Gravimetric dosing of minor additives — what went in and how much, signed in the moment.
Minor additives — dispersant, defoamer, biocide, drier, wetting agent — are grams that decide the behavior of hundreds of kilos. Today they are weighed correctly on the scale… and written down later, by hand, on the batch record. A swapped additive or an off-formula amount is invisible to the scale: it shows up at final QC or at the customer. iLEAN Connect captures every weighing from the scale in real time, Edge identifies the additive and the operator, and the agent checks against the formula at second zero. The record writes itself — and the person signs.
The scale weighs right; the record is filled in later — and that is where the gap opens.
The error is almost never in the weighing; it is in everything around it:
- Correct weighing, deferred record — the batch record is filled in at the end of the task or the shift, from memory or by copying the theoretical formula. The record says what should have gone in, not what did.
- A swapped additive — two white drums, similar labels: defoamer A instead of B is invisible to the scale, because the weight is plausible. It surfaces as foam, craters or a biocide gap weeks later.
- The order of addition is recorded nowhere — some additives go in before dispersion, some after, some diluted first. The real sequence, which changes the result, is not captured anywhere.
- Off-formula is found out late — at final QC if you are lucky, at the customer if you are not. And with no real dosing record, the investigation starts blind: nobody can say what actually went into that batch.
It is not an operator problem — it is a capture problem. A batch record copied from the formula is a wish; a record captured from the scale is a fact.
iLEAN does not replace your scale — it captures it, checks the formula and writes the real record.
Dosing does not fail for lack of a scale; it fails because the data of each weighing is neither tied to the batch nor checked against the formula in the moment. iLEAN acts as the putty between the scale, the formula and the batch record, without changing how people work.
Edge identifies the drum and the batch. Connect captures the weighing from the scale. The agent checks against the formula at second zero — and the record writes itself.
The iLEAN pieces applied to additive dosing:
- Connect — hooks into the scale (or the tank load cells) and captures every weighing in real time: value, time, stabilization. Analog dial? The operator photographs it and Connect reads it.
- Edge — the operator scans the QR on the additive drum and on the batch. If the drum is not the additive the formula calls for, the warning fires before it goes in, not at final QC. It works without a network.
- Dosing agent — checks every weighing against the formula: right additive, amount within tolerance, right order of addition. Off-tolerance means a warning at second zero; a missing addition holds the batch's next step until the supervisor validates. The batch record composes itself with what was actually dosed — and the person signs it.
Hand-written batch record vs. dosing captured with iLEAN
| Aspect | Scale + hand-filled record | With iLEAN Connect + Edge (weighing by weighing) |
|---|---|---|
| Dosing record | By hand, at the end of the shift | Captured from the scale in real time |
| Additive identity | Trust in the label | Drum QR checked against the formula |
| Off-tolerance amount | Found at final QC | Operator warned at second zero |
| Order of addition | Not recorded | Real sequence, with time, per batch |
| Batch record | Says the theoretical formula | Says what was actually dosed, signed |
| Deviation investigation | Blind, weeks later | Full per-batch trace in one click |
Impact estimate for your plant — to be validated with your numbers.
The block below is an estimate to be validated with the specific data of your plant. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.
- Pastes, dispersions, coatings or chemical preparations plant with manual dosing of minor additives on scales.
- Pilot: Connect on one scale + Edge at the additive station. First value expected within a few weeks.
- Indicative payback between 4 and 9 months, depending on the cost of an off-formula batch (rework, downgrade to second quality, complaint) and the hours now spent on records.
- Hard levers: ≥ 50% fewer undetected dosing deviations (estimate to validate); batch records with zero retyping hours; complaint investigations that start from facts, not guesses.
And the quality manager's reasonable doubt
"What if the scale sends a bad value, or a photo is misread?" — it happens, and that is why the system never blocks or approves alone. Hallucination is a problem of free generation, not of anchored tasks: where the AI merely receives a weight and compares it against a tolerance, the best models brought error below 1.5% [1]. Every capture stays reviewable, and both the exception and the batch are signed by a person — never by the system.
[1] OpenAI paper “Why Language Models Hallucinate”, 2025 — on the reliability of AI in anchored tasks.
What people ask about gravimetric additive dosing
Why do minor additives concentrate so much risk if they are only grams per batch?
Because their effect is wildly out of proportion to their weight. A dispersant stabilizes the whole pigment charge; a missing defoamer means craters in the film; a short biocide dose means a spoiled batch in storage; an overdosed drier wrecks drying behavior. They are the smallest lines of the formula and the ones a hand-written record captures worst — precisely because "it is just a splash of additive."
Why does the classic hand-filled batch record fail?
Because it is written after the fact. Filled in at the end of the task or shift, it records what the formula says, not what happened at the scale: a swapped drum, an amount at the edge of tolerance, an inverted order of addition — none of it reaches the paper. The record is not false out of malice; it is false by construction. When final QC finds foam or sedimentation, the investigation starts from a document that cannot answer the only question that matters: what actually went in?
How does iLEAN prevent the wrong additive from being dosed?
With identification before dosing: the operator scans the QR (or the label — Edge can read the printed text of drums without codes) on the additive drum and on the batch. The agent crosses the pair against the formula: if the drum is not the additive that batch step calls for, the warning fires before anything is poured, on the terminal or the earpiece. The plausible-weight trap — the scale cannot tell defoamer A from B — disappears, because identity is checked independently of weight.
What happens if a weighing falls outside the formula's tolerance?
The agent flags it at second zero: "dispersant, 412 g weighed, tolerance 380–400 g — adjust." The operator corrects on the spot, while adjusting is still trivial. If the deviation is confirmed — for instance, a deliberate correction ordered by the technical manager — it is recorded as a validated exception with his signature. And if an addition is missing when the batch moves on, the agent holds the next step until the supervisor decides. The person signs — the batch never advances on its own.
What does the plant gain beyond catching errors?
Three things: (1) a real batch record — what was dosed, when, by whom, in what order — composed automatically, with zero retyping hours; (2) investigations with facts — when a complaint arrives, the real dosing of that batch is one click away, and the difference between "we think" and "we know" is days of work; (3) the real order of addition captured, which turns the veteran's way of building a batch into a pattern the plant keeps. All as an estimate to validate with your numbers.
More cases in other sectors
- Pigment dispersion in a bead mill — stop at the right fineness, without over-milling.How iLEAN ties every Hegman gauge reading to its mill pass and its batch, and calls it when the batch re…
- Particle size of the milled batch — a distribution correlated with application, not a loose sample.How iLEAN captures the full particle size distribution from the analyzer, ties it to the batch and corre…
- Accelerated stability of the pigment batch — evidence tied to the batch, not a loose sheet.How iLEAN identifies every stability specimen, captures each checkpoint reading in the moment and compos…
- Void detection in XLPE insulation with AI — a dielectric failure is not a warning, it is a breakdown.How iLEAN Vision detects voids and bubbles in XLPE insulation as it leaves the extruder, before curing,…
- UV varnish defect detection on lacquered doors — a crater spotted after the UV tunnel is another full booth cycle.iLEAN Edge detects UV varnish defects (craters, dust, sags, low film build) on lacquered doors in line,…
- EN ISO 20471 high-visibility garment control with AI — one centimeter of band decides whether the garment is Class 2 or not.How iLEAN Vision controls retroreflective tape, fluorescent color and band position on EN ISO 20471…
Tell us your case and in 48h we'll send you the estimated ROI for your dosing station.
We work on your plant's real data, not ours. Diagnostic with no commitment.
Request estimated ROI in 48h See specialty chemicals