Troubleshooting
Fixing Ballscrew Thrust Bearing Axial Play and Preload
Short answer
Measure screw end, bearing housing and table movement separately under a reversing axial force. If 0.0002–0.0003 in of play remains between the screw end and the base after the locknut is correctly tightened, the matched back-to-back set has lost its preload: replace it as a set. Shimming the outer rings without the mounting drawing is not a controlled fix.
On this page
Play that drops from 0.0005 in to 0.0002–0.0003 in after tightening the locknut, and that lives between the screw end and the machine base, is a bearing set that has lost its preload. The fix is a new matched set, not more torque and not a shim. Prove it first: one dial indicator, moved between the screw end, the bearing housing and the table, under a reversing axial force, tells you in ten minutes whether the bearing, the housing mounting or something upstream owns the movement. Preload in a back-to-back angular-contact pair is ground into the ring faces at the factory, which is why the set is replaced as a set and why field shimming cannot hit the number.
Measure each interface separately #
Do not measure backlash at the table and call it bearing play. Mount a magnetic indicator base on the machine base and take three readings while you push, then pull, along the screw axis with a steady force. A spring balance reading a few tens of pounds is plenty on a mill-size axis; the force only has to fully reverse the load path, not imitate cutting. Disable the drives and block the axis before putting hands and indicators in the envelope, because a jogging axis will drive the screw you are working on.
- Screw end versus bearing housing: movement here is inside the support itself.
- Bearing housing versus machine base: movement here is the mounting, not the bearing.
- Table versus machine base with the screw restrained: this isolates the nut, the coupling and the drive train from the support.
Comparing the three numbers localizes the play before any part is ordered or opened.
What each reading means #
| Reading after the locknut is correctly torqued | What it tells you | Confirming check |
|---|---|---|
| Screw end moves 0.0002–0.0003 in against the housing, in both directions | The preloaded pair no longer preloads | Open and verify the clamping path; if it is clean and seated, replace the set |
| Housing moves against the base | The mounting owns the play, not the bearing | Bolt torque, mounting-face condition, fretting under the housing |
| Screw end is solid but the table still moves | The play is upstream: coupling, drive belt or nut flange | Indicator on the ballnut flange and on the coupling while reversing the force |
| Jog drive load well above the pre-work baseline | Over-preload or a bind introduced during the work | Back the locknut off, re-seat the stack, re-measure |
Why the locknut cannot restore worn preload #
A back-to-back pair is preloaded by geometry: the rings are ground so that when the inner-ring stack is clamped and the outer-ring stack seats, the balls carry load before any external force exists. The locknut's only job is to close the inner-ring stack — shaft shoulder, inner ring faces, spacers, nut face — and hold it closed.
Three things defeat it. A nut that bottoms on a thread runout or against another component feels tight while the rings stay loose. Burrs, paint or contamination under a ring face hold the stack long with the nut fully torqued. And wear in the raceways simply consumes the ground offset, and no clamp force you can apply recovers it. A reading that improves when you clean and retorque, then stops improving, is the signature of that third case, and it ends with a new set.
The shim question, with the arithmetic #
Matched sets carry the preload in ground ring faces; universal, unmatched bearings take it from spacers whose lengths the mounting drawing fixes. Either way the displacement involved is tiny, and that is the problem with adding a shim by feel.
Worked example, calculated and not measured. Every input except the unit conversion comes from the datasheet of the exact set; the stiffness here is an illustrative stand-in:
- Pair axial stiffness k = 200 N/µm (illustrative; substitute your datasheet value).
- Displacement to reach the specified preload class δ = 0.0002 in = 0.0002 × 25.4 = 0.00508 mm = 5.08 µm.
- Preload force F = k × δ = 200 N/µm × 5.08 µm ≈ 1,020 N, about 230 lbf.
- One tenth of a thousandth of shim error: 0.0001 in = 2.54 µm, so ΔF = 200 × 2.54 ≈ 510 N, about 115 lbf — half the target, in whichever direction the error happens to fall.
Real bearings stiffen as load rises, so the error grows with preload instead of averaging out, and a tenth of shim stock is at the edge of what a bench can verify without gauge blocks. The conclusion is not a blanket ban: supports designed for universal pairs are shimmed, to calculated lengths on a drawing. It is that without a drawing calling for a shim with a length on it, you are adjusting a 510-N-per-tenth quantity blind, so replace the matched set instead.
