The challenge
The most useful data in a factory is trapped in the machines: in PLCs, drives, robot controllers and sensors that only raise an alarm once a hard limit is crossed. Maintenance runs on the calendar or on failure, robot utilisation is a guess, and OEE is worked out in a spreadsheet at the end of the week.
Our approach
Read the line directly and let it report its own health. An edge gateway on site speaks the plant's own protocols, every reading lands in one place, and models look for the drift that comes before a failure rather than waiting for the alarm.
- Connected over OPC UA, Modbus TCP and MQTT to PLCs, drives, scales and discrete sensors
- Temperature, vibration, pressure, humidity, current and flow tracked per asset
- Robot fleets in view: arms, delta and SCARA robots, conveyors and AGVs, with cycles, utilisation, faults and emergency stops
- Anomaly detection over every stream, so drift is flagged before a limit is breached
- Predictive maintenance from vibration and current trends plus failure history, giving a recommended window
What we built
The connected asset layer of WAJD Forge: edge connectors with circuit breakers, a per site MQTT topic structure, live OEE, MTTR, MTBF and takt time line by line, and a robot fleet board. Anomaly detection uses rolling statistics and isolation forests, and maintenance risk turns into a planned window rather than a breakdown.
Where it stands
Built in WAJD Forge. The connectors are tested against a real Modbus TCP server and our own MQTT broker, and the plant data is simulated, so moving to real hardware is a connector setting, not a rewrite. WAJD Forge is open to design partners.