Skip to content
Zester
A machine on a dim factory floor lit in green
All applications

Ceramics A vitrified tile plant

Holding the kiln curve and the tile face in ceramics

Kiln sensors and surface vision tied together, so a defect on the sorting line points back to the firing zone that caused it.

Stage
Reference design
Sector
Ceramics
Capabilities
Industrial IoTComputer VisionEdge AI
Deployment
Edge gateway at the kiln + line-scan camera at sorting. Works offline.
Integrations
Kiln PLC (Modbus / OPC UA), WhatsApp, shift report
ZONE TEMPZ1Z2Z3Z4Z5Z6SORTINGDEFECT → ZONE
Signal path. Illustration, not a site photograph.

The knowledge at risk

It lives in the head of the kiln operator.

He knows which zone to nudge when the weather turns or a new glaze batch arrives. His corrections are made by feel and recorded nowhere.

Shop-floor constraints

  • A roller kiln is long and slow: cause and effect are an hour apart.
  • High line speed at sorting.
  • Existing PLCs vary in age and protocol.

What we build

  1. 01

    Read the kiln

    Zone temperatures, fan speeds and roller speed are logged from the PLC every few seconds.

  2. 02

    Log the operator's nudges

    Each setpoint change is captured with a reason chosen by voice or tap.

  3. 03

    Inspect the face

    A line-scan camera grades surface defects at sorting: pinholes, cracks, shade drift.

  4. 04

    Link cause to defect

    Tile position in the kiln is time-shifted to match its grade, so defects map to zone conditions.

Today: what it improves

  • Defects traced to a zone within the shift, not after the batch.
  • Consistent grading across shifts and sorters.
  • Alerts when a zone drifts outside the operator's usual band.

Over time: how it learns

Firing curves paired with the result on every tile, so each batch starts from what worked.

  • D1Zone temperature and roller-speed time series
  • D2Operator setpoint changes with stated reason
  • D3Surface images and defect grade per tile
  • D4Glaze batch, body mix and ambient context

Recorded on site and owned by the plant. The system uses it to keep learning your floor.

Architecture

  1. 01

    Capture

    PLC tap, ambient sensors, line-scan camera at sorting

  2. 02

    Edge

    Gateway: time alignment, defect inference, offline buffer

  3. 03

    Model

    Surface defect classifier; drift detection on zone curves

  4. 04

    Record

    Tile-level join of firing conditions and grade

  5. 05

    Surface

    Kiln panel, WhatsApp alerts, shift quality report

What we will measure

  • First-grade yield rises.

    PILOT MEASUREMENT REQUIRED: premium-grade share per shift before and after.

  • Defects reach the operator sooner.

    PILOT MEASUREMENT REQUIRED: time from defect onset to corrective setpoint.

  • Grading is consistent.

    PILOT MEASUREMENT REQUIRED: agreement between camera and manual sorters.

Run a plant with an expert like this?

Tell us whose judgement your floor cannot run without. That is where a pilot starts.