Process historian and SCADA
The continuous record of process values: temperatures, pressures, flows, speeds and states, sampled constantly.
01 What it holds
The data inside it
Time-series tags at very high frequency, with compression and interpolation rules that change what a query returns depending on how it is asked.
Tag, Sample / Event, Asset framework element, Unit of measure, Compression setting
02 Getting the data out
Extraction, and the keys that decide whether it joins
Native APIs, OPC UA, or an export to a time-series store. Aggregation must be specified — average, time-weighted average, minimum, maximum — because the raw signal is irregular by design.
Tag names are the only identifier and are usually built from a site naming convention. Mapping tags to an asset and to a product run is manual in most plants, and it is the entire value of the data.
03 What goes wrong
The failures that look like success
Each of these produces a number that is plausible, reconciles to something, and is wrong. They are worth reading before the first extract, not after the first dispute.
Compression is lossy on purpose
Historians store exceptions rather than every sample, so two different queries over the same period legitimately return different averages.
Time-weighted versus simple average
Irregular sampling makes a simple average wrong. The distinction matters most exactly where the process is unstable.
Tag naming carries the model
Without an asset framework, the relationship between a tag and a machine lives in a convention and in one engineer's head.
No product context
The historian knows the temperature; it does not know which batch or product was running. That join is what turns process data into quality and yield analysis.
04 Joins that matter
What this system is worth joining to
| Join to | What it lets you answer |
|---|---|
| MES | to attach process conditions to a batch and a work order |
| QMS | to correlate process excursions with quality failures |
| CMMS | condition data as the leading indicator of a failure |
05 Metrics it feeds
The numbers that come out of it
06 Questions
Frequently asked
Should historian data go into the warehouse?
Selectively. Raw high-frequency tags rarely belong there; features derived at the right grain — per batch, per run, per shift — almost always do.
What makes this data useful rather than merely large?
The asset and batch context. A tag series with no link to what was being made is telemetry; the same series joined to batch and quality result is a yield model.
Connect this system without replacing it
SCIKIQ reads it in place, resolves the entities inside it and governs what the fields mean.