Industry Trends · In-depth
Why Zero-Point Repeatability Drifts — and How to Get It Back
Almost every zero-point article tells you the interface holds 0.005 mm. Almost none tell you what happens in month eight. Here are the real failure modes, the maintenance intervals that prevent them, and a 20-cycle test to prove where you stand.
Every zero-point brochure quotes the same headline: repeat positioning accuracy below 0.005 mm. That number is real, and on day one you will measure it. The question nobody answers is what the interface reads in month eight, after 12,000 clamping cycles, three coolant changes and one operator who cleaned a locator with an air gun and a rag. This article is about that second number — the one that decides whether your setup sheets stay trustworthy.
Repeatability does not usually collapse. It drifts, and it drifts for a small number of identifiable reasons. If you can name the five, you can build a maintenance interval around each one and stop guessing.
The five things that actually cost you microns
| Failure mode | What you see | Typical cause | Fix |
|---|---|---|---|
| Chip on the taper | Sudden, repeatable offset on one station only | Chip fan or air blow-off not firing before load | Clean, verify blow-off timing, re-test |
| Coolant film / residue | Slow drift over weeks; part sits high | Coolant concentration too rich; no purge cycle | Purge cycle before clamp; check concentration |
| Pull-stud wear or wrong torque | Inconsistent pull-down; occasional loose pallet | Stud re-used across pallets, torque never re-checked | Torque audit; replace studs as a set |
| Seal degradation | Air consumption climbs; release gets sluggish | Normal wear, accelerated by hot coolant | Scheduled seal change (see intervals below) |
| Base-plate mounting shift | All stations drift together | Table bolts relaxed after thermal cycling | Re-torque to spec; re-map the grid |
Chips and coolant: the top two, by a wide margin
A single chip trapped on a ground locating taper erases the repeatability you paid for. This is not a marginal effect — a 0.05 mm chip under one contact face will tilt a pallet far beyond the interface's own 0.005 mm claim. Everything else on this list is a slower problem.
The design answer is that the interface should never rely on an operator noticing. Quality locators are sealed against chips and coolant, use inclined flange surfaces so swarf runs off rather than settling, and provide air blow-off through the seating face. The process answer is that blow-off must fire before the pallet lands, not after it is clamped.
Seat check: stop trusting, start measuring
Pneumatic seat checking works by feeding low-pressure air through vent holes in the seating surface and watching back-pressure. Fully seated means restricted flow; a chip or a partly engaged stud means the air escapes and the pressure never builds. The cell then refuses to start the cycle.
This matters most exactly where you can least afford it: unattended running. If a robot or pallet changer loads without a human ever looking at the interface, seat check is not a nice-to-have — it is the only thing standing between a chip and a scrapped part or a crashed spindle.
Maintenance intervals worth putting on a schedule
The intervals below are a defensible starting point for a shop running one or two shifts on water-miscible coolant. Treat them as a first draft: log what you actually find at each service and let the data pull the numbers apart. A cell running hot, chip-heavy cast iron will need tightening; a clean aluminium job shop can usually relax them.
| Task | Interval | Why this interval |
|---|---|---|
| Wipe seating face + verify blow-off fires | Every changeover | Contamination is the dominant failure mode |
| Inspect pull studs for wear and burrs | Monthly | Wear is gradual; catching it early avoids a bad batch |
| Torque audit on studs and base-plate bolts | Quarterly | Thermal cycling relaxes fasteners predictably |
| Full repeatability re-test (see below) | Quarterly, and after any crash | Gives you a number, not an opinion |
| Seal / wear-part replacement | Annually or per maker's cycle count | Seals degrade on a predictable curve |
| Re-map grid after any table work | Event-driven | Any table intervention invalidates the datum map |
The 20-cycle test: proving where you actually stand
Opinions about whether the interface is 'still good' are worth very little. A repeatability test takes about twenty minutes and gives you a defensible number you can trend over years. The method mirrors how positioning repeatability is defined in ISO 230-2: repeat the same approach many times and characterise the spread, rather than measuring once and declaring victory.
- 1Mount a test pallet carrying a ground reference block or a precision sphere.
- 2Set an indicator (or probe the sphere) in X, Y and Z and zero it with the pallet clamped.
- 3Unclamp, lift the pallet clear, blow off the interface, and re-clamp. That is one cycle.
- 4Repeat for 20 cycles, recording all three axes each time — do not stop at 5.
- 5Report the full spread (max minus min) per axis, not the average. The spread is your repeatability.
- 6Trend the result quarterly. A rising spread is your early warning, long before parts go out of tolerance.
What to demand at purchase so maintenance stays cheap
- Sealed design with inclined flange surfaces so chips run off instead of settling.
- Air blow-off through the seating face, plumbed so it fires before the pallet lands.
- Seat-check capability (vent holes or a gap sensor) — essential for any unattended cell.
- Pull studs available as matched replacement sets, with a published torque figure.
- Documented wear parts and a stated service interval, so maintenance is schedulable rather than reactive.
- One grid pitch across the fleet, so a fixture proven on one machine is proven on all of them.
See the interface being checked
Frequently asked questions
How often should I re-test zero-point repeatability?+
My repeatability got worse but only on one station — what is it?+
Does coolant concentration really affect a mechanical interface?+
Is seat check worth it if an operator loads every pallet by hand?+
How long do the seals and wear parts actually last?+
Sources & further reading
Related products
Discuss your application with our engineers
Send a drawing or describe the part and machine; we review it and come back with a configuration and quotation.
