PROJECT DESIGN REVIEW — MINUTES OF MEETING
Project: six-station machining cell with one robot load station
Review subject: automatic load and unload mechanism, gripper sizing, cycle timing, fault recovery
Date: 18 August, 10:30 to 12:20, meeting room 2
Document ref: DMR-2026-0818-LUL10
Attendees
| Name | Role |
|---|---|
| J. Okafor | Lead applications engineer (chair) |
| R. Hoffmann | Robot and gripper engineer |
| S. Velez | Production planner, cell owner |
| K. Lindqvist | Tooling and fixture designer |
| P. Fontaine | Controls and safety engineer |
1. Purpose of the review
Six machining stations share one loading robot, and the cell promises a cycle of 42 seconds per part with a two-minute in-machine time hidden by the robot serving the stations out of phase. The draft load station is a gravity chute feeding a pick cradle, a double gripper carrying a finished part out and a raw one in, and a photoelectric check that the part is seated. The review was called because the cell owner wanted the mechanism committed on paper before the robot program existed, after the previous cell lost four hours a shift to misloaded parts that the robot then crumpled against the fixture.
2. Review comment LU-01: the cycle budget, which the 42-second promise depends on
R. Hoffmann reduced the 42 second curve to its pieces and the room saw where the time went. The robot spends 3.8 seconds travelling from the last station to the load station, 2.6 seconds picking the finished part, 1.9 seconds placing it on the out-feed chute, 3.1 seconds picking the raw part, 3.4 seconds inserting it into the fixture, and 1.2 seconds releasing and withdrawing, with 2.0 seconds of position regains and handshakes, totalling 18 seconds of robot time per part at the load station alone. The six stations phase so the robot makes six load visits per cycle, and the review’s arithmetic said the robot utilization sits near 78 percent with zero fault time, leaving 14 seconds of slack across the cell. The number the room wrote down: the load mechanism must not add more than 3 seconds of waiting at the cradle, or the cell misses its 42 second promise.
3. Review comment LU-02: the gravity chute, which jams on the ninth part
S. Velez read the fault log from the trial chute, and the pattern was boring and predictable. A gravity chute with a 92 mm wide track feeding a 86 mm part, clearanced at 6 mm, ran clean for eight parts and then jammed on the ninth, a part that arrived across the track on its web. The trial ran 1,200 parts and logged 41 jams, one in every 29 parts, each costing a hand to clear. K. Lindqvist changed the approach in the review: the chute becomes a powered roller bed with a 1.5 degree working slope instead of the 18 degree gravity slope, so the part is carried not slid, and an optical presence check at the pick position stops the bed when the cradle is full. The number that fixed the trial: the part must rest supported across at least two rollers at every point of travel, and the 60 mm roller pitch on a 110 mm part gives that, instead of the chute’s single-line contact.
4. Review comment LU-03: the double gripper and its datum problem
R. Hoffmann stepped the room through the gripper and the failure the trial found. The double gripper carries the finished part on one jaw set and the raw part on the other, and the trial found the raw part grip drifting by 0.25 mm between the pick and the insert, enough to ring the fixture bore and alarm out once every forty parts. The cause was the two-jaw platten deflecting under the moment of the two parts hanging off it, one on each side. The review settled the gripper on a three-point nest against a machined datum face on each jaw, with the raw part pushed back to a hard stop before the robot leaves the cradle, so the part position is set by the nest and the stop, not by the grip force. The drift number dropped in the design review from 0.25 mm to a target under 0.05 mm at the fixture mouth.
5. Review comment LU-04: the seated check, one sensor is not a check
P. Fontaine pressed the room on how the cell knows a part is actually seated before the cycle starts. The draft used a single photoelectric sensor at the fixture mouth, and the review found the failure path: a part can trip the sensor while standing on its locating pins half rotated, and the robot withdraws, and the tool then crashes into the tilted part. The design principle adopted: a seated part must satisfy two independent checks, the photoelectric beam and a mechanical height probe at the fixture bar, sampled simultaneously in the handshake before the clamping cycle. R. Hoffmann logged the runtime cost at 0.3 seconds per station and the room took it, because the trial’s crumpled-part cost was a full stop and a rework, not a third of a second.
6. Review comment LU-05: what happens when the robot drops a part
The review moved to what the cell does when it fails, because a mechanism that only works when nothing goes wrong is a mechanism that stops the cell. The draft had the robot, on a dropped-part alarm, retrying the pick once and then stopping the whole cell for a hand. The review wrote the recovery ladder: first retry, then a repositioning sequence that backs the part on the bed and pulls it to the stop again, then a single operator call at the cell PLC with a graphic pointing to the exact station and the exact sensor. S. Velez insisted the recovery not reset the machine counters, because the previous cell quietly reset its cycle counter on every recovery and the plant lost the production record. The PLC sequence stores the fault and the corrected cycle count before any retry.
