KTL degrease bath conductivity — one out-of-band reading, one batch of paint lost in e-coat.

The problem

The bath that decides whether e-coat is scrap or quality.

In a KTL tunnel, alkaline degrease is the first barrier between a dirty part and an electrochemical process that only works on a clean surface. Bath conductivity is the chemical summary of that whole barrier: it rises when dragged-in salts and oils build up, it drops when dosing falls short. And in both cases the symptom is not visible in the bath: it is visible after the oven, on the painted part.

1. What the bath carries — actual conductivity, alkaline dose, upstream and downstream rinses. It usually sits in the chemical supplier's control system or on an old isolated panel.
2. What enters the bath — the previous rack, the accumulated load, where the parts come from (stamping, machining). Sometimes only on a shift report.
3. What leaves the bath — the e-coat line and the final oven, which absorb the cost of everything above. And a cosmetic rejection is rework or scrap, not an incident note.

The paint shop manager knows this — but cannot be reading conductivity every 30 seconds. Manual measurement once per shift works most of the time. The exception is the one that sends a whole batch to rework. The critical part is slow drift, the kind you notice when it is already too late.

How it fits the IRIS system

iLEAN does not replace your bath control — it seals the cracks between chemistry, line and planning.

The problem with the KTL degrease bath is not a lack of chemical control: it is that the bath data, the line data and the paint supplier data live on islands and reach the operator too late. iLEAN acts as the filler that closes those gaps, without asking you to change your dosing system or the tunnel panel.

Edge sees conductivity every second. Connect reads the dose wherever it lives — modern PLC, old panel, the chemical supplier's spreadsheet. The agent crosses it with the production order and, if something drifts, holds the rack before e-coat. The person signs off — never the other way round.

The three iLEAN pieces applied to KTL bath conductivity:

• Edge — terminal with vision + probe integration on the degrease tank itself. It reads conductivity in seconds, crosses it with the tunnel PLC reading and, if the drift exceeds the threshold, triggers a stack light or holds the rack before phosphating. It works with no network.
• Connect — captures the alkaline dose even when it lives on the chemical supplier's old isolated panel, or in a spreadsheet the process technician maintains. And it captures from outside: an email from the paint supplier with a formulation change, a message from the stamping supervisor about a drawing-oil change that soils more.
• Agent — crosses conductivity, dose, the bath's accumulated load and the rack's origin with the e-coat recipe. If the drift points to a coverage failure downstream, it proposes the dose or purge adjustment to the technician before the batch reaches the oven. The technician validates; nobody adjusts alone.

See the full IRIS architecture →

Before and after

Manual bath readings vs. online tracking with iLEAN

Aspect	Manual control + shift report	With iLEAN Edge + Connect + Agent
Bath reading	Manual log once per shift or by exception	Continuous, second by second, online
Reaction to drift	When the rejected batch comes back after the oven	Stack light and rack hold before phosphating
Dosing	Fixed dose from the chemical supplier's recipe	Adjustment suggested to the technician from real drift
Bath cleaning/purge	Scheduled by hours or by appearance	Scheduled by real accumulated load
Per-rack traceability	Whatever the MES holds + the shift notebook	Per-rack record with conductivity and dose curve
Operation with no network	n/a	Edge keeps reading and holding with the panel light

Impact estimate

Impact estimate for your plant — to be validated with your numbers.

The block below is an estimate to be validated with the actual data from your tunnel. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.

• Multi-stage KTL tunnel with alkaline degrease, phosphating and rinses, feeding an e-coat line with a final oven.
• Edge pilot on the degrease tank (conductivity probe + integration with the tunnel PLC + rack hold actuator). First value expected within a few weeks.
• Cosmetic e-coat rejection reduced by ≥ 30 % on the recurring bath incident (the other failure modes remain the operator's business).
• Indicative payback between 4 and 9 months, depending on how often coverage incidents occur and the average rework/scrap cost.
• The hard lever: a single avoided batch of badly deposited paint pays for a good part of the pilot.

And the IT manager's fair objection

"What if the AI gets it wrong and proposes a dose adjustment that ruins the bath?" — hallucination is a problem of free generation, not of anchored tasks. On tasks where the AI is limited to recontextualising a measured value (crossing conductivity with dose and load), the best models brought error below 1.5 % ^[1]. And even so, critical decisions are never taken alone: iLEAN proposes and the person signs off. The three safety rings exist precisely for this.

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