Night shift ran a high-speed pouch line after a “temporary” air isolation on the reject cylinder. The checkweigher still printed grams. Packs under target kept walking to the case packer. Morning QA found the reject bin empty and the mean fill 4 g light.
A checkweigher owns three jobs: weigh, decide, and remove. Calibration of the first does not certify the third.
What the weight display cannot prove
Dynamic vs static. A certified 100 g weight on a stopped belt is not the same as a pack at line speed with vibration and product shift. Dynamic verification with known packs belongs on the traveler—not only the annual metrology sticker.
Reject path. Pusher timing, air pressure, photo-eye confirmation, and a bin that actually receives product. If the HMI shows “reject” and the bin stays empty, the PLC lied or the actuator failed.
Zone honesty. Multi-lane or multi-head setups need per-lane stats. Averaging good lanes with a dead lane hides underfills.
Plant rules that hold
- Prove reject with a marked underweight pack every shift start and after any air or PLC change.
- Log reject counts against expected statistical rejects; zero rejects on a noisy fill process is a red flag, not a win.
- Lock out “reject bypass” as a controlled MoC—not a convenience toggle.
This is not metal-detector tramp control, not AI reject-cost thresholds, and not a scale in the QA lab. It is in-line net-content enforcement that fails when the gate is silent.
Ship only what the reject gate can still remove.