PROJECT DESIGN REVIEW — MINUTES OF MEETING
Project: gantry-style inspection machine, X-axis rail conversion
Review subject: guide rail mounting datum, reference edge and bolting sequence
Date: 6 May, 13:30 to 15:10, workshop office
Document ref: DMR-2026-0506-GAX03
Attendees
| Name | Role |
|---|---|
| H. Berg | Lead mechanism designer (chair) |
| D. Ilyin | Lead fitter, precision assembly |
| M. Chen | Laser tracker / alignment engineer |
| T. Weber | Machine design, baseplate |
| V. Santos | Procurement |
1. Purpose of the review
The gantry X-axis carries a 21 kg bridge on two parallel rails, each 1,100 mm long, size 25 mm wide rail with a 23 mm wide carriage. The machine is being rebuilt because the last straightness test on the previous bed showed 0.09 mm over 900 mm, against a 0.05 mm spec, and the customer’s laser was already marking it as out of tolerance. The review was called to fix the mounting datum before the new bed is machined, because the old build had bolted the rails to a milled surface without a controlling reference edge, and the assembly honesty had been left to the fitter’s eye.
2. Review comment GX-01: the missing reference edge
D. Ilyin opened with the workshop confession: the previous build had no datum edge at all. The cure was to clamp and shift, clamping the rail, tapping it with a dead-blow hammer, measuring, and repeating, which on a 900 mm travel costs a full day and still leaves a straightness that drifts with the bolt torque. The chair asked why the drawing did not carry a machined shoulder. T. Weber answered that the baseplate was a welded and stress-relieved fabrication, that machining a full-length datuming shoulder was seen as expensive, and that the previous supplier had promised a scraper would fix it by hand. The review wrote the first decision before anything else: the new bed carries a machined datum shoulder on each rail pocket, ground, not milled, and the fitter does not scrape to straightness any more.
3. Review comment GX-02: datum on the rail or on the bed
M. Chen raised the real design question, which is which surface is the datum. The catalogue rail is straight and true to a few microns per metre, so the datum can be the rail edge or the bed shoulder. If the datum is the rail edge, the rail must be pushed hard against the shoulder with a clamp force that overcomes the bolt friction, and the shoulder does the alignment. If the datum is the bed shoulder only, the rail is placed against it but the carriage then sees whatever the rail’s own edge does, which on a 25 mm rail is held to about 8 microns per metre of straightness. The review decided that the rail edge is the functional datum, because the carriage rides the rail groove, not the shoulder, and therefore the shoulder must be straight to half the rail straightness, about 4 microns per metre, or it becomes the weak link.
T. Weber was not satisfied with the microns talk and asked for a number the shop can grind to. The answer was written on the board: shoulder straightness 0.008 mm over the full 1,100 mm rail length, shoulder squareness to the rail plane within 0.01 mm over the width, and the two shoulders on opposite rails parallel within 0.015 mm along the travel. These are grinding tolerances, not mill tolerances, and the new bed goes to the grinder.
4. Review comment GX-03: bolting sequence and its effect on straightness
D. Ilyin described the assembly step that spoils a good datum: bolting the rail down from one end to the other in order. Every bolt pulls the rail into the bed locally, and the accumulated pull curls the rail away from the shoulder in the middle of a long rail. On the old 1,100 mm rail the middle of the two rails had visibly opened by about 0.03 mm against the shoulder after bolting, measured with a feeler gauge. The fix the fitter proposed, and the review accepted, is a centre-out sequence: first the centre bolt, then alternating bolts outward to each end, each tightened twice, once to 60 percent and again to full torque after the whole row is seated, so the rail relaxes as a beam instead of being nailed down like a plank.
5. Review comment GX-04: torque, bolt size and preload of the rail bolts
T. Weber questioned the bolting itself. The rails are held by M6 socket cap screws on a 60 mm pitch, sixteen per rail. The workshop had been torquing them to 9 N-m, which the catalogue calls light for a rail this size. H. Berg looked at the preload: an M6 12.9 bolt at 9 N-m develops roughly 7.5 kN of clamping force against a yield of about 19 kN, so the rail has plenty left, but the vertical clamping force also must not crush the raceway. The catalogue limit for the 25 mm rail is about 3 kN per bolt radial, so the 7.5 kN is axial clamping, not radial, and it is safe. The review set the assembly torque at 10 N-m plus or minus 0.5, applied with a torque wrench, and the fitter logged each rail bolt number in the assembly record so a loose bolt is traceable.
6. Review comment GX-05: ordering the rails and their matched sets
V. Santos asked whether the two rails could be bought separately at a lower price. M. Chen warned that the catalogue sells matched rail pairs or individual rails, and that two individual rails from different batches can have a height mismatch of up to 0.03 mm, which doubles the preload on the higher rail carriage. On a matched pair the maker grinds the rails in one pass and the height difference stays under 0.005 mm. The review made procurement order a matched pair with the height difference certificate included, and V. Santos noted the price difference was 42 euros against a rebuild budget measured in thousands, so the decision cost nothing in practice.
