Twenty Containers, One Curve: Holding Batch Spread Under Five Percent

Twenty containers of 20 tonnes arrived over six weeks from the same supplier, same grade, same purchase order. Absorption in 0.9 percent saline, measured on a composite sample from each container, ranged from 41 to 52 g/g. Every container was inside the certificate’s stated range of 40 to 55. The customer’s converting line was not, because it had been set up on the 52 end and could not run the 41 end without changing the compression ratio.

This is what batch consistency means in practice. Not “does it meet the spec” but “how far does it wander inside the spec”, and whether your process can absorb that wander without a changeover.

One container is usually one lot

A reactor batch of 20 to 30 tonnes often fills exactly one container, which means you cannot assume the twenty containers are twenty samples of one population. They are twenty populations. If the supplier blends into a silo before packing, the spread narrows and the lot number may cover several containers; if they pack straight from the dryer, each container carries its own history of neutralisation degree, crosslinker dose and drying curve.

The first question to ask is therefore not about standard deviation but about packing: “Does one lot number cover more than one container, and is there a blend silo between drying and packing?” The answer tells you how to sample.

Chart three numbers, not twelve

Free swell in distilled water is the least useful figure on the sheet, because it moves with everything and predicts little. Chart these instead: centrifuge retention capacity in 0.9 percent saline, absorption under load at 0.7 psi, and moisture at packing. The first two track the customer’s process; the third tracks storage behaviour and is the earliest warning of a dryer drifting.

The action limit matters more than the tolerance. A container that is 12 percent below the mean is a conversation; three consecutive lots drifting one direction is a process problem and worth a call before the fourth ships.

Where the drift comes from

Neutralisation degree is the big one. A 1 percent shift in how much caustic goes into the kettle moves capacity by 3 to 5 percent, and it is invisible on the certificate because the certificate is issued after the fact. Surface crosslinker age is the second: the same dose applied from a drum that has been open for three months behaves differently from a fresh one, and the effect shows up in AUL rather than in CRC.

The third is the weather at packing. Powder leaving the dryer at 60 degrees and packed into a bag in a monsoon-season warehouse picks up moisture in the bag, and that moisture reads as a capacity loss three weeks later when the customer tests.

The blend that hid the problem

When three low containers turned up, the supplier blended them with seventeen good ones and shipped a uniform 25-tonne lot. The customer’s incoming test passed, the line ran, and nobody learned anything. Six months later a different customer received an unblended shipment from the same production window and failed. Blending is a legitimate tool, but it hides the trend rather than fixing it, and it only works if the supplier tells you it happened.

For the next contract, ask for a per-container sample retained for twelve months and a control chart on CRC and AUL by lot. The polyacrylate granule grades page gives the nominal values to chart against. Five percent spread is achievable; twenty is a supplier that is not watching its dryer.