DataOps
How historian compression distorts machine learning
The series in a historian is not the series the sensor saw. Compression settings change what features and models learn.
Check data quality before calculating OEE
Availability, performance and quality must use consistent production boundaries.
In short
- OEE is availability multiplied by performance and quality, and all three inputs must cover the same shift and asset.
- Run time cannot exceed planned time, and good count cannot exceed total count.
- Performance above 100% should prompt an inspection of the ideal cycle time and the counts rather than silent clipping.
Fix the production boundary
OEE equals availability multiplied by performance and quality. Planned time, run time and counts must cover the same shift and asset. Document how planned stops are treated. Comparing lines with different boundaries may be misleading.
Keep units explicit
Availability is run time divided by planned production time. Performance is ideal cycle time multiplied by total count, divided by run time. Quality is good count divided by total count. Convert run time to seconds when the cycle uses seconds. Quality is undefined when total count is zero.
Do not hide impossible values
Run time cannot exceed planned time and good count cannot exceed total count. If performance exceeds 100%, investigate the ideal cycle and counts instead of silently clipping it. Duplicate counter readings and counter resets can look like unexpected production changes.
Connect the result to improvement
Low OEE does not diagnose a fault by itself. Interpret it alongside stop reasons, changeovers and quality losses. Start by reconciling one shift with operator records. Then inspect the table in Forge and reproduce the calculation in the lab using the same inputs.
Frequently asked
How do I calculate availability, performance and quality for OEE?
Availability is run time divided by planned production time. Performance is ideal cycle time multiplied by total count, divided by run time. Quality is good count divided by total count, which is undefined when total count is zero. Convert run time to seconds when the cycle is in seconds.
What should I check first when OEE results look wrong?
Reconcile one shift with operator records first. Then inspect the table in Forge and reproduce the calculation in the lab with the same inputs. A low OEE does not diagnose a fault by itself, so read it together with stop reasons, changeovers and quality losses.
Sources and further reading
Basic ratio calculations and illustrative examples authored by Kutup.
Pilot steps in this guide are proposed engineering workflows. Examples do not constitute measured field performance.
Questions on this topic
Keep reading.
Full library
DataOps
Prepare OPC UA data for AI
Turn a tag value into a usable observation with units, quality, timestamps and asset context.
DataOps
Time-Series Data in Industry: Storage, Trends, and Forecasting
From historians to machine learning, time-series data powers industrial analytics. A practical guide to collection, quality, and the forecasting methods that actually work on plant data.