7. Review comment GX-06: carriage preload, parallelism and the two-rail twist
The meeting turned to what happens after both rails are straight and parallel, which is twist. If the two rails are not coplanar, the bridge twists, and the twist shows up as pitch error at the tool. M. Chen explained that the signature is a difference in the vertical position of the two carriages on the same bridge, and that the fix is to set both rails in the same plane, not merely parallel in plan. The review wrote the check as a coplanarity target: both rails within 0.02 mm in the vertical direction along the whole travel, confirmed with the laser tracker before the bridge is bolted on. The fitter agreed, and noted the previous build had never checked coplanarity at all, only left-right position, which is another reason the old straightness drifted after a week.
8. Review comment GX-07: the heat of the rail and its reference to the bed
T. Weber asked whether the steel bed and rail grow together. The bed is 16 mm thick steel plate in the rail pocket, while the rail is 25 mm wide hardened steel, and both are plain carbon steel with the same expansion coefficient, but the bed also carries the bridge loads, so the rail and bed never sit at exactly the same temperature. In a machine that warms up by 6 degrees C, the rail sees nearly the same growth because it is clamped to the bed, and the datum shoulder keeps both aligned, so the earlier tear on the previous build came from the rails being unconstrained and floating during warm-up. The review kept the datum shoulder design and did not add expansion slots, because the differential growth over 1,100 mm at a 6 degree delta is only about 8 microns, well inside the straightness tolerance.
9. Risks and mitigation
Four risks went onto the board. Risk 1: the ground datum shoulder could be damaged in handling before the rails arrive; mitigation is a plywood cover and a straightness check on the bare bed the day before assembly. Risk 2: the rail bolts could be over-torqued by habit from the old 9 N-m routine; mitigation is a calibrated torque wrench on the bench, not in a drawer. Risk 3: the matched pair could arrive without the height difference certificate; mitigation is to reject the consignment unless the certificate travels with it. Risk 4: the coplanarity check needs the laser tracker, which is booked elsewhere on the planned day; mitigation is to book the tracker first and move the assembly window around it, not the other way round.
10. Actions and owners
| ID | Action | Owner | Deadline |
|---|---|---|---|
| ACT-1 | Machine and grind datum shoulders on both rail pockets, straightness 0.008 mm | T. Weber | 20 May |
| ACT-2 | Order matched 25 mm rail pair with height certificate | V. Santos | 12 May |
| ACT-3 | Write assembly procedure: centre-out bolting, 10 N-m, log each bolt | D. Ilyin | 26 May |
| ACT-4 | Laser tracker check: shoulder straightness, parallelism, coplanarity | M. Chen | 29 May |
| ACT-5 | Book tracker window around the assembly date | H. Berg | 14 May |
11. Decision summary
Grind datum shoulders into the new bed and push both rails against them, with the rail edge as the functional datum. Use a centre-out bolting sequence at 10 N-m. Buy a matched rail pair with a height difference certificate. Check straightness, parallelism and coplanarity with the laser tracker before fitting the bridge. Keep the 0.05 mm straightness specification and treat it as a build geometry, not a final adjustment.
12. Post-meeting addendum, 12 June
The new bed came off the grinder holding 0.006 mm straightness on both shoulders, and the rails went on in one shift. The centre-out sequence took the fitter about ninety minutes per rail, against the old full day of hammering, and the final laser tracker report showed straightness 0.021 mm over 900 mm travelled, well inside the 0.05 mm spec, with coplanarity 0.014 mm. The customer ran the acceptance plate three times and each pass recorded within 0.03 mm of the previous. The annual re-check is scheduled, and the fitter’s closing note in the record reads that a datum edge is cheaper than every dead-blow hammer ever sold.
Glossary
Datum shoulder: a machined reference surface that a rail is pushed against for alignment. Coplanarity: how far two rail tops lie out of a common plane. Straightness: deviation of the rail running surface from a true straight line over its length. Rail preload: the internal load chosen between carriage rollers and raceway. Height difference: the vertical mismatch between matched rails, controlled in a matched pair.
Appendix. Reference table of the agreed tolerances
The review closed with the tolerance sheet that now lives inside the assembly drawing, reproduced here exactly as agreed. Shoulder straightness: 0.008 mm over 1,100 mm. Shoulder squareness to rail plane: 0.01 mm over 23 mm rail width. Shoulder parallelism between left and right shoulders: 0.015 mm along travel. Coplanarity of the two rails: 0.02 mm vertical. Rail-to-shoulder contact after bolting: no feeler gauge of 0.03 mm to pass along a 300 mm radius of the seam. Bolt torque: 10 N-m plus or minus 0.5 on M6 socket head screws. Height difference of the matched pair: under 0.005 mm. Straightness of the assembled system at acceptance: 0.05 mm over 900 mm. The fitter keeps this sheet in the rail pocket box, and the laser tracker report is filed next to it.
One remark the chair asked to be captured word for word, because it is the whole lesson of the morning: a guide rail does not need to be set, it needs to be placed. Every machine that fails the straightness test fails because somebody assumed the rail could be nudged into position, and every design that passes leaves nothing for the fitter to nudge. The old machine was a monument to patience with a hammer; the new one is a machine a fitter can build before lunch and trust for a year.
