Eyeshadow and blush compaction with AI — the crack is born in the press and paid for by the customer on the shelf.
The eyeshadow or blush pan comes out of the press apparently fine and cracks weeks later in distribution. iLEAN Edge cross-references the force-time curve of every cycle with visual inspection of the pan, holds the ones that fall outside the tolerance band on the line, and leaves the root cause traced forever. The person signs.
The crack in the pan is not born in distribution — it is born in the press, and nobody sees it.
An eyeshadow or a blush is pressed powder: pigment + filler (mica, talc) + binder + emollient, compacted with a hydraulic press into a metal pan (godet). The process looks simple: a powder recipe, a tray of pans, a press cycle, done. In practice, the final quality of the pan depends on three things that break without warning:
- The press force-time curve — peak force and hold time determine the final density. An aging pump or a dirty valve shifts the curve without the operator noticing.
- Powder moisture and particle size — a new lot of talc or mica with slightly different moisture changes the packing density and the behavior under pressure.
- The specific SKU — a matte eyeshadow, a shimmer with large pearls, a blush with extra binder: each one has its own optimal force band. The operator learns one and uses it on all of them.
The pan comes out, passes the visual check and goes to the magazine. The crack appears weeks later on the retailer's shelf or in the end customer's bag. The return lands on the CFO's desk; the root cause is blamed on transport; the problem persists. Without the force-time curve captured and cross-referenced with the visual outcome, the information is lost forever.
iLEAN Edge does not change the press — it captures what the press was never telling anyone.
The problem with compaction is not a lack of technology in the press: it is that the force-time curve stays inside the PLC and is never cross-referenced with what happened to that pan afterwards. iLEAN acts as the putty that seals that crack between process and outcome, without asking you to change the press.
Edge captures the curve. The camera looks at the pan. The agent cross-references both and holds the suspect pan before the magazine. The person signs — never the other way round.
The iLEAN pieces applied to press force control in color cosmetics:
- Edge — a terminal with an analog input for the load cell and a top camera with grazing light, installed over the press. It captures the force-time curve of every cycle and the image of the pan immediately afterwards. The CNN is trained on real samples from your line: a good pan, a pan with a hairline crack, a pan with a halo, a pan that looks perfect but has a curve outside the band. It holds the pan on the line, before the magazine.
- Connect — captures the SKU in production, the formulation in force and the talc/mica lot from the ERP or the MES. Edge already knows which band to compare each cycle against, without anyone configuring anything by hand at changeover.
- Agent — cross-references the curve, the image, the formulation and the powder lot. When it detects a drift (the curve shifting cycle after cycle in the same direction), it warns before the band is broken. Flagged as assist and simplify: the operator gets one concrete suggestion, not another dashboard.
Blind compaction vs. force control + vision with iLEAN Edge
| Aspect | Press with a fixed recipe | With iLEAN Edge (curve + vision) |
|---|---|---|
| Force-time curve of the cycle | Lives in the PLC, never matched to the outcome | Captured, compared against the band per SKU |
| Detecting the crack | On the shelf, as a return | On the line, before the magazine |
| Change of talc/mica lot | Invisible until something breaks | Drifting curve → early warning |
| Root cause of a return | “It must have been transport” | Reproducible with the curve log and the photo |
| SKU changeover | Operator adjusts from memory | Band per SKU, nothing to reconfigure |
| Operation with no network | n/a | Edge keeps running on cabinet power |
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 press. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.
- Color cosmetics plant, one or two compaction presses, a broad catalog of eyeshadows and blushes, returns for breakage visible in the P&L.
- Edge pilot on one press (analog input to the load cell + top camera with grazing light + reject actuator + MES integration). First value expected within a few weeks: the force-time curve captured and cross-referenced with the visual outcome for the first time.
- Indicative payback between 4 and 9 months, dominated by a reduction in returns for breakage, recovery of powder waste and the elimination of stops caused by an undiagnosed root cause.
- The hard lever is a scrap reduction in the order of 30% or more on the target failure mode, once the curve stays inside the band per SKU.
And the production director's reasonable doubt
“What if the vision system gets it wrong and rejects good pans?” — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI merely classifies against learned patterns (pan inside the band / outside the band), the best models brought error below 1.5% [1]. And even so, what is critical is never decided alone: in a doubtful case, Edge holds the pan and the person signs.
[1] OpenAI paper “Why Language Models Hallucinate”, 2025 — on the reliability of AI in anchored tasks.
What people ask about eyeshadow and blush compaction with AI
Why do eyeshadows and blushes crack in distribution?
A crack in an eyeshadow or a blush almost never appears in the plant — it appears on the retail shelf or in the end customer's bag, weeks later. The cause is usually inconsistent compaction: the force-time curve of the hydraulic press drifted away from the standard because of a small pressure fluctuation, an aging pump, a variation in powder moisture or a change of talc/mica supplier. The pan comes out apparently fine, passes the visual check, but internally has heterogeneous compaction that breaks with the first knock in transit. Without in-line control of the curve, that information is lost forever.
How do you control press force without changing the machine?
iLEAN Edge connects to the press's existing load cell (nearly every modern press has one) or to one added if the press is old, and captures the force-time curve of every cycle. It cross-references that curve with the SKU in production (matte eyeshadow, shimmer blush, formulation with or without binding agents) and with the reference curve learned from pans that reached the customer in good shape. If the curve moves outside the tolerance band, it holds the pan on the line before it enters the magazine. It works with no network: if the plant loses WiFi, Edge keeps reading and holding, because what is critical cannot depend on connectivity.
Does it also detect a visible surface crack on the pan?
Yes. The force-time curve is the process data; the image of the pan is the outcome data. Edge combines both: a top camera with grazing light to catch the micro-fissure, and a CNN trained on real samples from your line — good pans, pans with a halo, pans with a hairline crack, pans with no visible defect but with a curve outside the band. The combination catches both failure modes: the one you can see (the crack) and the one you cannot see until it reaches the shelf (heterogeneous compaction).
What real savings come from reducing compaction waste in color cosmetics?
Waste in color cosmetics has two components: direct scrap (a broken pan thrown away, powder lost) and delayed scrap (a retailer return for breakage on the shelf, a whole batch withdrawn). The first is usually accounted for; the second shows up in the returns line of the P&L and gets blamed on transport. Once the force-time curve of every pan is traced, the root cause becomes visible for the first time: there is no more arguing about whether it was the loader or the press — it is in the log.
How much does it cost to implement Edge force control on a compaction line?
The order of magnitude of an Edge pilot on a cosmetic compaction press covers the terminal with vision and an analog input for the load cell, the camera with the right lighting, the reject actuator and the integration with MES, plus an annual license. A reasonable payback to present to the committee is in the order of several months — the hard lever is the reduction in returns for breakage on the shelf + the recovery of waste. We ask for your press's data and send you the estimated ROI in 48h, with your numbers.
Tell us your case and in 48h we'll send you the estimated ROI of this AI project for your compaction line.
We work on your plant's real data, not ours. Diagnostic with no commitment.
Request estimated ROI in 48h See cosmetics