The challenge
Each plant ran its own historian with incompatible schemas. Group-level reporting was assembled by hand each month and was obsolete by the time it circulated.
What we did
A streaming ingestion layer normalises telemetry into a governed lakehouse with per-plant lineage. Failure-prediction models run against the unified feed and surface directly in the maintenance scheduling system.
The result
- 26
- Plants unified
- -34%
- Unplanned downtime
- 1.2TB
- Ingested daily