How to Keep Instrument Calibration Records That Support Plant Data
A calibration record that supports plant data needs eight things: the instrument tag, its calibrated range, the method and reference used, the as-found values, the as-left values, the date, the next due date, and the person who did the work. Keep these records in one place, using the same tag names as the historian. Then link each calibration event to the historian so anyone reading a trend can tell which periods the reading was verified for and which periods are open to question.
Why do calibration records matter to people who use plant data?
Most people who read a trend, a dashboard or a production report never see the instrument behind it. They assume the number is right. A calibration record is the evidence that tells them whether that assumption holds.
When an instrument fails a check, someone usually has to ask which reports, batches or decisions depended on it, and over what period. That question can only be answered if the records show when the instrument was last confirmed good, what condition it was found in and what was changed. For the investigation side, see detecting sensor drift in plant data.
Records kept only to satisfy a maintenance schedule often lack the detail needed to answer it. The fields below are chosen so the record serves both maintenance and data users.
What should every calibration record contain?
| Field | What to record | Why it matters for data users |
|---|---|---|
| Instrument tag | The tag exactly as it appears in the control system and historian | Lets the record be matched to the data without guesswork |
| Location and service | Where the instrument is installed and what it measures | Confirms the right physical device was checked |
| Calibrated range and unit | Lower and upper range values with engineering unit | Shows whether the range matches the scaling downstream |
| Acceptance limit | The allowed error and who approved it | Defines what "pass" meant at the time |
| Method | Procedure followed, in situ or on the bench, test points used | Shows how far the result can be relied on |
| Reference instrument | Identity and calibration status of the reference | Supports traceability of the result |
| As-found values | Readings at each test point before any adjustment | Shows the condition the data was produced under |
| As-left values | Readings at each test point after adjustment | Shows the condition going forward |
| Result | Pass, fail, adjusted, replaced or removed from service | Drives the data quality decision |
| Date and time | When the work started and finished | Marks the boundary in the historian |
| Next due | The next scheduled check | Shows when confidence is due for renewal |
| Performed by and approved by | Names or roles of the technician and reviewer | Makes the record accountable |
| Notes | Repairs, range changes, replacements, anything unusual | Explains changes that affect the data |
The as-found values matter most to data users and are the field most often skipped. If a technician adjusts first and records only the as-left result, nobody can tell whether the readings before the check were accurate.
What does a usable record template look like?
The template below is illustrative. Adapt the fields and limits to your own procedures and the requirements of the measurement owner.
| Field | Entry |
|---|---|
| Instrument tag | TT-2104 |
| Location and service | Tank 2 outlet, product temperature |
| Calibrated range and unit | 0 to 150 °C |
| Acceptance limit | ±0.5 °C (set by measurement owner) |
| Method | In situ comparison at three points, site procedure reference |
| Reference instrument | REF-07, calibration current at time of use |
| As-found (reference / reading) | 20.0 / 20.3 · 80.0 / 80.6 · 140.0 / 140.9 °C |
| As-left (reference / reading) | 20.0 / 20.1 · 80.0 / 80.1 · 140.0 / 140.2 °C |
| Result | Failed as found, adjusted, passed as left |
| Start and finish | Date and time of isolation and return to service |
| Next due | Per agreed interval |
| Performed by / approved by | Technician role / maintenance lead role |
| Notes | Historian flag raised for interval since previous pass |
All values in this example are illustrative. They show how to lay out a record, not what limits or intervals to use.
How do you link calibration records to the historian?
The aim is that someone looking at a trend can see the calibration status of the reading without leaving the screen or asking maintenance. There are several ways to do this. Choose one based on what your historian and maintenance systems support.
Annotate calibration events on the tag. Most historians allow events or annotations against a tag. Add one for each calibration, showing the date, result and a reference to the full record. This is the simplest option and works with almost any system.
Write a quality status. Where the historian supports quality flags or a companion status tag, set the reading to a "suspect" or "under review" status for the interval between the last pass and a failed as-found result. Clear it only after a documented review.
Mark the out-of-service window. During calibration the reading may be isolated, simulated or held. Mark that window so calculations and reports do not treat test values as process data.
Share one tag identity. Use the same tag name in the calibration system, the control system and the historian. If names differ, keep a maintained cross-reference. See standardising engineering units in plant data for the related problem of units and scaling.
Align the clocks. A calibration window is only useful if its times match the historian's timestamps. The guide to aligning timestamps across plant data systems covers this.
Whatever method you use, keep the raw historian values. Marking a period as suspect is a label, not a deletion. Check that your historian retention rules keep annotations and quality flags for as long as the data itself.
What should happen when an instrument fails as found?
A failed as-found result means the data produced since the last good check may not be reliable. Treat that as a data question as well as a maintenance one.
- Record the as-found values before any adjustment.
- Flag the interval from the last passing check to the failed check as under review in the historian.
- Identify the reports, dashboards, batches and control decisions that used the reading in that interval.
- Ask the measurement owner to decide what treatment applies: accept with a note, exclude, or investigate further.
- Record the decision and who made it, and link it to the calibration record.
Do not assume the error was present for the whole interval, and do not apply a straight-line correction between the two results unless an approved technical assessment supports it. Where the start of the error cannot be established, record that uncertainty. If the instrument supports a safety function, involve a qualified assessor before deciding how operation and records should be handled.
What are the common mistakes?
Recording only as-left values. This hides the condition the data was produced under.
Using different tag names in different systems. Records that cannot be matched to the historian give data users no help.
Leaving the range unchecked against scaling. If the instrument was re-ranged but the controller or historian scaling was not updated, the calibration passes and the data is still wrong.
Letting test values enter reports. Simulated or held values during calibration should be marked, not averaged into a shift report.
Keeping records on paper only. Paper records can be valid, but they are hard to link to the historian. If you keep paper, add at least a dated event on the tag.
Ignoring overdue instruments. An instrument past its next due date has not failed, but its data no longer carries the same confidence. Show overdue status where data users can see it.
When does this approach not apply?
Not every signal needs this level of record. Status contacts, indication-only gauges and values that feed no decision may be checked by function rather than calibrated. Agree with the measurement owner which instruments are in scope, based on what depends on them. Where an instrument is subject to regulatory or customer requirements, follow those requirements and your quality system; this guide does not replace them.
Before startup on new or modified installations, a loop check confirms the signal path, which is different from calibration. See performing instrument loop checks before startup.
Frequently asked questions
Who should own calibration records?
Usually maintenance or the instrument team owns the records, while a named measurement owner sets acceptance limits and decides on data treatment after a failure. Make both roles explicit.
How often should instruments be calibrated?
Set intervals with the measurement owner based on the consequence of error, operating conditions, manufacturer guidance and previous as-found results. Review the interval after a failed check.
Do we need a separate calibration management system?
Not necessarily. What matters is that records are complete, use the same tag names as the historian and can be linked to it. A dedicated system can make that easier as the number of instruments grows.
Should the historian store calibration data itself?
Store the event, result and a reference to the full record against the tag. Keep the full record in the system built to manage it, so there is one authoritative copy.
If you want help connecting calibration records to your historian, see our services.