Shampoo viscosity out of range: correct it in line, not as scrap at the end of the shift.
The viscosity of a shampoo or hair mask moves with the homogenizer temperature, the surfactant batch and the mixing time. When the lab checks at the end, the product is already made: costly rework or scrap. iLEAN Edge reads the in-line viscometer, cross-references it with the recipe and the temperature, and proposes the dosing-pump correction at second zero. The operator signs; the batch comes out within specification.
Checking at the end is checking too late.
The classic process in a shampoo or hair-mask mixing operation is:
- Mixing in the homogenizer to a recipe — water, surfactants, thickener (NaCl, polymer), preservative, fragrance.
- Temperature that rises through shear, takes steam from the jacket, or is cooled per the protocol.
- Mixing time settled by experience.
- Sample to the lab at the end — viscosity, pH, appearance.
- Decision after the fact — if it's out of range: rework (add more salt or more thickener, with the risk of overshooting) or scrap.
The problem is not the operator's: it's the timing. By the time the lab measures, correcting under normal conditions is no longer possible — the tank is at process volume, the next batch is waiting, and the production manager has to choose between two bad options. The in-line viscometer exists; what's missing is cross-referencing what it measures with the recipe and the temperature so the correction arrives before the batch closes.
iLEAN doesn't add a fourth system — it seals the cracks between viscometer, MES and dosing pump.
The homogenizer doesn't fail for lack of information — the viscometer generates it and so does the PT100. It fails because that information lives in islands that only get cross-referenced when the lab analyzes at the end. iLEAN acts as the putty that fills the gaps at second zero, without asking you to change the homogenizer, the viscometer or the PLC.
Edge sees the viscosity the way the veteran formulator would if they could stand at the tank all shift. And it proposes the correction before the lab discovers it arrived too late. The person signs — the line doesn't resume on its own with changes outside the agreed range.
The specific pieces:
- iLEAN Edge — a terminal with a CNN trained on the product's characteristic curve (shampoo, mask, conditioner). It reads the in-line viscometer, the PT100, the rotor speed and the recipe from the MES. When it sees the trajectory drifting away from the target range, it proposes the correction: extra NaCl, more turns, a thickener adjustment. It reacts in milliseconds. It works without a network.
- Connect — if the dispenser is old and has no modern interface, Connect reads its panel and pushes the correction through whatever channel the operator uses (earpiece, tablet).
- Three safety rings — a correction within the agreed range can run on auto-approval; a correction outside the range is proposed and signed by the operator or the shift lead. The plant decides the autonomy level of each ring.
Viscosity checked at the end vs. viscosity corrected in line
| Aspect | Classic mixing | With iLEAN Edge + Connect |
|---|---|---|
| Moment of control | Sample to the lab at the end | Continuous curve from the in-line viscometer |
| Rework decision | After the fact, with the batch already made | Correction proposed before the batch closes |
| Cross-reference with temperature | In the operator's head | Automatic, every minute |
| Batch-by-batch traceability | Final result + manual logbook | Full curve + corrections in the PIF |
| Operation without network | n/a | Edge keeps reading and proposing |
| Veteran formulator's knowledge | Walks out the day they retire | Captured as a model in Edge |
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.
- Plant with a homogenizer of several thousand liters, 2-4 references per shift (shampoo, mask, conditioner), 1-3 documented deviations per month.
- Edge pilot on the highest-rotation homogenizer (the low-hanging fruit). First value — the continuous curve and the first suggested correction — within a few weeks.
- Indicative payback between 4 and 9 months, depending on the batch cost, the deviation frequency and the documented rework cost.
- Expected reduction of the % of out-of-spec batches ≥ 30%. A defensible floor; the ceiling is refined with your deviation Pareto.
And the production manager's reasonable doubt
“What if the AI proposes the wrong correction and ruins the batch?” — the correction is proposed; it does not apply itself outside the agreed range. The reliability of AI in tasks anchored to the source (reading a sensor and comparing it against a recipe) is below 1.5% error in the best models [1]. And the operator signs. The three rings are there precisely so critical corrections don't happen on their own.
[1] OpenAI paper “Why Language Models Hallucinate”, 2025 — on the reliability of AI in anchored tasks.
What people ask about in-line shampoo viscosity control
Why does shampoo viscosity drift out of range during mixing?
Because the viscosity of a shampoo or hair mask depends on three variables that change at the same time: raw material (a surfactant or thickener batch with slight variability), temperature of the mass (the homogenizer heats through shear), and the speed and time of the homogenizer itself. If the previous batch heated the tank more than usual, the next one starts from a different baseline. When the lab checks the viscosity at the end of the batch and sees it has drifted out of range, the product is already made: either it's reworked (adding sodium chloride or more thickener, with the risk of overshooting) or it goes to scrap. Both options cost money.
What in-line sensor is needed and how is it integrated?
An in-line viscometer (vibrational, capillary or rotational, depending on the manufacturer) installed in the homogenizer bypass or in the discharge line to the buffer tank. iLEAN Edge connects to that sensor — and to the tank's PT100 temperature probe, and to the recipe lookup in the MES — without asking anyone to change the PLC logic. If the machine only has an analog panel, Connect photographs and digitizes it. If it has a local computer, however old, Edge connects to it. The integration requires no construction work.
Can iLEAN Edge control the thickener dosing pump automatically?
Edge proposes the correction — more sodium chloride, more thickener, another homogenization pass — and hands it to the operator via earpiece or panel. The operator accepts and the correction is applied. In maximum-autonomy mode (a plant with low criticality and a high confidence threshold), the system can adjust the dosing without a signature for minor changes within an agreed range. What is critical always carries a human signature — the three rings are there so the plant decides where.
How is batch-by-batch traceability guaranteed for cosmetic GMP?
Every batch leaves the homogenizer with its complete viscosity curve, the corrections applied, the temperature logged minute by minute and the operator's signature. The file enters the batch's PIF automatically — the ISO 22716 GMP auditor sees it without anyone reconstructing anything. If the safety assessor needs to reprocess the batch because of a later observation, the data is there; it doesn't have to be rebuilt from the shift lead's memory.
How much is saved in scrap and rework per deviated batch?
It depends on the batch cost and the deviation frequency. In a mid-sized shampoo or hair-mask mixing operation, the cost of a typical deviated batch (raw material + tank time + rework) is in the order of several thousand euros — and reworking a viscous product already packed costs far more. If your history is 1-2 deviated batches per month, the payback of an Edge pilot is a few months. We'll refine it with your data in 48h.
Tell us your case and in 48h we'll send you the estimated ROI of this AI project for your hair-care mixing line.
We work on your plant's real data, not ours. Diagnostic with no commitment.
Request estimated ROI in 48h See cosmetics