Most disputes about a new machine are not about whether it works, but about whether it works at the rate and quality that was promised. That argument is settled by two tests and one sampling plan, agreed before the order is placed.
Factory acceptance test (FAT)
The FAT happens at the supplier’s plant, before shipment, using material close to what will run in production. A useful FAT covers:
- Rated speed run of at least one hour with continuous output counting.
- Changeover from one size to another, timed, including tooling change.
- Quality sample pulled at the start, middle and end of the run.
- Safety functions: guards, interlocks, emergency stop, motor protection.
- Utilities: actual air consumption, power draw at rated speed.
Record the counter reading at start and end. A video of the run is worth keeping: it is the reference when the same machine behaves differently six months later.
Site acceptance test (SAT)
The SAT is not a repeat of the FAT. It confirms the machine works with the buyer’s material, the buyer’s operators and the buyer’s utilities:
- Voltage and air pressure measured at the machine, not at the wall.
- First-off sample approved by quality before the run continues.
- Ramp from 50% to 100% of rated speed, watching reject rate at each step.
- An eight-hour run at rated speed, which is when thermal drift and material variation show up.
How many pieces to inspect
Inspecting every piece for eight hours is not realistic, and inspecting ten pieces proves nothing. A workable plan for a converting machine:
| Stage | Sample | What is measured |
|---|---|---|
| First 100 pieces after start-up | 100% visual | Seal integrity, cut position, print registration |
| Steady running | 20 pieces per hour, random | Dimensions, seal strength, appearance |
| End of run | 20 pieces | Repeat of the first sample, to detect drift |
Classify defects before the test, not after: critical (unsafe or unusable), major (fails the customer’s specification), minor (cosmetic). Agree acceptance limits in numbers — for example zero critical, under 1% major, under 3% minor over the sampled lot.
Where the AQL idea fits
ISO 2859-1 sampling tables are designed for lot inspection of incoming goods, not for machine acceptance, and applying them literally to a production run usually produces an argument. The useful part is the discipline: define the lot, define the sample size, define the defect classes and the accept/reject numbers, and write them into the purchase order. The numbers themselves can be simpler than the standard.
Documents to leave with
- Signed FAT and SAT checklists with counter readings.
- Sample pieces, labelled with date, material and machine settings.
- Parameter sheet for each product size, so the run can be repeated.
- List of open items with an owner and a date.