Steam curing chamber control — a ramp that drifts two degrees costs three megapascals.

The problem

The curing curve is followed by the chamber lead's eye — and on the odd days it shows.

The ideal steam curing curve for a concrete piece is well known: EHE-08 and EN 13369 set the range, and the lab confirms it with cylinder tests at 24 hours and 28 days. But between the curve in the manual and what happens on the shop floor there is a world of difference:

1. Chambers are loaded mixed — beams, slabs and joint covers in the same load. Every thickness responds differently. The "average" curve is nobody's curve.
2. The shift change lands right at the critical moment — opening the valve and starting the plateau depend on the incoming operator reading the outgoing operator's handover correctly. Holiday Mondays are the odd days.
3. The chamber is opened when there is a load waiting — not when the curve has finished cleanly. Thermal shock, microcracking, and a week later a non-conforming cylinder test.

The classic system works because good people watch the chamber and because cylinder tests catch the worst batches. But microcracking is not caught by the 24-hour cylinder test: it is caught by the site inspector when the piece cracks on site. What fails is not the chamber: it is knowing which curve each piece has seen.

How it fits the IRIS system

iLEAN does not replace the chamber lead — it holds the curve with the same patience all night long.

The curing problem is not one of judgment: the chamber lead knows perfectly well how the ramp should run. The problem is one of constant presence and memory per load. Nobody can watch thermocouples for eight hours straight, nor remember the exact curve seen by Thursday's two-in-the-morning load. iLEAN acts as the putty that seals the crack between the curve in the manual, the steam solenoid valve and the piece's dossier — without asking you to change the chamber, the boiler or the MES.

Edge tracks the actual curve against the target. Connect reads the load sheet (which pieces are inside and what thickness). The agent decides whether the load is released or held. The person signs — Edge does not open the chamber on its own.

The three iLEAN pieces applied to steam curing:

• Edge — an industrial terminal connected to the chamber's thermocouples, the hygrometer and the steam solenoid valve. It tracks the actual curve, predicts the drift and proposes the correction (more steam, wait for the plateau, delay opening). It works with no network. If the plant loses WiFi, Edge keeps controlling.
• Connect — captures the load sheet (which pieces, which concrete, which mix design, which is the most sensitive thickness) from the MES or from the lab spreadsheet, plus the operator's reports dictated by voice whenever a manual adjustment is made.
• Agent — cross-references the actual curve with the expected strength threshold for that mix design, and if the curve drifted beyond the limit, it holds the load before the piece is labeled. The quality manager decides: an extra cylinder test, waiting for 28 days, or reclassification. The chamber does not open on its own.

See the full IRIS architecture →

Before and after

Curing by the operator's eye vs. curing held steady with iLEAN Edge

Aspect	Manual control + 24-hour cylinder test	With iLEAN Edge + Connect + Agent
Curve tracking	A one-off reading at the shift change	Actual curve logged every minute, point by point
Shift change	Verbal handover, lost a couple of Mondays a year	Chamber status visible, handover recorded
Mixed load	Average curve, most sensitive piece at the limit	Curve tuned to the most sensitive declared piece
Microcracking from thermal shock	Shows up on site months later	Opening held back if the drop is too abrupt
Dossier per load	Non-existent or rebuilt by hand	Actual curve per load + pieces inside, archived
Operation with no network	n/a	Edge keeps controlling on the panel's power alone

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.

• A plant with two steam curing chambers and mixed production of beams, hollow-core slabs and joint covers — multi-SKU loads.
• Edge pilot on one chamber (terminal + calibrated thermocouples + integration with the solenoid valve and the load MES). First value expected within a few weeks: the actual curve logged per load and a preventive alert at shift changes.
• Indicative payback between 4 and 9 months, depending on how often cylinder tests come back non-conforming and how often the site inspector raises claims for late cracking.
• The hard lever is a single load held and rescued before it reaches the customer with doubtful strength, plus a reduction ≥ 30% in scrap from microcracking detected in the yard.

And the chamber lead's reasonable doubt

"What if the AI decides to open or close the solenoid valve without me knowing?" — the answer is: it does not decide alone. iLEAN's operating rule in critical OT is assist and simplify, never replace. Edge proposes the correction, the chamber lead validates it with one click, and the chamber only opens once the person signs off. In tasks where the AI simply reads thermocouples and compares them with a target curve, the best models brought error below 1.5% ^[1]. Safety in critical OT is not delegated: the three rings are there precisely for this.

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