PROJECT DESIGN REVIEW — MINUTES OF MEETING
Project: three-axis pick-and-place cell, Z-axis rebuild
Review subject: ball screw preload class for the vertical drive
Date: 22 April, 10:00 to 11:40, meeting room 1
Document ref: DMR-2026-0422-ZAX01
Attendees
| Name | Role |
|---|---|
| E. Novak | Lead mechanism designer (chair) |
| S. El-Gharbi | Senior fitter, machine assembly |
| L. Fontaine | Servo and drive engineer |
| K. Ono | Metrology / CMM engineer |
| R. Muller | Procurement |
1. Purpose of the review
The Z-axis caries a 14 kg head assembly on a 32 mm diameter ball screw with a 10 mm lead and a 625 mm travel. The customer complaint was position overshoot on rapid moves: the axis overshoots 0.11 mm after a 400 mm move at 55 m/min and settles in about 240 ms, which is twice the 120 ms specified. The draft had an un-preloaded ground ball screw, class C5, and the review was called to decide whether to keep it or move to a preloaded nut, and if so, what preload class.
2. Review comment ZA-01: why the un-preloaded nut overshoots
E. Novak opened with the mechanics, and the cause was not what the service report said. The service report blamed the servo tuning. That was wrong. An un-preloaded ball nut has clearance between the balls and the groove, nominally 0.02 to 0.035 mm, which in an axis with a 14 kg moving mass means the nut can rock about 0.04 mm before the balls take load. At 55 m/min the head changes direction instantly at the reversal, the nut clearance closes with the full momentum behind it, and the reversal hammer lands as a position spike on the encoder, which the servo then corrects over the next 200 ms. The overshoot is therefore a mechanical clearance, not a tuning gain problem, and no amount of P and I tuning will remove it.
S. El-Gharbi confirmed from the build floor: the same screw design on a horizontal axis had shown 0.03 mm of lost motion from day one, and the team had tuned around it. The chair asked whether the answer was simply a preloaded nut, and the review moved to the numbers.
3. Review comment ZA-02: preload classes and the available catalogue
L. Fontaine presented the preload families available on the 32×10 screw. The catalogue offers four: zero-clearance (called P0 in this maker, equivalent to a few microns of internal clearance removed), light preload P3, medium preload P5, and heavy preload P7, with the design preload force in newtons given per class. P0 keeps the nut free-running with near-zero clearance, P3 sets about 3 to 5 percent of the dynamic load rating as preload, P5 about 7 to 10 percent, and P7 about 12 to 15 percent. On the 32×10 screw with a dynamic rating of 19.4 kN, the numbers work out to roughly 0.6 to 1.0 kN for P3, 1.4 to 1.9 kN for P5, and 2.3 to 2.9 kN for P7.
K. Ono asked how these numbers translate to stiffness, because that is what the axis actually feels. The assembled stiffness of the nut and screw is a series combination: screw shaft stiffness plus nut-plus-preload stiffness. The shaft on a 32 mm screw at 300 mm bearing span contributes about 260 N per micron of deflection in bending and about 530 N per micron in axial compression for the shorter columns, but the nut assembly stiffness is the weak link. An un-preloaded nut measures roughly 90 N per micron, a P3 nut about 220 N per micron, P5 about 320 N per micron, and P7 about 480 N per micron. The review wrote these values on the board and stopped talking in classes, because the numbers mattered more than the letters.
4. Review comment ZA-03: does a heavier preload always help
E. Novak cautioned that stiffness is not free. A heavier preload shortens the screw life, raises the breakaway torque, and heats the nut. L. Fontaine pulled the life trade: doubling the preload roughly halves the remaining fatigue life of the ball track because the preload is a standing load on the balls. On the 19.4 kN rated screw, a P7 preload of 2.6 kN reduces the calculated L10 life at the customer’s 4.2 kN mean axial load from 9,400 hours to about 5,100 hours, which still passes the 4,000-hour target but leaves less margin than P5. The breakaway torque also climbs: P3 adds about 0.06 N-m, P5 about 0.11 N-m, P7 about 0.17 N-m of drag, and against a 2.2 kW servomotor that is small, but the drag heats the nut, and the review asked the fitter how hot the nut runs in normal duty.
5. Review comment ZA-04: breakaway torque and the servo budget
L. Fontaine moved the discussion to the drive. The vertical axis hangs on the screw, so any preload drag fights gravity on the up stroke and adds to it on the down stroke. The 2.2 kW motor with a 10 mm lead screw develops roughly 7.0 N-m of continuous torque at the screw input. Gravity on the 14 kg head plus the 6 kg counterweight spring offset leaves a nett downward load of about 80 N at the screw, requiring only 0.13 N-m to hold, so the axis is far inside the budget. The real constraint is the demanded acceleration: the customer wants 9.5 m/s-squared on short moves, which at a 10 mm lead translates to a servo peak of about 13 N-m on the motor, and the preload drag of 0.11 N-m is under two percent of that peak. The conclusion was written as a line: speed and acceleration consume the budget, preload drag does not.
