The press control knows the hit, the weld controller knows the spot, the peel test knows the nugget, and the OEM's scorecard knows the ppm. Today those four live in four places and meet in a meeting. Pragmatos puts them on one read-only record, time-stamped by one clock, so the die change, the suspect window, the weld finding and the performance-test count are settled from readings.
Press controls keep the last few hundred events on the HMI. Weld controllers keep per-spot readings in a ring buffer that is overwritten in days. Peel and chisel results are on a shift sheet. Robot alarms are on the pendant. The launch file is a spreadsheet assembled the week before the OEM visit.
A read-only edge in each shop reads what the controls already expose. One site server holds the record. Pragmatos Live shows it line by line and cell by cell, Pulse shows the trend per press and per cell, Orbis proposes a hold or a check and a named person approves it, Decisions keeps the chain. Nothing writes back to a press, a robot or a weld controller.
It does not start a press, move a robot, change a weld schedule or dress a cap. It does not replace the peel test, the OEM's audit or the launch team. It makes no forecast of die life or electrode life. The ppm on the OEM's scorecard stays the OEM's figure; this record shows the plant's side of it.
Twelve weeks, sheet 6. On the plant's own servers, or on infrastructure the plant operates. Open-source core; the record exports in full.
| Station | Equipment, lettered | What it records today | Who reads it |
|---|---|---|---|
| Press line 1 | Transfer press, maker A, press control H; die ID read by the control at die set-down | Last few hundred events on the HMI; strokes per minute, slide position, cushion pressure, die ID; overwritten | Setter, at the press |
| Press line 2 | Tandem line of five presses, maker B, control I; crossbar transfer | Same, per press; no shared clock between the five | Setter, at each press |
| Weld cells 1–10 | Robots C, D and E; resistance-spot weld controllers F (cells 1–6) and G (7–10); cell PLC | Per-spot current, voltage, time and electrode force in a ring buffer; cap-dress counter; robot alarms on the pendant | Cell maintenance, when called |
| Weld cells 11–12 | Robots D; laser source J; seam tracking on the head | Laser power, feed, seam position per weld; a text log on the source | Process engineer, weekly |
| Test bench | Peel and chisel bench per shop and shift; coupon per cell per shift | Nugget diameter and failure mode on a shift sheet, per ISO 10447 | Quality technician |
| Launch | One part family in production part approval; the performance test under series conditions is the gate | Count, cycle time and first-pass yield on a tally sheet; the OEM witnesses | Launch manager |
| Scorecard | Monthly from the OEM: ppm, delivery, incidents, a rating per line | The plant's rebuttal is assembled from the four records above, after the fact | Quality manager |
Three shifts, about 230 production days. The press shop and the body shop each have their own supervisor; the launch team sits between them and borrows line time from both. Every quantity on the OEM's scorecard is produced on this floor, and no two of the records that explain it share a clock.
Three in four European suppliers report a profitability below five per cent; one in four reports a loss; three in four say their portfolio is under review. S1
Employment in the German automotive industry fell 5.8 % in a year to 691,500; at suppliers it fell 7.6 % to 219,500, the lowest since 2005. S2
A fifth of Germany's workforce is 55 or older. The plant's setters and weld technicians are in that fifth more often than its planners are. S2
Cost-down letters arrive yearly. The plant's answer has to come from its own floor: minutes on the press, parts not sorted, hours not reworked, launches not re-run. Each of those is a reading before it is a euro.
