Pharma pure gas control with AI — the gas is not in the batch, but it decides the batch.

The problem

The gas has a certificate. The analyzer has the data. And the batch sees neither.

A typical pharma plant uses medicinal nitrogen (tank blanketing, aseptic filling, aerosol), medicinal oxygen (controlled oxidation processes, BBC) and medicinal compressed air (critical instrumentation, actuators, filling). The crack is rarely in the quality of the gas: it is in the traceability from gas to batch.

1. The online analyzer (Servomex, AMETEK, Michell) reads purity every second at every point of use — and the data lives in the utilities SCADA, on an island.
2. The supplier certificate for the cylinders arrives by email or on paper — and gets filed in a folder belonging to the quality lead.
3. The batch MES knows which point of use consumed gas and at what time — but it does not know whether the gas purity at that moment was within Ph.Eur., nor which cylinder was connected.

The quality lead knows this. An FDA/EMA audit always asks about gas-cylinder-batch traceability. The classic system works 99% of the time. That 1% is the OOS caused by gas — and that OOS is a lost batch plus a root cause that takes weeks to close.

How it fits the IRIS system

iLEAN does not touch the PSA — it cross-checks the analyzer, the certificate and the batch at second zero.

The pure gas problem is not one of instrumentation: it is information on islands between the analyzer, the supplier certificate, the PSA SCADA and the batch MES. iLEAN acts as the putty that fills those gaps without asking you to change the PSA, the utilities SCADA or the validated MES.

Connect reads the analyzer and the supplier's email. Edge reads the cylinder code when it enters the plant. The agent cross-checks with the MES and, if there is a deviation, holds the batch. The person signs — never the other way round.

The iLEAN pieces applied to pure gas control:

• Edge — a machine-vision terminal in the cylinder receiving area and in the utilities room. It reads the QR code, barcode or printed label on every cylinder and records entry, exit and connection to a point of use. It works with no network.
• Connect — captures the online analyzers (OPC UA, Modbus, historical CSV) at second zero. Captures the supplier's email with the quality certificate and extracts the parameters from the PDF. Captures the SCADA of the onsite PSA / generator. And captures the alerts from the compressor's coalescing filter and catalyst.
• Agent — cross-checks online purity with the connected cylinder, the supplier certificate and the batch in the MES. If N2 drops below 99.5%, if residual O2 rises, if hydrocarbons appear in medical air: it holds the batch at the next step and prepares the root cause cross-checked with the maintenance schedule.

See the full IRIS architecture →

Before and after

Manual gas-batch traceability vs. cross-checked traceability with iLEAN

Aspect	Utilities SCADA + certificate folder	With iLEAN Edge + Connect + Agent
Online purity reading	Analyzer on an island, alarm in the local SCADA	Second zero, tied to every batch
Supplier certificate	PDF in a folder, manual search	Extracted from the email, cross-checked by cylinder batch
Cylinder-point-batch traceability	Spreadsheet kept by the gas lead	Edge reads the cylinder, the agent links it to the MES
Purity deviation	Spotted when reviewing logs, days later	Immediate detection, batch held
Filter maintenance	Separate schedule, not tied to the gas	Cross-checked with purity events
OOS file for a gas issue	Manual reconstruction, weeks	Cross-checked root cause, ready for CAPA

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 an onsite N2 PSA + O2 and medical compressed air cylinders from a supplier, 10-30 critical points of use (aseptic filling, lyophilization, blanketing).
• Edge pilot in the cylinder receiving area (code/batch reading) + Connect over analyzers and certificates + an agent that cross-checks with the MES. First value expected within a few weeks.
• Indicative payback between 4 and 9 months, depending on the frequency of purity incidents over the past few years and the average cost of a held batch in your portfolio.
• The hard lever is a ≥30% reduction in OOS investigation time for utility issues + a single batch rescued by in-line detection. One batch of sterile injectable pays for the pilot several times over.

And the utilities manager's reasonable doubt

“The analyzer already has alarms, why would I want another system?” — the analyzer has alarms for the utilities operator. iLEAN does not add one more alarm: it ties the alarm to the batch. The difference is that when a deviation fires, it is not just “there is a problem with N2 at PU-14”, it is “there is a problem with N2 at PU-14, it is affecting batch XYZ-2026-Q at the blanketing step, and the connected cylinder is from supplier X with a February certificate”. That is something a person decides in minutes, instead of reconstructing it over weeks. And in pharma, decision time is product.

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