Net weight and closure control in drum and jug filling — every container verified before it hits the pallet.

The problem

The two defects of drum filling travel in silence — and the customer discovers them.

A specialty chemicals filling line — resins, additives, inks, solvents, surfactants — fills 55-gallon drums, 5-gallon jugs and jerrycans of every format. The scale is usually there, and usually calibrated. And still:

1. The weight is seen, but never recorded per unit — the operator glances at the display, accepts the fill, and the reading dies there, or lands on a paper sheet aggregated by batch. When a customer claims short weight six months later, there is no record for that specific drum.
2. Closures are checked by eye while the line keeps pushing — cross-threaded cap, missing gasket, unseated seal. At line speed, some slip through. And with a chemical product, a drum dripping on the truck is not a stain: it is recovery, cleanup, a carrier claim, and sometimes dangerous goods involved.
3. Nobody sees the filling head drift coming — a head slowly going short (or long) gram by gram trips no alarm until the accumulated total is money: product given away in one direction, complaints in the other.
4. The sign-off belongs to the shift, not the container — the filling report is signed by the supervisor at the end; if one specific container went out wrong, there is no way to know which one or when.

The filling supervisor knows it: the classic system works almost every time. The problem is that its failures are invisible at the plant — they show up in transit and in the complaint, which is the most expensive possible place to discover them.

How it fits the IRIS system

iLEAN does not change your filling line — it captures the scale you already have and watches the closure nobody watches.

Filling control does not fail for lack of a scale; it fails because each container's data is not tied to the container. iLEAN acts as the putty that fills the gap between the scale, the line and the ERP, without forcing you to change anything.

Connect captures the net weight of every fill. Edge identifies the container and checks the closure with a camera. The agent holds the pallet if anything is off. The person signs — never the other way round.

The iLEAN pieces applied to drum and jug filling:

• Connect — hooks into the scale or filling head you already have (serial port, ethernet or, if there is no digital output, reading the display with a camera) and captures the net weight of every container in real time, with the tare of that container SKU.
• Edge — a terminal at the line exit. It identifies the drum or jug by its batch label and, with a camera over the conveyor, checks the closure: cap present and seated, gasket, seal. It works without a network. If the plant loses its connection, it keeps capturing locally.
• Filling agent — cross-references weight + SKU tolerance + closure photo + batch + operator. If a container falls out of tolerance or a closure looks doubtful, it warns at second zero through the earpiece or on the panel and blocks consolidation onto the pallet until it is fixed or the quality manager validates the exception. The person signs — the pallet never ships on its own.

See the full IRIS architecture →

Before and after

Scale + eyeball check vs. filling traced with iLEAN

Aspect	Scale + visual check	With iLEAN Connect + Edge (unit by unit)
Weight record per container	Display glanced at and forgotten, or a per-batch sheet	Net weight captured and tied to each container
Closure verification	By eye, with the line pushing	Camera over the conveyor: cap, gasket, seal per unit
Container out of tolerance	If the operator spots it, it gets reworked; if not, it travels	Warning at second zero, rework before palletizing
Filling head drift	Discovered when the accumulated total is already money	Trend per head visible shift by shift
Answering a weight complaint	Manual reconstruction, no container data	Container record: weight, time, line, batch, who validated
Batch filling report	Paper signed at end of shift	Assembles itself from data already captured

Impact estimate

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.

• Specialty chemicals plant with one or more drum, jug or jerrycan filling lines, an inline scale already installed and visual closure checks.
• Pilot on one line (Connect to the scale + Edge with a camera at the line exit). First value expected within a few weeks.
• Indicative payback between 4 and 10 months, depending on containers/day and the average cost of a leak incident in transit or a weight complaint ending in a credit note.
• Hard levers: ≥ 40% reduction in closure incidents discovered outside the plant (estimate to be validated); an end to systematic overfilling from head drift; per-container records that turn a weight complaint into a thirty-second lookup.

And the quality manager's reasonable doubt

“What if the camera rejects a closure that is actually fine?” — false positives exist, and that is why the system never decides alone. Hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI merely receives a digital weight from the scale and compares it against a tolerance, or checks the presence and seating of a cap against a pattern, the best models brought error below 1.5% ^[1]. And even so, what is critical is never decided alone: iLEAN warns and the operator or the quality manager sign. The three safety rings exist precisely for this.

^[1] OpenAI paper “Why Language Models Hallucinate”, 2025 — on the reliability of AI in anchored tasks.