The returnable bottle washer starts to talk

In a plant running returnable glass, the bottle washer is at once the most critical and the oldest asset on the floor. Fifteen to twenty-five years old, a proprietary panel with no network, and on its screen everything that determines container safety: caustic soda concentration, temperature per zone, contact time, water and steam consumption.

‹ See all cases of beverage bottling

Fixed industrial camera pointed at the panel of a returnable bottle washer, showing caustic concentration and temperature by zone
The problem

The most critical machine on the floor is also the oldest.

Replacement costs a fortune and weeks of line downtime, so the sensible decision is to keep it. The price of keeping it is that its data vanishes the moment the screen changes. Today someone writes those parameters onto a sheet by hand every couple of hours, which is a very poor sample of a continuous process.

  • Its screen holds everything that determines container safety: caustic soda concentration, temperature by zone, contact time, water and steam consumption.
  • A proprietary panel with no network, fifteen or twenty-five years old. Replacing it costs a fortune and weeks of line downtime, so the sensible decision is to keep it.
  • The price of keeping it is that its data is lost the moment the screen changes.
  • Today somebody writes those parameters on a sheet every two hours: a very poor sample of a continuous process, and the only record you can show.
How it fits the IRIS system

Connect in photo mode on the panel — an external camera and nothing else.

An external camera and nothing else. Connect in panel-photo mode solves this without touching the machine: a fixed industrial camera aimed at the screen, periodic capture, and a grounded language model that reads each frame and structures the time series. No integration, no PLC to touch, no manufacturer warranty to void. The washer stays exactly as it is, but now it talks. And what it says can be cross-checked. iLEAN automatically links each stretch of the series to the batch and format running at that moment. That enables something impossible today: correlating a drop in caustic concentration with empty bottle inspector rejects, or with an off-taste complaint that will arrive days later.

The machine is not touched: no port is opened, nothing is installed on its control system and the warranty stands. That is the difference between a case approved in two weeks and an integration project that never rises up the list.

See the full IRIS architecture →

Before and after

Replacing the washer versus reading its panel

AspectReplacing itWith iLEAN Connect
InvestmentHundreds of thousandsIndustrial camera and commissioning
Line downtimeWeeksNone: installed while running
Hygiene recordA note every two hoursContinuous series per batch
Correlating rejects with a parameterImpossibleDirect, batch by batch
Water and steam consumptionShows up on the billAttributable to the batch
The machine's control systemUntouched

Impact estimate

Estimated impact — to validate with your own numbers.

The block below is an estimate to be validated against your plant's actual data. We put it forward so the committee has an order of magnitude; we refine it during the assessment.

  • Estimated payback 5-10 months, with three non-overlapping sources of return.
  • Replacement capex in the hundreds of thousands deferred indefinitely.
  • Reduced bottle rejects attributable to caustic or temperature drift.
  • Complaints about poorly sanitized containers cut before they escalate.

The return has three legs: avoiding or indefinitely deferring a replacement CAPEX in the hundreds of thousands, reducing bottle rejects, and cutting container-related complaints before they escalate. Estimated payback between five and ten months. Estimate to be validated.

And the fair question from the production manager

«Does a camera watching a screen count as a hygiene record?» — here the task is anchored: fixed screen, fields in known positions and expected physical ranges, where the best models drop below 1.5% error [1]. And the system discards any frame whose value falls outside the possible range instead of accepting it, which is what would invalidate the series.

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

Frequently asked questions

What people ask about reading the washer panel

Does the washer have to be networked?

No, and that is the point. The camera looks at the panel from outside: no ports, no software on the machine and no effect on the manufacturer's warranty.

What if the panel cycles through screens?

That is designed for: capture cadence is set to the rotation and the system identifies which screen it is looking at before extracting any value.

Does it replace the manual hygiene log?

It feeds and extends it. The record goes from a note every two hours to a continuous series, which is what lets you demonstrate the real contact time of each batch.

Does it work for the other old machines on the floor?

Yes: the same pattern serves the pasteurizer, the blow molder or the cooling tunnel. You start with the washer because that is where the food safety risk sits.

How much history is needed?

To investigate a specific batch, having it captured. To detect caustic drift, some weeks — the time the trend takes to draw itself.

Let's talk

Tell us how often the washer parameters are written down today.

We work on your plant's real data, not ours. Assessment with no commitment.

Request estimated ROI within 48h ‹ See all cases of beverage bottling See food