Coriolis flow in aseptic UHT filling — dosing is measured in grams per unit, not at end of shift.

The problem

UHT dosing is decided between three realities that are almost never in the same place.

A correct carton of UHT milk is the intersection of three pieces of data that, in most plants, each live on their own island:

1. Real mass flow — read by the Coriolis meter installed on the filler pipework. It sits in the transmitter signal, almost never cross-checked against anything else.
2. The batch recipe — grammage per SKU, expected density, tolerance. It lives in the ERP, in a MES, or in a spreadsheet that R&D updates in the morning and nobody forwards to packaging.
3. Process state — SIP/CIP phase, first pack after a recipe change, product temperature leaving the plate heat exchanger. Often only in the shift leader's head.

The packaging operator has three screens, two checklists, and the SKU changeover is the most vulnerable moment. Quality control weighs a sample every so many units and reconstructs what happened. The classic system works 99% of the time. That 1% is the claim that lands — and, if it shows a pattern, it is a penalty.

How it fits into the IRIS system

iLEAN does not replace your Coriolis meter or your filler — it seals the cracks between the systems you already have.

The UHT dosing problem is not a lack of information: it is information living on islands that, at the critical moment (the SKU changeover, carryover after SIP, density drift from temperature), does not reach the decision maker in time. iLEAN acts as the filler compound that closes those gaps, without asking you to change the filling machine, the ERP or the flow meter.

Edge reads the Coriolis meter at second zero. Connect captures the recipe wherever it lives — ERP, MES, a forgotten spreadsheet. The agent cross-checks it against the set point and, if something does not add up, holds the pack. The person signs — never the other way around.

The three iLEAN pieces applied to Coriolis flow in aseptic UHT filling:

• Edge — a terminal running a neural network over the Coriolis signal (mass flow, density, temperature) and vision over the filler itself. It cross-checks data and view in milliseconds and fires the pack hold (stack light, rejector) when dosing goes out of tolerance. It works without a network. If the plant loses WiFi, Edge keeps reading and keeps holding.
• Connect — captures the batch recipe whether it comes from a modern ERP, a vertical MES or the spreadsheet that R&D updates every morning in a shared folder. It also captures what arrives from outside (a lab email with the newly validated grammage, a supplier message about a packaging change) at second zero, with nobody having to forward anything.
• Agent — cross-checks the Coriolis signal, the batch recipe, the SIP/CIP phase and the SKU changeover history. If there is a deviation, it does not send an email at 10pm: it holds the pack and alerts the packaging supervisor on whatever channel they use. The person validates and signs; the filler does not restart on its own.

See the full IRIS architecture →

Before and after

UHT dosing by sampling vs. dosing cross-checked with iLEAN

Aspect	Sampling + manual control	With iLEAN Edge + Connect + Agent
Flow source	Coriolis reads, filler executes — never cross-checked	Coriolis cross-checked with recipe and process phase
Control frequency	Weighed every N units	Every filling cycle, in milliseconds
SKU change / carryover after SIP	Assumption: "if we purged, it is fine"	First pack at steady state identified by the agent
Temperature drift	Detected in later analytics	Coriolis density compensated in line
Operation without network	n/a	Edge keeps running as long as the cabinet has power
Dossier for the IFS/BRC auditor	Rebuilt by hand, weeks	Per-batch dossier, automatic, with Coriolis history

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.

• UHT dairy plant with 2 aseptic fillers, multi-SKU carton + bottle, recipe varying by density and grammage (whole, semi-skimmed, fortified, plant-based).
• Edge pilot on one filler (terminal running a neural network over the Coriolis signal + vision over the pack + integration with the ERP/MES recipe). First value expected within a few weeks.
• Indicative payback between 4 and 9 months, depending on how often weight non-conformities are recorded and the average cost of a claim or rework in your chain.
• The hard lever: a reduction of ≥30% in dosing scrap, elimination of out-of-tolerance weight claims (EU regulation on nominal quantities) and operator hours freed from manual checking.

And the packaging supervisor's fair objection

"What if the AI gets it wrong and lets an under-filled pack through?" — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI is limited to cross-checking a signal against a piece of system data (reading the Coriolis meter and comparing it to the recipe set point), the best models brought error below 1.5% ^[1]. And even so, what is critical is never decided alone: iLEAN holds the pack and the person signs. The three safety rings exist precisely for this.

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