Tunnel kiln temperature in glazed ceramic firing — stable shade and flatness between loads.

The problem

The SCADA draws the setpoint; the tile pays for the real curve.

The glazed ceramic firing defect is born from a crossover that is almost never in the same place:

1. The real zone-by-zone curve of the tunnel — not the PLC's, the one each thermocouple measures. A side burner losing gas, an infiltration through the muffle, a drifting thermocouple — and the SCADA setpoint and reality no longer match.
2. The glaze recipe — composition, particle size, viscosity, thickness. It usually lives in the kiln master's spreadsheet or in the ERP, but it is rarely cross-checked against the real curve.
3. Final flatness and shade — exit profilometry and colorimetry from the sorting machine. They find it once the tile is already fired and the defect is scrap or a downgrade to seconds.

The kiln master carries half of that in his head and compensates by eye between one load and the next. When he retires, his judgement leaves with him — and the kiln gives no warning. It is the pattern the ceramic sector saw in the Marmaris case decades ago: star product at the trade fair, massive orders, and in high-volume production the stresses between the glaze and the body warped the tile as soon as it touched water. The solution back then — lasers at the kiln exit plus a semi-expert system that anticipated the deformation 30 minutes ahead by reading every thermocouple — is exactly what iLEAN Edge industrialises today.

How it fits the IRIS system

iLEAN Edge does not touch the kiln — it seals the crack between curve, recipe and flatness.

The firing problem is not a lack of data: what is missing is data cross-checked at second zero. The SCADA has the setpoint, the thermocouples measure, the glaze lives in a recipe, flatness and shade get measured at the end. Four islands. iLEAN Edge acts as the filler between all of them, without asking you to change the tunnel kiln, the SCADA or the sorting machine.

Edge measures the real curve per zone. Connect captures the glaze recipe wherever it comes from — the ERP or the kiln master's spreadsheet. The agent cross-checks it against exit flatness and shade, anticipates the drift and alerts with margin. The person signs off — never the other way round.

The three iLEAN pieces applied to the glaze tunnel kiln:

• Edge — terminal with its own thermocouples in the critical zones of the tunnel (preheating, firing, cooling), in parallel to the burner control loop. It also reads exit laser flatness and the colorimetric shade from the sorting machine. It works with no network.
• Connect — captures the glaze and body recipe wherever it lives: the ERP, the MES or the spreadsheet the kiln master updates every morning. And it captures what comes from outside: a change of frit supplier, a new specification from the distributor.
• Agent — cross-checks real curve, recipe, flatness and shade. It anticipates when thermal drift is pushing the batch towards a downgrade and proposes the curve adjustment to the kiln master through whichever channel he uses. The master decides; the tunnel does not readjust itself.

See the full IRIS architecture →

Before and after

Tunnel run by the master's eye vs. tunnel with an auditable iLEAN curve

Aspect	Classic firing + SCADA + the master's eye	With iLEAN Edge + Connect + Agent
Tunnel curve	The one the SCADA draws — the setpoint	Real curve per zone, measured in parallel
Cross-check with glaze recipe	The master does it by eye	Recipe linked to the batch from ERP/spreadsheet
Shade drift alert	The sorting machine detects it once already fired	Anticipation with margin to adjust the curve
Flatness / edge curl / crowning	Defect found at the end, tile already scrap	Trend detected before the defect is born
The master's knowledge	It leaves with him the day he retires	Captured as a pattern in the system
Operation with no network	Depends on the SCADA server	Edge keeps measuring on the panel's power

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 set it out so the committee has an order of magnitude; we refine it during the diagnostic.

• Glazed wall tile or glazed porcelain stoneware factory, 1-2 tunnel kilns, 8-12 SKUs in rotation, documentable historical downgrades from shade or flatness.
• Edge pilot on one tunnel (plant thermocouples in the critical zones + integration with the glaze recipe + alert to the kiln master). First value expected within a few weeks.
• Expected reduction in downgrades and scrap from shade or flatness drift of ≥ 30% over the baseline.
• Indicative payback between 4 and 9 months, depending on incident frequency and the average downgrade cost of each SKU.
• Hard lever: a single load rescued per firing on an expensive SKU pays for weeks of project.

And the plant manager's reasonable doubt

“What if the AI proposes adjusting the curve and gets it wrong?” — hallucination is a problem of free generation, not of anchored tasks. On tasks where the AI limits itself to checking a measured value against a recipe and a history (reading the real curve and comparing it with the theoretical one + flatness + shade), the best models brought the error below 1.5% ^[1]. And even so, what is critical is not decided alone: iLEAN proposes the adjustment and the kiln master signs off. The three safety rings are there precisely for this.

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