7. Review comment LU-06: the part keep the orientation through the whole path
K. Lindqvist asked whether the part could get turned around between the bed and the fixture, and the trial said yes. The raw part leaves the roller bed in one orientation, is picked, and arrives at the fixture, and the review added a datum face on the bed side rail that the part rides against for the last 400 mm of travel, so the orientation at pick is the orientation at insert. The rail adds an alignment strip and removes the rotate-in-hand gamble, and the review logged it as the cheapest insurance in the mechanism, a machined strip of steel rather than another sensor.
8. Disagreement: a second robot or a buffer magazine
The budget argument spilled into the review when S. Velez asked whether the 78 percent robot utilization was healthy enough, and R. Hoffmann answered no if a station alarms mid-cycle. The room debated a second robot over a buffer magazine for the finished parts. A second robot, at 28 thousand euros, would guarantee the cell keeps its 42 seconds even with one station down, but the cell owner baulked at the spend for a promise that a clean load station might never need. K. Lindqvist argued the buffer magazine instead: a ten-pocket rotary indexer at the out-feed that holds finished parts when the conveyor blocks, at three thousand euros, so the robot never waits on the conveyor and the mechanism gains ten parts of elastic storage at the cost of a rotary table. The decision went to the buffer magazine, with the note that a second robot can be added later on the same bed plate, and the cell keeps the 3 second slack target as its buffer for now.
9. Risks and mitigation
Four risks were logged. Risk 1: the roller bed stops when the cradle is full but the parts keep arriving from upstream; mitigation is a back-pressure stop and a part-backup alarm before the bed end. Risk 2: the two-sensor seated check can itself drift out of alignment; mitigation is a weekly probe-verify cycle that touches both sensors against a gauge part. Risk 3: the datumed gripper nest wears as the raw parts slide into the stop; mitigation is a hardened insert in the nest, replaceable, logged in the spare list. Risk 4: the buffer magazine can trap a part when the indexer and the robot disagree; mitigation is a position teach check on each pocket at commissioning and a photo on the magazine, not on the pockets alone.
10. Actions and owners
| ID | Action | Owner | Deadline |
|---|---|---|---|
| ACT-1 | Replace gravity chute with powered roller bed, 1.5 degree slope | K. Lindqvist | 24 August |
| ACT-2 | Add datumed three-point nest and hard stop to both gripper jaws | R. Hoffmann | 26 August |
| ACT-3 | Add height probe beside the photoelectric seated check | P. Fontaine | 27 August |
| ACT-4 | Implement fault recovery ladder with counter preservation | P. Fontaine | 29 August |
| ACT-5 | Fit ten-pocket buffer magazine at the out-feed | K. Lindqvist | 31 August |
11. Decision summary
Feed the pick cradle with a powered roller bed at 1.5 degrees, part supported over two rollers throughout. Set the raw part against a machined datum nest and hard stop before the robot leaves the cradle. Verify seating with the photoelectric beam plus a height probe before clamping. Recover from a dropped part through retry, then reposition, then an operator call, preserving the cycle counters. Buffer finished parts in a ten-pocket magazine so the cell never waits on the conveyor.
12. Post-meeting addendum, 9 September
The rebuilt load station ran 3,400 parts over two shifts with nine jams, one in 378 parts instead of one in 29, and all nine cleared through the recovery ladder without an operator hand. The 42 second cycle held at 42.6 seconds average, inside the check against the 3 second slack. The magazine caught the conveyor block twice and the robot never waited, and the fixture mouth inspected clean on 40 consecutive parts at the 0.05 mm insert-target the review had set.
Glossary
Load and unload mechanism: the station that removes a finished part and feeds a raw part into a machining cell. Utilization: the share of robot time actually spent moving and gripping. Roller bed: a powered conveyor of small rollers carrying parts under controlled forward slope. Indexer: a rotary table that steps through fixed positions to present parts. Recovery ladder: an ordered set of automated retry steps before an operator intervenes.
Appendix. The timing sheet and the mechanism bill the review signed
Robot time per load visit: travel 3.8, unload pick 2.6, finished place 1.9, raw pick 3.1, insert 3.4, release 1.2, regains and handshake 2.0, total 18.0 seconds, across six stations giving robot utilization 78 percent at a 42.6 second average cell cycle. Feed bed parameters: powered rollers at 60 mm pitch on 1.5 degree slope, optical full-stop at the cradle, back-pressure stop upstream, alignment rail over the last 400 mm. Gripper: double-jaw with a three-point nest and hard stop per jaw, hardened nest insert, 0.05 mm insert drift target. Seating check: photoelectric beam plus height probe, sampled simultaneously, 0.3 seconds added per station. Fault recovery: retry, then reposition, then operator call, with PLC counters preserved across every recovery. Buffer: ten-pocket indexer at the out-feed, teach-checked at commissioning.
The chair’s closing remark was written in the minutes as agreed: an automatic load and unload mechanism is judged by what it does when the ninth part differs, not by how clean the first eight are, and the cell that absorbs its own faults is the cell that keeps its promise. The minutes were signed by all five attendees and issued as controlled drawing number LUL-2217, with the timing sheet and the mechanism bill as appendices.