Set it by the document, verify it by movement, load and heat #
Preload classes — light, medium, heavy — are defined by the bearing manufacturer for that part family, not by an interchangeable industry number, so the machine's parts document is the authority on which class the support was built around. Verification, in order:
- Movement: repeat the screw-end-versus-housing indicator test with the same force as the baseline. Zero repeatable movement is the pass condition.
- Drive load: record the idle jog load or motor current before opening the support and compare after. Rolling resistance is not an acceptance number — seals, lubricant and the rest of the axis are in it — but a step change that appears with the new preload is your over-preload flag.
- Temperature: run the axis and watch the housing. A support that warms for a while and then stabilizes is normal friction; one that keeps creeping is over-preloaded or binding.
One arrangement check while you are in there: the non-drive end of a ballscrew should float axially — a deep-groove or cylindrical roller bearing, free in the housing — so the screw's growth with temperature is not fought by a second preloaded support. Angular-contact geometry loads the screw axially by design, so finding it at both ends means checking the drawing before you duplicate the arrangement. Where you are redesigning the support rather than repairing it, the choice between holding preload in ground bearing geometry and letting a spring hold it is worth making deliberately; a comparison of angular-bearing preload against spring preload lays out that trade-off.
Repair sequence #
- Baseline the three indicator readings and the idle jog drive load, and write them down.
- Identify the set: bearing designation from the machine parts document, confirmed against the pairing marks and orientation on the installed parts.
- Open the clamping path: shaft shoulder, spacer faces, housing shoulder, locknut face and threads. Deburr, clean, and check that the nut does not bottom on the thread runout.
- Reassemble to the documented stack, torque the locknut to the machine or bearing document's value, and stake or lock it.
- Re-measure with the same indicators and the same force. Zero movement ends the job. If movement survives, order the matched set — same designation, same preload class — and replace both bearings together, applying fitting force only to the ring being seated, never through the balls.
- After the swap, re-run the movement, drive-load and temperature checks, and repeat them after the first warm-up and cool-down cycle.
- If the movement read housing-to-base, correct the mounting first — faces, bolts, fit — and only then judge the bearing.
The decision for each finding #
- Cleaned, verified clamping path and the retorque removes the movement: run the axis, confirm drive load and housing temperature against the baseline, and log the repair.
- The clamping path is clean and 0.0002–0.0003 in of screw-to-housing movement survives the retorque: the set has lost its ground preload. Replace the matched pair as a set and do not shim.
- The movement reads housing-to-base: the bearing is likely fine. Correct the mounting and re-test before touching bearings.
A matched set replaced as a set, with force applied only to the ring being fitted, is the repair that holds. Everything above it is diagnosis.
Questions asked
How much axial play is acceptable at a ballscrew thrust bearing?
None. A correctly preloaded back-to-back set shows no repeatable movement on a dial indicator resolving 0.0001 in when you reverse a steady axial force at the screw end. Any movement that survives a verified clamping path means the preload is gone, and the set is replaced rather than re-torqued. Screw-and-nut backlash is a separate measurement handled by the machine builder's compensation.
Why back-to-back instead of face-to-face for a screw support?
In a back-to-back pair the contact lines spread outward, so the effective span between pressure centers is longer and the support resists moment loads and misalignment better, which suits a screw end that also carries the radial weight of the screw. Face-to-face gives a shorter span. Either orientation can be preloaded; the machine drawing decides, and rings from two sets or two orientations must never be mixed.
Can I judge preload by how hard the screw is to turn?
Only as a change detector, not an acceptance test. Seal drag, lubricant, temperature and the rest of the axis all sit inside that torque, so no universal rolling-resistance number exists. Record the idle jog drive load or motor current before the work and compare it after: a step change that appears with the new preload and grows as the housing warms is your over-preload flag. The number that matters is in the bearing datasheet.
Are tapered roller bearings a valid substitute at the screw support?
For a new, slow, heavily loaded leadscrew they work and were long used on manual machines, because they carry thrust at low cost. They run hotter at preload and slower for a given size than angular contact, which is why preloaded CNC ball screw supports use ground angular-contact sets. In an existing support, swapping types changes every spacer, shoulder and fit: treat it as a redesign, not a repair.
What if the play is between the bearing housing and the machine base?
Then the bearing is not your problem yet. Check the housing bolts for torque and stretch, the mounting face for paint, burrs and fretting, and the housing fit in its bore. Clean, re-torque and re-measure screw-to-base movement with the same indicator setup. Chase the bearing only once the housing itself is rigid; otherwise you will preload a support that still walks.