Machine data for OEE

What you will be able to do

• You can derive availability from a state machine and defend the boundaries.
• You can separate planned from unplanned downtime, correctly, at the source.
• You can get trustworthy counts, including across a counter rollover.
• You can collect downtime reasons as events instead of inferring them.
• You can state which OEE inputs must come from ERP or MES, and why.

Syllabus

1. 1
   Define the three OEE terms precisely

   Availability, performance and quality, with the denominators written out, because most disagreements about an OEE number are disagreements about a denominator.

   Asset One shift of real machine data with OEE computed three defensible ways, and the gap between them

   8 min
2. 2
   Take availability from a state machine

   Not from 'is the spindle turning'. MachineryItemState and the production state machines give you exact durations with transition events.

   Asset A state timeline built from transition events, next to one inferred from 1 Hz sampling

   9 min
3. 3
   Split planned from unplanned with operation mode

   The single most common OEE modelling error. Without operation mode you cannot separate setup and maintenance from breakdown, and the resulting number is wrong in a way nobody can debug.

   Asset The same shift with and without operation mode, showing a 14-point availability swing

   10 min
4. 4
   Count from a counter

   MachineryOperationCounterType, ProductionStatisticsType, PackMLCountDataType, and the rollover and reset handling that edge-detecting a boolean will never give you.

   Asset A counter across a rollover, with the naive delta calculation and the correct one

   9 min
5. 5
   Get performance right

   Ideal cycle time, where it actually lives, and why comparing cycle times across program revisions without the revision is meaningless.

   Asset Cycle-time data for two program revisions, with the misleading comparison and the fair one

   8 min
6. 6
   Get quality from the result model

   Machinery Result, ResultEvaluationEnum, and scrap versus rework versus rejected-at-inspection.

   Asset Result data from a machine mapped into a quality calculation, with the definitions stated

   8 min
7. 7
   Collect reasons as events

   Part 9 conditions with ConditionClassId, filtered server-side. Timestamped reasons attribute cleanly; reasons inferred from sampled state do not.

   Asset A downtime report built from condition events, next to one built from sampled state

   9 min
8. 8
   Decide where aggregation happens

   Server-side aggregates against downstream computation: cost, fidelity and what you can still re-derive later if you choose wrong.

   Asset The same shift aggregated at the server and at the platform, with data volume and cost for each

   8 min
9. 9
   Name the boundary honestly

   Planned production time, shift calendars, orders and standard cycle times are not in the machine. Machinery Jobs and ISA-95 Job Control V2 are the intended bridge, and this is where projects stall.

   Asset A data-source map for a full OEE calculation, with each input traced to machine, MES or ERP

   15 min

Lab

Last verified 7 September 2026 against OPC UA 1.05.07. OPC UA Connect is an educational resource and is not affiliated with the OPC Foundation.