R existing interface
S survey confirms a path first
H entered by hand
| Category | What the OEM sees | What the plant knows, and cannot show | Rating | Src |
|---|---|---|---|---|
| Quality at the dock | A ppm figure per month, against a target in the supplier manual. A rating per line. | Which lots went out on which date. Not which shift, which die set or which cap-dress interval made them; the controller memory is gone by the time the complaint arrives. | S7 | |
| Containment | Controlled shipping invoked: 100 % inspection by the plant's own people, then, at the next level, by a third party the OEM chooses and the plant pays. | The defect is real and the window is not: without a time-stamped record of the cushion pressure or the weld current, the suspect window is "since the last good audit", which is the whole fortnight. | S7 | |
| Launch readiness | The performance test is a production run under series conditions, with the series equipment, tooling, people and cycle time, re-run after a failure once the cause is removed. | The line made rate on the second shift and the tally sheet says it did not. The re-run costs a day of line time and a second OEM visit; the dispute is about the count, not the capacity. | S3 | |
| Die change | Nothing, until a delivery slips. Then: a corrective-action request. | Median 42 minutes from last part to first good part, fastest crew 31. The spread is the record the plant does not have: which crew did the external work before the stop and which did it during. | S5 | |
| Weld quality | Audit findings cluster on five clauses: problem solving, control plan, nonconformity and corrective action, and monitoring and measuring. An oversight-body ranking puts them at the top every year. | A peel coupon per cell per shift gives nugget size and failure mode for one spot in thousands. Between coupons the only record is the controller's own per-spot reading, and it is overwritten. | S4 S6 | |
| Cost-down | A yearly letter asking for a percentage. | Which line, which crew and which hour the percentage is hiding in. The plant answers with an average because an average is what the records allow. | S1 |
Every line on the scorecard is argued from a record the plant assembles after the fact, from four memories that do not share a clock and two that are overwritten within days. The plant is not short of readings. It is short of a record that was written at the time, by the equipment, and can be shown to the OEM's auditor without a spreadsheet in between.
That is the whole of what this reference adds: the record. The readings exist; the press control, the weld controller, the laser source and the robot controller write them already. The peel bench and the launch tally are entered by hand, against the same clock, and stay beside the readings they confirm or contradict.
Sources S1 to S7 are listed on sheet 7.
Each line is a category on an OEM scorecard. The middle column is what the plant could put in its rebuttal if it had the record. The rating cell is empty because the rating belongs to the OEM and the plant's own numbers go in a plant's own copy.
No ppm target. The targets live in each OEM's supplier manual and differ by part and by year; a figure here would be invented.
No defect rate from the standards. ISO 10447 describes a test, VDA 2 describes an approval; neither says how often a plant fails them.
PPA production process and product approval (VDA 2), the German-market counterpart of PPAP. Performance test the run at series rate the OEM witnesses; often called run@rate. Controlled shipping OEM-imposed 100 % inspection, level 1 by the plant, level 2 by a third party. Peel test destructive check of a spot weld's nugget.
The edge has no write path. Pragmatos does not start a press, change a weld schedule, move a robot, set a force or dress a cap. It does not decide the hold; the shift leader does. It does not replace the coupon; the coupon decides the nugget. The edge keeps about 72 hours of readings if the link to the site server drops, and a reading older than three poll cycles is shown as out of contact, not as a value.
Press control: stroke count, slide position, cushion pressure, die ID, state. R
Weld controller: per-spot current, voltage, time, force; cap-dress counter. S
Robot: state, program, alarm list. RS
Laser source: power, feed, seam position. S
Every reading with the clock, the die or program, the shift and the lot it fell in; every proposal, approval and piece of work against the same clock. Retention set per record class; the plant chooses.
Lines and cells now; the trend per press and per cell; the open holds and checks; the launch file as it stands; the rebuttal as a list of records, not a narrative.
Nine roles. The installer who connects a press control is never the person who approves the connection. Figures about people are shown for groups of five or more.
On the plant's own servers, or on infrastructure the plant operates. The OEM's auditor sees the record on the plant's screen, not on a vendor's.
The press control writes the die ID at set-down and the first good stroke after it; the record between the two is the changeover. Pragmatos Live shows it as a timer at the line, Pulse shows the fortnight as this figure, and the press-shop supervisor sees for the first time that the fastest crew is not faster at the work, it starts the work earlier. The figure is the input to C-07 on sheet 7: a third of the spread held, not all of it, because a crew on a bad night will still do the external work late.
| On the scorecard | Today | With one record |
|---|---|---|
| Die change | A median from memory; the spread is an argument between crews. | Every change timed by the press control, split internal and external, by crew; the external share is coached from the figure, not the anecdote. |
| Suspect window | "Since the last good audit": the fortnight, sorted at the plant's cost. | From the first out-of-limit reading to the approved hold: cell, robot, spots, assemblies, on the record the OEM's engineer can read. |
| Press drift | Cushion pressure drift found at the end-of-line audit; the run is scrapped back to the last audit. | Pulse shows the drift against its limits as it starts; Orbis proposes a stop and a check; the setter approves; the scrap is the parts since onset. |
| Weld rework | A missed cap dress is found at the next coupon; the shift's assemblies are re-welded by hand. | The controller's per-spot reading raises the fault at the first bad spot; the hold is 38 assemblies, the coupon confirms, the rest of the shift stands. |
| Performance test | A tally sheet and a stopwatch; the OEM's witness has their own. | Count, cycle time and stop causes from the press control and the cell PLCs, on one clock, printed from Decisions for the launch file; the dispute moves from the count to the cause. |
| Audit | A week assembling evidence for 10.2.3 and 8.5.1.1 from four systems and a shift sheet. | The chain reading → proposal → approval → work → result is the evidence; the coupon is on it; the auditor reads it on the plant's screen. |
Crew colours: A primer red, B zinc grey, C amber. Internal work solid; external-done-late hatched. The dashed line is the plant median, the dotted line the fastest crew's median.
