How to Set Up a Shaft Alignment and Know It Held

Why this matters

Most alignment jobs are done well and proved badly. The tech shims until the indicator reads inside tolerance, torques the bolts, closes the guard, and writes "aligned" in the notes. Six months later somebody re-reads it, finds it still inside tolerance, and writes "checked, good." Both entries are true and neither one tells you the machine moved four mils in between, which is the only number in the whole exercise that predicts anything. This is a walk through one job, and it treats the record as part of the work rather than paperwork after it.

The job

A 25 hp motor and a hard-piped end-suction pump on a grouted base. The pump was pulled for a mechanical seal and is going back. Nothing about the motor changed. That last fact is the trap the shop walked into last time, and it is where the story starts.

Isolate before anything, and keep it isolated through the readings

Open and lock the motor disconnect and prove the terminals dead with a live-dead-live check against a known live source before and after (29 CFR 1910.333(b)(2); NFPA 70E-2021, 120.5). Then the mechanical side, under 29 CFR 1910.147: confirm zero rotation, block the rotor, and isolate, lock, relieve and drain the suction and discharge lines to a container before a flange or a foot bolt is loosened, because a pressurized line releases into the machine the moment a restraint lets go and a hot one scalds while it does it. Read casing surfaces with a non-contact infrared thermometer and let them cool below the roughly 140 F contact-burn threshold before working bare-handed.

Three acts later in this job carry their own hazards. Turning the shafts by hand for readings is done with a bar on a barring point with the disconnect still locked, never by gripping a hub, and the rotor gets re-blocked whenever a hand enters the coupling gap. Jack bolts and hydraulic jacks under a machine foot hold real energy: back a loaded jack bolt off gradually rather than cracking it loose, and keep fingers out from between the foot and the base by moving the machine with a bar. Cut shim stock has edges that open a hand, so use pre-cut slotted shims and handle them with cut-resistant gloves.

Fix the foundation before you measure anything

The pump went back on and the first reading looked reasonable. It got ignored, because the base had not been checked yet and a reading taken before the base is right is a reading of the wrong machine.

Under each foot: the base surface clean down to bare metal, no paint film, no rust scale, no weld spatter, no burrs. Under the shims the same. The shims themselves clean, dry, flat stainless, slotted, and few in number - coupling and machine manufacturers commonly limit a foot to about four or five pieces, because a tall stack of thin shims behaves like a spring and lets the foot move under load. Get the limit from the machine documentation where it is published and stay at or under it by using one thick shim instead of five thin ones. A shim with a bent corner or a burr is a soft foot you built yourself.

Then soft foot, corrected before any alignment reading is trusted, for reasons a sibling article covers in full: bolting a machine down on a non-coplanar foot bends the housing and moves the bearing bores, so the alignment you measured is not the alignment that exists once the bolts are tight.

Then pipe strain, cold. With the coupling readings set up, loosen the suction flange bolts and watch the indicators. Movement means the pipe is pulling the pump, and the fix is at the pipe hangers, not at the pump feet. This is the cold check only. A set can pass it and still develop strain once the line heats and grows, which is a different failure and is covered separately in the references below.

Vertical first, horizontal second, and the reason is bolt torque

Vertical correction is shims, which means breaking the hold-down bolts loose and re-torquing them. Horizontal correction is sliding the machine with jack bolts and re-torquing. Do vertical first, because loosening every bolt to change shims will move the machine sideways and undo a horizontal correction, while a horizontal move does not disturb a shim stack.

Torque the hold-down bolts to the machine manufacturer's specification and use the same torque every time you re-read. A bolt at hand-tight and a bolt at spec produce different readings on the same machine, so a re-read at a different torque is not comparable to the baseline. Write the torque on the record.

The vertical correction, worked

Reverse dial, two measurement stations 6.0 inches apart, bracket sag measured on a pipe and already subtracted. Pump stationary, motor movable. Let x be inches from the near station toward the motor.

Slope is 6 minus 3 over 6.0, which is 0.5 mil per inch. The motor centerline is y equals 3 plus 0.5x, in mils. At the front foot, y equals 3 plus 0.5 times 9.0, which is 7.5 mils. At the back foot, y equals 3 plus 0.5 times 24.0, which is 15 mils.

Shim stock does not come in half mils. Take 7 mils out of the front feet and 15 out of the back, and check what the rounding left behind. The front foot drops from 7.5 to 0.5 and the back from 15 to 0, so the new slope is minus 0.5 over 15.0, which is minus 0.033 mil per inch. Projecting back: station 1 now reads 0.8 mils and station 2 reads 0.6 mils.

