- Agree downtime reasons and state semantics with operators before instrumenting.
- Time alignment across PLCs, gateways and MES is essential for credible OEE.
- Edge buffering protects the data path from ordinary network failures.
Define states with the people who run the line
OEE depends on availability, performance and quality definitions that match actual production. A machine can be running but not producing good parts; a line can be stopped for a planned changeover or an unplanned fault. We agree state semantics, reason codes and shift calendars with operators and supervisors before building dashboards.
Signals are mapped to an asset hierarchy with stable identifiers. A tag without context is a number. The asset model connects line, cell, machine and sensor so downtime can be analysed at the right level.
- Agreed state and reason-code taxonomy
- Shift and calendar definitions
- Asset hierarchy and stable tag IDs
- Quality and performance calculations documented
Engineer time and quality
PLCs, gateways, MES and cloud services have different clocks. We define time synchronisation, sequence numbers and latency expectations. Late events are handled according to policy rather than silently shifting a downtime window. Quality flags distinguish good data, estimated data and missing data.
Edge buffering with store-and-forward prevents ordinary network interruptions from becoming permanent gaps. The buffer has a clear retention and drop policy, and drops are visible to operations. This is more trustworthy than a dashboard that quietly interpolates missing hours.
- Clock sync and sequence policy
- Late and out-of-order event handling
- Explicit data-quality flags
- Visible buffer retention and drops
Close the loop
An OEE dashboard should lead to action: a reason code review, a maintenance task, a changeover improvement or a quality investigation. Alerts need context and an owner. Reports should compare shifts and lines with the same definitions, not silently change the formula between releases.
Pilot-to-fleet rollout needs device provisioning, remote diagnostics, signed updates and a support model. The plant should be able to operate the system without depending on the original project team for every change.
- Action-oriented downtime and quality workflows
- Context-rich alerts with owners
- Versioned metric definitions
- Fleet provisioning, updates and support
Questions we hear in the room
Can you read data from older machines?
Often, through PLC, Modbus, OPC UA or sensor gateways. We assess the controller, network and safety boundary before proposing an approach.
Do we need to replace our MES?
Not necessarily. We can integrate operational data with the existing MES or build the missing visibility layer around it.