Not a measured plant. The bars are modelled from the inputs on sheet 7 for a line of this kind; a plant's own fortnight replaces them in the first week of the press-shop phase.
Not a forecast. Nothing on this sheet says what the next change will take.
Of the five results, the die-change minutes are the one the plant can check against its own press control within a month, without a complaint, a launch or an audit to wait for. It is the figure to argue with first.
The cell figure is on the next sheet as a dial and is named the weakest of the five: it rests on one assumed OEE point and on the plant acting on the stop causes the record shows.
| Phase | Line running | Line stopped | Who |
|---|---|---|---|
| Survey, weeks 1–2 | Walk every control with the plant's own list: which interface, which tag, which clock. The S badges on sheet 4 turn into R or H. | None. | Plant maintenance lead, Crius engineer |
| Press shop, weeks 3–6 | Edge installed and read-only connection to control H approved by the press-shop supervisor; die-change timer live; first fortnight figure. | One changeover window on line 1 to confirm the OPC UA read and the die ID. | Press-shop supervisor approves; setter crews read |
| Body shop, weeks 5–9 | Edge installed; cells 7–10 (controller G) read through the cell PLC; Pulse trend per cell; Orbis rules written with the weld engineer; first coupon entered against the record. | One stop per cell, inside a scheduled cap change, to confirm the per-spot read. | Body-shop supervisor approves; weld engineer writes the rules |
| All cells, launch, weeks 9–12 | Cells 1–6 and the two laser cells; the performance-test count read from the press control and the cell PLCs; the launch file printed from Decisions; roles set for the OEM's auditor's screen. | One Saturday morning for the two laser cells. | Launch manager, quality manager |
Each dial is scaled from zero. The grey arc is today's reading; the red segment is the modelled movement. The euro figure below each dial is one operating year and comes from sheet 7. The five do not overlap: press lines and weld cells are different equipment, internal scrap and OEM-side containment are different episodes, and launch re-runs are outside series production. MODELLED
Open-source core under Apache-2.0; the plant's record exports in full, per company, in open formats. Monthly releases, each one supported for twelve months, installable as an offline bundle with rollback and a software bill of materials.
Retention per record class; the plant chooses. For a part on an OEM programme that is commonly the life of the part plus one year, which the record can hold.
Not the adapters; a weld-controller read is a week's work for anyone. The moat is the record itself after a year: ten thousand die changes and a hundred million spots, each with its shift, die, crew and coupon beside it, on one clock, in a form the plant owns and can take elsewhere. The OEM's engineer learns to ask for it by name.
The ppm on the OEM's scorecard; energy per hit; electrode-cap life; the quality manager's hours on rebuttals; audit preparation hours; Saturday shifts. Each is on the record; none is in the total.
