Varnish and resin filtration — the filter change is called by the real pressure curve, not the clock.

The problem

The gauge tells the truth about the filter. The change schedule never listens to it.

Between the reactor or dispersion tank and packaging, the varnish passes through filters — bag, cartridge, plate — whose job is to hold back gels, skins, particles and grinding residue. The life of each element depends on the product, the lot and the upstream process. On the floor, however, the change is decided like this:

1. Fixed interval by hours or batches — inherited, conservative and blind: the same for the pigmented varnish that loads the filter fast and for the clear one that barely touches it.
2. Changing too early costs hard cash — elements discarded at half life, line stops for the change, and the product held up in every discarded bag or cartridge.
3. Changing too late costs complaints — at the end of the filter's life differential pressure spikes, retention efficiency drops, and haze, gels or particles reach the packaged lot.
4. A burst warns nobody — a burst or unseated element lets varnish through unfiltered; differential pressure drops at once, but if nobody is watching the gauge, it is discovered at QC or at the customer.

The striking part is that the data already exists: almost every plant has gauges at the inlet and outlet of each filter. What does not exist is continuous monitoring, a memory of how each reference fouls, or the link between the curve and the lot that was passing through.

How it fits the IRIS system

iLEAN does not change your filtration equipment — it listens to its gauges and learns how each product fouls.

The change decision does not fail for lack of technical judgment; it fails because nobody watches the pressure curve continuously and because every product fouls its own way. iLEAN acts as the putty between the filter gauges, the production schedule and the line operator, without forcing you to change anything.

Connect captures the differential pressure. The agent learns each product's fouling signature and flags the change window — or the sudden drop that betrays a burst. Edge logs the change and ties the curve to the lot. The person decides — never the other way round.

The iLEAN pieces applied to the filtration endpoint:

• Connect — reads existing pressure transmitters from the PLC (OPC UA, Modbus, 4-20 mA); if there are only analog gauges, it digitizes the photo the operator takes on his rounds with Edge. Flow and temperature, where available, enrich the curve.
• Filtration agent — ties each curve to the product and lot in process, learns the fouling signature of every reference and computes the real change window. It catches the sudden drop (burst or unseated element) and the sudden rise (anomalous batch or upstream problem) the moment they happen.
• Edge — a terminal in the filtration room: shows the state of every filter, guides the change, logs the element installed (consumable reference and lot via QR) and ties all the evidence to the product lot. It works without a network.

See the full IRIS architecture →

Before and after

Filter changes by the clock vs. a pressure endpoint with iLEAN

Aspect	Fixed interval + unwatched gauge	With iLEAN (pressure endpoint)
Change criterion	Hours or batches, same for every product	Real differential-pressure curve, per product
Filters discarded half-used	Frequent — the interval is conservative	Every element used up to its real window
End of life stretched too far	Haze and gels slip into packaging	Warning ahead of collapse, with lead time
Burst or channeled filter	Discovered at QC or in the complaint	Anomalous drop caught at once, lot flagged
Curve-to-lot link	Does not exist	Every lot tied to the curve of its filter
Change record	A sheet, if anyone fills it in	Element, date, operator and pressures, automatic

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.

• Varnish, resin, ink or coatings plant with finishing filtration through bag, cartridge or plate filters before packaging.
• Pilot on one filtration line (pressure readings + per-product curves + Edge in the room). First value expected within a few weeks.
• Indicative payback between 5 and 10 months, depending on annual filter element consumption, the cost of change downtime and the history of haze or particle complaints.
• Hard levers: ≥ 20% reduction in filter consumption by using real element life; bursts and channeling caught the moment they happen; the filtration curve archived per lot as evidence against claims.

And the production manager's reasonable doubt

“What if the system stretches a filter and a hazy lot slips through?” — a healthy doubt, and the answer is the architecture: iLEAN does not stretch filters to save money, it warns based on the curve, and the threshold for each product is set by your quality team, not by the model. Hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI merely follows a measured pressure against a product's fouling signature, 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 decide. The three safety rings exist precisely for this.

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