Define observable states first
Describe running, waiting, setup, maintenance and connection loss with engineering staff. A source may indicate power or rotation rather than productive output. Keep measured signals, inferred states and human-confirmed reasons separate. Agree sampling, clock accuracy and the meaning of missing intervals before drawing utilisation charts.
Collect and retain technical events
Adapters, gateways, queues, history and area views can form the solution. OPC UA, Modbus, MQTT, HTTP or files depend on actual interfaces, documentation and licences. A protocol label does not establish model compatibility. Verify registers, units, counters, timestamps and polling constraints. Use manual recording or an approved alternative when safe reading is unavailable.
Downtime, alerts and meaningful metrics
Define downtime boundaries, reasons and correction rights. Alerts need recipients, repetition rules and acknowledgement rather than reacting to every signal change. OEE requires agreed calendars, planned time, good output and speed standards. A connected sensor alone cannot establish a valid calculation; expose uncertain intervals and data quality.
Monitoring is not protective control
Applications and networks can fail. They do not replace safety PLCs, emergency stops or protective interlocks. Separate read access from control networks as agreed. Parameter writes or remote commands require a separate engineering project and authority. Protect credentials and site details; show connection loss as unknown rather than running.
Bench test and controlled pilot
Inventory models, sources and permissions, then verify one channel safely. Compare events against an engineer’s observations. Test gateway restarts, outages and queue capacity. Document maintenance, backup and signal-map changes before expanding.
- Define each signal’s source, units and meaning.
- Distinguish outages from downtime.
- Deduplicate repeated events.
- Reconcile clocks and intervals.
- Show stale data and unconfirmed reasons.
Project questions
Can older machinery be connected?
Sometimes through an existing interface or approved external sensor. Assessment, documentation and safe testing are prerequisites.
Can monitoring stop dangerous equipment?
Ordinary monitoring software is not a protective system. Safety stopping remains a separately engineered responsibility.
Is OEE available immediately?
Not reliably without agreed time, performance and quality definitions and verified source data. State history can be a useful first step.
Illustrative image