| Row | Input | Value | Kind |
|---|---|---|---|
| C-01 | Production days a year | 230 | ASSUMED |
| C-02 | Press lines (one transfer line, one tandem line) | 2 | ASSUMED |
| C-03 | Die changes a line a day, all shifts | 4.5 | ASSUMED |
| C-04 | Die changes a year = C-02 x C-03 x C-01 | 2,070 | DERIVED |
| C-05 | Die-change minutes, last part to first good part, median crew today | 42 min | ASSUMED |
| C-06 | The same, fastest crew today | 31 min | ASSUMED |
| C-07 | Share of the crew spread held once every change is on one record | 0.33 | ASSUMED |
| C-08 | Minutes back per change = (C-05 - C-06) x C-07 | 3.7 min | DERIVED |
| C-09 | Press hours back a year = C-04 x C-08 / 60 | 128 h | DERIVED |
| C-10 | Marginal cost of a press-line hour: crew overtime, energy, maintenance (not the absorbed rate) | €900 | ASSUMED |
| C-11 | Containment episodes a year: OEM-side sort or controlled shipping | 3 | ASSUMED |
| C-12 | Cost of one episode: third-party sort, 100 % inspection, expedited freight | €35,000 | ASSUMED |
| C-13 | Episodes where the suspect window can be bounded by the readings | 2 | ASSUMED |
| C-14 | Share of the sort that falls away when the window is bounded | 0.6 | ASSUMED |
| C-15 | Press runs a year stopped for a parameter drift: cushion pressure, slide position, blank lubrication | 18 | ASSUMED |
| C-16 | Parts scrapped per such run, drift seen at the end-of-line audit, today | 340 | ASSUMED |
| C-17 | The same, drift seen at onset on the press readings | 90 | ASSUMED |
| C-18 | Cost of one scrapped part: blank net of scrap return, plus processing | €6.4 | ASSUMED |
| C-19 | Weld rework hours a year: manual re-weld after a missed cap dress or a stuck electrode | 1,650 h | ASSUMED |
| C-20 | Share that falls away when the controller readings raise the fault at the first bad spot | 0.15 | ASSUMED |
| C-21 | Loaded cost of a rework hour | €62 | ASSUMED |
| C-22 | Weld cells that pace the line (of 12) | 4 | ASSUMED |
| C-23 | Planned hours a cell a year, three shifts | 5,500 h | ASSUMED |
| C-24 | OEE of those cells today | 0.61 | ASSUMED |
| C-25 | OEE points back once every stop has a cause and a length | 1 pt | ASSUMED |
| C-26 | Cell hours back a year = C-22 x C-23 x C-25 / 100 | 220 h | DERIVED |
| C-27 | Marginal value of a pacing-cell hour: Saturday shift avoided | €220 | ASSUMED |
| C-28 | Part-family launches a year | 2 | ASSUMED |
| C-29 | Performance-test re-runs per launch today | 1.4 | ASSUMED |
| C-30 | Share of re-runs caused by a disputed or incomplete count, not a capacity shortfall | 0.4 | ASSUMED |
| C-31 | Cost of one re-run: eight hours of line time, launch team, OEM visit | €18,000 | ASSUMED |
| Result | Formula | A year | Half | Double |
|---|---|---|---|---|
| R1 | Die-change minutes held = C-09 × C-10 | €115,200 | €56,700 | €227,700 |
| R2 | Containment = C-13 × C-12 × C-14 | €42,000 | €21,000 | €84,000 |
| R3 | Scrap and rework = C-15 × (C-16 − C-17) × C-18 + C-19 × C-20 × C-21 | €44,145 | €22,072 | €88,290 |
| R4 | Weld-cell OEE = C-26 × C-27 (the weakest figure) | €48,400 | €24,200 | €96,800 |
| R5 | Performance-test re-runs = C-28 × C-29 × C-30 × C-31 | €20,160 | €10,080 | €40,320 |
| Total | five figures, no overlap MODELLED | €269,905 | €134,052 | €537,110 |
ASSUMED a stated input for a plant of this size; replace it with the plant's own figure. DERIVED arithmetic on the rows above. Half / Double the result with the movement input (C-07, C-14, C-17 and C-20 together, C-25, C-30) halved or doubled, the frame inputs unchanged; the double is capped where it would exceed what exists (C-11 episodes, one re-run per launch).
R4, weld-cell OEE. It rests on an assumed one point on four cells and on the plant acting on the stop causes the record shows; the record itself moves no OEE. Halve C-25 before quoting it. No figure on these sheets rests on a forecast of die, tool or electrode life; none is made.
R1 is press lines, R4 is weld cells: different equipment. R3 is internal scrap and rework; R2 is OEM-side containment: different episodes, and a lot that is scrapped inside is not sorted outside. R5 is launch, outside series production.
The ppm itself, energy per hit, cap life, quality-manager hours, audit preparation, Saturday shifts: shown on the record and left out of the total, because their value depends on contracts and rosters this sheet cannot see.
Frame inputs (lines, cells, shifts, days, launches) describe a plant of this size. Rate inputs (C-10, C-12, C-18, C-21, C-27, C-31) are marginal costs, not absorbed rates. Movement inputs (C-07, C-14, C-17, C-20, C-25, C-30) are the ones to argue about; the half and double columns show what the argument is worth.
The seven sheets above as an A4 reference, with the calculation sheet and the sources.