Magnetic flow meter in reagent dosing — discharge is adjusted in the reactor, not at the inspection.

The problem

Operating at the legal limit is where the saving is — and where the risk concentrates.

Correct reagent dosing in an industrial wastewater plant is the intersection of five realities that are almost never in the same place:

1. Reagent flow — read by the magnetic flow meter on the dosing pump. It sits in the transmitter, almost never cross-checked with anything outside.
2. Influent flow — recorded by another mag meter upstream, in another cabinet, sometimes in another SCADA.
3. Inlet pH/conductivity — the equalisation tank probe, with a curve that shifts by shift or by water origin.
4. Outlet pH/conductivity — the indicator the environmental manager looks at when they walk into the room.
5. The environmental permit — the legal limit, locked inside a PDF that very few people cross-check against today's data.

The plant operator has three screens, a notebook and the caution of someone who knows the cost of an out-of-spec discharge is not measured in reagent, it is measured in fines and reputational damage. So they over-dose — and over-pay for reagent. Operating at the legal limit without overshooting requires cross-checking the five readings in real time. No classic system does it on its own.

How it fits the IRIS system

iLEAN does not replace the SCADA or the pump — it seals the cracks between the five readings you already have.

The dosing problem is not a lack of sensors: it is information living on islands that, at the critical moment (a change in water origin, a load peak, the weekend restart), is not cross-checked in time. iLEAN acts as the filler that closes those gaps, without asking you to change SCADA, pump or flow meter.

Edge reads the mag meter and the probe. Connect captures the environmental permit and the environmental team's emails. The agent cross-checks against the influent curve and, if the dose is going to overshoot, holds the pump. The person signs off — never the other way round.

The three iLEAN pieces applied to magnetic flow in reagent dosing:

• Edge — neural-network terminal on the mag flow meter signal (reagent flow) and on the inlet and outlet pH/conductivity probes. It anticipates the drift and holds the dosing pump via dry contact before the discharge leaves the limit. It works with no network. As long as the cabinet has power, the reading and the hold keep running.
• Connect — captures the environmental permit whether it comes from the permit PDF, the compliance spreadsheet or the environmental manager's email. And it captures what arrives from outside (a heads-up from the upstream process about a special discharge, a message from the lab with the new analysis) at second zero, without anyone forwarding anything.
• Agent — cross-checks reagent flow, influent flow, both probes and the environmental permit curve. If the kinetics indicate the set-point is going to be overshot, it does not send an email at 10 pm: it holds the pump and alerts the operator on whichever channel they use. The person validates and signs off; the pump does not resume on its own.

See the full IRIS architecture →

Before and after

Dosing with a safety margin vs. cross-checked dosing with iLEAN

Aspect	Dosing with margin + sampling	With iLEAN Edge + Connect + Agent
Dosing strategy	High safety margin — chronic over-dosing	Operation at the legal limit with anticipation
Control frequency	Sample every N hours, manual adjustment	Cross-checked reading at second zero
Change in influent origin	Late reaction — discharge out of band	The agent sees it in the flow and compensates
Reagent over-dosing	High consumption "just in case"	Consumption cut by 30% or more (estimate to be validated)
Operation with no network	n/a	Edge keeps running on the cabinet's power
Evidence file for the environmental permit	Rebuilt by hand, weeks	Per-shift dossier, automatic, with cross-checked 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 lay it out so the committee has an order of magnitude; we refine it during the diagnostic.

• Industrial wastewater plant with one acid-base neutralisation + coagulation line, influent varying by shift and by origin, environmental permit with a narrow margin.
• Edge pilot on the reagent line (neural-network terminal on the pump's mag meter + pH/conductivity probes + integration with the existing SCADA). First value expected within a few weeks.
• Indicative payback between 4 and 9 months, depending on historical reagent consumption and the current margin against the legal limit.
• Hard levers: reagent consumption cut by 30% or more, elimination of the risk of a fine for out-of-spec discharge, and operator hours freed from manual sampling.

And the environmental manager's reasonable doubt

"What if the AI gets it wrong and lets an out-of-spec discharge through?" — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI merely cross-checks one signal against another (the mag meter reading against the pH curve and the limit in the PDF), the best models pushed the error below 1.5% ^[1]. And even so, nothing critical is decided alone: iLEAN holds the pump and the person signs off. The three safety rings exist for exactly this.

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