real-time operational insight across plants
machine, quality, and energy data from six plants unified into one live view — with forecasts that flag trouble a shift ahead.
- client
- a packaging manufacturer, 6 plants across 3 countries
- sector
- manufacturing ↗
- duration
- 7 months
- team
- 6 engineers, 1 data engineer, 1 ml engineer
each plant ran its own historian, mes, and reporting. oee was calculated four different ways, and group leadership saw performance a week late, averaged into meaninglessness.
maintenance was calendar-based, so lines either stopped unexpectedly or were serviced when nothing was wrong.
28 weeks, 4 phases, one team.
edge gateways and a common data model across six historians.
one oee definition, live dashboards per line, plant, and group.
failure models for the three highest-cost assets per line.
alerts wired into maintenance work orders, team trained.
- four oee definitions
- weekly, averaged reporting
- calendar-based maintenance
- plant data locked on-site
- one oee, live
- line-level visibility for everyone
- condition-based maintenance
- cloud platform with edge buffering
for the first time, the plants argue about what to do — not about whose numbers are right.
plant managers and group leadership look at the same numbers, live. predicted failures arrive as work orders with enough lead time to schedule the fix between runs.
the model and data platform are now the base for the group's energy and esg reporting.
- opc ua
- mqtt
- go
- timescaledb
- kafka
- python
- grafana