MDI inhaler control with AI — one gram of propellant out of range is one delivered dose out of range.

The problem

Three plant-floor realities that almost never get cross-checked in time.

On a pressurized MDI line — corticosteroid suspension or bronchodilator solution, HFA-134a / HFA-152a / HFC-227 propellant — the guarantee of delivered dose (USP <601>) rests on three pieces of data that in most plants each live on their own island:

1. The propellant fill — how many grams the gasser is calling for and how that varies canister by canister with vessel pressure, temperature and API batch.
2. The crimping of the metering valve — height, symmetry, head pressure. If the curve drifts, the metering chamber changes and the delivered dose changes with it.
3. The leak test — water bath or vacuum decay. The one check the operator reviews by eye or by curve, and almost always on a sample basis.

The quality manager knows everything adds up for 99% of the batch. It is that 1% that triggers the OOS investigation into delivered dose or canister integrity — and, in the worst case, the doubt over F-gas reporting for a high-GWP propellant. Caught late it is a deviation; caught in line, it is nothing.

How it fits the IRIS system

iLEAN does not add a fourth system — it seals the cracks between the three you already have.

The MDI problem is not a lack of information: it is information living on islands that, at the critical moment (vessel pressure dropping, a new propellant batch, a crimping head drifting), does not reach the person who decides in time. iLEAN acts as the putty that fills those gaps, without asking you to change your Pamasol/Coster/OPTIMA P gasser.

Edge sees crimping, weight and bubbles. Connect reads the gasser wherever it lives — modern PLC, legacy panel, forgotten spreadsheet. The agent cross-checks against the recipe, USP <601> and F-gas reporting and, if something does not add up, holds the run. The person signs — never the other way round.

The three iLEAN pieces applied to the MDI inhaler:

• Edge — a terminal with machine vision (CNN) over the carousel and over the leak-test bath. It verifies crimp height and symmetry, canister weight (cross-checked with the in-line balance) and the presence of bubbles in the water bath. It fires the actuator (selective reject gate, stack light, line stop) before cartoning. It works with no network. If the plant loses its Internet connection, Edge keeps verifying.
• Connect — captures the propellant gasser whether the data comes from the PLC over OPC-UA, from an old terminal via OCR, or from the setpoint spreadsheet. It also captures vessel pressure, product temperature, the vacuum decay curve from the leak-test bench and whatever arrives from outside (an email from the validation lead with a change to the crimping range, a WhatsApp from the supplier about a deviation in a valve or propellant batch).
• Agent — cross-checks fill, crimping, leak testing, the MES recipe and the USP/GMP/F-gas specification. If there is a deviation, it does not send an email at 10 p.m.: it holds the run and alerts the quality manager on whatever channel they use. The person validates and signs; the line does not restart on its own. And in parallel it assembles the F-gas reporting dataset for the environmental lead.

See the full IRIS architecture →

Before and after

MDI controlled by manual IPC vs. cross-checked with iLEAN

Aspect	Manual control + sampling IPC	With iLEAN Edge + Connect + Agent
Fill verification	Periodic weighing on a balance	Every canister verified in line
Valve crimping	Gauge check every N units	Height/symmetry verified canister by canister
Leak test (water bath)	Operator watches for bubbles by eye	CNN vision on every canister in the bath
F-gas reporting	Reconciled after the fact from delivery notes	Consumption traceable live, canister by canister
Operation with no network	n/a	Edge keeps running on the cabinet's own power
File for an FDA/EMA auditor	Rebuilt from paper, days of work	Per-batch dossier, automatic, with images and traces

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.

• Pharma plant with one MDI line (corticosteroid suspension + bronchodilator), 80-150 canisters per minute, HFA-134a or HFA-152a propellant.
• Edge pilot on one line (cameras over the carousel and the water bath + actuator + integration with the gasser and the MES). First value expected within a few weeks.
• Reduction of OOS deviations caused by fill, crimping or leaks of ≥ 30% in the first pilot batch, depending on initial severity and current spread.
• Indicative payback between 4 and 9 months, considering: a single OOS deviation avoided, the saving on reconciled F-gas reporting, and the brand cost of a recall for out-of-spec delivered dose.
• The hard lever is one single major deviation avoided: investigation, rework, risk of a scrapped batch, the auditor's attention.

And the quality manager's reasonable doubt

"What if the AI gets it wrong and lets an under-filled canister through?" — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI merely recontextualizes a piece of data from one system into another (reading the canister weight and comparing it with the specification), the best models brought error below 1.5% ^[1]. And even so, what is critical is never decided alone: iLEAN holds the canister 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.

← Nasal spray control with AI · Transdermal patch control with AI → · Suppository manufacturing control →