Allot Tech
Filling Accuracy Troubleshooting: Separate Bias, Spread and Supply Drift
Increasing the target dose can hide underfills while wasting product. First establish whether the measurement is reliable, whether the average is biased and whether individual doses scatter. Then compare the pattern by filling head, time and supply condition.
Confirm what the machine is controlling
An overflow filler controls a visual level. A volumetric filler meters volume, while a net-weight filler controls mass. Container internal dimensions and liquid density can make these results differ. Identify the contractual quantity and measurement method before diagnosing the machine. Use density measured for the actual product at the relevant temperature when converting mass and volume.
Validate the measurement before changing calibration
- Check scale resolution and repeatability for the tolerance being evaluated, using the site-approved verification procedure.
- Weigh empty containers individually where their mass variation matters; do not assume one average tare is adequate.
- Keep the same sampling, foam-settling and weighing procedure. Do not include leaked liquid or a changed closure weight unintentionally.
- Label every result with filling head, time, product temperature, tank level and recipe version.
Use the pattern to choose the investigation
| Observed pattern | First investigation | Avoid |
|---|---|---|
| One head consistently low | Its valve, hose, air ingress, dose setting and assembly | Recalibrating every head together |
| All heads drift as the tank empties | Product supply, suction conditions, temperature and air | Adding a fixed offset without a low-level trial |
| Wide spread around a reasonable mean | Variable feed, entrained air, unstable cutoff or measurement noise | Reporting only the average |
| Only first bottles after a stop are wrong | Priming, product drain-back and restart sequence | Discarding startup data from the report |
Example: a correct average can conceal defects
Illustrative data, not a machine performance claim: five measured doses of 498, 500, 502, 496 and 504 g average 500 g. With an agreed illustrative band of 498–502 g, two doses are outside it. Reporting “average 500 g” misses those failures. Keep average bias, minimum, maximum, spread and out-of-limit count separately. Actual limits and any statutory sampling requirements must be defined for the product and destination.
Sample across real operating conditions
As an initial diagnostic exercise, collect ten consecutive containers from each head, then expand according to risk and the acceptance agreement. Repeat at startup, stable running, low tank level, restart and after format change. Preserve baseline readings and change one factor per trial. Finish with a sustained run demonstrating both quantity conformity and the required good output; single-head bench accuracy alone does not verify line capacity.
Connect correction to the buying decision
Ask which product supply, fill range, head configuration, scale or meter and calibration procedure are included. Require evidence with actual liquid and packaging rather than a universal accuracy percentage. Record how recipes are controlled and what maintenance or cleaning can disturb the product path. Keep unresolved items and retest responsibility in the acceptance record.