Color-change CIP in multipigment production — a trace of blue in a yellow is an off-shade batch.

The problem

The probe says clean — and the shade says otherwise.

A multipigment plant alternates, on the same equipment train (reactor, safety filter, bead mill), batches of pigments that differ widely in composition and tinting strength. The CIP between two different batches is validated today with three levers that, taken separately, never see the whole truth:

1. Cleaning time and recipe from the PLC — the CIP ran for the planned minutes with the planned chemicals. That is what the PLC signs off.
2. Conductivity of the final rinse — it is within spec. That is what the conductivity meter signs off.
3. The operator's eye at the end of the shift — the reactor sight glass looks clean. That is what the person signs off.

And even so, a trace of phthalocyanine blue in the disperser disc seal can push the ΔE of the next yellow batch out of tolerance. The classic system works 99% of the time. That 1% is the off-shade batch that turns up in quality control when the batch is already packed — and the cost is not just the rework (rarely 100% recoverable), it is the brand damage with a customer who received a pigment that is not exactly the one on the data sheet.

How it fits the IRIS system

iLEAN Vision does not replace the CIP — it seals the crack between the probe, the clock and the eye.

The problem with color changes in multipigment production is not a lack of cleaning: it is a lack of cross-checked truth. The CIP PLC, the conductivity meter and the operator's eye each live on their own island and, all three at once, do not fit in the shift lead's head at the end of a long changeover. iLEAN acts as the putty that covers the dead zone between the CIP sign-off and the reality of the next batch, without asking you to change your CIP system or your probe.

Vision reads the critical zones of the equipment after the rinse. Connect captures the conductivity curve and the CIP recipe. The agent cross-references them against the planned color change and, if it does not add up, holds the next charge. The person signs — never the other way round.

The iLEAN pieces applied to color-change CIP in multipigment production:

• Edge — Vision — Edge cameras over the critical zones of the train (reactor sight glass, bottom of the safety filter, disperser disc seal, bead mill discharge, rinse tank drain). A CNN trained on the clean-equipment signature detects traces that a tired eye at the end of the shift does not catch. It works with no network. If the plant loses WiFi, Edge keeps holding the equipment.
• Connect — captures the full curve from the rinse conductivity meter, the recipe and timings from the CIP PLC, and the sheet for the planned color change (which pigment is leaving, which one is coming in). At second zero, with nobody having to resend anything.
• Agent — cross-references the three sources — visual reading of the critical zones, conductivity curve, CIP recipe — against the planned color change. If conductivity comes back in spec but Vision detects a visible trace in the disc seal, the agent proposes repeating the CIP in intensive mode and alerts the shift lead. The person validates and signs; the line does not restart on its own.

See the full IRIS architecture →

Before and after

CIP validated by probe vs. CIP validated by cross-checked truth with iLEAN

Aspect	Conductivity + time + eye	With iLEAN Vision + Connect + agent
CIP validation	Conductivity in spec, time completed	Three sources cross-referenced against the planned change
Organic pigment with no conductive signature	Passes the rinse, leaves a trace	Vision detects the trace in the critical zones
Detecting the off-shade batch	QC of the next batch — with the product already packed	Before the next batch is charged
Deciding to repeat an intensive CIP	The shift lead's instinct	A proposal with data + an image of the suspect spot
Traceability for audit / customer	Rebuilt by hand	Per-color-change dossier with photo + curves
Operation with no network	n/a	Edge keeps holding on the panel's own 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 data of your plant. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.

• Organic multipigment plant, a reactor + safety filter + bead mill train shared across families (phthalocyanine, monoazo, quinacridone), several color changes per shift.
• Vision pilot with 4-6 Edge cameras over the critical zones plus integration with the CIP PLC and the existing conductivity meter. First value expected within a few weeks: the first automatic holds on equipment that is not fully clean.
• Indicative payback between 4 and 9 months, depending on the historical frequency of off-shade batches and the average cost of rework per batch.
• Expected reduction in batches with ΔE out of tolerance of ≥ 30% in the first quarter.
• The hard lever is a single color-change run saved: a batch not reworked, equipment not in quarantine, a customer not complaining about a shade that is not the one on the data sheet.

And the shift lead's reasonable doubt

"What if Vision holds the equipment because of a reflection or a shadow and stops a perfectly valid color change?" — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI merely compares an image against the clean-equipment signature learned on that very machine, the best models brought error below 1.5% ^[1]. And even so, what is critical is never decided alone: iLEAN holds and the person signs the restart. The three safety rings exist precisely for this.

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