Under a mil of residual position and a thirtieth of a mil per inch of residual slope is comfortably inside any tolerance a coupling of this size carries at 1,780 rpm, and the point of running the check is that you know it rather than assume it. The alternative rounding, 8 mils at the front, lands the same magnitude in the other direction. What you must not do is round both feet the same way "to keep it simple", because the difference between the two foot corrections is the entire angular correction.

The record is the deliverable

The job is not finished when the guard closes. Write down, on the equipment record:

• Both station readings, vertical and horizontal, after final torque, with the station spacing in inches.
• The distances from station 1 to the front and back feet.
• The bracket sag figure and the span it was measured at.
• The shim thickness now under each foot, and the count of pieces.
• The hold-down bolt torque used.
• The machine condition at the reading: cold, hours since shutdown, casing full or drained.

Every one of those exists so the next reading is comparable. A coupling reading with no station spacing attached cannot be projected to the feet by anyone else. A reading with no machine condition attached cannot be compared to anything, because the same machine reads differently full and drained.

Six months later: the check that does the work

The re-check is run under the conditions written on the record: same stations, same spacing, same brackets and sag figure, same bolt torque, cold, casing full, and taken the morning after a shutdown rather than an hour after one.

Station 1 reads +1 mil. Station 2 reads +7 mils. Slope is 7 minus 1 over 6.0, which is 1.0 mil per inch.

Run it against the baseline rather than against the tolerance. The baseline after correction was 0.8 and 0.6 mils with a slope of minus 0.033 mil per inch. Station 2 has moved from 0.6 to 7 mils, and the slope has gone from effectively flat to 1.0 mil per inch, in six months, on a machine nobody touched.

Suppose that 1.0 mil per inch is still inside the coupling's published installation tolerance at this speed. That is exactly the case the pass-and-move-on habit gets wrong. A machine that moves six mils at station 2 in six months is not a machine that is in tolerance, it is a machine that is on its way out of tolerance and will get there sometime in the next half year, and the only thing that told you is the delta. A tech who records "checked, in tolerance" has thrown that information away and will meet the same set again as an emergency.

What to do with the delta. Over six mils of movement at a station has a cause, and the short list is short: the hold-down bolts have relaxed, the base or grout has settled or cracked, the shim stack has taken a set, or the piping has been re-hung or re-supported since commissioning. Check bolt torque first because it is one wrench and it either moved or it did not, then re-run the soft-foot check at the recorded torque, then walk the pipe supports looking for a hanger that has been adjusted or a spring can that has bottomed, working from a properly set and secured ladder or a work platform and using fall protection where the exposure meets the trigger in 29 CFR 1910.28 for general industry or 29 CFR 1926.501 on construction work. If all of those are clean and the movement only shows up on a machine that has been running hot, you are in different territory and the thermal articles in the references apply.

The failure mode of skipping the baseline is what happened to this set the first time. The pump was pulled, the motor was not touched, and the motor was therefore assumed not to have moved. It had - the base had settled under it while the pump was out and the piping was unsupported. With no recorded baseline for the motor, nobody could tell, so the correction was made entirely at the pump and the whole set was carried out of position together. It read fine at the coupling and the pump inboard bearing failed twice.

How to verify the job before you close the guard

Four confirmations, in order: soft foot re-checked at final bolt torque and inside limit; both stations re-read after final torque, not before; the twelve-plus-six against three-plus-nine validity sum closing at both stations; and the record filled in with station spacing, foot distances, sag, shims, torque and machine condition. The fourth is the one that gets skipped and the only one that makes the next visit worth anything.

References

• 29 CFR 1910.147 for mechanical isolation, stored energy, rotor blocking and draining a pressurized line before restraint is removed; 29 CFR 1910.333(b)(2) with NFPA 70E-2021, 120.5, for the motor circuit; 29 CFR 1910.28 and 29 CFR 1926.501 for fall protection duties in general industry and construction when pipe supports are inspected at height.
• Machine and coupling manufacturer documentation for hold-down bolt torque, maximum shim pieces per foot, and the installation alignment tolerance at operating speed.
• See related: "How to Tell Angular Misalignment From Parallel Offset"; "Why Soft Foot Defeats an Alignment"; "The Alignment That Was Perfect Cold and Wrong Hot"; "Shimming, Leveling, and Alignment".