the base and use a bar rather than a hand to steady anything. Machining or grinding a foot or base flat is abrasive wheel work under 29 CFR 1910.215 with the guard in place and a face shield over safety glasses, and the airborne route needs its own control: grinding steel puts metal particulate in your breathing zone, and grinding or chipping concrete or cementitious grout releases respirable crystalline silica, which is regulated under 29 CFR 1910.1053 in general industry and requires the exposure controls that standard specifies, not a nuisance dust mask. Pouring an epoxy chock brings a chemical route: the hazard communication standard, 29 CFR 1910.1200, requires the safety data sheet be available before you mix, amine hardeners in these systems are skin and respiratory sensitizers, so mix with local exhaust or outdoors, wear the chemical-resistant glove type the sheet names, and use respiratory protection selected under a 29 CFR 1910.134 program where the sheet calls for it.

Soft foot is frame distortion, not a gap

A motor frame or a pump casing is a welded or cast structure with real but finite stiffness. It is not a rigid body. Torque a hold-down bolt on a foot that is not sitting flat on the base and you do not simply close a gap: you pull that corner of the frame down and twist the whole structure to reach it.

The bearing bores are machined into that structure. When the frame twists, the bores move relative to each other and relative to the base, and the shaft centerline moves with them. That is the mechanism, and it is why the coupling reading changes when nothing you did was at the coupling.

Worse than the magnitude is the unpredictability. A twisted frame does not respond linearly to shim changes, because a shim added at one foot alters the load path through all four. That is what produces the non-converging job: each correction moves the frame somewhere the arithmetic did not predict, and the tech loses confidence in a measurement method that was never the problem.

The consequence people miss: internal clearances move too

Everything above is about what soft foot does to the alignment. It also does something to the machine that no coupling reading will ever show, and this is the part worth carrying away.

A twisted motor frame runs an uneven air gap between rotor and stator. An eccentric air gap produces an unbalanced radial magnetic pull toward the narrow side, and that pull is a steady bearing load the machine was not rated for, on a machine that measures fine at the coupling. A twisted pump casing changes wear ring clearance and works the mechanical seal faces out of square. The vibration signature that goes with an eccentric air gap is a subject the library's vibration articles own; the point here is mechanical, and it is that a machine can be inside alignment tolerance and still be running with its own internals distorted, because the distortion and the alignment are two different consequences of the same twisted frame.

That is also why "the alignment is fine, so the feet must be fine" is backwards. Correct the feet because the machine needs its designed shape, then align.

Four kinds, and they need four different fixes

The angular row is the one that gets handled wrong most often, because a flat shim closes the widest part of the gap and the indicator reading improves. What actually happened is that the foot now bears harder on its contact edge, the frame is still being twisted, and the reader has spent the correction without removing the cause. Partial shims - a shim tucked under half a foot - are the same mistake with a worse load concentration.

How to measure it, and what the number means

Indicator on the foot, not on the coupling. Every hold-down bolt at the machine manufacturer's specified torque. Loosen exactly one bolt, read the lift at that foot, retorque that bolt to spec, then move to the next. One at a time, because loosening two lets the frame relax and redistributes the error across both.

The number you record is the lift at that foot, in a single loosen-and-retorque cycle, with all other bolts at specification torque. A widely used action level is about 2 mils of lift, and it tightens where the machine builder publishes a smaller limit or where the machine is high speed or precision, so take the published figure when there is one. Then, with the bolt loose, put a feeler gauge at all four corners of the foot: uniform gap says parallel, a gap that tapers across the foot says angular, and no gap at all next to a foot that moved says springing or induced.

Worked example: four feet that will not settle

A 25 hp motor, four feet, all bolts at specification torque, indicator on each foot in turn.

First pass.

Front right is angular at 6 mils across a 4.0 inch foot width, which is a slope of 1.5 mils per inch across the foot. A flat 6 mil shim under that foot would leave the inboard edge 6 mils proud and the frame twisted exactly as before.

Correct the easy two first and re-measure everything. Back left gets a single 5 mil shim, handled with cut-resistant gloves because a shim edge will open a hand. Back right gets its seven-piece stack pulled out, the metal cleaned, and a stack rebuilt to the same total thickness in two pieces. Now re-measure all four feet, because that is the step shops skip and it is the whole reason the job does not converge.

Second pass: front left 3 mils, front right 4 mils, back left 0, back right 1 mil.

Front left went from 1 mil to 3 and front right went from 6 to 4, and nobody touched either of them. The frame relaxed when the back two stopped fighting it, and the error redistributed. Anyone who had corrected front right's 6 mils on the first pass would now be dealing with an over-correction they did not cause.

Now handle the angular foot at its current value. Front right measures 4 mils across the 4.0 inch width, a slope of 1.0 mil per inch, corrected with a tapered or stepped shim set to that slope rather than a flat shim. Third pass: all four feet under 2 mils. The frame is now sitting the way it was machined.

What the correction did to the coupling. Before any of this, the coupling read plus 2 mils at station 1 and plus 4 mils at station 2. After the feet were corrected and nothing else was changed, it reads minus 3 mils at station 1 and minus 1 mil at station 2. Position moved 5 mils at both stations and the slope stayed at 0.33 mil per inch across the 6.0 inch station spacing, purely from correcting the feet.

That is the article's whole point in one comparison. Anyone who had aligned the set before the feet were fixed did careful, correct arithmetic on a machine whose shape was about to change by 5 mils. The alignment was not wrong; it described a machine that no longer existed.

Failure mode. The visible version is the non-converging job described at the top. The invisible version is worse: the shop gets the coupling inside tolerance on a twisted frame, closes it out, and the motor runs an eccentric air gap and a steady magnetic side pull for the rest of its life, with a coupling reading on file that says everything is fine.

What changes the answer

A machine on a fabricated steel skid rather than a grouted concrete base is more flexible, so the skid deflects along with the frame and a foot correction can move the base as well as the machine. On those, expect more redistribution between passes and more passes. A machine with more than four feet, or with a centre foot, has a redundant support and the loosen-one-at-a-time rule matters more, not less. And on any machine that is bolted to a base that is itself out of flat, the durable fix is the base surface or an epoxy chock, not a shim library.

How to verify

Re-check soft foot at final bolt torque after the alignment is complete, not only before it, because a shim change made for alignment can create a soft foot that was not there. Then confirm two things agree: every foot under the limit, and the coupling reading stable across a full loosen-and-retorque cycle of all four bolts. A set whose coupling reading shifts when the bolts are re-torqued still has frame distortion in it, whatever the individual foot numbers say.

References

• 29 CFR 1910.147 for mechanical isolation and stored energy; 29 CFR 1910.333(b)(2) with NFPA 70E-2021, 120.5, for the motor circuit; 29 CFR 1910.215 for abrasive wheel work; 29 CFR 1910.1053 for respirable crystalline silica in general industry; 29 CFR 1910.1200 for the safety data sheet; 29 CFR 1910.134 for respiratory protection program requirements.
• Machine manufacturer documentation for hold-down bolt torque, permissible soft foot, and maximum shim pieces per foot.
• See related: "How to Set Up a Shaft Alignment and Know It Held"; "What Misalignment Does to a Bearing in Mechanical Terms"; "Shimming, Leveling, and Alignment".