The Machine That Passed FAT But Failed on Day 3

A beverage packaging client in Guangzhou conducted their factory acceptance test (FAT) at our facility last September. The machine — a ZCF-800 configured for 350 g/m² beverage cartons — ran a 2-hour test with 98.3% first-pass yield at 55 ppm. The customer signed off. We crated it, shipped it 400 km, installed it in their plant, and commissioned it.

Three days into production, their quality manager called: scrap rate was 7.2%, carton side seams were inconsistent, and the feeder was jamming roughly every 18 minutes. The machine that had sailed through FAT was limping through real production.

What changed? Three things: their board was 380 g/m² instead of 350 (a “minor” material substitution by purchasing), their factory humidity was 78% versus our conditioned 50%, and — critically — the FAT test had been run with our technician adjusting parameters between runs, masking the machine’s sensitivity to setup drift.

FAT is not a demonstration. It’s a forensic examination. Here are the indexes that separate a real acceptance test from a performance done for your benefit.

Index 1: Sustained Throughput (Not Peak Speed)

The most common mistake: accepting a 10-minute speed burst as proof of capacity. A real throughput test requires a minimum 4-hour continuous run — no pauses, no parameter adjustments, no technician interventions. If the machine stops, the clock keeps running. If a carton jams, it counts as a defect.

What to measure. Net good output per hour. Take the total number of cartons that pass visual and dimensional inspection after the 4-hour run, divide by 4. This is your sustained throughput.

Acceptance criteria. Sustained throughput must be at least 85% of nameplate speed. For a machine rated at 50 ppm: 50 × 60 × 0.85 = 2,550 cartons per hour minimum. Anything below 80% is a functional failure regardless of what the spec sheet says.

How to do it. Run YOUR material — not the supplier’s demo stock — at YOUR specified board thickness and grammage. 4 hours. No exceptions. If the supplier says “our other customers accept a 1-hour test,” politely decline and mention that a machine represents a 5–10 year investment and you expect a 4-hour test to validate it.

Index 2: First-Pass Yield (Unassisted)

FPY measures the percentage of cartons that exit the machine meeting all quality specifications without any manual rework, adjustment, or sorting.

Defect categories to track. Define these BEFORE the test, not during:

• Side seam gap. Maximum acceptable: 0.5 mm. Measure with a feeler gauge on 50 random cartons per hour. Any carton exceeding 0.5 mm on any point of the seam is a defect.
• Flap alignment. Both side flaps must close within 1.5 mm of the centerline. Measure from the flap edge to a scored reference line.
• Glue squeeze-out. No visible adhesive on the exterior surface when viewed from 30 cm under standard factory lighting (500 lux).
• Board damage. No scratches, creases, or dents that penetrate the printed surface. Surface scuffing that removes ink or coating is a defect.
• Dimensional accuracy. Carton length and width must be within ±0.5 mm of the specified dimension after forming.

Acceptance criteria. FPY must exceed 96% at sustained throughput. A 4% defect rate on a 50 ppm machine means 2 defective cartons every minute, 120 per hour, 960 per 8-hour shift. That’s a full pallet of scrap every shift.

How to do it. Designate one person as the quality inspector for the duration of the test. They do nothing else. Every 30 minutes, they pull 20 consecutive cartons from the discharge conveyor and perform the full quality check. That’s 160 cartons inspected over 4 hours. Record every defect by category. Calculate FPY as (good cartons / 160) × 100.

Index 3: Mean Time Between Stoppages (MTBS)

Some stoppages are unavoidable — magazine reloads, glue tank refills. Those are planned. What matters is unscheduled stoppages: blank jams, sensor faults, glue clogs, vacuum failures, mis-feeds.

Acceptance criteria. Over 4 hours, there should be zero unscheduled stoppages where the root cause is the machine (as opposed to defective incoming material). In practice, 1–2 material-related stops is acceptable if the incoming board has known quality issues — but the machine must auto-detect the issue (sensor alarm) rather than crashing mechanically.

If the machine stops more than 3 times in 4 hours for machine-related reasons, the FAT has failed. Do not accept “we just need to adjust the sensor sensitivity” as an excuse — if it’s that easy, adjust it now and restart the 4-hour clock.

How to do it. Assign one person to log every stop. Timestamp, duration, root cause category. During the test, the supplier’s technician can only intervene if the machine has been stopped for more than 5 minutes and the operator cannot clear the fault. Every technician intervention counts as a machine-related stop and resets the “zero unscheduled stops” counter.