6. Review comment ZA-05: installed length, thermal growth and preload retention
K. Ono raised the thermal question, and it was the sharpest comment of the day. A P5 nut preloaded to 1.6 kN holds that value only if the screw and nut stay at the same temperature. The Z-axis sits between two bearings with a fixed-fixed mount, and the screw temperature rises about 8 degrees C above the housing in a continuous run. Steel at 11.7 microns per metre per degree C grows about 29 microns over the 300 mm span at that delta, and that growth attempts to push the nut toward the fixed end, raising or relaxing the preload depending on which end is fixed. The review noted that a fixed-fixed screw with a preloaded nut needs a vented or spring-loaded support at one end, or the preload changes with every shift warm-up. The workshop had not thought of this, and the minutes record it as a finding, not an afterthought.
S. El-Gharbi added that the machine greases the nut only at commissioning, and asked whether the preload stays put through the year. The answer from the catalogue was that a ground screw with a matched set of balls keeps its preload for roughly two years of normal duty, after which the preload drops below 60 percent of nameplate. The review therefore added a nut preload check at the annual service, and the fitter agreed to measure the drag torque with a torque wrench at the motor coupling and log it against the commissioning value of 0.11 N-m.
7. Review comment ZA-06: procurement and the lead time argument
R. Muller spoke for the supply chain. A 32×10 ground ball screw with a P5 preloaded nut came back quoted at 38 working days and 1,180 euros from the German maker, versus the fitted un-preloaded C5 screw already on the machine at zero marginal cost. The price difference looked large on paper, 1,180 euros against a nil line, but the review computed the cost of the overshoot complaint: one service visit, one servo-tuning session, one half-day of line downtime, and the customer’s quality report, all of which had already been invoiced at roughly 1,600 euros on the previous axis. E. Novak put the decision in one sentence, that a 1,180 euro part is cheap when the lack of it has already cost more, and the procurement line was closed.
8. Disagreement that split the room: fixed-fixed versus fixed-simple mounting
The genuine dispute was not preload class but mounting. E. Novak wanted to keep the fixed-fixed bearing arrangement for rigidity, because bending rigidity of the screw at 300 mm span is what keeps the moving head from nodding. S. El-Gharbi argued for a fixed-simple arrangement with one spring-supported end, accepting a small bending penalty to keep the preload stable through thermal growth. K. Ono split the difference with a calculation: on a 300 mm span, a fixed-simple support reduces axial stiffness by about 12 percent but removes the thermal preload drift almost completely, provided the spring sets at the nut preload force. The room voted to keep fixed-fixed on this axis because the 12 percent bending stiffness matters more to the pick head than the thermal drift, and to fit the preload check into service instead. The chair overrode nobody; the vote was eight to two.
9. Risks and mitigation
Three risks were logged and each got an owner. Risk 1: the P5 nut could arrive with a wrong ball set and measure too stiff or too soft; mitigation is to verify drag torque at the commissioning rig, target 0.11 N-m plus or minus 0.02, before the axis is coupled. Risk 2: the existing C5 screw already in the machine has a plain nut, so the team must order a matching screw and nut pair, not a nut alone, because the ball track pitch of the old screw will not match a new P5 nut; the review caught this before procurement wasted a lead time. Risk 3: the thermal drift measured in production could still push the overshoot back above 0.08 mm; mitigation is the annual preload check plus a quarterly log of positioning deviation, and if the deviation grows by more than 0.02 mm between logs, the nut is re-preloaded by shims or replaced.
10. Actions and owners
| ID | Action | Owner | Deadline |
|---|---|---|---|
| ACT-1 | Order 32×10 ground ball screw with P5 preloaded nut, matched pair | R. Muller | 30 April |
| ACT-2 | Verify drag torque on commissioning rig at 0.11 N-m | S. El-Gharbi | 12 May |
| ACT-3 | Retune servo position gains after mechanical change, log reversal error | L. Fontaine | 14 May |
| ACT-4 | Set up quarterly positioning deviation log on CMM | K. Ono | 19 May |
| ACT-5 | Add annual nut preload check to service schedule | S. El-Gharbi | 31 May |
11. Decision summary
Adopt a 32×10 ground ball screw with a P5 preloaded nut, preload force 1.4 to 1.9 kN, equivalent nut stiffness about 320 N per micron. Keep the fixed-fixed mounting and compensate the thermal drift by measuring and logging, not by changing the bearing arrangement. Retune the servo after the mechanical fix, and treat the 0.04 mm nut rock as the root cause, not as a tuning artifact. The 0.05 mm positioning specification stays.
12. Post-meeting addendum, 18 May
Commissioning is done. The new P5 nut measured 0.10 N-m drag against the 0.11 N-m target, within tolerance. Reversal error after the retune dropped from 0.11 mm to 0.028 mm on a 400 mm move at 55 m/min, and the settle time came down from 240 ms to 110 ms, inside the 120 ms figure. The customer signed the acceptance on the fourth move, which was faster than the review dared to promise. The annual preload log is on the service sheet, and the fitter’s note reads that a preloaded nut is a flower that needs its torque checked once a year.
Glossary
Ball screw preload: removing internal clearance between the nut and the screw ballsets so the nut pushes against the track under a set force. Dynamic load rating: the axial load at which a screw reaches one million revolutions of L10 life. Breakaway torque: the additional torque required to start the screw turning against the preload drag. Lost motion: the axial play between nut and screw when the direction reverses. Fixed-fixed mounting: both screw ends held in rigid bearing blocks, highest axial stiffness and highest thermal sensitivity.
