ANPR + ADR code at the chemical terminal gate — a truck that does not add up does not get in.

The problem

The tanker is already at the gate and nobody has yet cross-checked the three data points that matter.

At a well-run chemical terminal gate there are four sources of truth about every truck, and they are almost never in the same place at the same time:

1. The plate of the tractor unit and of the trailer/tanker — to know which carrier it is and whether the tanker is the one that was authorised.
2. The ADR panel on the tanker — the Kemler code on top and the UN number below: what hazard it carries and which exact product.
3. The carrier's delivery note — what was loaded at origin, in what quantity and for which bund of the terminal.
4. The history — whether that tanker is compatible with its previous load, which cleanings are declared for it, what past incidents there are with that carrier.

The officer in the gatehouse knows this, but cannot check all four at once in thirty seconds. The queue of trucks builds pressure, the driver shows paperwork and the barrier opens because "the panel looked like a match". The classic system works 99% of the time. That 1% is the incompatible mix in the bund, the ADR penalty or the picture that ends up in the press — and each of those, on its own, pays for several years of reinforced control.

How it fits into the IRIS system

iLEAN does not replace the gate officer — it puts the four cross-checked data points in front of them before the barrier opens.

The problem with access control at a chemical terminal is not a lack of data: it is information living in islands that, at the critical moment (thirty seconds at the gate, with a queue of trucks behind), does not reach the person deciding in time. iLEAN acts as the filler that closes those gaps, without asking you to change your TMS, your ERP or your gatehouse. See the category: full IRIS architecture.

Edge sees the plate and the ADR panel before the barrier. Connect reads the delivery note wherever it lives — TMS, carrier email, WhatsApp. The agent cross-checks against the tanker's history and, if something does not add up, holds access. The person signs off — never the other way round.

The three iLEAN pieces applied to ANPR + ADR control at a chemical terminal:

• iLEAN Vision (Edge) — one or two terminals with cameras in the truck waiting area. The CNN is trained to do ANPR and ADR panel reading in the same pass. It reads the Kemler and the UN number, identifies them as structured entities and triggers the barrier decision. It works with no network. If the line goes down, Edge keeps reading and keeps holding against the last valid synchronised list.
• Connect — captures the carrier's delivery note whether it comes from the TMS, from the email that lands at 06:00, or from the traffic manager's WhatsApp. It digitises it at second zero, without anyone forwarding anything. And it also captures what arrives from outside (a last-minute tanker swap by the carrier, a customs incident, an updated safety data sheet for the product).
• Agent — cross-checks the ANPR reading, the ADR reading, the delivery note and the history of the tanker and the carrier. If there is a deviation (a UN number different from the one declared, an unauthorised plate, a tanker with an incompatible previous load), it does not send an email after the fact: it holds access and presents the cross-check to the HSE person. The person validates and signs off — the barrier never opens on its own.

See the full IRIS architecture →

Before and after

Manual ANPR + ADR control vs. cross-checked control with iLEAN Vision

Aspect	Manual gatehouse + checklist	With iLEAN Vision + Connect + Agent
Plate reading	Officer types it from the gatehouse	ANPR in milliseconds, tractor unit + trailer
ADR panel verification	Visual from a few metres, no record	Structured reading of Kemler and UN number
Cross-check with the declared delivery note	Paper against paper, by eye	Automatic cross-check against TMS/ERP in seconds
Mislabelled or unauthorised tanker	Detected late or never	Barrier held before entry
Traceability for ADR audit	Reconstructed by hand, weeks	Dossier per entry with photo, reading and sign-off
Operation with no network	n/a	Edge keeps reading and holding on cabinet power

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 figures of your terminal. We set it out so the committee has an order of magnitude; we refine it during the diagnostic.

• Chemical / petrochemical terminal with 60-150 tanker entries a day, ADR control done manually at the gatehouse today.
• Edge pilot at the gate (two cameras + integration with the TMS/ERP holding the delivery note). First expected value within a few weeks: an ANPR + ADR read rate that is genuinely useful at the gatehouse.
• Indicative payback between 4 and 9 months, depending on entry volume, number of gatehouses equipped and the average cost of one documented wrong entry.
• Expected reduction in entries with unvalidated ADR identity of ≥ 30% versus the previous manual control, with a floor that is reasonably defensible before the committee.
• The hard lever is a single wrong unloading avoided: incompatible mix in the bund, terminal shutdown, ADR penalty, reputational damage. One of those pays for the pilot.

And the IT manager's reasonable doubt

"What if the AI gets the ADR reading wrong and lets a mislabelled tanker in?" — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI simply reads a visible data point (plate, ADR panel) and compares it with a declared one (delivery note), the best models brought the error below 1.5% ^[1]. And even so, what is critical is never decided alone: the system holds and the person signs off. The three safety rings are there for exactly this.

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