Surface treatment baths — pH and conductivity tracked continuously, tied to the batch that went through each bath.

The problem

The bath drifts every minute. The log sheet measures it twice per shift.

On a pretreatment or finishing line — degreasing, pickling, phosphating, anodizing, passivation — the final quality of the part is decided in tanks whose chemistry changes with every rack that enters. The drift is physical and never rests:

1. Drag-in and depletion — every load drags in solution from the previous stage and consumes active components; the degreaser loses alkalinity, the pickling bath loses acid, the phosphating bath unbalances its free-to-total acid ratio.
2. Control in jumps — strips, a handheld pH meter or a lab titration once or twice per shift. If the bath left its window mid-morning, everything treated until the next round ran out of spec without anyone knowing.
3. The data dies on the sheet — the reading lands on paper or in a line-side Excel: no time series, no trend alarm, no link to the batches that were passing through.
4. The claim arrives with no defense — poor adhesion, premature corrosion, staining: the problem surfaces weeks later at the customer, and reconstructing the bath state of that day is impossible.

The paradox is that many lines already have probes installed — the phosphating pH is usually right on the panel — but the data stays on the local indicator: nobody historizes it per batch, nobody watches the trend, and dosing still goes by eye after the round.

How it fits the IRIS system

iLEAN does not change your chemistry — it historizes your probes and ties every batch to the real state of its bath.

Bath control does not fail for lack of probes or judgment; it fails because the measurement is not historized, the trend warns nobody and the batch is never linked to the bath state. iLEAN acts as the putty between the line's probes, the operator's round and the quality record, without forcing you to change anything.

Connect captures pH, conductivity and temperature continuously. Tracer ties each batch to the bath state during its treatment window. The agent warns of drift before the window is breached. The person doses and signs — never the other way round.

The iLEAN pieces applied to bath control:

• Connect — reads the existing probes (from the PLC over OPC UA/Modbus or straight from the transmitter over 4-20 mA) and digitizes the round's manual measurements — lab titrations included — from the Edge terminal. A complete time series per bath, without touching the line.
• Tracer — the piece no probe provides: it crosses the time window of each batch or rack with the series of every bath, leaving each part linked to the real pH, conductivity and temperature of every stage it went through.
• Bath agent — watches each tank's trend against its operating window, warns of drift with margin to dose, flags batches that passed through an out-of-window bath and can hold their progress until quality signs off. The person releases — the batch never moves on by itself.

See the full IRIS architecture →

Before and after

Rounds with a log sheet vs. baths traced per batch with iLEAN

Aspect	Strips + pH meter + log sheet	With iLEAN (continuous bath tied to the batch)
Measurement frequency	One or two rounds per shift	Continuous, with the round as a cross-check
Drift between rounds	Invisible until the next reading	Trend watched, warning before the window is breached
Batch-to-bath-state link	Does not exist	Every batch tied to the real state of every stage
Dosing	By eye, after the round	Trend-triggered warning, executed by the person
Batch treated out of window	Passes without anyone knowing	Flagged and held until quality signs off
Answer to a claim	Memory and argument	That part's bath state, in seconds

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.

• Surface treatment line — paint pretreatment, phosphating, anodizing, galvanizing, passivation — with 3 to 10 tanks and control currently done by rounds and log sheets.
• Pilot on the 2–3 most critical baths of one line (probe readings + Tracer + operating window validated with quality). First value expected within a few weeks.
• Indicative payback between 5 and 12 months, depending on the reject rate from treatment defects, the cost of bath chemistry and the weight of field claims.
• Hard levers: ≥ 25% reduction in batches treated out of window; a documented answer to claims; dosing and renewal by real state, saving chemicals, water and effluent.

And the quality manager's reasonable doubt

“What if a miscalibrated probe flags good batches as suspect?” — it can happen, and that is why the system never decides alone. The agent cross-checks the probe against the round's measurements and also warns when the two diverge — which is exactly the signature of a probe out of calibration. Hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI merely compares probe readings against a window defined by quality, the best models brought error below 1.5% ^[1]. And even so, what is critical is never decided alone: iLEAN warns and the operator or the quality manager sign. The three safety rings exist precisely for this.

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