Color changeover on an ink line — cleaning validated before startup, not when the first batch comes out contaminated.

The problem

The previous color's residue gives no warning — it shows up in the next color's first batch.

On a multiproduct line for inks, coatings or pigments, the color changeover is the highest quality-risk moment of the week: cross-contamination is not visible while cleaning, it is visible while manufacturing. On the floor:

1. The circuit has dead spots — pumps, hoses, gaskets, elbows, tank bottoms, the mill. Residue survives a “correct” cleaning and waits for the next color.
2. Validation is by eye — someone looks at the rinse solvent and decides from experience. Black to white forgives almost nothing; light to dark forgives almost everything.
3. The wash goes unrecorded — how many cycles, how much solvent, who decided to start: nothing gets written down. When the batch comes out gray, there is no way to know which stage was cut short.
4. Production pressure pushes — the changeover is line downtime. The temptation of “good enough” exists on every shift, and without an objective criterion, it often wins.

The quality manager knows it; the line lead knows it. The classic system works 99% of the time. That 1% is the batch of white with a gray cast nobody can explain — and then the whole cleaning gets repeated with the batch already made and lost, and the schedule rebuilt around it.

How it fits the IRIS system

iLEAN does not change your cleaning procedure — it guides it, validates it and puts it on record.

The changeover does not fail for lack of a procedure; it fails because the procedure lives in the operator's head and the validation has neither an objective criterion nor a record. iLEAN acts as the putty that fills the gap between the cleaning procedure, the line and the scheduling ERP, without forcing you to change anything.

Edge guides the wash sequence and watches the rinse with the camera. The agent holds the startup until the cleaning passes validation. The person signs — the line never starts on its own.

The iLEAN pieces applied to the color changeover:

• Edge — a terminal at the line. It identifies the transition from the production orders (which color you are leaving, which you are entering) and guides the wash sequence that applies — yellow→orange is not black→white — with hands-free voice confirmation of each stage. At the end, the camera evaluates the clarity of the rinse solvent against that transition's standard. It works without a network.
• Connect — if the line has meters or flow sensors, it captures the solvent consumed and the real time per stage; if not, the times come from the voice confirmations. Solvent consumption per changeover stops being an estimate.
• Changeover agent — cross-references transition + executed sequence + rinse result. If a stage is missing or the validation fails, it holds the next order's startup until the rinse is repeated or the quality manager signs off the exception. Every changeover is fully recorded: stages, cycles, solvent, result and signatures.

See the full IRIS architecture →

Before and after

Cleaning at the shift's discretion vs. changeover validated with iLEAN Edge

Aspect	Eyeball + experience	With iLEAN Edge (validated, traced changeover)
Wash sequence	In the operator's head	Guided per transition, confirmed stage by stage
Rinse validation	By eye, per shift and hurry	Camera against that transition's standard + signature
Changeover record	None, or one line in the log	Stages, times, solvent and operator, automatic
Startup on doubtful cleaning	Depends on production pressure	Held until validation or a signed exception
Contaminated first batch	Caught in QC, batch already made	Prevented before startup
Changeover duration	Estimated after the fact	Measured per stage — real input for SMED

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.

• Ink, coating or pigment plant with multiproduct lines and frequent color changeovers — especially if it makes whites and light colors, or low-migration inks for food packaging.
• Edge pilot on one line, starting with the most punishing transition (dark→light). First value expected within a few weeks.
• Indicative payback between 4 and 9 months, depending on changeovers/month, the average cost of a contaminated first batch (product, repeated cleaning, delay) and today's changeover time.
• Hard levers: ≥ 40% reduction in first batches out of tolerance from cross-contamination; shorter, more consistent changeovers with duration measured per stage; a per-changeover record ready for audits.

And the quality manager's reasonable doubt

“What if the camera passes a cleaning that is not clean?” — false negatives exist, and that is why the system never decides alone. Hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI merely compares the rinse against the standard defined for that transition, 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.