Index 4: Glue Bond Strength

Visual inspection catches squeeze-out and missing beads. It does NOT catch weak bonds that will fail in the customer’s hands, during transport, or on a retail shelf after 3 weeks.

How to measure. Perform a lap-shear test on 10 random cartons pulled from the middle of the 4-hour run (not the beginning, when glue temperature is most stable, or the end, when the tank might be low). Cut a 25 mm × 25 mm section centered on the side seam. Mount it in a tensile tester with the seam in the center of the gauge length. Pull at 50 mm/min until failure.

Acceptance criteria. Bond strength must exceed 80% of the board’s internal fiber-tear strength. If the board itself tears before the glue bond fails, that’s a 100% fiber-tear bond — excellent. If the glue peels cleanly from the board surface, the bond is inadequate. Minimum acceptable: 65% fiber tear across all 10 samples, with no single sample below 50%.

If the supplier’s facility doesn’t have a tensile tester, bring a portable one. A basic digital force gauge with a manual test stand costs $400–$600. For an $80,000+ machine investment, bringing your own test equipment is not unreasonable.

Index 5: Changeover Accuracy

If your operation requires format changes, the FAT must include at least two complete changeovers from Format A to Format B and back to Format A.

What to measure.

• Changeover time. From last good carton of Format A to first good carton of Format B. The supplier’s quoted changeover time is only valid if it includes the adjustment AND the quality verification run.
• First-carton accuracy after changeover. The first carton of Format B after changeover must meet all dimensional specs. If it doesn’t, the changeover procedure is incomplete — it relies on operator trial-and-error rather than stored recipes or indexed mechanical stops.

Acceptance criteria. Changeover time within 120% of quoted spec (e.g., if quoted at 10 minutes, 12 minutes is acceptable; 18 minutes is not). First-carton accuracy: dimensions within spec on the first carton; glue bead position within 0.5 mm of target on the third carton (first two are allowed for thermal stabilization after nozzle adjustment).

Index 6: Safety System Verification

This is non-negotiable and must be tested with the machine under full power.

Tests to perform.
1. Open each interlocked guard door during operation. The machine must stop within the specified stopping time — typically under 200 ms for a machine running at 50+ ppm.
2. Press each emergency stop button during operation. All motion must cease immediately, and the machine must NOT restart when the E-stop is released — it must require a deliberate reset.
3. Verify that the safety PLC (if equipped) is an independent controller, not a software routine on the main motion PLC. Ask to see the safety circuit diagram and the SIL/PL rating of the safety components.
4. Test the light curtain (if equipped) by passing a test object through the protected zone at the machine’s maximum speed. Response time must be fast enough that the dangerous motion stops before the object reaches the hazard point. Calculate this: stopping time (ms) × approach speed (2,000 mm/s per ISO 13855) = minimum safety distance. If the light curtain is closer than this distance, it’s installed wrong.

The Acceptance Test Checklist

Print this and bring it to FAT:

• [ ] 4-hour continuous run on YOUR material — confirmed
• [ ] Sustained throughput ≥ 85% of nameplate speed — confirmed
• [ ] FPY ≥ 96% on 160-carton sample — confirmed
• [ ] Zero machine-caused unscheduled stops in 4 hours — confirmed
• [ ] Glue bond strength ≥ 65% fiber tear on all 10 samples — confirmed
• [ ] 2 format changeovers completed within 120% of quoted time — confirmed
• [ ] First carton after changeover within dimensional spec — confirmed
• [ ] All safety interlocks tested and stopping time verified — confirmed
• [ ] Noise level under 85 dBA at operator position (use a sound level meter — ear protection is required above 85 dBA) — confirmed
• [ ] All documentation delivered: electrical schematics, pneumatic diagram, PLC program backup, spare parts list with part numbers, maintenance schedule — confirmed

Do not sign the acceptance certificate until every box is checked. Once you sign, the machine is yours — mechanical issues become warranty claims with response times, freight costs, and production losses. A rigorous FAT is your best leverage to get everything right before the machine leaves the supplier’s floor.

The Guangzhou beverage client eventually resolved their issues — the board substitution was the root cause, not the machine — but their quality manager now brings his own board stock to FAT, runs the full 4 hours, and doesn’t let the supplier’s technician touch the controls during the test. He hasn’t had a post-installation surprise since.