Plus or Minus Eight Grams: The Fill Tolerance That Decides Reject Rate

The spec said 800 g ±8 g. The checkweigher was set to reject outside that window. In the first month it rejected 0.4% of packs and everyone was happy. In the third month it rejected 3.8%, the same product, the same machine, the same operators, and the quality manager was asked to explain a problem that had not changed.

Fill weight tolerance is one of those specifications that looks like a number and behaves like a system. It interacts with the pump, the gel viscosity, the freeze expansion, and the headspace, and a change in any one of them shows up at the checkweigher as if the filler had drifted.

What the tolerance is really protecting

Two things, and they pull in opposite directions.

Under-fill loses cooling capacity. A pack at 780 g instead of 800 g has 2.5% less thermal mass and, more importantly, more headspace, which means the pack is a worse conductor and the gel can slosh. In a 48-hour lane that 2.5% can be the difference between 7.6 °C and 8.2 °C at hour 44.

Over-fill costs money and creates a burst risk. Every extra gram is polymer, water, film stress and freezer energy. At 810 g average on a million packs, you are giving away 10 tonnes of product a year, and the packs sit fuller against the seal, which raises the failure rate at high temperature.

A ±8 g window on 800 g is ±1%, which is achievable on a good piston filler and difficult on a gravity filler with a variable-viscosity gel. That is the first thing to check: is the tolerance matched to the machine, or was it copied from another product?

Why the reject rate moved without the filler moving

Three mechanisms, all common.

Viscosity drift. Gel viscosity changes with temperature and with mix time. A batch mixed at 19 °C is thicker than one at 26 °C, and a volumetric filler delivers fewer grams of thicker gel for the same stroke. The machine has not changed; the material has.

Freeze expansion. Water expands about 9% on freezing. A pack filled to 800 g at 20 °C and then frozen is under internal pressure, and if it is weighed after freezing the reading is the same but the pack is tighter. If you checkweigh after the freezer, you are measuring a different object than the one you filled.

Headspace and squeeze. Packs are often squeezed or pressed during packing, which changes the fill level in the pouch without changing the mass. A pack that reads 795 g may be correctly filled but badly distributed, with gel pushed against one seal.

The numbers that make it manageable

Set the target in the middle of the window, not at the nominal weight. If the window is 792–808 g, target 800 g and tune the filler to a standard deviation of 2.5 g. That puts 99.7% of packs inside the window at three sigma, which is a realistic capability for a piston filler.

• Measure 30 consecutive packs per shift, not 5.
• Calculate the mean and the standard deviation, and plot both.
• Recalibrate when the mean moves 3 g, not when the checkweigher starts rejecting.
• Record the gel temperature at fill, because it explains most of the drift.
• Checkweigh before freezing, and inspect after freezing for headspace, not weight.

A filler running at 1.2 g standard deviation is capable of ±5 g. One at 4 g cannot hold ±8 g without constant rejects, no matter how carefully the operator watches it. The number to fix is the standard deviation, not the reject rate.

The counter-intuitive finding

A plant tightened its tolerance from ±10 g to ±5 g to reduce complaints, and complaints went up. The tighter window forced the operators to over-fill, because the filler’s natural variation was 4 g and the only way to stay inside ±5 g was to run the mean at 803 g. Over-filled packs bulged, the pallet stack became unstable, and cartons started crushing at the corners.

The right answer was to keep ±10 g but reduce the standard deviation to 2.5 g, so the mean could sit at 800 g with margin on both sides. Tolerance without capability is just a way of generating rejects.

What to write on the batch sheet

Target weight, window, machine standard deviation from the last capability study, gel temperature at fill, and the mean of the 30-pack check. Five fields. With them, a shift that starts rejecting can be diagnosed in ten minutes instead of argued about for a week.

Before the next run, take thirty packs off the line and weigh them. If the standard deviation is above 3 g, the tolerance you have written is not the tolerance you have. SAP dosing for gel packs includes dose-per-pack tables that keep the polymer fraction correct when the fill weight is adjusted.