Masterbatch color blending in plastics — batch color matches the master, without relying on the operator's eye.

The problem

Half a percent of dosing or the wrong bag — and the whole batch comes out a different color.

In injection molding, extrusion or blow molding, the final color depends on a chain that is almost never controlled end to end: right bag, right percentage, controlled regrind, right dispersion. On the floor:

1. The wrong masterbatch — references that differ by one digit, near-identical bags on the same shelf. The mistake is invisible at the hopper; it shows up in the part.
2. The actual percentage is not the recipe's — an out-of-calibration doser or a badly loaded hopper turns a nominal 2% into an actual 1.6%. Nobody records it.
3. Regrind shifts the base — reground material already carries color, and its share varies shift to shift. The recipe states one number; the machine runs another.
4. Approval is by eye against a master part — one that yellows over the months, under workstation light, with a different observer every shift.

The quality manager knows it; the shift lead knows it. The classic system works 99% of the time. That 1% is the full order the customer rejects because it does not match the parts from the previous order — and that material never goes back to being virgin: regrind at best.

How it fits the IRIS system

iLEAN does not touch the recipe — it verifies that what enters the machine is what the recipe says.

Color control in plastics processing does not fail for lack of a recipe; it fails because nobody verifies the whole chain: masterbatch identity, actual percentage, regrind and a measured result. iLEAN acts as the putty that fills the gap between the warehouse, the doser, the spectrophotometer and the ERP, without forcing you to change anything.

Edge verifies the masterbatch is the recipe's. Connect captures the doser's actual percentage. The agent measures the first parts against the digital master and holds the batch if the ΔE does not land. The person signs — never the other way round.

The iLEAN pieces applied to masterbatch blending:

• Edge — a terminal at the machine. Before loading, the operator scans the bag label and Edge checks it against the production order's recipe: if it does not match, it warns at second zero. The regrind share is declared by voice, hands-free. It works without a network. If the plant loses its connection, it keeps capturing.
• Connect — captures the actually dosed percentage from the gravimetric or volumetric doser, continuously. And at startup, it captures the spectrophotometer reading (X-Rite, Konica Minolta) on the first parts or plaques, over Bluetooth or USB. If the instrument only shows the value on its screen, a photo is enough: Connect reads it.
• Color agent — cross-references masterbatch identity + actual percentage + declared regrind + the first parts' ΔE against the digital master. If something does not add up, it warns and holds the order's release until the adjustment lands in tolerance or the quality manager validates. The person signs — the batch is never released on its own.

See the full IRIS architecture →

Before and after

Recipe + operator's eye vs. blending verified and measured with iLEAN

Aspect	Recipe + master part by eye	With iLEAN (chain verified from bag to part)
Masterbatch identity	Trusted to be the loaded bag	Verified against the recipe before loading
Dosing percentage	The recipe's, presumably	The doser's actual value, captured continuously
Color reference	Aging master part	Digital L*a*b* master values per reference
Startup approval	Operator's eye at the workstation	First parts' ΔE against the agreed tolerance
Off-color batch	Caught by final QC or the customer	Held at startup, with material still recoverable
Record per batch	Machine log	Masterbatch, actual %, regrind and ΔE tied to the batch

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.

• Plastics processor (injection, extrusion, blow molding) coloring with masterbatch, with gravimetric or volumetric dosers at the machines and demanding brand colors.
• Pilot on one critical machine (bag scanning + doser capture + spectrophotometer at startup). First value expected within a few weeks.
• Indicative payback between 4 and 9 months, depending on batches/month, the average cost of a batch rejected for color, and today's startup scrap.
• Hard levers: ≥ 30% reduction in color rejections; shorter startups with the ΔE measured on the first parts; a per-batch color dossier for the automotive or consumer goods customer.

And the quality manager's reasonable doubt

“What if the system mismatches a bag or misreads a value?” — false positives 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 cross-checks a scanned label against a recipe, or a ΔE against an agreed tolerance, 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.